{"global_step": 1, "acc_step": 0, "speed/wps": 2876.5455804482713, "speed/FLOPS": 45180101128790.2, "speed/curr_iter_time": 1.5755, "speed/data_load_time": 0.2067, "optim/grad_norm": 2.017601251602173, "optim/lr": 0.0, "optim/total_tokens": 524288, "memory/max_active_gib": 55.36551237106323, "memory/max_active_pct": 69.94922576842406, "memory/max_reserved_gib": 59.232421875, "memory/max_reserved_pct": 74.83470978786407, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 99706, "loss/out": 11.955284118652344, "created_at": "2025-01-15T05:32:11.855321+00:00"} {"global_step": 2, "acc_step": 0, "speed/wps": 12978.34508821607, "speed/FLOPS": 203842743725466.28, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0482797622680664, "optim/lr": 6.000000000000001e-07, "optim/total_tokens": 1048576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487147, "loss/out": 11.962308883666992, "created_at": "2025-01-15T05:32:21.960834+00:00"} {"global_step": 3, "acc_step": 0, "speed/wps": 12966.948243545607, "speed/FLOPS": 203663740619010.94, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9791204929351807, "optim/lr": 1.2000000000000002e-06, "optim/total_tokens": 1572864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 11.899374008178711, "created_at": "2025-01-15T05:32:32.076425+00:00"} {"global_step": 4, "acc_step": 0, "speed/wps": 12960.038267125816, "speed/FLOPS": 203555209944035.47, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.048267364501953, "optim/lr": 1.8e-06, "optim/total_tokens": 2097152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 11.852862358093262, "created_at": "2025-01-15T05:32:42.194191+00:00"} {"global_step": 5, "acc_step": 0, "speed/wps": 12950.11063194536, "speed/FLOPS": 203399282791526.1, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.0544726848602295, "optim/lr": 2.4000000000000003e-06, "optim/total_tokens": 2621440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 11.731986999511719, "created_at": "2025-01-15T05:32:52.322683+00:00"} {"global_step": 6, "acc_step": 0, "speed/wps": 12942.0042481836, "speed/FLOPS": 203271960895209.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0005, "optim/grad_norm": 2.037708044052124, "optim/lr": 3e-06, "optim/total_tokens": 3145728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491917, "loss/out": 11.552703857421875, "created_at": "2025-01-15T05:33:02.456534+00:00"} {"global_step": 7, "acc_step": 0, "speed/wps": 12925.640932399976, "speed/FLOPS": 203014952535275.03, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0748043060302734, "optim/lr": 3.6e-06, "optim/total_tokens": 3670016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 11.274547576904297, "created_at": "2025-01-15T05:33:12.605610+00:00"} {"global_step": 8, "acc_step": 0, "speed/wps": 12925.45656590015, "speed/FLOPS": 203012056806049.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.0337302684783936, "optim/lr": 4.2000000000000004e-06, "optim/total_tokens": 4194304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 11.071268081665039, "created_at": "2025-01-15T05:33:22.752397+00:00"} {"global_step": 9, "acc_step": 0, "speed/wps": 12929.255195661673, "speed/FLOPS": 203071719506318.16, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9211119413375854, "optim/lr": 4.800000000000001e-06, "optim/total_tokens": 4718592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436080, "loss/out": 10.845499038696289, "created_at": "2025-01-15T05:33:32.897050+00:00"} {"global_step": 10, "acc_step": 0, "speed/wps": 12934.244176709986, "speed/FLOPS": 203150078309260.28, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.924033522605896, "optim/lr": 5.4e-06, "optim/total_tokens": 5242880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 10.703878402709961, "created_at": "2025-01-15T05:33:43.043258+00:00"} {"global_step": 11, "acc_step": 0, "speed/wps": 12935.808854641864, "speed/FLOPS": 203174653726273.22, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.094818353652954, "optim/lr": 6e-06, "optim/total_tokens": 5767168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 10.424379348754883, "created_at": "2025-01-15T05:33:53.182583+00:00"} {"global_step": 12, "acc_step": 0, "speed/wps": 12926.678794263293, "speed/FLOPS": 203031253582010.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0015, "optim/grad_norm": 1.6356112957000732, "optim/lr": 6.6e-06, "optim/total_tokens": 6291456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 10.212084770202637, "created_at": "2025-01-15T05:34:03.329247+00:00"} {"global_step": 13, "acc_step": 0, "speed/wps": 9496.824595230888, "speed/FLOPS": 149160680272637.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8007370233535767, "optim/lr": 7.2e-06, "optim/total_tokens": 6815744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 85301, "loss/out": 10.16910171508789, "created_at": "2025-01-15T05:34:17.168155+00:00"} {"global_step": 14, "acc_step": 0, "speed/wps": 12981.964556963834, "speed/FLOPS": 203899592455821.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.001, "optim/grad_norm": 1.6211696863174438, "optim/lr": 7.8e-06, "optim/total_tokens": 7340032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 10.006213188171387, "created_at": "2025-01-15T05:34:27.268190+00:00"} {"global_step": 15, "acc_step": 0, "speed/wps": 12967.42391445662, "speed/FLOPS": 203671211684308.72, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4518533945083618, "optim/lr": 8.400000000000001e-06, "optim/total_tokens": 7864320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 9.914666175842285, "created_at": "2025-01-15T05:34:37.384252+00:00"} {"global_step": 16, "acc_step": 0, "speed/wps": 12952.142098503076, "speed/FLOPS": 203431189765342.62, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.404903531074524, "optim/lr": 9e-06, "optim/total_tokens": 8388608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 9.863900184631348, "created_at": "2025-01-15T05:34:47.510264+00:00"} {"global_step": 17, "acc_step": 0, "speed/wps": 12942.840058078204, "speed/FLOPS": 203285088438126.97, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.394577145576477, "optim/lr": 9.600000000000001e-06, "optim/total_tokens": 8912896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 9.758782386779785, "created_at": "2025-01-15T05:34:57.642203+00:00"} {"global_step": 18, "acc_step": 0, "speed/wps": 12954.84950451805, "speed/FLOPS": 203473713297173.5, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3222562074661255, "optim/lr": 1.0199999999999999e-05, "optim/total_tokens": 9437184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 9.635913848876953, "created_at": "2025-01-15T05:35:07.764879+00:00"} {"global_step": 19, "acc_step": 0, "speed/wps": 12945.693338547888, "speed/FLOPS": 203329903128721.6, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3060349225997925, "optim/lr": 1.08e-05, "optim/total_tokens": 9961472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 9.640716552734375, "created_at": "2025-01-15T05:35:17.893377+00:00"} {"global_step": 20, "acc_step": 0, "speed/wps": 12950.651362046718, "speed/FLOPS": 203407775700808.94, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.23166024684906, "optim/lr": 1.1400000000000001e-05, "optim/total_tokens": 10485760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 9.606426239013672, "created_at": "2025-01-15T05:35:28.018166+00:00"} {"global_step": 21, "acc_step": 0, "speed/wps": 12951.438420314618, "speed/FLOPS": 203420137532437.72, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2755461931228638, "optim/lr": 1.2e-05, "optim/total_tokens": 11010048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 9.495306015014648, "created_at": "2025-01-15T05:35:38.143668+00:00"} {"global_step": 22, "acc_step": 0, "speed/wps": 12961.978661601264, "speed/FLOPS": 203585686505653.8, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2105661630630493, "optim/lr": 1.26e-05, "optim/total_tokens": 11534336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 9.41015911102295, "created_at": "2025-01-15T05:35:48.261662+00:00"} {"global_step": 23, "acc_step": 0, "speed/wps": 12962.908963844024, "speed/FLOPS": 203600298180745.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2491904497146606, "optim/lr": 1.32e-05, "optim/total_tokens": 12058624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 9.360716819763184, "created_at": "2025-01-15T05:35:58.380678+00:00"} {"global_step": 24, "acc_step": 0, "speed/wps": 12965.318935254238, "speed/FLOPS": 203638150093390.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2704106569290161, "optim/lr": 1.38e-05, "optim/total_tokens": 12582912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370920, "loss/out": 9.267946243286133, "created_at": "2025-01-15T05:36:08.494710+00:00"} {"global_step": 25, "acc_step": 0, "speed/wps": 12964.605463394382, "speed/FLOPS": 203626944037418.66, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1714509725570679, "optim/lr": 1.44e-05, "optim/total_tokens": 13107200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 9.270057678222656, "created_at": "2025-01-15T05:36:18.608470+00:00"} {"global_step": 26, "acc_step": 0, "speed/wps": 12966.36825612955, "speed/FLOPS": 203654631119655.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1896551847457886, "optim/lr": 1.5e-05, "optim/total_tokens": 13631488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 9.241281509399414, "created_at": "2025-01-15T05:36:28.726066+00:00"} {"global_step": 27, "acc_step": 0, "speed/wps": 12962.648177541714, "speed/FLOPS": 203596202173509.9, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2128711938858032, "optim/lr": 1.56e-05, "optim/total_tokens": 14155776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 9.159324645996094, "created_at": "2025-01-15T05:36:38.842849+00:00"} {"global_step": 28, "acc_step": 0, "speed/wps": 12963.114706870132, "speed/FLOPS": 203603529657691.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1426650285720825, "optim/lr": 1.62e-05, "optim/total_tokens": 14680064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421973, "loss/out": 9.214592933654785, "created_at": "2025-01-15T05:36:48.962154+00:00"} {"global_step": 29, "acc_step": 0, "speed/wps": 12961.262717564334, "speed/FLOPS": 203574441620743.62, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1780290603637695, "optim/lr": 1.6800000000000002e-05, "optim/total_tokens": 15204352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 9.135016441345215, "created_at": "2025-01-15T05:36:59.080796+00:00"} {"global_step": 30, "acc_step": 0, "speed/wps": 12966.79813003991, "speed/FLOPS": 203661382880126.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1765724420547485, "optim/lr": 1.74e-05, "optim/total_tokens": 15728640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 9.078001976013184, "created_at": "2025-01-15T05:37:09.202267+00:00"} {"global_step": 31, "acc_step": 0, "speed/wps": 12961.756025868148, "speed/FLOPS": 203582189705531.7, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1642777919769287, "optim/lr": 1.8e-05, "optim/total_tokens": 16252928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 9.221593856811523, "created_at": "2025-01-15T05:37:19.318175+00:00"} {"global_step": 32, "acc_step": 0, "speed/wps": 12971.89504073641, "speed/FLOPS": 203741436866504.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1437002420425415, "optim/lr": 1.8599999999999998e-05, "optim/total_tokens": 16777216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341330, "loss/out": 8.985284805297852, "created_at": "2025-01-15T05:37:29.427436+00:00"} {"global_step": 33, "acc_step": 0, "speed/wps": 12961.011218243197, "speed/FLOPS": 203570491478308.88, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1638574600219727, "optim/lr": 1.9200000000000003e-05, "optim/total_tokens": 17301504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 8.955659866333008, "created_at": "2025-01-15T05:37:39.547755+00:00"} {"global_step": 34, "acc_step": 0, "speed/wps": 12960.451853518864, "speed/FLOPS": 203561705886665.56, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1884366273880005, "optim/lr": 1.98e-05, "optim/total_tokens": 17825792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 8.951687812805176, "created_at": "2025-01-15T05:37:49.668386+00:00"} {"global_step": 35, "acc_step": 0, "speed/wps": 12958.011012628205, "speed/FLOPS": 203523369126410.9, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1711724996566772, "optim/lr": 2.0399999999999998e-05, "optim/total_tokens": 18350080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308327, "loss/out": 8.932416915893555, "created_at": "2025-01-15T05:37:59.787862+00:00"} {"global_step": 36, "acc_step": 0, "speed/wps": 12961.710450568195, "speed/FLOPS": 203581473882818.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1566264629364014, "optim/lr": 2.1000000000000002e-05, "optim/total_tokens": 18874368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 8.856736183166504, "created_at": "2025-01-15T05:38:09.907265+00:00"} {"global_step": 37, "acc_step": 0, "speed/wps": 12966.28783295294, "speed/FLOPS": 203653367963153.84, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1871349811553955, "optim/lr": 2.16e-05, "optim/total_tokens": 19398656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 8.865288734436035, "created_at": "2025-01-15T05:38:20.023160+00:00"} {"global_step": 38, "acc_step": 0, "speed/wps": 12962.633495841474, "speed/FLOPS": 203595971577232.84, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1454787254333496, "optim/lr": 2.22e-05, "optim/total_tokens": 19922944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 8.763337135314941, "created_at": "2025-01-15T05:38:30.139468+00:00"} {"global_step": 39, "acc_step": 0, "speed/wps": 12963.093923919083, "speed/FLOPS": 203603203232886.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.190330147743225, "optim/lr": 2.2800000000000002e-05, "optim/total_tokens": 20447232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 8.777400016784668, "created_at": "2025-01-15T05:38:40.255840+00:00"} {"global_step": 40, "acc_step": 0, "speed/wps": 12967.147292754615, "speed/FLOPS": 203666866960360.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2085877656936646, "optim/lr": 2.34e-05, "optim/total_tokens": 20971520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 8.772250175476074, "created_at": "2025-01-15T05:38:50.375857+00:00"} {"global_step": 41, "acc_step": 0, "speed/wps": 12966.181118067403, "speed/FLOPS": 203651691859234.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1585661172866821, "optim/lr": 2.4e-05, "optim/total_tokens": 21495808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 8.663045883178711, "created_at": "2025-01-15T05:39:00.488668+00:00"} {"global_step": 42, "acc_step": 0, "speed/wps": 12963.228744440488, "speed/FLOPS": 203605320774437.66, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.226742148399353, "optim/lr": 2.46e-05, "optim/total_tokens": 22020096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 8.641132354736328, "created_at": "2025-01-15T05:39:10.603681+00:00"} {"global_step": 43, "acc_step": 0, "speed/wps": 12967.669468594408, "speed/FLOPS": 203675068449469.44, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1452211141586304, "optim/lr": 2.52e-05, "optim/total_tokens": 22544384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 8.55803108215332, "created_at": "2025-01-15T05:39:20.721608+00:00"} {"global_step": 44, "acc_step": 0, "speed/wps": 12964.516210141845, "speed/FLOPS": 203625542192440.75, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.1599105596542358, "optim/lr": 2.58e-05, "optim/total_tokens": 23068672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 8.469062805175781, "created_at": "2025-01-15T05:39:30.836324+00:00"} {"global_step": 45, "acc_step": 0, "speed/wps": 12964.586680865244, "speed/FLOPS": 203626649031988.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1968845129013062, "optim/lr": 2.64e-05, "optim/total_tokens": 23592960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 8.50183391571045, "created_at": "2025-01-15T05:39:40.952292+00:00"} {"global_step": 46, "acc_step": 0, "speed/wps": 12964.78948298762, "speed/FLOPS": 203629834317999.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0926963090896606, "optim/lr": 2.7e-05, "optim/total_tokens": 24117248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 8.467103958129883, "created_at": "2025-01-15T05:39:51.067934+00:00"} {"global_step": 47, "acc_step": 0, "speed/wps": 12963.647079277522, "speed/FLOPS": 203611891297905.6, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1346487998962402, "optim/lr": 2.76e-05, "optim/total_tokens": 24641536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 8.456869125366211, "created_at": "2025-01-15T05:40:01.187228+00:00"} {"global_step": 48, "acc_step": 0, "speed/wps": 12961.266130272099, "speed/FLOPS": 203574495222008.62, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1829123497009277, "optim/lr": 2.82e-05, "optim/total_tokens": 25165824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 8.377923965454102, "created_at": "2025-01-15T05:40:11.303341+00:00"} {"global_step": 49, "acc_step": 0, "speed/wps": 12955.471308477447, "speed/FLOPS": 203483479582804.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1536345481872559, "optim/lr": 2.88e-05, "optim/total_tokens": 25690112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 8.261479377746582, "created_at": "2025-01-15T05:40:21.427058+00:00"} {"global_step": 50, "acc_step": 0, "speed/wps": 12956.566665689137, "speed/FLOPS": 203500683672992.16, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2447164058685303, "optim/lr": 2.94e-05, "optim/total_tokens": 26214400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 8.256531715393066, "created_at": "2025-01-15T05:40:31.552516+00:00"} {"global_step": 51, "acc_step": 0, "speed/wps": 12958.447578587135, "speed/FLOPS": 203530225994700.38, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.202534794807434, "optim/lr": 3e-05, "optim/total_tokens": 26738688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 8.279088020324707, "created_at": "2025-01-15T05:40:41.672204+00:00"} {"global_step": 52, "acc_step": 0, "speed/wps": 12950.882122052151, "speed/FLOPS": 203411400103791.75, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1960328817367554, "optim/lr": 3.0600000000000005e-05, "optim/total_tokens": 27262976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 8.19095230102539, "created_at": "2025-01-15T05:40:51.799210+00:00"} {"global_step": 53, "acc_step": 0, "speed/wps": 12956.473571373126, "speed/FLOPS": 203499221498834.22, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2194690704345703, "optim/lr": 3.12e-05, "optim/total_tokens": 27787264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 8.185504913330078, "created_at": "2025-01-15T05:41:01.952942+00:00"} {"global_step": 54, "acc_step": 0, "speed/wps": 12952.085514881695, "speed/FLOPS": 203430301041814.78, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3778324127197266, "optim/lr": 3.18e-05, "optim/total_tokens": 28311552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 8.162140846252441, "created_at": "2025-01-15T05:41:12.079126+00:00"} {"global_step": 55, "acc_step": 0, "speed/wps": 12946.982032694359, "speed/FLOPS": 203350143841143.25, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1515536308288574, "optim/lr": 3.24e-05, "optim/total_tokens": 28835840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 8.056032180786133, "created_at": "2025-01-15T05:41:22.206534+00:00"} {"global_step": 56, "acc_step": 0, "speed/wps": 12947.165053269211, "speed/FLOPS": 203353018430767.9, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4397169351577759, "optim/lr": 3.2999999999999996e-05, "optim/total_tokens": 29360128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346784, "loss/out": 8.08599853515625, "created_at": "2025-01-15T05:41:32.338009+00:00"} {"global_step": 57, "acc_step": 0, "speed/wps": 12947.632867759803, "speed/FLOPS": 203360366100187.16, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2148303985595703, "optim/lr": 3.3600000000000004e-05, "optim/total_tokens": 29884416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 8.222200393676758, "created_at": "2025-01-15T05:41:42.465897+00:00"} {"global_step": 58, "acc_step": 0, "speed/wps": 12916.924776336622, "speed/FLOPS": 202878053327046.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.3238085508346558, "optim/lr": 3.4200000000000005e-05, "optim/total_tokens": 30408704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 8.090154647827148, "created_at": "2025-01-15T05:41:52.617502+00:00"} {"global_step": 59, "acc_step": 0, "speed/wps": 12938.565552466702, "speed/FLOPS": 203217951453712.1, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3070601224899292, "optim/lr": 3.48e-05, "optim/total_tokens": 30932992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 7.9205498695373535, "created_at": "2025-01-15T05:42:02.752943+00:00"} {"global_step": 60, "acc_step": 0, "speed/wps": 12938.773622203325, "speed/FLOPS": 203221219474843.53, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2083542346954346, "optim/lr": 3.54e-05, "optim/total_tokens": 31457280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 7.91619873046875, "created_at": "2025-01-15T05:42:12.887991+00:00"} {"global_step": 61, "acc_step": 0, "speed/wps": 12924.49116596664, "speed/FLOPS": 202996893873494.84, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.346136450767517, "optim/lr": 3.6e-05, "optim/total_tokens": 31981568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428908, "loss/out": 7.814106464385986, "created_at": "2025-01-15T05:42:23.037174+00:00"} {"global_step": 62, "acc_step": 0, "speed/wps": 12931.451666057952, "speed/FLOPS": 203106218092158.34, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.08980393409729, "optim/lr": 3.66e-05, "optim/total_tokens": 32505856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 7.776556491851807, "created_at": "2025-01-15T05:42:33.177025+00:00"} {"global_step": 63, "acc_step": 0, "speed/wps": 12929.292272424052, "speed/FLOPS": 203072301847820.16, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1661349534988403, "optim/lr": 3.7199999999999996e-05, "optim/total_tokens": 33030144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315344, "loss/out": 7.766096115112305, "created_at": "2025-01-15T05:42:43.320322+00:00"} {"global_step": 64, "acc_step": 0, "speed/wps": 12929.728739450875, "speed/FLOPS": 203079157162244.94, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2133196592330933, "optim/lr": 3.7800000000000004e-05, "optim/total_tokens": 33554432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 7.75492525100708, "created_at": "2025-01-15T05:42:53.463007+00:00"} {"global_step": 65, "acc_step": 0, "speed/wps": 12921.12804215283, "speed/FLOPS": 202944071392582.56, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.147181510925293, "optim/lr": 3.8400000000000005e-05, "optim/total_tokens": 34078720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 7.646509647369385, "created_at": "2025-01-15T05:43:03.610627+00:00"} {"global_step": 66, "acc_step": 0, "speed/wps": 12925.004306909621, "speed/FLOPS": 203004953457133.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.277360439300537, "optim/lr": 3.9e-05, "optim/total_tokens": 34603008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 7.704341888427734, "created_at": "2025-01-15T05:43:13.759747+00:00"} {"global_step": 67, "acc_step": 0, "speed/wps": 12922.774741389208, "speed/FLOPS": 202969935066891.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2008103132247925, "optim/lr": 3.96e-05, "optim/total_tokens": 35127296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 7.595422744750977, "created_at": "2025-01-15T05:43:23.907349+00:00"} {"global_step": 68, "acc_step": 0, "speed/wps": 12923.203248539534, "speed/FLOPS": 202976665360518.16, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2328848838806152, "optim/lr": 4.02e-05, "optim/total_tokens": 35651584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 7.5578107833862305, "created_at": "2025-01-15T05:43:34.053677+00:00"} {"global_step": 69, "acc_step": 0, "speed/wps": 12921.54792909528, "speed/FLOPS": 202950666293998.88, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2148462533950806, "optim/lr": 4.0799999999999996e-05, "optim/total_tokens": 36175872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 7.670148849487305, "created_at": "2025-01-15T05:43:44.201941+00:00"} {"global_step": 70, "acc_step": 0, "speed/wps": 12917.654756957187, "speed/FLOPS": 202889518675790.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1656428575515747, "optim/lr": 4.1400000000000003e-05, "optim/total_tokens": 36700160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 7.558905601501465, "created_at": "2025-01-15T05:43:54.356212+00:00"} {"global_step": 71, "acc_step": 0, "speed/wps": 12915.055843956628, "speed/FLOPS": 202848699175836.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2178263664245605, "optim/lr": 4.2000000000000004e-05, "optim/total_tokens": 37224448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292430, "loss/out": 7.539411544799805, "created_at": "2025-01-15T05:44:04.509306+00:00"} {"global_step": 72, "acc_step": 0, "speed/wps": 12911.603586606463, "speed/FLOPS": 202794476730253.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2793490886688232, "optim/lr": 4.2600000000000005e-05, "optim/total_tokens": 37748736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 7.480741024017334, "created_at": "2025-01-15T05:44:14.691756+00:00"} {"global_step": 73, "acc_step": 0, "speed/wps": 12916.023937309088, "speed/FLOPS": 202863904412236.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8604750633239746, "optim/lr": 4.32e-05, "optim/total_tokens": 38273024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 7.51432991027832, "created_at": "2025-01-15T05:44:24.845523+00:00"} {"global_step": 74, "acc_step": 0, "speed/wps": 12912.39309678202, "speed/FLOPS": 202806877072461.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4733731746673584, "optim/lr": 4.38e-05, "optim/total_tokens": 38797312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 7.51400089263916, "created_at": "2025-01-15T05:44:35.000639+00:00"} {"global_step": 75, "acc_step": 0, "speed/wps": 12910.192299513032, "speed/FLOPS": 202772310527139.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2652056217193604, "optim/lr": 4.44e-05, "optim/total_tokens": 39321600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 7.473238945007324, "created_at": "2025-01-15T05:44:45.156943+00:00"} {"global_step": 76, "acc_step": 0, "speed/wps": 12910.674378869993, "speed/FLOPS": 202779882245886.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.250699758529663, "optim/lr": 4.4999999999999996e-05, "optim/total_tokens": 39845888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 7.505756378173828, "created_at": "2025-01-15T05:44:55.315329+00:00"} {"global_step": 77, "acc_step": 0, "speed/wps": 12904.862387016132, "speed/FLOPS": 202688596927308.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3762165307998657, "optim/lr": 4.5600000000000004e-05, "optim/total_tokens": 40370176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 7.391574859619141, "created_at": "2025-01-15T05:45:05.481506+00:00"} {"global_step": 78, "acc_step": 0, "speed/wps": 12904.524793996276, "speed/FLOPS": 202683294565030.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2051877975463867, "optim/lr": 4.6200000000000005e-05, "optim/total_tokens": 40894464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 7.373611927032471, "created_at": "2025-01-15T05:45:15.645438+00:00"} {"global_step": 79, "acc_step": 0, "speed/wps": 12902.119357579078, "speed/FLOPS": 202645513880681.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2501698732376099, "optim/lr": 4.68e-05, "optim/total_tokens": 41418752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 7.462690353393555, "created_at": "2025-01-15T05:45:25.811246+00:00"} {"global_step": 80, "acc_step": 0, "speed/wps": 12905.894087239332, "speed/FLOPS": 202704801197018.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1767730712890625, "optim/lr": 4.740000000000001e-05, "optim/total_tokens": 41943040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 7.396706581115723, "created_at": "2025-01-15T05:45:35.970722+00:00"} {"global_step": 81, "acc_step": 0, "speed/wps": 12901.9703758345, "speed/FLOPS": 202643173917660.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1750850677490234, "optim/lr": 4.8e-05, "optim/total_tokens": 42467328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474863, "loss/out": 7.320599555969238, "created_at": "2025-01-15T05:45:46.138015+00:00"} {"global_step": 82, "acc_step": 0, "speed/wps": 12899.364528909606, "speed/FLOPS": 202602245510896.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2139102220535278, "optim/lr": 4.8599999999999995e-05, "optim/total_tokens": 42991616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485284, "loss/out": 7.335229873657227, "created_at": "2025-01-15T05:45:56.305056+00:00"} {"global_step": 83, "acc_step": 0, "speed/wps": 12897.762344527351, "speed/FLOPS": 202577081003541.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.144783616065979, "optim/lr": 4.92e-05, "optim/total_tokens": 43515904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 7.340839385986328, "created_at": "2025-01-15T05:46:06.471216+00:00"} {"global_step": 84, "acc_step": 0, "speed/wps": 12897.572009232885, "speed/FLOPS": 202574091526193.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2189514636993408, "optim/lr": 4.9800000000000004e-05, "optim/total_tokens": 44040192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 7.262691974639893, "created_at": "2025-01-15T05:46:16.639901+00:00"} {"global_step": 85, "acc_step": 0, "speed/wps": 12899.883856336068, "speed/FLOPS": 202610402261759.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2123125791549683, "optim/lr": 5.04e-05, "optim/total_tokens": 44564480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375384, "loss/out": 7.317293167114258, "created_at": "2025-01-15T05:46:26.807988+00:00"} {"global_step": 86, "acc_step": 0, "speed/wps": 12895.984952412533, "speed/FLOPS": 202549164617985.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.169638991355896, "optim/lr": 5.1000000000000006e-05, "optim/total_tokens": 45088768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 7.233262062072754, "created_at": "2025-01-15T05:46:36.978145+00:00"} {"global_step": 87, "acc_step": 0, "speed/wps": 12896.465167699647, "speed/FLOPS": 202556707058954.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2641834020614624, "optim/lr": 5.16e-05, "optim/total_tokens": 45613056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 7.2652130126953125, "created_at": "2025-01-15T05:46:47.146477+00:00"} {"global_step": 88, "acc_step": 0, "speed/wps": 12889.32365410595, "speed/FLOPS": 202444539774497.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.200156569480896, "optim/lr": 5.2199999999999995e-05, "optim/total_tokens": 46137344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 7.202808856964111, "created_at": "2025-01-15T05:46:57.325329+00:00"} {"global_step": 89, "acc_step": 0, "speed/wps": 12894.984956757911, "speed/FLOPS": 202533458312091.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.279550552368164, "optim/lr": 5.28e-05, "optim/total_tokens": 46661632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 7.210537910461426, "created_at": "2025-01-15T05:47:07.496182+00:00"} {"global_step": 90, "acc_step": 0, "speed/wps": 12893.318179684988, "speed/FLOPS": 202507279287769.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.217459797859192, "optim/lr": 5.3400000000000004e-05, "optim/total_tokens": 47185920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 7.230947017669678, "created_at": "2025-01-15T05:47:17.665527+00:00"} {"global_step": 91, "acc_step": 0, "speed/wps": 12890.17602696312, "speed/FLOPS": 202457927461502.2, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.212580680847168, "optim/lr": 5.4e-05, "optim/total_tokens": 47710208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 7.270803451538086, "created_at": "2025-01-15T05:47:27.839586+00:00"} {"global_step": 92, "acc_step": 0, "speed/wps": 12890.411093842717, "speed/FLOPS": 202461619509862.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2821608781814575, "optim/lr": 5.4600000000000006e-05, "optim/total_tokens": 48234496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 7.150531768798828, "created_at": "2025-01-15T05:47:38.011999+00:00"} {"global_step": 93, "acc_step": 0, "speed/wps": 12888.280697807615, "speed/FLOPS": 202428158712659.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.188724398612976, "optim/lr": 5.52e-05, "optim/total_tokens": 48758784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 7.219142436981201, "created_at": "2025-01-15T05:47:48.187698+00:00"} {"global_step": 94, "acc_step": 0, "speed/wps": 12888.727897515058, "speed/FLOPS": 202435182598582.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2091894149780273, "optim/lr": 5.5799999999999994e-05, "optim/total_tokens": 49283072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 7.138932228088379, "created_at": "2025-01-15T05:47:58.363183+00:00"} {"global_step": 95, "acc_step": 0, "speed/wps": 12887.20331645139, "speed/FLOPS": 202411236957983.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1946660280227661, "optim/lr": 5.64e-05, "optim/total_tokens": 49807360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 7.0806989669799805, "created_at": "2025-01-15T05:48:08.540163+00:00"} {"global_step": 96, "acc_step": 0, "speed/wps": 12886.873934447087, "speed/FLOPS": 202406063560987.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2290871143341064, "optim/lr": 5.7e-05, "optim/total_tokens": 50331648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 7.201218128204346, "created_at": "2025-01-15T05:48:18.721182+00:00"} {"global_step": 97, "acc_step": 0, "speed/wps": 12888.354684434438, "speed/FLOPS": 202429320774302.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1744956970214844, "optim/lr": 5.76e-05, "optim/total_tokens": 50855936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 7.0410871505737305, "created_at": "2025-01-15T05:48:28.898558+00:00"} {"global_step": 98, "acc_step": 0, "speed/wps": 12888.270299665852, "speed/FLOPS": 202427995395554.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.203914999961853, "optim/lr": 5.8200000000000005e-05, "optim/total_tokens": 51380224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 7.105074405670166, "created_at": "2025-01-15T05:48:39.075278+00:00"} {"global_step": 99, "acc_step": 0, "speed/wps": 12885.110440952254, "speed/FLOPS": 202378365472356.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1779556274414062, "optim/lr": 5.88e-05, "optim/total_tokens": 51904512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 7.066592216491699, "created_at": "2025-01-15T05:48:49.258487+00:00"} {"global_step": 100, "acc_step": 0, "speed/wps": 12884.376177692033, "speed/FLOPS": 202366832858871.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.126652479171753, "optim/lr": 5.940000000000001e-05, "optim/total_tokens": 52428800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458985, "loss/out": 7.109035015106201, "created_at": "2025-01-15T05:48:59.435068+00:00"} {"global_step": 101, "acc_step": 0, "speed/wps": 12886.946493616422, "speed/FLOPS": 202407203202448.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.183781623840332, "optim/lr": 6e-05, "optim/total_tokens": 52953088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 7.040726661682129, "created_at": "2025-01-15T05:49:09.610973+00:00"} {"global_step": 102, "acc_step": 0, "speed/wps": 12880.73777637733, "speed/FLOPS": 202309686766536.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1172465085983276, "optim/lr": 6.0599999999999996e-05, "optim/total_tokens": 53477376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 6.952965259552002, "created_at": "2025-01-15T05:49:19.793376+00:00"} {"global_step": 103, "acc_step": 0, "speed/wps": 12886.63884079891, "speed/FLOPS": 202402371092190.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1894807815551758, "optim/lr": 6.120000000000001e-05, "optim/total_tokens": 54001664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 6.97404146194458, "created_at": "2025-01-15T05:49:29.970052+00:00"} {"global_step": 104, "acc_step": 0, "speed/wps": 12882.316376281333, "speed/FLOPS": 202334480847252.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1662925481796265, "optim/lr": 6.18e-05, "optim/total_tokens": 54525952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315105, "loss/out": 6.971856117248535, "created_at": "2025-01-15T05:49:40.148567+00:00"} {"global_step": 105, "acc_step": 0, "speed/wps": 12881.636906847198, "speed/FLOPS": 202323808846100.97, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.998971164226532, "optim/lr": 6.24e-05, "optim/total_tokens": 55050240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 7.005993366241455, "created_at": "2025-01-15T05:49:50.331739+00:00"} {"global_step": 106, "acc_step": 0, "speed/wps": 12878.941755235835, "speed/FLOPS": 202281477786518.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2675118446350098, "optim/lr": 6.3e-05, "optim/total_tokens": 55574528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.927533149719238, "created_at": "2025-01-15T05:50:00.513091+00:00"} {"global_step": 107, "acc_step": 0, "speed/wps": 12872.492885390131, "speed/FLOPS": 202180189423915.5, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9943126440048218, "optim/lr": 6.36e-05, "optim/total_tokens": 56098816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 6.939741134643555, "created_at": "2025-01-15T05:50:10.699112+00:00"} {"global_step": 108, "acc_step": 0, "speed/wps": 12880.00475752414, "speed/FLOPS": 202298173698174.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1524373292922974, "optim/lr": 6.42e-05, "optim/total_tokens": 56623104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494018, "loss/out": 6.865457534790039, "created_at": "2025-01-15T05:50:20.880994+00:00"} {"global_step": 109, "acc_step": 0, "speed/wps": 12882.768464114377, "speed/FLOPS": 202341581507904.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1531689167022705, "optim/lr": 6.48e-05, "optim/total_tokens": 57147392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 6.836931228637695, "created_at": "2025-01-15T05:50:31.058639+00:00"} {"global_step": 110, "acc_step": 0, "speed/wps": 12881.386457864655, "speed/FLOPS": 202319875200677.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1371043920516968, "optim/lr": 6.54e-05, "optim/total_tokens": 57671680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 6.9054789543151855, "created_at": "2025-01-15T05:50:41.247643+00:00"} {"global_step": 111, "acc_step": 0, "speed/wps": 12878.187253122287, "speed/FLOPS": 202269627294030.28, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9091846942901611, "optim/lr": 6.599999999999999e-05, "optim/total_tokens": 58195968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467150, "loss/out": 6.950311183929443, "created_at": "2025-01-15T05:50:51.429094+00:00"} {"global_step": 112, "acc_step": 0, "speed/wps": 12874.023944772496, "speed/FLOPS": 202204236815411.62, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9333502054214478, "optim/lr": 6.66e-05, "optim/total_tokens": 58720256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 6.874694347381592, "created_at": "2025-01-15T05:51:01.615021+00:00"} {"global_step": 113, "acc_step": 0, "speed/wps": 12879.753571661433, "speed/FLOPS": 202294228479034.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0388554334640503, "optim/lr": 6.720000000000001e-05, "optim/total_tokens": 59244544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 6.797607421875, "created_at": "2025-01-15T05:51:11.800691+00:00"} {"global_step": 114, "acc_step": 0, "speed/wps": 12880.79473213166, "speed/FLOPS": 202310581334924.12, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1586110591888428, "optim/lr": 6.78e-05, "optim/total_tokens": 59768832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.826437473297119, "created_at": "2025-01-15T05:51:21.982459+00:00"} {"global_step": 115, "acc_step": 0, "speed/wps": 12878.089918819103, "speed/FLOPS": 202268098525047.44, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8082816004753113, "optim/lr": 6.840000000000001e-05, "optim/total_tokens": 60293120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 6.787317752838135, "created_at": "2025-01-15T05:51:32.172032+00:00"} {"global_step": 116, "acc_step": 0, "speed/wps": 12877.951083464386, "speed/FLOPS": 202265917925021.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9718465209007263, "optim/lr": 6.9e-05, "optim/total_tokens": 60817408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373755, "loss/out": 6.813703536987305, "created_at": "2025-01-15T05:51:42.354077+00:00"} {"global_step": 117, "acc_step": 0, "speed/wps": 12866.157464401722, "speed/FLOPS": 202080682931511.8, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0756932497024536, "optim/lr": 6.96e-05, "optim/total_tokens": 61341696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 6.812175750732422, "created_at": "2025-01-15T05:51:52.546134+00:00"} {"global_step": 118, "acc_step": 0, "speed/wps": 12873.882523848613, "speed/FLOPS": 202202015605469.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.035384178161621, "optim/lr": 7.02e-05, "optim/total_tokens": 61865984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.8048014640808105, "created_at": "2025-01-15T05:52:02.732045+00:00"} {"global_step": 119, "acc_step": 0, "speed/wps": 12861.81673369344, "speed/FLOPS": 202012505790949.2, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.171999454498291, "optim/lr": 7.08e-05, "optim/total_tokens": 62390272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 6.716472625732422, "created_at": "2025-01-15T05:52:12.927659+00:00"} {"global_step": 120, "acc_step": 0, "speed/wps": 12869.685879839368, "speed/FLOPS": 202136101544510.94, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8240696787834167, "optim/lr": 7.14e-05, "optim/total_tokens": 62914560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 6.792017936706543, "created_at": "2025-01-15T05:52:23.117008+00:00"} {"global_step": 121, "acc_step": 0, "speed/wps": 12877.942210302135, "speed/FLOPS": 202265778559815.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.162576675415039, "optim/lr": 7.2e-05, "optim/total_tokens": 63438848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 6.805678367614746, "created_at": "2025-01-15T05:52:33.300539+00:00"} {"global_step": 122, "acc_step": 0, "speed/wps": 12873.264431418043, "speed/FLOPS": 202192307614499.22, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9940779209136963, "optim/lr": 7.26e-05, "optim/total_tokens": 63963136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 6.7787041664123535, "created_at": "2025-01-15T05:52:43.488786+00:00"} {"global_step": 123, "acc_step": 0, "speed/wps": 12876.247163344262, "speed/FLOPS": 202239155518103.66, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2170310020446777, "optim/lr": 7.32e-05, "optim/total_tokens": 64487424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 6.641376495361328, "created_at": "2025-01-15T05:52:53.674422+00:00"} {"global_step": 124, "acc_step": 0, "speed/wps": 12877.464220635615, "speed/FLOPS": 202258271075276.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1769654750823975, "optim/lr": 7.38e-05, "optim/total_tokens": 65011712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 6.663385391235352, "created_at": "2025-01-15T05:53:03.860363+00:00"} {"global_step": 125, "acc_step": 0, "speed/wps": 12873.28754967203, "speed/FLOPS": 202192670718445.88, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1728745698928833, "optim/lr": 7.439999999999999e-05, "optim/total_tokens": 65536000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422265, "loss/out": 6.721560001373291, "created_at": "2025-01-15T05:53:14.046484+00:00"} {"global_step": 126, "acc_step": 0, "speed/wps": 12876.545697524038, "speed/FLOPS": 202243844407626.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1579703092575073, "optim/lr": 7.500000000000001e-05, "optim/total_tokens": 66060288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350412, "loss/out": 6.767038345336914, "created_at": "2025-01-15T05:53:24.229504+00:00"} {"global_step": 127, "acc_step": 0, "speed/wps": 12875.055066081317, "speed/FLOPS": 202220431992475.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1683578491210938, "optim/lr": 7.560000000000001e-05, "optim/total_tokens": 66584576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 6.698503017425537, "created_at": "2025-01-15T05:53:34.421781+00:00"} {"global_step": 128, "acc_step": 0, "speed/wps": 12874.853281899434, "speed/FLOPS": 202217262694618.8, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1801871061325073, "optim/lr": 7.62e-05, "optim/total_tokens": 67108864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 6.68381404876709, "created_at": "2025-01-15T05:53:44.607639+00:00"} {"global_step": 129, "acc_step": 0, "speed/wps": 12874.891957721957, "speed/FLOPS": 202217870151557.7, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9541498422622681, "optim/lr": 7.680000000000001e-05, "optim/total_tokens": 67633152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 6.681320667266846, "created_at": "2025-01-15T05:53:54.800213+00:00"} {"global_step": 130, "acc_step": 0, "speed/wps": 12875.720180425888, "speed/FLOPS": 202230878527220.2, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1495108604431152, "optim/lr": 7.74e-05, "optim/total_tokens": 68157440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 6.705558776855469, "created_at": "2025-01-15T05:54:04.983826+00:00"} {"global_step": 131, "acc_step": 0, "speed/wps": 12872.030091638157, "speed/FLOPS": 202172920612095.5, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.132981777191162, "optim/lr": 7.8e-05, "optim/total_tokens": 68681728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.664605140686035, "created_at": "2025-01-15T05:54:15.170282+00:00"} {"global_step": 132, "acc_step": 0, "speed/wps": 12874.15202802535, "speed/FLOPS": 202206248538902.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8124447464942932, "optim/lr": 7.86e-05, "optim/total_tokens": 69206016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 6.639238357543945, "created_at": "2025-01-15T05:54:25.357781+00:00"} {"global_step": 133, "acc_step": 0, "speed/wps": 12873.899411101556, "speed/FLOPS": 202202280842982.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.108591079711914, "optim/lr": 7.92e-05, "optim/total_tokens": 69730304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507176, "loss/out": 6.823969841003418, "created_at": "2025-01-15T05:54:35.545204+00:00"} {"global_step": 134, "acc_step": 0, "speed/wps": 12870.87111878447, "speed/FLOPS": 202154717350832.75, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7109673619270325, "optim/lr": 7.98e-05, "optim/total_tokens": 70254592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 6.616881370544434, "created_at": "2025-01-15T05:54:45.733615+00:00"} {"global_step": 135, "acc_step": 0, "speed/wps": 12875.516061940638, "speed/FLOPS": 202227672565921.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8253679275512695, "optim/lr": 8.04e-05, "optim/total_tokens": 70778880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423026, "loss/out": 6.634981155395508, "created_at": "2025-01-15T05:54:55.917817+00:00"} {"global_step": 136, "acc_step": 0, "speed/wps": 12876.98179400875, "speed/FLOPS": 202250693902177.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.095055103302002, "optim/lr": 8.1e-05, "optim/total_tokens": 71303168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 6.645674705505371, "created_at": "2025-01-15T05:55:06.100305+00:00"} {"global_step": 137, "acc_step": 0, "speed/wps": 12869.921200354804, "speed/FLOPS": 202139797576570.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9613890647888184, "optim/lr": 8.159999999999999e-05, "optim/total_tokens": 71827456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505844, "loss/out": 6.686657428741455, "created_at": "2025-01-15T05:55:16.289953+00:00"} {"global_step": 138, "acc_step": 0, "speed/wps": 12870.688944646789, "speed/FLOPS": 202151856055666.97, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1823759078979492, "optim/lr": 8.22e-05, "optim/total_tokens": 72351744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 6.639095306396484, "created_at": "2025-01-15T05:55:26.478556+00:00"} {"global_step": 139, "acc_step": 0, "speed/wps": 12864.175003450699, "speed/FLOPS": 202049545658089.16, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.655198872089386, "optim/lr": 8.280000000000001e-05, "optim/total_tokens": 72876032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 6.577792644500732, "created_at": "2025-01-15T05:55:36.671141+00:00"} {"global_step": 140, "acc_step": 0, "speed/wps": 12861.254784138655, "speed/FLOPS": 202003679600991.66, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.075697898864746, "optim/lr": 8.34e-05, "optim/total_tokens": 73400320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 6.6437578201293945, "created_at": "2025-01-15T05:55:46.869148+00:00"} {"global_step": 141, "acc_step": 0, "speed/wps": 12856.923614706584, "speed/FLOPS": 201935652633410.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1590038537979126, "optim/lr": 8.400000000000001e-05, "optim/total_tokens": 73924608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.511228084564209, "created_at": "2025-01-15T05:55:57.067433+00:00"} {"global_step": 142, "acc_step": 0, "speed/wps": 12871.986461334112, "speed/FLOPS": 202172235338216.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7710729241371155, "optim/lr": 8.46e-05, "optim/total_tokens": 74448896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.50341796875, "created_at": "2025-01-15T05:56:07.259471+00:00"} {"global_step": 143, "acc_step": 0, "speed/wps": 12869.41563999232, "speed/FLOPS": 202131857056364.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.078412652015686, "optim/lr": 8.520000000000001e-05, "optim/total_tokens": 74973184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.5736799240112305, "created_at": "2025-01-15T05:56:17.447812+00:00"} {"global_step": 144, "acc_step": 0, "speed/wps": 12871.453651579133, "speed/FLOPS": 202163866828856.88, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7484143376350403, "optim/lr": 8.58e-05, "optim/total_tokens": 75497472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418245, "loss/out": 6.454092979431152, "created_at": "2025-01-15T05:56:27.636884+00:00"} {"global_step": 145, "acc_step": 0, "speed/wps": 12873.022666768782, "speed/FLOPS": 202188510368463.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0726317167282104, "optim/lr": 8.64e-05, "optim/total_tokens": 76021760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 6.643899440765381, "created_at": "2025-01-15T05:56:37.826861+00:00"} {"global_step": 146, "acc_step": 0, "speed/wps": 12869.093431655197, "speed/FLOPS": 202126796331669.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6649344563484192, "optim/lr": 8.7e-05, "optim/total_tokens": 76546048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 6.50813627243042, "created_at": "2025-01-15T05:56:48.015877+00:00"} {"global_step": 147, "acc_step": 0, "speed/wps": 12874.41584092099, "speed/FLOPS": 202210392082945.47, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0175423622131348, "optim/lr": 8.76e-05, "optim/total_tokens": 77070336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467150, "loss/out": 6.541326522827148, "created_at": "2025-01-15T05:56:58.201565+00:00"} {"global_step": 148, "acc_step": 0, "speed/wps": 12874.016280585109, "speed/FLOPS": 202204116438817.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.843481719493866, "optim/lr": 8.82e-05, "optim/total_tokens": 77594624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 6.5324859619140625, "created_at": "2025-01-15T05:57:08.386248+00:00"} {"global_step": 149, "acc_step": 0, "speed/wps": 12866.17234593274, "speed/FLOPS": 202080916666405.8, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.956148624420166, "optim/lr": 8.88e-05, "optim/total_tokens": 78118912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 6.473970890045166, "created_at": "2025-01-15T05:57:18.585038+00:00"} {"global_step": 150, "acc_step": 0, "speed/wps": 12866.33583684962, "speed/FLOPS": 202083484515915.47, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1759501695632935, "optim/lr": 8.94e-05, "optim/total_tokens": 78643200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 6.61909818649292, "created_at": "2025-01-15T05:57:28.776144+00:00"} {"global_step": 151, "acc_step": 0, "speed/wps": 12863.112117592194, "speed/FLOPS": 202032851575123.12, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8723154664039612, "optim/lr": 8.999999999999999e-05, "optim/total_tokens": 79167488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 6.611944198608398, "created_at": "2025-01-15T05:57:38.975763+00:00"} {"global_step": 152, "acc_step": 0, "speed/wps": 12862.086875487548, "speed/FLOPS": 202016748739039.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7970246076583862, "optim/lr": 9.06e-05, "optim/total_tokens": 79691776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 6.448464393615723, "created_at": "2025-01-15T05:57:49.171472+00:00"} {"global_step": 153, "acc_step": 0, "speed/wps": 12863.89775320023, "speed/FLOPS": 202045191061923.78, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9657597541809082, "optim/lr": 9.120000000000001e-05, "optim/total_tokens": 80216064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 6.498696804046631, "created_at": "2025-01-15T05:57:59.366413+00:00"} {"global_step": 154, "acc_step": 0, "speed/wps": 12867.140382784957, "speed/FLOPS": 202096121015391.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.996684193611145, "optim/lr": 9.18e-05, "optim/total_tokens": 80740352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 6.446409702301025, "created_at": "2025-01-15T05:58:09.562181+00:00"} {"global_step": 155, "acc_step": 0, "speed/wps": 12858.060775989006, "speed/FLOPS": 201953513313974.3, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7526047229766846, "optim/lr": 9.240000000000001e-05, "optim/total_tokens": 81264640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.4259419441223145, "created_at": "2025-01-15T05:58:19.763464+00:00"} {"global_step": 156, "acc_step": 0, "speed/wps": 12865.646323764393, "speed/FLOPS": 202072654765421.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7444785833358765, "optim/lr": 9.3e-05, "optim/total_tokens": 81788928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 6.391111373901367, "created_at": "2025-01-15T05:58:29.955992+00:00"} {"global_step": 157, "acc_step": 0, "speed/wps": 12864.743275000978, "speed/FLOPS": 202058471143673.25, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1654837131500244, "optim/lr": 9.36e-05, "optim/total_tokens": 82313216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388267, "loss/out": 6.4291300773620605, "created_at": "2025-01-15T05:58:40.148565+00:00"} {"global_step": 158, "acc_step": 0, "speed/wps": 12862.316785883257, "speed/FLOPS": 202020359797733.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.814704418182373, "optim/lr": 9.42e-05, "optim/total_tokens": 82837504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364786, "loss/out": 6.496053695678711, "created_at": "2025-01-15T05:58:50.343398+00:00"} {"global_step": 159, "acc_step": 0, "speed/wps": 12867.010806201537, "speed/FLOPS": 202094085837092.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0553228855133057, "optim/lr": 9.480000000000001e-05, "optim/total_tokens": 83361792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 6.478477478027344, "created_at": "2025-01-15T05:59:00.543704+00:00"} {"global_step": 160, "acc_step": 0, "speed/wps": 12873.497293363103, "speed/FLOPS": 202195965031332.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.040545105934143, "optim/lr": 9.54e-05, "optim/total_tokens": 83886080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 6.4290771484375, "created_at": "2025-01-15T05:59:10.729610+00:00"} {"global_step": 161, "acc_step": 0, "speed/wps": 12871.39849590331, "speed/FLOPS": 202163000533176.22, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7226749658584595, "optim/lr": 9.6e-05, "optim/total_tokens": 84410368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 6.279353141784668, "created_at": "2025-01-15T05:59:20.916648+00:00"} {"global_step": 162, "acc_step": 0, "speed/wps": 12872.411908268505, "speed/FLOPS": 202178917566946.16, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8552564382553101, "optim/lr": 9.66e-05, "optim/total_tokens": 84934656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 6.390967845916748, "created_at": "2025-01-15T05:59:31.107720+00:00"} {"global_step": 163, "acc_step": 0, "speed/wps": 12869.617928099042, "speed/FLOPS": 202135034269053.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8685490489006042, "optim/lr": 9.719999999999999e-05, "optim/total_tokens": 85458944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.521334171295166, "created_at": "2025-01-15T05:59:41.301775+00:00"} {"global_step": 164, "acc_step": 0, "speed/wps": 12853.492731812961, "speed/FLOPS": 201881765903039.03, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.8149117827415466, "optim/lr": 9.779999999999999e-05, "optim/total_tokens": 85983232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 6.463631629943848, "created_at": "2025-01-15T05:59:51.503060+00:00"} {"global_step": 165, "acc_step": 0, "speed/wps": 12865.998169954713, "speed/FLOPS": 202078180993328.03, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7061933875083923, "optim/lr": 9.84e-05, "optim/total_tokens": 86507520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477322, "loss/out": 6.345087051391602, "created_at": "2025-01-15T06:00:01.696589+00:00"} {"global_step": 166, "acc_step": 0, "speed/wps": 12869.970814218535, "speed/FLOPS": 202140576830476.62, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0237987041473389, "optim/lr": 9.900000000000001e-05, "optim/total_tokens": 87031808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 6.300676345825195, "created_at": "2025-01-15T06:00:11.885398+00:00"} {"global_step": 167, "acc_step": 0, "speed/wps": 12855.387695431906, "speed/FLOPS": 201911528910627.44, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0179940462112427, "optim/lr": 9.960000000000001e-05, "optim/total_tokens": 87556096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 6.361686706542969, "created_at": "2025-01-15T06:00:22.086304+00:00"} {"global_step": 168, "acc_step": 0, "speed/wps": 12870.193611733554, "speed/FLOPS": 202144076171605.88, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8443593978881836, "optim/lr": 0.0001002, "optim/total_tokens": 88080384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 6.294613838195801, "created_at": "2025-01-15T06:00:32.276674+00:00"} {"global_step": 169, "acc_step": 0, "speed/wps": 12871.240151097862, "speed/FLOPS": 202160513510418.1, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9101779460906982, "optim/lr": 0.0001008, "optim/total_tokens": 88604672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 6.442990779876709, "created_at": "2025-01-15T06:00:42.467061+00:00"} {"global_step": 170, "acc_step": 0, "speed/wps": 12869.941985734931, "speed/FLOPS": 202140124039527.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7382975816726685, "optim/lr": 0.0001014, "optim/total_tokens": 89128960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 6.355175018310547, "created_at": "2025-01-15T06:00:52.656611+00:00"} {"global_step": 171, "acc_step": 0, "speed/wps": 12870.457477815184, "speed/FLOPS": 202148220551007.84, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9933062195777893, "optim/lr": 0.00010200000000000001, "optim/total_tokens": 89653248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 6.369632720947266, "created_at": "2025-01-15T06:01:02.844988+00:00"} {"global_step": 172, "acc_step": 0, "speed/wps": 12869.03496837089, "speed/FLOPS": 202125878085452.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.045178771018982, "optim/lr": 0.0001026, "optim/total_tokens": 90177536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 6.198520660400391, "created_at": "2025-01-15T06:01:13.035853+00:00"} {"global_step": 173, "acc_step": 0, "speed/wps": 12870.281962923988, "speed/FLOPS": 202145463848458.88, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8328323364257812, "optim/lr": 0.0001032, "optim/total_tokens": 90701824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 6.234740257263184, "created_at": "2025-01-15T06:01:23.228810+00:00"} {"global_step": 174, "acc_step": 0, "speed/wps": 12873.506009673163, "speed/FLOPS": 202196101932959.2, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7060696482658386, "optim/lr": 0.0001038, "optim/total_tokens": 91226112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 6.410252571105957, "created_at": "2025-01-15T06:01:33.414576+00:00"} {"global_step": 175, "acc_step": 0, "speed/wps": 12868.290838199187, "speed/FLOPS": 202114190498563.8, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7696258425712585, "optim/lr": 0.00010439999999999999, "optim/total_tokens": 91750400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.384557723999023, "created_at": "2025-01-15T06:01:43.604582+00:00"} {"global_step": 176, "acc_step": 0, "speed/wps": 12852.54124520601, "speed/FLOPS": 201866821498397.25, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7782318592071533, "optim/lr": 0.00010500000000000002, "optim/total_tokens": 92274688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 6.347922325134277, "created_at": "2025-01-15T06:01:53.808974+00:00"} {"global_step": 177, "acc_step": 0, "speed/wps": 12854.418841956352, "speed/FLOPS": 201896311735449.7, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5820114612579346, "optim/lr": 0.0001056, "optim/total_tokens": 92798976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 6.353024005889893, "created_at": "2025-01-15T06:02:04.010278+00:00"} {"global_step": 178, "acc_step": 0, "speed/wps": 12853.740532426147, "speed/FLOPS": 201885657952182.84, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7689597010612488, "optim/lr": 0.0001062, "optim/total_tokens": 93323264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 6.360607624053955, "created_at": "2025-01-15T06:02:14.211270+00:00"} {"global_step": 179, "acc_step": 0, "speed/wps": 12858.816422683292, "speed/FLOPS": 201965381783675.44, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.725888192653656, "optim/lr": 0.00010680000000000001, "optim/total_tokens": 93847552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 6.290322303771973, "created_at": "2025-01-15T06:02:24.411971+00:00"} {"global_step": 180, "acc_step": 0, "speed/wps": 12858.274845076668, "speed/FLOPS": 201956875563157.78, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5685674548149109, "optim/lr": 0.0001074, "optim/total_tokens": 94371840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 6.406412601470947, "created_at": "2025-01-15T06:02:34.613964+00:00"} {"global_step": 181, "acc_step": 0, "speed/wps": 12865.547040776451, "speed/FLOPS": 202071095389666.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7076352834701538, "optim/lr": 0.000108, "optim/total_tokens": 94896128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.48013973236084, "created_at": "2025-01-15T06:02:44.807682+00:00"} {"global_step": 182, "acc_step": 0, "speed/wps": 12863.891076981981, "speed/FLOPS": 202045086202742.1, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9099478125572205, "optim/lr": 0.00010860000000000001, "optim/total_tokens": 95420416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 6.279152870178223, "created_at": "2025-01-15T06:02:55.000402+00:00"} {"global_step": 183, "acc_step": 0, "speed/wps": 12867.995848060187, "speed/FLOPS": 202109557273071.88, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.081868052482605, "optim/lr": 0.00010920000000000001, "optim/total_tokens": 95944704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369048, "loss/out": 6.46730375289917, "created_at": "2025-01-15T06:03:05.190060+00:00"} {"global_step": 184, "acc_step": 0, "speed/wps": 12866.153184144789, "speed/FLOPS": 202080615704194.97, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0408529043197632, "optim/lr": 0.0001098, "optim/total_tokens": 96468992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 6.269406795501709, "created_at": "2025-01-15T06:03:15.384843+00:00"} {"global_step": 185, "acc_step": 0, "speed/wps": 12867.515447049844, "speed/FLOPS": 202102011915064.28, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8848649263381958, "optim/lr": 0.0001104, "optim/total_tokens": 96993280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 6.252609729766846, "created_at": "2025-01-15T06:03:25.576031+00:00"} {"global_step": 186, "acc_step": 0, "speed/wps": 12864.913490368997, "speed/FLOPS": 202061144609928.4, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0007580518722534, "optim/lr": 0.000111, "optim/total_tokens": 97517568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 6.3094024658203125, "created_at": "2025-01-15T06:03:35.769095+00:00"} {"global_step": 187, "acc_step": 0, "speed/wps": 12866.687986601946, "speed/FLOPS": 202089015511680.22, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.098418951034546, "optim/lr": 0.00011159999999999999, "optim/total_tokens": 98041856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 6.378536701202393, "created_at": "2025-01-15T06:03:45.960709+00:00"} {"global_step": 188, "acc_step": 0, "speed/wps": 12871.41059022861, "speed/FLOPS": 202163190491174.4, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9859809875488281, "optim/lr": 0.00011220000000000002, "optim/total_tokens": 98566144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 6.233794212341309, "created_at": "2025-01-15T06:03:56.149990+00:00"} {"global_step": 189, "acc_step": 0, "speed/wps": 12869.802390381708, "speed/FLOPS": 202137931502680.66, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9069012999534607, "optim/lr": 0.0001128, "optim/total_tokens": 99090432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 6.285008907318115, "created_at": "2025-01-15T06:04:06.340548+00:00"} {"global_step": 190, "acc_step": 0, "speed/wps": 12868.030718875249, "speed/FLOPS": 202110104967139.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9111846089363098, "optim/lr": 0.0001134, "optim/total_tokens": 99614720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 6.214839458465576, "created_at": "2025-01-15T06:04:16.530363+00:00"} {"global_step": 191, "acc_step": 0, "speed/wps": 12864.432458713185, "speed/FLOPS": 202053589346767.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.142890453338623, "optim/lr": 0.000114, "optim/total_tokens": 100139008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.25375509262085, "created_at": "2025-01-15T06:04:26.729247+00:00"} {"global_step": 192, "acc_step": 0, "speed/wps": 12871.582840891097, "speed/FLOPS": 202165895924526.0, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7663801312446594, "optim/lr": 0.0001146, "optim/total_tokens": 100663296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 6.340198516845703, "created_at": "2025-01-15T06:04:36.916675+00:00"} {"global_step": 193, "acc_step": 0, "speed/wps": 12867.796529882911, "speed/FLOPS": 202106426707205.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1469414234161377, "optim/lr": 0.0001152, "optim/total_tokens": 101187584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 6.313780784606934, "created_at": "2025-01-15T06:04:47.109062+00:00"} {"global_step": 194, "acc_step": 0, "speed/wps": 12863.413003460382, "speed/FLOPS": 202037577401143.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8492510318756104, "optim/lr": 0.00011580000000000001, "optim/total_tokens": 101711872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.273621559143066, "created_at": "2025-01-15T06:04:57.304054+00:00"} {"global_step": 195, "acc_step": 0, "speed/wps": 12864.004311425775, "speed/FLOPS": 202046864705282.3, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8538012504577637, "optim/lr": 0.00011640000000000001, "optim/total_tokens": 102236160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 6.269147872924805, "created_at": "2025-01-15T06:05:07.499513+00:00"} {"global_step": 196, "acc_step": 0, "speed/wps": 12863.32199071057, "speed/FLOPS": 202036147920843.12, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1400163173675537, "optim/lr": 0.000117, "optim/total_tokens": 102760448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 6.3018035888671875, "created_at": "2025-01-15T06:05:17.693156+00:00"} {"global_step": 197, "acc_step": 0, "speed/wps": 12865.805604794408, "speed/FLOPS": 202075156492873.2, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.6449283361434937, "optim/lr": 0.0001176, "optim/total_tokens": 103284736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 6.384906768798828, "created_at": "2025-01-15T06:05:27.889964+00:00"} {"global_step": 198, "acc_step": 0, "speed/wps": 12866.893281598606, "speed/FLOPS": 202092239951707.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7028873562812805, "optim/lr": 0.0001182, "optim/total_tokens": 103809024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 6.199958801269531, "created_at": "2025-01-15T06:05:38.081088+00:00"} {"global_step": 199, "acc_step": 0, "speed/wps": 12861.072131106794, "speed/FLOPS": 202000810784134.72, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7123018503189087, "optim/lr": 0.00011880000000000001, "optim/total_tokens": 104333312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.167636394500732, "created_at": "2025-01-15T06:05:48.278356+00:00"} {"global_step": 200, "acc_step": 0, "speed/wps": 12869.71386806825, "speed/FLOPS": 202136541138105.38, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6785900592803955, "optim/lr": 0.0001194, "optim/total_tokens": 104857600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.138528823852539, "created_at": "2025-01-15T06:05:58.468223+00:00"} {"global_step": 201, "acc_step": 0, "speed/wps": 12864.17219785926, "speed/FLOPS": 202049501592420.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7309972643852234, "optim/lr": 0.00012, "optim/total_tokens": 105381888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 6.1965765953063965, "created_at": "2025-01-15T06:06:08.664191+00:00"} {"global_step": 202, "acc_step": 0, "speed/wps": 12868.530423659437, "speed/FLOPS": 202117953517441.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.724270224571228, "optim/lr": 0.0001206, "optim/total_tokens": 105906176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 6.120830535888672, "created_at": "2025-01-15T06:06:18.857717+00:00"} {"global_step": 203, "acc_step": 0, "speed/wps": 12868.115927150551, "speed/FLOPS": 202111443280192.0, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7487147450447083, "optim/lr": 0.00012119999999999999, "optim/total_tokens": 106430464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 6.217230796813965, "created_at": "2025-01-15T06:06:29.051382+00:00"} {"global_step": 204, "acc_step": 0, "speed/wps": 12865.553240037767, "speed/FLOPS": 202071192757584.6, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9226841926574707, "optim/lr": 0.00012179999999999999, "optim/total_tokens": 106954752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 6.321343421936035, "created_at": "2025-01-15T06:06:39.248351+00:00"} {"global_step": 205, "acc_step": 0, "speed/wps": 12867.137700598048, "speed/FLOPS": 202096078887960.8, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.6857073307037354, "optim/lr": 0.00012240000000000002, "optim/total_tokens": 107479040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 6.163058757781982, "created_at": "2025-01-15T06:06:49.438410+00:00"} {"global_step": 206, "acc_step": 0, "speed/wps": 12865.92555362171, "speed/FLOPS": 202077040454032.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5971896648406982, "optim/lr": 0.000123, "optim/total_tokens": 108003328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 6.239707946777344, "created_at": "2025-01-15T06:06:59.630112+00:00"} {"global_step": 207, "acc_step": 0, "speed/wps": 12867.369700960035, "speed/FLOPS": 202099722772447.7, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5596458911895752, "optim/lr": 0.0001236, "optim/total_tokens": 108527616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.212241172790527, "created_at": "2025-01-15T06:07:09.823122+00:00"} {"global_step": 208, "acc_step": 0, "speed/wps": 12871.066400544209, "speed/FLOPS": 202157784519214.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5498385429382324, "optim/lr": 0.0001242, "optim/total_tokens": 109051904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 6.26035737991333, "created_at": "2025-01-15T06:07:20.012502+00:00"} {"global_step": 209, "acc_step": 0, "speed/wps": 12870.237620636675, "speed/FLOPS": 202144767391903.97, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5414058566093445, "optim/lr": 0.0001248, "optim/total_tokens": 109576192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 6.314898490905762, "created_at": "2025-01-15T06:07:30.204913+00:00"} {"global_step": 210, "acc_step": 0, "speed/wps": 12865.475316310225, "speed/FLOPS": 202069968858365.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.505725085735321, "optim/lr": 0.00012539999999999999, "optim/total_tokens": 110100480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 6.0735297203063965, "created_at": "2025-01-15T06:07:40.399520+00:00"} {"global_step": 211, "acc_step": 0, "speed/wps": 12866.281502521077, "speed/FLOPS": 202082631120622.5, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6604998707771301, "optim/lr": 0.000126, "optim/total_tokens": 110624768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 6.186749458312988, "created_at": "2025-01-15T06:07:50.593763+00:00"} {"global_step": 212, "acc_step": 0, "speed/wps": 12859.979535722056, "speed/FLOPS": 201983650072234.03, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6250146627426147, "optim/lr": 0.00012660000000000001, "optim/total_tokens": 111149056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 6.122768878936768, "created_at": "2025-01-15T06:08:00.795822+00:00"} {"global_step": 213, "acc_step": 0, "speed/wps": 12865.809299095048, "speed/FLOPS": 202075214516941.22, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5743023753166199, "optim/lr": 0.0001272, "optim/total_tokens": 111673344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 6.220978736877441, "created_at": "2025-01-15T06:08:10.988056+00:00"} {"global_step": 214, "acc_step": 0, "speed/wps": 12863.658494233583, "speed/FLOPS": 202041433171076.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47958698868751526, "optim/lr": 0.0001278, "optim/total_tokens": 112197632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 6.194436550140381, "created_at": "2025-01-15T06:08:21.184805+00:00"} {"global_step": 215, "acc_step": 0, "speed/wps": 12867.383294737956, "speed/FLOPS": 202099936281409.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5760096907615662, "optim/lr": 0.0001284, "optim/total_tokens": 112721920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 6.18904972076416, "created_at": "2025-01-15T06:08:31.380092+00:00"} {"global_step": 216, "acc_step": 0, "speed/wps": 12864.349399347222, "speed/FLOPS": 202052284785289.12, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6503686904907227, "optim/lr": 0.000129, "optim/total_tokens": 113246208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 6.063530921936035, "created_at": "2025-01-15T06:08:41.575548+00:00"} {"global_step": 217, "acc_step": 0, "speed/wps": 12851.246902756488, "speed/FLOPS": 201846492071614.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.81917405128479, "optim/lr": 0.0001296, "optim/total_tokens": 113770496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 6.199063301086426, "created_at": "2025-01-15T06:08:51.785080+00:00"} {"global_step": 218, "acc_step": 0, "speed/wps": 12859.558206091388, "speed/FLOPS": 201977032511416.78, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0064767599105835, "optim/lr": 0.0001302, "optim/total_tokens": 114294784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 6.169016361236572, "created_at": "2025-01-15T06:09:01.985951+00:00"} {"global_step": 219, "acc_step": 0, "speed/wps": 12856.863137236602, "speed/FLOPS": 201934702751639.7, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1712992191314697, "optim/lr": 0.0001308, "optim/total_tokens": 114819072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 6.1769609451293945, "created_at": "2025-01-15T06:09:12.184536+00:00"} {"global_step": 220, "acc_step": 0, "speed/wps": 12858.77074717404, "speed/FLOPS": 201964664387037.9, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9118359684944153, "optim/lr": 0.0001314, "optim/total_tokens": 115343360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 6.3010101318359375, "created_at": "2025-01-15T06:09:22.388972+00:00"} {"global_step": 221, "acc_step": 0, "speed/wps": 12854.31954550026, "speed/FLOPS": 201894752148159.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0568301677703857, "optim/lr": 0.00013199999999999998, "optim/total_tokens": 115867648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467150, "loss/out": 6.148226737976074, "created_at": "2025-01-15T06:09:32.600123+00:00"} {"global_step": 222, "acc_step": 0, "speed/wps": 12865.960450854962, "speed/FLOPS": 202077588563034.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9044404625892639, "optim/lr": 0.00013260000000000002, "optim/total_tokens": 116391936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 6.0703654289245605, "created_at": "2025-01-15T06:09:42.795564+00:00"} {"global_step": 223, "acc_step": 0, "speed/wps": 12861.879876646577, "speed/FLOPS": 202013497537795.72, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.798576295375824, "optim/lr": 0.0001332, "optim/total_tokens": 116916224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 6.185610771179199, "created_at": "2025-01-15T06:09:52.992223+00:00"} {"global_step": 224, "acc_step": 0, "speed/wps": 12861.227292239133, "speed/FLOPS": 202003247802931.84, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0361871719360352, "optim/lr": 0.0001338, "optim/total_tokens": 117440512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.155908584594727, "created_at": "2025-01-15T06:10:03.188160+00:00"} {"global_step": 225, "acc_step": 0, "speed/wps": 12862.978710781705, "speed/FLOPS": 202030756237844.22, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.915643572807312, "optim/lr": 0.00013440000000000001, "optim/total_tokens": 117964800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 6.125782012939453, "created_at": "2025-01-15T06:10:13.383217+00:00"} {"global_step": 226, "acc_step": 0, "speed/wps": 12866.846014739178, "speed/FLOPS": 202091497560728.88, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8085552453994751, "optim/lr": 0.000135, "optim/total_tokens": 118489088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 6.0485453605651855, "created_at": "2025-01-15T06:10:23.577903+00:00"} {"global_step": 227, "acc_step": 0, "speed/wps": 12868.163024499676, "speed/FLOPS": 202112183008778.56, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7042124271392822, "optim/lr": 0.0001356, "optim/total_tokens": 119013376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 6.155132293701172, "created_at": "2025-01-15T06:10:33.768911+00:00"} {"global_step": 228, "acc_step": 0, "speed/wps": 12868.899707722823, "speed/FLOPS": 202123753631106.9, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.915107786655426, "optim/lr": 0.0001362, "optim/total_tokens": 119537664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350412, "loss/out": 6.103359699249268, "created_at": "2025-01-15T06:10:43.959357+00:00"} {"global_step": 229, "acc_step": 0, "speed/wps": 12863.097206694072, "speed/FLOPS": 202032617378978.4, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8051996827125549, "optim/lr": 0.00013680000000000002, "optim/total_tokens": 120061952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 6.100149154663086, "created_at": "2025-01-15T06:10:54.153423+00:00"} {"global_step": 230, "acc_step": 0, "speed/wps": 12866.276922240373, "speed/FLOPS": 202082559181020.1, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5969291925430298, "optim/lr": 0.0001374, "optim/total_tokens": 120586240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 5.952427864074707, "created_at": "2025-01-15T06:11:04.344350+00:00"} {"global_step": 231, "acc_step": 0, "speed/wps": 12854.938788295567, "speed/FLOPS": 201904478207188.2, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.909939706325531, "optim/lr": 0.000138, "optim/total_tokens": 121110528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 6.055379867553711, "created_at": "2025-01-15T06:11:14.545093+00:00"} {"global_step": 232, "acc_step": 0, "speed/wps": 12860.916062891534, "speed/FLOPS": 201998359518353.88, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.690605878829956, "optim/lr": 0.0001386, "optim/total_tokens": 121634816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467434, "loss/out": 6.097685813903809, "created_at": "2025-01-15T06:11:24.742225+00:00"} {"global_step": 233, "acc_step": 0, "speed/wps": 12852.106948079845, "speed/FLOPS": 201860000265244.03, "speed/curr_iter_time": 1.2884, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5547188520431519, "optim/lr": 0.0001392, "optim/total_tokens": 122159104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 6.061798095703125, "created_at": "2025-01-15T06:11:34.948351+00:00"} {"global_step": 234, "acc_step": 0, "speed/wps": 12845.289153864282, "speed/FLOPS": 201752917438457.6, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5794796943664551, "optim/lr": 0.0001398, "optim/total_tokens": 122683392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 6.1006879806518555, "created_at": "2025-01-15T06:11:45.156991+00:00"} {"global_step": 235, "acc_step": 0, "speed/wps": 12855.697949157984, "speed/FLOPS": 201916401871728.78, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4613018333911896, "optim/lr": 0.0001404, "optim/total_tokens": 123207680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 6.043722152709961, "created_at": "2025-01-15T06:11:55.360465+00:00"} {"global_step": 236, "acc_step": 0, "speed/wps": 12857.094109478447, "speed/FLOPS": 201938330488086.97, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5407420992851257, "optim/lr": 0.000141, "optim/total_tokens": 123731968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 6.0990447998046875, "created_at": "2025-01-15T06:12:05.560002+00:00"} {"global_step": 237, "acc_step": 0, "speed/wps": 12851.541530343742, "speed/FLOPS": 201851119602733.16, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5433853268623352, "optim/lr": 0.0001416, "optim/total_tokens": 124256256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 6.135260581970215, "created_at": "2025-01-15T06:12:15.764529+00:00"} {"global_step": 238, "acc_step": 0, "speed/wps": 12863.258549080529, "speed/FLOPS": 202035151482865.38, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5671433806419373, "optim/lr": 0.0001422, "optim/total_tokens": 124780544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472483, "loss/out": 6.119638919830322, "created_at": "2025-01-15T06:12:25.960712+00:00"} {"global_step": 239, "acc_step": 0, "speed/wps": 12863.757253072676, "speed/FLOPS": 202042984314353.12, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6576930284500122, "optim/lr": 0.0001428, "optim/total_tokens": 125304832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.952840328216553, "created_at": "2025-01-15T06:12:36.156450+00:00"} {"global_step": 240, "acc_step": 0, "speed/wps": 12861.99411126349, "speed/FLOPS": 202015291749429.12, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7875961661338806, "optim/lr": 0.00014340000000000002, "optim/total_tokens": 125829120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 6.0410308837890625, "created_at": "2025-01-15T06:12:46.355473+00:00"} {"global_step": 241, "acc_step": 0, "speed/wps": 12866.862732964491, "speed/FLOPS": 202091760143430.66, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5493472218513489, "optim/lr": 0.000144, "optim/total_tokens": 126353408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 6.076478958129883, "created_at": "2025-01-15T06:12:56.548515+00:00"} {"global_step": 242, "acc_step": 0, "speed/wps": 12857.946003900513, "speed/FLOPS": 201951710660611.16, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5203214287757874, "optim/lr": 0.0001446, "optim/total_tokens": 126877696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 6.029601097106934, "created_at": "2025-01-15T06:13:06.750508+00:00"} {"global_step": 243, "acc_step": 0, "speed/wps": 12863.063210257771, "speed/FLOPS": 202032083418230.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7033182978630066, "optim/lr": 0.0001452, "optim/total_tokens": 127401984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 6.14574670791626, "created_at": "2025-01-15T06:13:16.947565+00:00"} {"global_step": 244, "acc_step": 0, "speed/wps": 12862.177911902523, "speed/FLOPS": 202018178591033.7, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0149893760681152, "optim/lr": 0.0001458, "optim/total_tokens": 127926272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469009, "loss/out": 6.040755271911621, "created_at": "2025-01-15T06:13:27.142342+00:00"} {"global_step": 245, "acc_step": 0, "speed/wps": 12863.399382798587, "speed/FLOPS": 202037363469933.28, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0488420724868774, "optim/lr": 0.0001464, "optim/total_tokens": 128450560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 6.071481704711914, "created_at": "2025-01-15T06:13:37.337306+00:00"} {"global_step": 246, "acc_step": 0, "speed/wps": 12857.769781086012, "speed/FLOPS": 201948942839153.88, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6714775562286377, "optim/lr": 0.000147, "optim/total_tokens": 128974848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 6.018745422363281, "created_at": "2025-01-15T06:13:47.535194+00:00"} {"global_step": 247, "acc_step": 0, "speed/wps": 12854.998586600199, "speed/FLOPS": 201905417421733.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9122664332389832, "optim/lr": 0.0001476, "optim/total_tokens": 129499136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.073256969451904, "created_at": "2025-01-15T06:13:57.736555+00:00"} {"global_step": 248, "acc_step": 0, "speed/wps": 12857.197508154355, "speed/FLOPS": 201939954506376.78, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.9737152457237244, "optim/lr": 0.0001482, "optim/total_tokens": 130023424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 5.999083518981934, "created_at": "2025-01-15T06:14:07.943499+00:00"} {"global_step": 249, "acc_step": 0, "speed/wps": 12859.578052074647, "speed/FLOPS": 201977344219855.12, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.8736701607704163, "optim/lr": 0.00014879999999999998, "optim/total_tokens": 130547712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 6.032811641693115, "created_at": "2025-01-15T06:14:18.140976+00:00"} {"global_step": 250, "acc_step": 0, "speed/wps": 12860.684914528778, "speed/FLOPS": 201994729015685.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0093532800674438, "optim/lr": 0.0001494, "optim/total_tokens": 131072000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 6.031797409057617, "created_at": "2025-01-15T06:14:28.340095+00:00"} {"global_step": 251, "acc_step": 0, "speed/wps": 12860.672389068526, "speed/FLOPS": 201994532286120.6, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9952459931373596, "optim/lr": 0.00015000000000000001, "optim/total_tokens": 131596288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 6.056903839111328, "created_at": "2025-01-15T06:14:38.538896+00:00"} {"global_step": 252, "acc_step": 0, "speed/wps": 12865.610853113276, "speed/FLOPS": 202072097650104.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.898265540599823, "optim/lr": 0.0001506, "optim/total_tokens": 132120576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 6.020937919616699, "created_at": "2025-01-15T06:14:48.732556+00:00"} {"global_step": 253, "acc_step": 0, "speed/wps": 12861.509536715, "speed/FLOPS": 202007680840269.9, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6342220902442932, "optim/lr": 0.00015120000000000002, "optim/total_tokens": 132644864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.0132904052734375, "created_at": "2025-01-15T06:14:58.927132+00:00"} {"global_step": 254, "acc_step": 0, "speed/wps": 12864.573134692848, "speed/FLOPS": 202055798856336.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6866869926452637, "optim/lr": 0.0001518, "optim/total_tokens": 133169152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 6.041234016418457, "created_at": "2025-01-15T06:15:09.121937+00:00"} {"global_step": 255, "acc_step": 0, "speed/wps": 12862.976243119776, "speed/FLOPS": 202030717479822.72, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7383437156677246, "optim/lr": 0.0001524, "optim/total_tokens": 133693440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 6.070797920227051, "created_at": "2025-01-15T06:15:19.319200+00:00"} {"global_step": 256, "acc_step": 0, "speed/wps": 12864.226060490055, "speed/FLOPS": 202050347579051.97, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5660769939422607, "optim/lr": 0.000153, "optim/total_tokens": 134217728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284202, "loss/out": 6.031733512878418, "created_at": "2025-01-15T06:15:29.513807+00:00"} {"global_step": 257, "acc_step": 0, "speed/wps": 12865.825287760994, "speed/FLOPS": 202075465640910.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.052424669265747, "optim/lr": 0.00015360000000000002, "optim/total_tokens": 134742016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 5.957012176513672, "created_at": "2025-01-15T06:15:39.709086+00:00"} {"global_step": 258, "acc_step": 0, "speed/wps": 12865.641409916709, "speed/FLOPS": 202072577586691.5, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7724617719650269, "optim/lr": 0.0001542, "optim/total_tokens": 135266304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314286, "loss/out": 5.994149208068848, "created_at": "2025-01-15T06:15:49.900333+00:00"} {"global_step": 259, "acc_step": 0, "speed/wps": 12866.743293655776, "speed/FLOPS": 202089884184960.47, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6651655435562134, "optim/lr": 0.0001548, "optim/total_tokens": 135790592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 5.918166160583496, "created_at": "2025-01-15T06:16:00.093078+00:00"} {"global_step": 260, "acc_step": 0, "speed/wps": 12866.214560643331, "speed/FLOPS": 202081579706444.75, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7997202277183533, "optim/lr": 0.0001554, "optim/total_tokens": 136314880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 6.04630184173584, "created_at": "2025-01-15T06:16:10.285609+00:00"} {"global_step": 261, "acc_step": 0, "speed/wps": 12861.587224348606, "speed/FLOPS": 202008901031309.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5394229292869568, "optim/lr": 0.000156, "optim/total_tokens": 136839168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 5.984713077545166, "created_at": "2025-01-15T06:16:20.482752+00:00"} {"global_step": 262, "acc_step": 0, "speed/wps": 12868.74768556627, "speed/FLOPS": 202121365914237.9, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6723588705062866, "optim/lr": 0.0001566, "optim/total_tokens": 137363456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 5.9136552810668945, "created_at": "2025-01-15T06:16:30.675085+00:00"} {"global_step": 263, "acc_step": 0, "speed/wps": 12850.347840701923, "speed/FLOPS": 201832371066606.9, "speed/curr_iter_time": 1.2889, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.5397984385490417, "optim/lr": 0.0001572, "optim/total_tokens": 137887744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 5.9678635597229, "created_at": "2025-01-15T06:16:40.880191+00:00"} {"global_step": 264, "acc_step": 0, "speed/wps": 12818.236293566635, "speed/FLOPS": 201328015092957.4, "speed/curr_iter_time": 1.3008, "speed/data_load_time": 0.0043, "optim/grad_norm": 0.6423426866531372, "optim/lr": 0.00015780000000000001, "optim/total_tokens": 138412032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 260158, "loss/out": 5.928365707397461, "created_at": "2025-01-15T06:16:51.110469+00:00"} {"global_step": 265, "acc_step": 0, "speed/wps": 12822.891881629845, "speed/FLOPS": 201401137500938.5, "speed/curr_iter_time": 1.2896, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6625920534133911, "optim/lr": 0.0001584, "optim/total_tokens": 138936320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481564, "loss/out": 5.99961519241333, "created_at": "2025-01-15T06:17:01.336542+00:00"} {"global_step": 266, "acc_step": 0, "speed/wps": 12849.330161259346, "speed/FLOPS": 201816387012523.16, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.6178664565086365, "optim/lr": 0.000159, "optim/total_tokens": 139460608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 5.965884208679199, "created_at": "2025-01-15T06:17:11.541001+00:00"} {"global_step": 267, "acc_step": 0, "speed/wps": 12863.54891537799, "speed/FLOPS": 202039712084572.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8775327205657959, "optim/lr": 0.0001596, "optim/total_tokens": 139984896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 5.987802028656006, "created_at": "2025-01-15T06:17:21.739574+00:00"} {"global_step": 268, "acc_step": 0, "speed/wps": 12862.925400559927, "speed/FLOPS": 202029918927555.28, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9776220321655273, "optim/lr": 0.00016020000000000002, "optim/total_tokens": 140509184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 6.032313823699951, "created_at": "2025-01-15T06:17:31.935072+00:00"} {"global_step": 269, "acc_step": 0, "speed/wps": 12859.26500145802, "speed/FLOPS": 201972427329744.34, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.001, "optim/grad_norm": 0.614384114742279, "optim/lr": 0.0001608, "optim/total_tokens": 141033472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 5.858595848083496, "created_at": "2025-01-15T06:17:42.136081+00:00"} {"global_step": 270, "acc_step": 0, "speed/wps": 12863.976870088807, "speed/FLOPS": 202046433701376.88, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.0500034093856812, "optim/lr": 0.0001614, "optim/total_tokens": 141557760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 6.065915584564209, "created_at": "2025-01-15T06:17:52.330998+00:00"} {"global_step": 271, "acc_step": 0, "speed/wps": 12864.42995051276, "speed/FLOPS": 202053549952032.8, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.9125368595123291, "optim/lr": 0.000162, "optim/total_tokens": 142082048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 5.852790832519531, "created_at": "2025-01-15T06:18:02.527461+00:00"} {"global_step": 272, "acc_step": 0, "speed/wps": 12865.233407820293, "speed/FLOPS": 202066169353113.66, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8118820786476135, "optim/lr": 0.0001626, "optim/total_tokens": 142606336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 5.950486660003662, "created_at": "2025-01-15T06:18:12.720161+00:00"} {"global_step": 273, "acc_step": 0, "speed/wps": 12860.279614624718, "speed/FLOPS": 201988363223751.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.1237839460372925, "optim/lr": 0.00016319999999999998, "optim/total_tokens": 143130624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.8542585372924805, "created_at": "2025-01-15T06:18:22.917710+00:00"} {"global_step": 274, "acc_step": 0, "speed/wps": 12863.38262066509, "speed/FLOPS": 202037100197593.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6885740160942078, "optim/lr": 0.00016380000000000002, "optim/total_tokens": 143654912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 6.033660888671875, "created_at": "2025-01-15T06:18:33.120190+00:00"} {"global_step": 275, "acc_step": 0, "speed/wps": 12862.887663030568, "speed/FLOPS": 202029326207799.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.694474458694458, "optim/lr": 0.0001644, "optim/total_tokens": 144179200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 5.9280266761779785, "created_at": "2025-01-15T06:18:43.321561+00:00"} {"global_step": 276, "acc_step": 0, "speed/wps": 12865.225767032252, "speed/FLOPS": 202066049344037.94, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6618914008140564, "optim/lr": 0.000165, "optim/total_tokens": 144703488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 5.989768981933594, "created_at": "2025-01-15T06:18:53.515338+00:00"} {"global_step": 277, "acc_step": 0, "speed/wps": 12866.856704187689, "speed/FLOPS": 202091665453206.56, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8977233171463013, "optim/lr": 0.00016560000000000001, "optim/total_tokens": 145227776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.905142784118652, "created_at": "2025-01-15T06:19:03.706229+00:00"} {"global_step": 278, "acc_step": 0, "speed/wps": 12865.08758993903, "speed/FLOPS": 202063879082913.66, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8004688024520874, "optim/lr": 0.0001662, "optim/total_tokens": 145752064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.924827575683594, "created_at": "2025-01-15T06:19:13.906224+00:00"} {"global_step": 279, "acc_step": 0, "speed/wps": 12863.089308622957, "speed/FLOPS": 202032493328918.44, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6038275361061096, "optim/lr": 0.0001668, "optim/total_tokens": 146276352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 6.000880241394043, "created_at": "2025-01-15T06:19:24.102313+00:00"} {"global_step": 280, "acc_step": 0, "speed/wps": 12868.003412133145, "speed/FLOPS": 202109676077231.8, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6655195355415344, "optim/lr": 0.0001674, "optim/total_tokens": 146800640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 5.913969039916992, "created_at": "2025-01-15T06:19:34.295873+00:00"} {"global_step": 281, "acc_step": 0, "speed/wps": 12859.3894906823, "speed/FLOPS": 201974382604077.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5244166851043701, "optim/lr": 0.00016800000000000002, "optim/total_tokens": 147324928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.017884731292725, "created_at": "2025-01-15T06:19:44.495085+00:00"} {"global_step": 282, "acc_step": 0, "speed/wps": 12863.90187322284, "speed/FLOPS": 202045255772540.38, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4919401705265045, "optim/lr": 0.0001686, "optim/total_tokens": 147849216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464032, "loss/out": 5.92171573638916, "created_at": "2025-01-15T06:19:54.692982+00:00"} {"global_step": 283, "acc_step": 0, "speed/wps": 12864.635186394122, "speed/FLOPS": 202056773463573.25, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5972114205360413, "optim/lr": 0.0001692, "optim/total_tokens": 148373504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416907, "loss/out": 5.92332649230957, "created_at": "2025-01-15T06:20:04.887147+00:00"} {"global_step": 284, "acc_step": 0, "speed/wps": 12861.047863792, "speed/FLOPS": 202000429632609.12, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6055188775062561, "optim/lr": 0.0001698, "optim/total_tokens": 148897792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 5.884049892425537, "created_at": "2025-01-15T06:20:15.086367+00:00"} {"global_step": 285, "acc_step": 0, "speed/wps": 12862.088327198282, "speed/FLOPS": 202016771540151.3, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5215648412704468, "optim/lr": 0.00017040000000000002, "optim/total_tokens": 149422080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394754, "loss/out": 5.954205513000488, "created_at": "2025-01-15T06:20:25.282410+00:00"} {"global_step": 286, "acc_step": 0, "speed/wps": 12864.58897329507, "speed/FLOPS": 202056047623349.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.6768203377723694, "optim/lr": 0.000171, "optim/total_tokens": 149946368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371737, "loss/out": 5.874295711517334, "created_at": "2025-01-15T06:20:35.475211+00:00"} {"global_step": 287, "acc_step": 0, "speed/wps": 12867.400057887986, "speed/FLOPS": 202100199569715.97, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6065649390220642, "optim/lr": 0.0001716, "optim/total_tokens": 150470656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308039, "loss/out": 5.910484313964844, "created_at": "2025-01-15T06:20:45.665038+00:00"} {"global_step": 288, "acc_step": 0, "speed/wps": 12865.844546193643, "speed/FLOPS": 202075768121059.25, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5145668983459473, "optim/lr": 0.0001722, "optim/total_tokens": 150994944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 6.080188751220703, "created_at": "2025-01-15T06:20:55.858261+00:00"} {"global_step": 289, "acc_step": 0, "speed/wps": 12867.789614313857, "speed/FLOPS": 202106318088690.9, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5332404971122742, "optim/lr": 0.0001728, "optim/total_tokens": 151519232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.045726776123047, "created_at": "2025-01-15T06:21:06.047811+00:00"} {"global_step": 290, "acc_step": 0, "speed/wps": 12863.338262417932, "speed/FLOPS": 202036403490366.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5871739387512207, "optim/lr": 0.00017339999999999999, "optim/total_tokens": 152043520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 6.027541160583496, "created_at": "2025-01-15T06:21:16.245648+00:00"} {"global_step": 291, "acc_step": 0, "speed/wps": 12861.907083352953, "speed/FLOPS": 202013924856505.28, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4751407206058502, "optim/lr": 0.000174, "optim/total_tokens": 152567808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 5.95528507232666, "created_at": "2025-01-15T06:21:26.441075+00:00"} {"global_step": 292, "acc_step": 0, "speed/wps": 12862.176971509964, "speed/FLOPS": 202018163820876.3, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6039879322052002, "optim/lr": 0.00017460000000000002, "optim/total_tokens": 153092096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 5.8230977058410645, "created_at": "2025-01-15T06:21:36.635303+00:00"} {"global_step": 293, "acc_step": 0, "speed/wps": 12864.079175440616, "speed/FLOPS": 202048040547509.34, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6272299289703369, "optim/lr": 0.0001752, "optim/total_tokens": 153616384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.988805294036865, "created_at": "2025-01-15T06:21:46.827998+00:00"} {"global_step": 294, "acc_step": 0, "speed/wps": 12865.793315173447, "speed/FLOPS": 202074963467488.28, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5982754230499268, "optim/lr": 0.0001758, "optim/total_tokens": 154140672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 5.97502326965332, "created_at": "2025-01-15T06:21:57.019168+00:00"} {"global_step": 295, "acc_step": 0, "speed/wps": 12868.874108235961, "speed/FLOPS": 202123351555988.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5695486068725586, "optim/lr": 0.0001764, "optim/total_tokens": 154664960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402220, "loss/out": 5.821798324584961, "created_at": "2025-01-15T06:22:07.210580+00:00"} {"global_step": 296, "acc_step": 0, "speed/wps": 12868.703773926612, "speed/FLOPS": 202120676221596.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7743252515792847, "optim/lr": 0.000177, "optim/total_tokens": 155189248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347841, "loss/out": 5.829671859741211, "created_at": "2025-01-15T06:22:17.401270+00:00"} {"global_step": 297, "acc_step": 0, "speed/wps": 12864.9696676974, "speed/FLOPS": 202062026952066.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6073839664459229, "optim/lr": 0.0001776, "optim/total_tokens": 155713536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 5.938418865203857, "created_at": "2025-01-15T06:22:27.594026+00:00"} {"global_step": 298, "acc_step": 0, "speed/wps": 12864.47656227903, "speed/FLOPS": 202054282053873.34, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6695929765701294, "optim/lr": 0.0001782, "optim/total_tokens": 156237824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 5.9847612380981445, "created_at": "2025-01-15T06:22:37.790719+00:00"} {"global_step": 299, "acc_step": 0, "speed/wps": 12865.001231645021, "speed/FLOPS": 202062522707237.5, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6119787096977234, "optim/lr": 0.0001788, "optim/total_tokens": 156762112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 5.957975387573242, "created_at": "2025-01-15T06:22:47.984227+00:00"} {"global_step": 300, "acc_step": 0, "speed/wps": 12862.005714211873, "speed/FLOPS": 202015473989677.62, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5751529335975647, "optim/lr": 0.0001794, "optim/total_tokens": 157286400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 5.960322856903076, "created_at": "2025-01-15T06:22:58.182767+00:00"} {"global_step": 301, "acc_step": 0, "speed/wps": 12866.18459511001, "speed/FLOPS": 202081109056566.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5308234691619873, "optim/lr": 0.00017999999999999998, "optim/total_tokens": 157810688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.94257926940918, "created_at": "2025-01-15T06:23:08.377161+00:00"} {"global_step": 302, "acc_step": 0, "speed/wps": 12865.286583111098, "speed/FLOPS": 202067004544126.25, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4959542751312256, "optim/lr": 0.0001806, "optim/total_tokens": 158334976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 5.8554840087890625, "created_at": "2025-01-15T06:23:18.571045+00:00"} {"global_step": 303, "acc_step": 0, "speed/wps": 12866.773032407074, "speed/FLOPS": 202090351272914.94, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5080233216285706, "optim/lr": 0.0001812, "optim/total_tokens": 158859264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 5.883642196655273, "created_at": "2025-01-15T06:23:28.774325+00:00"} {"global_step": 304, "acc_step": 0, "speed/wps": 12870.27700669457, "speed/FLOPS": 202145386004065.3, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5582484602928162, "optim/lr": 0.0001818, "optim/total_tokens": 159383552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.863152503967285, "created_at": "2025-01-15T06:23:38.963703+00:00"} {"global_step": 305, "acc_step": 0, "speed/wps": 12867.187557694628, "speed/FLOPS": 202096861962173.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5362023115158081, "optim/lr": 0.00018240000000000002, "optim/total_tokens": 159907840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 5.868289947509766, "created_at": "2025-01-15T06:23:49.156660+00:00"} {"global_step": 306, "acc_step": 0, "speed/wps": 12868.594253827925, "speed/FLOPS": 202118956057949.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5599290132522583, "optim/lr": 0.000183, "optim/total_tokens": 160432128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 5.832460403442383, "created_at": "2025-01-15T06:23:59.350298+00:00"} {"global_step": 307, "acc_step": 0, "speed/wps": 12864.235434494103, "speed/FLOPS": 202050494810666.78, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6029773354530334, "optim/lr": 0.0001836, "optim/total_tokens": 160956416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 5.855600833892822, "created_at": "2025-01-15T06:24:09.550802+00:00"} {"global_step": 308, "acc_step": 0, "speed/wps": 12865.748460432053, "speed/FLOPS": 202074258962137.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8329247832298279, "optim/lr": 0.0001842, "optim/total_tokens": 161480704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 5.941386699676514, "created_at": "2025-01-15T06:24:19.745404+00:00"} {"global_step": 309, "acc_step": 0, "speed/wps": 12866.337058974423, "speed/FLOPS": 202083503711064.88, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8831405639648438, "optim/lr": 0.00018480000000000002, "optim/total_tokens": 162004992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391020, "loss/out": 6.011287212371826, "created_at": "2025-01-15T06:24:29.938784+00:00"} {"global_step": 310, "acc_step": 0, "speed/wps": 12863.206080676435, "speed/FLOPS": 202034327394479.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0549235343933105, "optim/lr": 0.0001854, "optim/total_tokens": 162529280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372985, "loss/out": 5.927837371826172, "created_at": "2025-01-15T06:24:40.132953+00:00"} {"global_step": 311, "acc_step": 0, "speed/wps": 12863.23690351939, "speed/FLOPS": 202034811509583.56, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0085440874099731, "optim/lr": 0.000186, "optim/total_tokens": 163053568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 5.914446830749512, "created_at": "2025-01-15T06:24:50.327154+00:00"} {"global_step": 312, "acc_step": 0, "speed/wps": 12864.635729678414, "speed/FLOPS": 202056781996599.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9032462239265442, "optim/lr": 0.0001866, "optim/total_tokens": 163577856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 5.880931854248047, "created_at": "2025-01-15T06:25:00.522135+00:00"} {"global_step": 313, "acc_step": 0, "speed/wps": 12865.96515323339, "speed/FLOPS": 202077662420349.9, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6943271160125732, "optim/lr": 0.0001872, "optim/total_tokens": 164102144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295368, "loss/out": 5.885444641113281, "created_at": "2025-01-15T06:25:10.713146+00:00"} {"global_step": 314, "acc_step": 0, "speed/wps": 12869.828490209278, "speed/FLOPS": 202138341436337.56, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7789410352706909, "optim/lr": 0.0001878, "optim/total_tokens": 164626432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 5.913005828857422, "created_at": "2025-01-15T06:25:20.901811+00:00"} {"global_step": 315, "acc_step": 0, "speed/wps": 12866.997768458165, "speed/FLOPS": 202093881061417.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7612534165382385, "optim/lr": 0.0001884, "optim/total_tokens": 165150720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.795870304107666, "created_at": "2025-01-15T06:25:31.095818+00:00"} {"global_step": 316, "acc_step": 0, "speed/wps": 12865.601502979187, "speed/FLOPS": 202071950793399.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8340765237808228, "optim/lr": 0.000189, "optim/total_tokens": 165675008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 5.897266387939453, "created_at": "2025-01-15T06:25:41.290402+00:00"} {"global_step": 317, "acc_step": 0, "speed/wps": 12860.916779800265, "speed/FLOPS": 201998370778390.62, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8303471207618713, "optim/lr": 0.00018960000000000003, "optim/total_tokens": 166199296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 5.963875770568848, "created_at": "2025-01-15T06:25:51.489782+00:00"} {"global_step": 318, "acc_step": 0, "speed/wps": 12863.99155491688, "speed/FLOPS": 202046664346780.84, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7286644577980042, "optim/lr": 0.0001902, "optim/total_tokens": 166723584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.896829128265381, "created_at": "2025-01-15T06:26:01.682380+00:00"} {"global_step": 319, "acc_step": 0, "speed/wps": 12864.333279122555, "speed/FLOPS": 202052031595009.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5775015950202942, "optim/lr": 0.0001908, "optim/total_tokens": 167247872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.915894508361816, "created_at": "2025-01-15T06:26:11.874614+00:00"} {"global_step": 320, "acc_step": 0, "speed/wps": 12865.890133182354, "speed/FLOPS": 202076484127360.06, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.655509889125824, "optim/lr": 0.0001914, "optim/total_tokens": 167772160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 5.876511573791504, "created_at": "2025-01-15T06:26:22.072302+00:00"} {"global_step": 321, "acc_step": 0, "speed/wps": 12865.434899981658, "speed/FLOPS": 202069334064387.12, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6880348324775696, "optim/lr": 0.000192, "optim/total_tokens": 168296448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 5.763254642486572, "created_at": "2025-01-15T06:26:32.266420+00:00"} {"global_step": 322, "acc_step": 0, "speed/wps": 12865.212585900385, "speed/FLOPS": 202065842316249.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5037867426872253, "optim/lr": 0.0001926, "optim/total_tokens": 168820736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 5.849250316619873, "created_at": "2025-01-15T06:26:42.462412+00:00"} {"global_step": 323, "acc_step": 0, "speed/wps": 12864.03164587932, "speed/FLOPS": 202047294030436.72, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5558505654335022, "optim/lr": 0.0001932, "optim/total_tokens": 169345024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 5.760216236114502, "created_at": "2025-01-15T06:26:52.654877+00:00"} {"global_step": 324, "acc_step": 0, "speed/wps": 12864.696046279481, "speed/FLOPS": 202057729351703.06, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4664349853992462, "optim/lr": 0.00019380000000000002, "optim/total_tokens": 169869312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 5.805147171020508, "created_at": "2025-01-15T06:27:02.850282+00:00"} {"global_step": 325, "acc_step": 0, "speed/wps": 12866.95026202932, "speed/FLOPS": 202093134907671.34, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6108608245849609, "optim/lr": 0.00019439999999999998, "optim/total_tokens": 170393600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 5.869358539581299, "created_at": "2025-01-15T06:27:13.040563+00:00"} {"global_step": 326, "acc_step": 0, "speed/wps": 12865.194264313077, "speed/FLOPS": 202065554550544.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40304264426231384, "optim/lr": 0.00019500000000000002, "optim/total_tokens": 170917888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333025, "loss/out": 5.748166084289551, "created_at": "2025-01-15T06:27:23.234235+00:00"} {"global_step": 327, "acc_step": 0, "speed/wps": 12867.15842640298, "speed/FLOPS": 202096404415207.5, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7804552316665649, "optim/lr": 0.00019559999999999998, "optim/total_tokens": 171442176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 5.852210998535156, "created_at": "2025-01-15T06:27:33.424454+00:00"} {"global_step": 328, "acc_step": 0, "speed/wps": 12863.7611049274, "speed/FLOPS": 202043044813024.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8513786196708679, "optim/lr": 0.0001962, "optim/total_tokens": 171966464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.7921223640441895, "created_at": "2025-01-15T06:27:43.624796+00:00"} {"global_step": 329, "acc_step": 0, "speed/wps": 12864.535804946796, "speed/FLOPS": 202055212541378.6, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5531588792800903, "optim/lr": 0.0001968, "optim/total_tokens": 172490752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 5.806337833404541, "created_at": "2025-01-15T06:27:53.820970+00:00"} {"global_step": 330, "acc_step": 0, "speed/wps": 12865.30129797894, "speed/FLOPS": 202067235661346.06, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7050641775131226, "optim/lr": 0.0001974, "optim/total_tokens": 173015040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 5.781750679016113, "created_at": "2025-01-15T06:28:04.013347+00:00"} {"global_step": 331, "acc_step": 0, "speed/wps": 12864.496691554605, "speed/FLOPS": 202054598211806.78, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5308874845504761, "optim/lr": 0.00019800000000000002, "optim/total_tokens": 173539328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 5.840012073516846, "created_at": "2025-01-15T06:28:14.207546+00:00"} {"global_step": 332, "acc_step": 0, "speed/wps": 12864.555792120786, "speed/FLOPS": 202055526467411.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5563783049583435, "optim/lr": 0.00019859999999999998, "optim/total_tokens": 174063616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.83128023147583, "created_at": "2025-01-15T06:28:24.400648+00:00"} {"global_step": 333, "acc_step": 0, "speed/wps": 12860.973003322742, "speed/FLOPS": 201999253846070.34, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5708152651786804, "optim/lr": 0.00019920000000000002, "optim/total_tokens": 174587904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 5.788722515106201, "created_at": "2025-01-15T06:28:34.599776+00:00"} {"global_step": 334, "acc_step": 0, "speed/wps": 12864.452705441401, "speed/FLOPS": 202053907349455.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6031139492988586, "optim/lr": 0.00019980000000000003, "optim/total_tokens": 175112192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 5.789869785308838, "created_at": "2025-01-15T06:28:44.795048+00:00"} {"global_step": 335, "acc_step": 0, "speed/wps": 12862.991886160868, "speed/FLOPS": 202030963175278.84, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5450162291526794, "optim/lr": 0.0002004, "optim/total_tokens": 175636480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 5.870997905731201, "created_at": "2025-01-15T06:28:54.994934+00:00"} {"global_step": 336, "acc_step": 0, "speed/wps": 12865.321439244082, "speed/FLOPS": 202067552007592.12, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6407616138458252, "optim/lr": 0.000201, "optim/total_tokens": 176160768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375916, "loss/out": 5.757988929748535, "created_at": "2025-01-15T06:29:05.186549+00:00"} {"global_step": 337, "acc_step": 0, "speed/wps": 12864.027275322036, "speed/FLOPS": 202047225384828.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7407321333885193, "optim/lr": 0.0002016, "optim/total_tokens": 176685056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 5.787464141845703, "created_at": "2025-01-15T06:29:15.380681+00:00"} {"global_step": 338, "acc_step": 0, "speed/wps": 12866.10424780152, "speed/FLOPS": 202079847091678.38, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47354525327682495, "optim/lr": 0.0002022, "optim/total_tokens": 177209344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 5.798498630523682, "created_at": "2025-01-15T06:29:25.573782+00:00"} {"global_step": 339, "acc_step": 0, "speed/wps": 12864.375743989838, "speed/FLOPS": 202052698564102.75, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8880510330200195, "optim/lr": 0.0002028, "optim/total_tokens": 177733632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 5.7783308029174805, "created_at": "2025-01-15T06:29:35.770453+00:00"} {"global_step": 340, "acc_step": 0, "speed/wps": 12861.21872653429, "speed/FLOPS": 202003113266766.78, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8310443162918091, "optim/lr": 0.0002034, "optim/total_tokens": 178257920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.77065896987915, "created_at": "2025-01-15T06:29:45.966436+00:00"} {"global_step": 341, "acc_step": 0, "speed/wps": 12863.649873556593, "speed/FLOPS": 202041297771498.12, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.0319985151290894, "optim/lr": 0.00020400000000000003, "optim/total_tokens": 178782208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 5.771039962768555, "created_at": "2025-01-15T06:29:56.160192+00:00"} {"global_step": 342, "acc_step": 0, "speed/wps": 12858.885138049047, "speed/FLOPS": 201966461052919.4, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.9384749531745911, "optim/lr": 0.00020459999999999999, "optim/total_tokens": 179306496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 5.735695838928223, "created_at": "2025-01-15T06:30:06.362384+00:00"} {"global_step": 343, "acc_step": 0, "speed/wps": 12865.137751030119, "speed/FLOPS": 202064666931777.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6045452356338501, "optim/lr": 0.0002052, "optim/total_tokens": 179830784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 5.785257816314697, "created_at": "2025-01-15T06:30:16.557195+00:00"} {"global_step": 344, "acc_step": 0, "speed/wps": 12863.490709387932, "speed/FLOPS": 202038797879514.84, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0321872234344482, "optim/lr": 0.0002058, "optim/total_tokens": 180355072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481042, "loss/out": 5.898383140563965, "created_at": "2025-01-15T06:30:26.752664+00:00"} {"global_step": 345, "acc_step": 0, "speed/wps": 12866.229389376016, "speed/FLOPS": 202081812612068.38, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7952282428741455, "optim/lr": 0.0002064, "optim/total_tokens": 180879360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 5.729846000671387, "created_at": "2025-01-15T06:30:36.944627+00:00"} {"global_step": 346, "acc_step": 0, "speed/wps": 12862.79802141513, "speed/FLOPS": 202027918263048.97, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5779751539230347, "optim/lr": 0.00020700000000000002, "optim/total_tokens": 181403648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469529, "loss/out": 5.789381980895996, "created_at": "2025-01-15T06:30:47.141864+00:00"} {"global_step": 347, "acc_step": 0, "speed/wps": 12866.850173117224, "speed/FLOPS": 202091562873770.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7212643623352051, "optim/lr": 0.0002076, "optim/total_tokens": 181927936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 5.756413459777832, "created_at": "2025-01-15T06:30:57.333883+00:00"} {"global_step": 348, "acc_step": 0, "speed/wps": 12867.294139842545, "speed/FLOPS": 202098535981265.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8065808415412903, "optim/lr": 0.00020820000000000002, "optim/total_tokens": 182452224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385632, "loss/out": 5.786073207855225, "created_at": "2025-01-15T06:31:07.527045+00:00"} {"global_step": 349, "acc_step": 0, "speed/wps": 12860.905577229005, "speed/FLOPS": 201998194826615.03, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6717178225517273, "optim/lr": 0.00020879999999999998, "optim/total_tokens": 182976512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428890, "loss/out": 5.891584873199463, "created_at": "2025-01-15T06:31:17.725328+00:00"} {"global_step": 350, "acc_step": 0, "speed/wps": 12866.260859525188, "speed/FLOPS": 202082306894005.62, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5540449023246765, "optim/lr": 0.0002094, "optim/total_tokens": 183500800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429413, "loss/out": 5.917880058288574, "created_at": "2025-01-15T06:31:27.918023+00:00"} {"global_step": 351, "acc_step": 0, "speed/wps": 12864.507290082127, "speed/FLOPS": 202054764676245.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6062077879905701, "optim/lr": 0.00021000000000000004, "optim/total_tokens": 184025088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402562, "loss/out": 5.82557487487793, "created_at": "2025-01-15T06:31:38.112075+00:00"} {"global_step": 352, "acc_step": 0, "speed/wps": 12865.358647709987, "speed/FLOPS": 202068136417678.94, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7107548713684082, "optim/lr": 0.0002106, "optim/total_tokens": 184549376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 5.814599990844727, "created_at": "2025-01-15T06:31:48.303604+00:00"} {"global_step": 353, "acc_step": 0, "speed/wps": 12862.427685034776, "speed/FLOPS": 202022101621299.97, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6042987704277039, "optim/lr": 0.0002112, "optim/total_tokens": 185073664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 5.793617248535156, "created_at": "2025-01-15T06:31:58.502641+00:00"} {"global_step": 354, "acc_step": 0, "speed/wps": 12866.224043406339, "speed/FLOPS": 202081728646268.44, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5773522257804871, "optim/lr": 0.0002118, "optim/total_tokens": 185597952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 5.7283854484558105, "created_at": "2025-01-15T06:32:08.697496+00:00"} {"global_step": 355, "acc_step": 0, "speed/wps": 12863.728378345055, "speed/FLOPS": 202042530797077.8, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5679392218589783, "optim/lr": 0.0002124, "optim/total_tokens": 186122240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 5.705313682556152, "created_at": "2025-01-15T06:32:18.895148+00:00"} {"global_step": 356, "acc_step": 0, "speed/wps": 12865.340018752686, "speed/FLOPS": 202067843824305.66, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7263593673706055, "optim/lr": 0.00021299999999999997, "optim/total_tokens": 186646528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 5.8507161140441895, "created_at": "2025-01-15T06:32:29.086526+00:00"} {"global_step": 357, "acc_step": 0, "speed/wps": 12867.723440294183, "speed/FLOPS": 202105278734779.28, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7415958046913147, "optim/lr": 0.00021360000000000001, "optim/total_tokens": 187170816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.79336404800415, "created_at": "2025-01-15T06:32:39.278292+00:00"} {"global_step": 358, "acc_step": 0, "speed/wps": 12864.264608993777, "speed/FLOPS": 202050953036274.12, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5641395449638367, "optim/lr": 0.00021420000000000003, "optim/total_tokens": 187695104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 5.664345741271973, "created_at": "2025-01-15T06:32:49.474496+00:00"} {"global_step": 359, "acc_step": 0, "speed/wps": 12862.547287575318, "speed/FLOPS": 202023980143550.3, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7065340280532837, "optim/lr": 0.0002148, "optim/total_tokens": 188219392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 5.781903266906738, "created_at": "2025-01-15T06:32:59.671410+00:00"} {"global_step": 360, "acc_step": 0, "speed/wps": 12864.94830802444, "speed/FLOPS": 202061691469051.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7264686226844788, "optim/lr": 0.0002154, "optim/total_tokens": 188743680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.787658214569092, "created_at": "2025-01-15T06:33:09.863594+00:00"} {"global_step": 361, "acc_step": 0, "speed/wps": 12869.13379919086, "speed/FLOPS": 202127430359287.8, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6017849445343018, "optim/lr": 0.000216, "optim/total_tokens": 189267968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.745023727416992, "created_at": "2025-01-15T06:33:20.053768+00:00"} {"global_step": 362, "acc_step": 0, "speed/wps": 12864.20231158485, "speed/FLOPS": 202049974569861.3, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47495561838150024, "optim/lr": 0.0002166, "optim/total_tokens": 189792256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 5.7414350509643555, "created_at": "2025-01-15T06:33:30.249624+00:00"} {"global_step": 363, "acc_step": 0, "speed/wps": 12867.132502099497, "speed/FLOPS": 202095997238397.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.588802695274353, "optim/lr": 0.00021720000000000002, "optim/total_tokens": 190316544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 5.721882343292236, "created_at": "2025-01-15T06:33:40.445030+00:00"} {"global_step": 364, "acc_step": 0, "speed/wps": 12868.788629868986, "speed/FLOPS": 202122009000775.44, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7479929327964783, "optim/lr": 0.0002178, "optim/total_tokens": 190840832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 5.7779951095581055, "created_at": "2025-01-15T06:33:50.634794+00:00"} {"global_step": 365, "acc_step": 0, "speed/wps": 12866.954647513008, "speed/FLOPS": 202093203787718.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5588903427124023, "optim/lr": 0.00021840000000000002, "optim/total_tokens": 191365120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 5.794157981872559, "created_at": "2025-01-15T06:34:00.824922+00:00"} {"global_step": 366, "acc_step": 0, "speed/wps": 12865.823144802283, "speed/FLOPS": 202075431982799.34, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7824045419692993, "optim/lr": 0.00021899999999999998, "optim/total_tokens": 191889408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 5.730737686157227, "created_at": "2025-01-15T06:34:11.016791+00:00"} {"global_step": 367, "acc_step": 0, "speed/wps": 12867.128198842276, "speed/FLOPS": 202095929649829.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6076114177703857, "optim/lr": 0.0002196, "optim/total_tokens": 192413696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.6313066482543945, "created_at": "2025-01-15T06:34:21.212811+00:00"} {"global_step": 368, "acc_step": 0, "speed/wps": 12867.071188602342, "speed/FLOPS": 202095034225671.16, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6183912754058838, "optim/lr": 0.0002202, "optim/total_tokens": 192937984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 5.686939239501953, "created_at": "2025-01-15T06:34:31.405696+00:00"} {"global_step": 369, "acc_step": 0, "speed/wps": 12864.589132045829, "speed/FLOPS": 202056050116748.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5481594800949097, "optim/lr": 0.0002208, "optim/total_tokens": 193462272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 5.569628715515137, "created_at": "2025-01-15T06:34:41.598017+00:00"} {"global_step": 370, "acc_step": 0, "speed/wps": 12863.955018693998, "speed/FLOPS": 202046090495194.44, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.599443256855011, "optim/lr": 0.00022140000000000001, "optim/total_tokens": 193986560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.8057684898376465, "created_at": "2025-01-15T06:34:51.791389+00:00"} {"global_step": 371, "acc_step": 0, "speed/wps": 12862.907775577847, "speed/FLOPS": 202029642102992.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6929612755775452, "optim/lr": 0.000222, "optim/total_tokens": 194510848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 5.78136682510376, "created_at": "2025-01-15T06:35:01.985392+00:00"} {"global_step": 372, "acc_step": 0, "speed/wps": 12862.355496922713, "speed/FLOPS": 202020967807803.2, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6354056596755981, "optim/lr": 0.00022260000000000002, "optim/total_tokens": 195035136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 5.741458415985107, "created_at": "2025-01-15T06:35:12.184253+00:00"} {"global_step": 373, "acc_step": 0, "speed/wps": 12863.724481830537, "speed/FLOPS": 202042469596962.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6224060654640198, "optim/lr": 0.00022319999999999998, "optim/total_tokens": 195559424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.763352870941162, "created_at": "2025-01-15T06:35:22.379353+00:00"} {"global_step": 374, "acc_step": 0, "speed/wps": 12867.716023613275, "speed/FLOPS": 202105162245614.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6402902007102966, "optim/lr": 0.0002238, "optim/total_tokens": 196083712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.698797225952148, "created_at": "2025-01-15T06:35:32.570182+00:00"} {"global_step": 375, "acc_step": 0, "speed/wps": 12865.580837466885, "speed/FLOPS": 202071626213131.75, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6784331798553467, "optim/lr": 0.00022440000000000003, "optim/total_tokens": 196608000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 5.650041103363037, "created_at": "2025-01-15T06:35:42.762405+00:00"} {"global_step": 376, "acc_step": 0, "speed/wps": 12866.066902111934, "speed/FLOPS": 202079260526305.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1426310539245605, "optim/lr": 0.000225, "optim/total_tokens": 197132288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 5.800615310668945, "created_at": "2025-01-15T06:35:52.953274+00:00"} {"global_step": 377, "acc_step": 0, "speed/wps": 12866.441243909141, "speed/FLOPS": 202085140078629.72, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7153489589691162, "optim/lr": 0.0002256, "optim/total_tokens": 197656576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420398, "loss/out": 5.664709091186523, "created_at": "2025-01-15T06:36:03.146683+00:00"} {"global_step": 378, "acc_step": 0, "speed/wps": 12868.2362991018, "speed/FLOPS": 202113333887094.75, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.714324414730072, "optim/lr": 0.0002262, "optim/total_tokens": 198180864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 5.703269958496094, "created_at": "2025-01-15T06:36:13.336345+00:00"} {"global_step": 379, "acc_step": 0, "speed/wps": 12862.389654192335, "speed/FLOPS": 202021504294659.6, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6875813603401184, "optim/lr": 0.0002268, "optim/total_tokens": 198705152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 5.720309257507324, "created_at": "2025-01-15T06:36:23.532865+00:00"} {"global_step": 380, "acc_step": 0, "speed/wps": 12858.617542674941, "speed/FLOPS": 201962258099854.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7289099097251892, "optim/lr": 0.00022740000000000003, "optim/total_tokens": 199229440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 5.651811599731445, "created_at": "2025-01-15T06:36:33.731391+00:00"} {"global_step": 381, "acc_step": 0, "speed/wps": 12865.301090992076, "speed/FLOPS": 202067232410332.94, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5607112050056458, "optim/lr": 0.000228, "optim/total_tokens": 199753728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.70732307434082, "created_at": "2025-01-15T06:36:43.922871+00:00"} {"global_step": 382, "acc_step": 0, "speed/wps": 12864.019706516243, "speed/FLOPS": 202047106506333.2, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5855141282081604, "optim/lr": 0.00022860000000000003, "optim/total_tokens": 200278016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 5.807519912719727, "created_at": "2025-01-15T06:36:54.118689+00:00"} {"global_step": 383, "acc_step": 0, "speed/wps": 12866.402105990606, "speed/FLOPS": 202084525363838.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.597615122795105, "optim/lr": 0.0002292, "optim/total_tokens": 200802304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 5.784581184387207, "created_at": "2025-01-15T06:37:04.309310+00:00"} {"global_step": 384, "acc_step": 0, "speed/wps": 12866.238735717834, "speed/FLOPS": 202081959409209.84, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4537712335586548, "optim/lr": 0.0002298, "optim/total_tokens": 201326592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.66307258605957, "created_at": "2025-01-15T06:37:14.502232+00:00"} {"global_step": 385, "acc_step": 0, "speed/wps": 12865.74774886527, "speed/FLOPS": 202074247786003.7, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7437506914138794, "optim/lr": 0.0002304, "optim/total_tokens": 201850880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.711275577545166, "created_at": "2025-01-15T06:37:24.694506+00:00"} {"global_step": 386, "acc_step": 0, "speed/wps": 12866.35384527616, "speed/FLOPS": 202083767363000.47, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7187018394470215, "optim/lr": 0.000231, "optim/total_tokens": 202375168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.741419315338135, "created_at": "2025-01-15T06:37:34.901949+00:00"} {"global_step": 387, "acc_step": 0, "speed/wps": 12868.910380579651, "speed/FLOPS": 202123921262989.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5398826003074646, "optim/lr": 0.00023160000000000002, "optim/total_tokens": 202899456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.6792216300964355, "created_at": "2025-01-15T06:37:45.090888+00:00"} {"global_step": 388, "acc_step": 0, "speed/wps": 12863.563302941517, "speed/FLOPS": 202039938061027.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5160946249961853, "optim/lr": 0.0002322, "optim/total_tokens": 203423744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 5.722184181213379, "created_at": "2025-01-15T06:37:55.287099+00:00"} {"global_step": 389, "acc_step": 0, "speed/wps": 12866.813760638759, "speed/FLOPS": 202090990965760.0, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5891109108924866, "optim/lr": 0.00023280000000000002, "optim/total_tokens": 203948032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 5.6718010902404785, "created_at": "2025-01-15T06:38:05.480473+00:00"} {"global_step": 390, "acc_step": 0, "speed/wps": 12867.453901670742, "speed/FLOPS": 202101045260313.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6129079461097717, "optim/lr": 0.00023339999999999998, "optim/total_tokens": 204472320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420398, "loss/out": 5.643631935119629, "created_at": "2025-01-15T06:38:15.671726+00:00"} {"global_step": 391, "acc_step": 0, "speed/wps": 12865.002644029457, "speed/FLOPS": 202062544890675.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.46873703598976135, "optim/lr": 0.000234, "optim/total_tokens": 204996608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401696, "loss/out": 5.731378078460693, "created_at": "2025-01-15T06:38:25.866388+00:00"} {"global_step": 392, "acc_step": 0, "speed/wps": 12867.429487915817, "speed/FLOPS": 202100661808744.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5327884554862976, "optim/lr": 0.0002346, "optim/total_tokens": 205520896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 5.73551607131958, "created_at": "2025-01-15T06:38:36.058086+00:00"} {"global_step": 393, "acc_step": 0, "speed/wps": 12862.514560115485, "speed/FLOPS": 202023466113821.4, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7335048317909241, "optim/lr": 0.0002352, "optim/total_tokens": 206045184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312082, "loss/out": 5.655545234680176, "created_at": "2025-01-15T06:38:46.255417+00:00"} {"global_step": 394, "acc_step": 0, "speed/wps": 12865.327898198704, "speed/FLOPS": 202067653454350.0, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7886712551116943, "optim/lr": 0.0002358, "optim/total_tokens": 206569472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 5.706343650817871, "created_at": "2025-01-15T06:38:56.449658+00:00"} {"global_step": 395, "acc_step": 0, "speed/wps": 12867.094636358075, "speed/FLOPS": 202095402504895.56, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1608622074127197, "optim/lr": 0.0002364, "optim/total_tokens": 207093760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.932650089263916, "created_at": "2025-01-15T06:39:06.639870+00:00"} {"global_step": 396, "acc_step": 0, "speed/wps": 12861.732638478215, "speed/FLOPS": 202011184960035.16, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7601967453956604, "optim/lr": 0.00023700000000000001, "optim/total_tokens": 207618048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.665111064910889, "created_at": "2025-01-15T06:39:16.834184+00:00"} {"global_step": 397, "acc_step": 0, "speed/wps": 12859.172373103573, "speed/FLOPS": 201970972474153.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7911771535873413, "optim/lr": 0.00023760000000000003, "optim/total_tokens": 208142336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 5.737153053283691, "created_at": "2025-01-15T06:39:27.033610+00:00"} {"global_step": 398, "acc_step": 0, "speed/wps": 12865.993147628938, "speed/FLOPS": 202078102110800.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4934275448322296, "optim/lr": 0.0002382, "optim/total_tokens": 208666624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505601, "loss/out": 5.774774551391602, "created_at": "2025-01-15T06:39:37.224622+00:00"} {"global_step": 399, "acc_step": 0, "speed/wps": 12863.692810098075, "speed/FLOPS": 202041972148883.1, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6043302416801453, "optim/lr": 0.0002388, "optim/total_tokens": 209190912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 5.7705302238464355, "created_at": "2025-01-15T06:39:47.417646+00:00"} {"global_step": 400, "acc_step": 0, "speed/wps": 12864.394788773683, "speed/FLOPS": 202052997688603.3, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5712223052978516, "optim/lr": 0.0002394, "optim/total_tokens": 209715200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 5.726345062255859, "created_at": "2025-01-15T06:39:57.614566+00:00"} {"global_step": 401, "acc_step": 0, "speed/wps": 12867.927541201514, "speed/FLOPS": 202108484419992.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46867889165878296, "optim/lr": 0.00024, "optim/total_tokens": 210239488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 5.567038536071777, "created_at": "2025-01-15T06:40:07.803984+00:00"} {"global_step": 402, "acc_step": 0, "speed/wps": 12867.445535902249, "speed/FLOPS": 202100913864423.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7570277452468872, "optim/lr": 0.0002406, "optim/total_tokens": 210763776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.694471836090088, "created_at": "2025-01-15T06:40:17.996026+00:00"} {"global_step": 403, "acc_step": 0, "speed/wps": 12866.812863092706, "speed/FLOPS": 202090976868565.88, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.960928201675415, "optim/lr": 0.0002412, "optim/total_tokens": 211288064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 5.758240699768066, "created_at": "2025-01-15T06:40:28.189294+00:00"} {"global_step": 404, "acc_step": 0, "speed/wps": 12864.348845504126, "speed/FLOPS": 202052276086422.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8330236077308655, "optim/lr": 0.00024180000000000002, "optim/total_tokens": 211812352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418499, "loss/out": 5.779189109802246, "created_at": "2025-01-15T06:40:38.384825+00:00"} {"global_step": 405, "acc_step": 0, "speed/wps": 12862.594963694497, "speed/FLOPS": 202024728962515.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7145732045173645, "optim/lr": 0.00024239999999999998, "optim/total_tokens": 212336640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 5.677924156188965, "created_at": "2025-01-15T06:40:48.584918+00:00"} {"global_step": 406, "acc_step": 0, "speed/wps": 12864.884965598092, "speed/FLOPS": 202060696589204.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7260547876358032, "optim/lr": 0.00024300000000000002, "optim/total_tokens": 212860928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 5.614537239074707, "created_at": "2025-01-15T06:40:58.779834+00:00"} {"global_step": 407, "acc_step": 0, "speed/wps": 12866.49494792448, "speed/FLOPS": 202085983573987.7, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6201658844947815, "optim/lr": 0.00024359999999999999, "optim/total_tokens": 213385216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 5.705810546875, "created_at": "2025-01-15T06:41:08.973777+00:00"} {"global_step": 408, "acc_step": 0, "speed/wps": 12866.171936607841, "speed/FLOPS": 202080910237395.8, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.4812955856323242, "optim/lr": 0.0002442, "optim/total_tokens": 213909504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 5.669589042663574, "created_at": "2025-01-15T06:41:19.169022+00:00"} {"global_step": 409, "acc_step": 0, "speed/wps": 12868.161659661206, "speed/FLOPS": 202112161572114.88, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5862284898757935, "optim/lr": 0.00024480000000000004, "optim/total_tokens": 214433792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 5.685440540313721, "created_at": "2025-01-15T06:41:29.359194+00:00"} {"global_step": 410, "acc_step": 0, "speed/wps": 12870.065278867443, "speed/FLOPS": 202142060527595.75, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4439340829849243, "optim/lr": 0.0002454, "optim/total_tokens": 214958080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.6528544425964355, "created_at": "2025-01-15T06:41:39.552959+00:00"} {"global_step": 411, "acc_step": 0, "speed/wps": 12868.786022303602, "speed/FLOPS": 202121968045377.9, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5422499775886536, "optim/lr": 0.000246, "optim/total_tokens": 215482368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 5.603452205657959, "created_at": "2025-01-15T06:41:49.741906+00:00"} {"global_step": 412, "acc_step": 0, "speed/wps": 12868.964445284586, "speed/FLOPS": 202124770423473.12, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7308642864227295, "optim/lr": 0.0002466, "optim/total_tokens": 216006656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 5.605785369873047, "created_at": "2025-01-15T06:41:59.938708+00:00"} {"global_step": 413, "acc_step": 0, "speed/wps": 12864.135383109973, "speed/FLOPS": 202048923366194.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5606611967086792, "optim/lr": 0.0002472, "optim/total_tokens": 216530944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 5.714973449707031, "created_at": "2025-01-15T06:42:10.136141+00:00"} {"global_step": 414, "acc_step": 0, "speed/wps": 12867.89005197799, "speed/FLOPS": 202107895600221.94, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6509976983070374, "optim/lr": 0.0002478, "optim/total_tokens": 217055232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 5.682011604309082, "created_at": "2025-01-15T06:42:20.327490+00:00"} {"global_step": 415, "acc_step": 0, "speed/wps": 12863.9847186896, "speed/FLOPS": 202046556974437.47, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6304953098297119, "optim/lr": 0.0002484, "optim/total_tokens": 217579520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.7264885902404785, "created_at": "2025-01-15T06:42:30.522140+00:00"} {"global_step": 416, "acc_step": 0, "speed/wps": 12866.704096260044, "speed/FLOPS": 202089268535997.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.633730947971344, "optim/lr": 0.00024900000000000004, "optim/total_tokens": 218103808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 5.743963718414307, "created_at": "2025-01-15T06:42:40.714719+00:00"} {"global_step": 417, "acc_step": 0, "speed/wps": 12862.791996446773, "speed/FLOPS": 202027823632641.72, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44816854596138, "optim/lr": 0.0002496, "optim/total_tokens": 218628096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412087, "loss/out": 5.700103759765625, "created_at": "2025-01-15T06:42:50.908959+00:00"} {"global_step": 418, "acc_step": 0, "speed/wps": 12868.113561055843, "speed/FLOPS": 202111406117423.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5031763911247253, "optim/lr": 0.0002502, "optim/total_tokens": 219152384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 5.582259178161621, "created_at": "2025-01-15T06:43:01.099320+00:00"} {"global_step": 419, "acc_step": 0, "speed/wps": 12864.992156406213, "speed/FLOPS": 202062380168141.34, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5066996812820435, "optim/lr": 0.00025079999999999997, "optim/total_tokens": 219676672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 5.652157783508301, "created_at": "2025-01-15T06:43:11.292904+00:00"} {"global_step": 420, "acc_step": 0, "speed/wps": 12864.336990215996, "speed/FLOPS": 202052089882831.28, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4659786522388458, "optim/lr": 0.0002514, "optim/total_tokens": 220200960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 5.551867961883545, "created_at": "2025-01-15T06:43:21.488609+00:00"} {"global_step": 421, "acc_step": 0, "speed/wps": 12866.648646033598, "speed/FLOPS": 202088397613994.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5809281468391418, "optim/lr": 0.000252, "optim/total_tokens": 220725248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 5.670897483825684, "created_at": "2025-01-15T06:43:31.682743+00:00"} {"global_step": 422, "acc_step": 0, "speed/wps": 12866.24330185654, "speed/FLOPS": 202082031126692.75, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.880528450012207, "optim/lr": 0.0002526, "optim/total_tokens": 221249536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 5.619828701019287, "created_at": "2025-01-15T06:43:41.875875+00:00"} {"global_step": 423, "acc_step": 0, "speed/wps": 12869.504380624108, "speed/FLOPS": 202133250849929.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8802422881126404, "optim/lr": 0.00025320000000000003, "optim/total_tokens": 221773824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 5.637925148010254, "created_at": "2025-01-15T06:43:52.066274+00:00"} {"global_step": 424, "acc_step": 0, "speed/wps": 12864.642365511674, "speed/FLOPS": 202056886221479.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8626874685287476, "optim/lr": 0.0002538, "optim/total_tokens": 222298112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.700468063354492, "created_at": "2025-01-15T06:44:02.261468+00:00"} {"global_step": 425, "acc_step": 0, "speed/wps": 12871.486306824907, "speed/FLOPS": 202164379724364.78, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.033583402633667, "optim/lr": 0.0002544, "optim/total_tokens": 222822400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 5.668054103851318, "created_at": "2025-01-15T06:44:12.448374+00:00"} {"global_step": 426, "acc_step": 0, "speed/wps": 12865.959402876646, "speed/FLOPS": 202077572103095.44, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7906351089477539, "optim/lr": 0.000255, "optim/total_tokens": 223346688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 5.547576904296875, "created_at": "2025-01-15T06:44:22.641023+00:00"} {"global_step": 427, "acc_step": 0, "speed/wps": 12864.807733204905, "speed/FLOPS": 202059483548340.78, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.782408595085144, "optim/lr": 0.0002556, "optim/total_tokens": 223870976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 5.906179428100586, "created_at": "2025-01-15T06:44:32.833917+00:00"} {"global_step": 428, "acc_step": 0, "speed/wps": 12865.118377110532, "speed/FLOPS": 202064362637748.1, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6211192011833191, "optim/lr": 0.0002562, "optim/total_tokens": 224395264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384291, "loss/out": 5.702404022216797, "created_at": "2025-01-15T06:44:43.026923+00:00"} {"global_step": 429, "acc_step": 0, "speed/wps": 12862.378525464303, "speed/FLOPS": 202021329502693.38, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5942365527153015, "optim/lr": 0.0002568, "optim/total_tokens": 224919552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372547, "loss/out": 5.697174072265625, "created_at": "2025-01-15T06:44:53.220894+00:00"} {"global_step": 430, "acc_step": 0, "speed/wps": 12867.225327517044, "speed/FLOPS": 202097455189135.62, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.554115355014801, "optim/lr": 0.0002574, "optim/total_tokens": 225443840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 5.629344940185547, "created_at": "2025-01-15T06:45:03.410916+00:00"} {"global_step": 431, "acc_step": 0, "speed/wps": 12866.190272742146, "speed/FLOPS": 202081198231581.53, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.66166752576828, "optim/lr": 0.000258, "optim/total_tokens": 225968128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 5.740896224975586, "created_at": "2025-01-15T06:45:13.605977+00:00"} {"global_step": 432, "acc_step": 0, "speed/wps": 12857.498335863456, "speed/FLOPS": 201944679418928.8, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6309328675270081, "optim/lr": 0.0002586, "optim/total_tokens": 226492416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349068, "loss/out": 5.723753452301025, "created_at": "2025-01-15T06:45:23.804132+00:00"} {"global_step": 433, "acc_step": 0, "speed/wps": 12865.034448045531, "speed/FLOPS": 202063044416451.66, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4927001893520355, "optim/lr": 0.0002592, "optim/total_tokens": 227016704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 5.558496475219727, "created_at": "2025-01-15T06:45:33.998302+00:00"} {"global_step": 434, "acc_step": 0, "speed/wps": 12868.507668422784, "speed/FLOPS": 202117596115181.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6685022115707397, "optim/lr": 0.00025979999999999997, "optim/total_tokens": 227540992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 5.658364295959473, "created_at": "2025-01-15T06:45:44.193207+00:00"} {"global_step": 435, "acc_step": 0, "speed/wps": 12866.16473813974, "speed/FLOPS": 202080797175562.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.616023063659668, "optim/lr": 0.0002604, "optim/total_tokens": 228065280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.609832763671875, "created_at": "2025-01-15T06:45:54.384006+00:00"} {"global_step": 436, "acc_step": 0, "speed/wps": 12867.546203489565, "speed/FLOPS": 202102494987213.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5814716815948486, "optim/lr": 0.000261, "optim/total_tokens": 228589568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 5.588252544403076, "created_at": "2025-01-15T06:46:04.583938+00:00"} {"global_step": 437, "acc_step": 0, "speed/wps": 12865.47649475609, "speed/FLOPS": 202069987367476.8, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4979701638221741, "optim/lr": 0.0002616, "optim/total_tokens": 229113856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 5.615085124969482, "created_at": "2025-01-15T06:46:14.781607+00:00"} {"global_step": 438, "acc_step": 0, "speed/wps": 12863.831680344285, "speed/FLOPS": 202044153296927.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.432381808757782, "optim/lr": 0.00026220000000000003, "optim/total_tokens": 229638144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 5.605197906494141, "created_at": "2025-01-15T06:46:24.977642+00:00"} {"global_step": 439, "acc_step": 0, "speed/wps": 12863.854359081275, "speed/FLOPS": 202044509497655.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3897249400615692, "optim/lr": 0.0002628, "optim/total_tokens": 230162432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.60236120223999, "created_at": "2025-01-15T06:46:35.172668+00:00"} {"global_step": 440, "acc_step": 0, "speed/wps": 12864.41067867631, "speed/FLOPS": 202053247261359.1, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3937879502773285, "optim/lr": 0.0002634, "optim/total_tokens": 230686720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 5.497933387756348, "created_at": "2025-01-15T06:46:45.364993+00:00"} {"global_step": 441, "acc_step": 0, "speed/wps": 12863.762390055268, "speed/FLOPS": 202043064997723.72, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44765394926071167, "optim/lr": 0.00026399999999999997, "optim/total_tokens": 231211008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.525851726531982, "created_at": "2025-01-15T06:46:55.559043+00:00"} {"global_step": 442, "acc_step": 0, "speed/wps": 12866.666261271497, "speed/FLOPS": 202088674285511.78, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4359782636165619, "optim/lr": 0.0002646, "optim/total_tokens": 231735296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 5.519183158874512, "created_at": "2025-01-15T06:47:05.753934+00:00"} {"global_step": 443, "acc_step": 0, "speed/wps": 12865.221486219156, "speed/FLOPS": 202065982107985.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4585653841495514, "optim/lr": 0.00026520000000000005, "optim/total_tokens": 232259584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.603172779083252, "created_at": "2025-01-15T06:47:15.951337+00:00"} {"global_step": 444, "acc_step": 0, "speed/wps": 12857.532378067785, "speed/FLOPS": 201945214098526.7, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.41933730244636536, "optim/lr": 0.0002658, "optim/total_tokens": 232783872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 5.606285572052002, "created_at": "2025-01-15T06:47:26.150449+00:00"} {"global_step": 445, "acc_step": 0, "speed/wps": 12865.184299715353, "speed/FLOPS": 202065398042843.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.513712465763092, "optim/lr": 0.0002664, "optim/total_tokens": 233308160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.422854900360107, "created_at": "2025-01-15T06:47:36.345394+00:00"} {"global_step": 446, "acc_step": 0, "speed/wps": 12866.769115226562, "speed/FLOPS": 202090289748212.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5897712111473083, "optim/lr": 0.000267, "optim/total_tokens": 233832448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.650152683258057, "created_at": "2025-01-15T06:47:46.536700+00:00"} {"global_step": 447, "acc_step": 0, "speed/wps": 12866.470196688419, "speed/FLOPS": 202085594821813.56, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6234804391860962, "optim/lr": 0.0002676, "optim/total_tokens": 234356736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 5.510210037231445, "created_at": "2025-01-15T06:47:56.727374+00:00"} {"global_step": 448, "acc_step": 0, "speed/wps": 12869.835991707916, "speed/FLOPS": 202138459257681.8, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7306691408157349, "optim/lr": 0.00026819999999999996, "optim/total_tokens": 234881024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.468038558959961, "created_at": "2025-01-15T06:48:06.915394+00:00"} {"global_step": 449, "acc_step": 0, "speed/wps": 12862.050714478091, "speed/FLOPS": 202016180780695.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7115294337272644, "optim/lr": 0.00026880000000000003, "optim/total_tokens": 235405312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 5.548253059387207, "created_at": "2025-01-15T06:48:17.109701+00:00"} {"global_step": 450, "acc_step": 0, "speed/wps": 12865.838997068075, "speed/FLOPS": 202075680964416.9, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.696819007396698, "optim/lr": 0.00026940000000000004, "optim/total_tokens": 235929600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.564143180847168, "created_at": "2025-01-15T06:48:27.308208+00:00"} {"global_step": 451, "acc_step": 0, "speed/wps": 12868.440378626621, "speed/FLOPS": 202116539236466.72, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.599216639995575, "optim/lr": 0.00027, "optim/total_tokens": 236453888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 5.610869407653809, "created_at": "2025-01-15T06:48:37.501922+00:00"} {"global_step": 452, "acc_step": 0, "speed/wps": 12868.862125757752, "speed/FLOPS": 202123163354702.44, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.658406138420105, "optim/lr": 0.0002706, "optim/total_tokens": 236978176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.569410800933838, "created_at": "2025-01-15T06:48:47.693599+00:00"} {"global_step": 453, "acc_step": 0, "speed/wps": 12868.89211317815, "speed/FLOPS": 202123634348346.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5819175243377686, "optim/lr": 0.0002712, "optim/total_tokens": 237502464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 5.5508713722229, "created_at": "2025-01-15T06:48:57.885696+00:00"} {"global_step": 454, "acc_step": 0, "speed/wps": 12869.997238382351, "speed/FLOPS": 202140991858279.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5446634292602539, "optim/lr": 0.0002718, "optim/total_tokens": 238026752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 5.583423137664795, "created_at": "2025-01-15T06:49:08.077206+00:00"} {"global_step": 455, "acc_step": 0, "speed/wps": 12870.564840365283, "speed/FLOPS": 202149906827388.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6249670386314392, "optim/lr": 0.0002724, "optim/total_tokens": 238551040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 5.5128021240234375, "created_at": "2025-01-15T06:49:18.268684+00:00"} {"global_step": 456, "acc_step": 0, "speed/wps": 12867.215863277033, "speed/FLOPS": 202097306540241.03, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.658424437046051, "optim/lr": 0.000273, "optim/total_tokens": 239075328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.500619888305664, "created_at": "2025-01-15T06:49:28.458700+00:00"} {"global_step": 457, "acc_step": 0, "speed/wps": 12866.298309850241, "speed/FLOPS": 202082895102822.72, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43445447087287903, "optim/lr": 0.00027360000000000004, "optim/total_tokens": 239599616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 5.579320907592773, "created_at": "2025-01-15T06:49:38.653843+00:00"} {"global_step": 458, "acc_step": 0, "speed/wps": 12868.246009575829, "speed/FLOPS": 202113486403432.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5393295884132385, "optim/lr": 0.0002742, "optim/total_tokens": 240123904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482092, "loss/out": 5.577887535095215, "created_at": "2025-01-15T06:49:48.843340+00:00"} {"global_step": 459, "acc_step": 0, "speed/wps": 12862.255850940299, "speed/FLOPS": 202019402730721.44, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5266751646995544, "optim/lr": 0.0002748, "optim/total_tokens": 240648192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 5.434761047363281, "created_at": "2025-01-15T06:49:59.038483+00:00"} {"global_step": 460, "acc_step": 0, "speed/wps": 12868.084298501484, "speed/FLOPS": 202110946508796.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4418547749519348, "optim/lr": 0.0002754, "optim/total_tokens": 241172480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 5.48153829574585, "created_at": "2025-01-15T06:50:09.228146+00:00"} {"global_step": 461, "acc_step": 0, "speed/wps": 12863.748059659503, "speed/FLOPS": 202042839919166.2, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45091500878334045, "optim/lr": 0.000276, "optim/total_tokens": 241696768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 5.563556671142578, "created_at": "2025-01-15T06:50:19.420932+00:00"} {"global_step": 462, "acc_step": 0, "speed/wps": 12866.1261594281, "speed/FLOPS": 202080191243883.53, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5461782217025757, "optim/lr": 0.0002766, "optim/total_tokens": 242221056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.441958427429199, "created_at": "2025-01-15T06:50:29.615227+00:00"} {"global_step": 463, "acc_step": 0, "speed/wps": 12866.278961841084, "speed/FLOPS": 202082591215751.97, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5608537793159485, "optim/lr": 0.0002772, "optim/total_tokens": 242745344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 5.486669540405273, "created_at": "2025-01-15T06:50:39.806010+00:00"} {"global_step": 464, "acc_step": 0, "speed/wps": 12868.28831084487, "speed/FLOPS": 202114150802991.28, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5595384836196899, "optim/lr": 0.00027780000000000003, "optim/total_tokens": 243269632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 5.448694229125977, "created_at": "2025-01-15T06:50:49.997620+00:00"} {"global_step": 465, "acc_step": 0, "speed/wps": 12870.0898685735, "speed/FLOPS": 202142446742719.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7013726234436035, "optim/lr": 0.0002784, "optim/total_tokens": 243793920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.4633026123046875, "created_at": "2025-01-15T06:51:00.190461+00:00"} {"global_step": 466, "acc_step": 0, "speed/wps": 12864.978963993006, "speed/FLOPS": 202062172963163.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6158109903335571, "optim/lr": 0.000279, "optim/total_tokens": 244318208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 5.479556083679199, "created_at": "2025-01-15T06:51:10.382169+00:00"} {"global_step": 467, "acc_step": 0, "speed/wps": 12863.735957390014, "speed/FLOPS": 202042649836393.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5618063807487488, "optim/lr": 0.0002796, "optim/total_tokens": 244842496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 5.487341403961182, "created_at": "2025-01-15T06:51:20.576764+00:00"} {"global_step": 468, "acc_step": 0, "speed/wps": 12866.19627266504, "speed/FLOPS": 202081292468615.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36434701085090637, "optim/lr": 0.0002802, "optim/total_tokens": 245366784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 5.468110084533691, "created_at": "2025-01-15T06:51:30.769958+00:00"} {"global_step": 469, "acc_step": 0, "speed/wps": 12867.229380262983, "speed/FLOPS": 202097518843079.66, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47287899255752563, "optim/lr": 0.0002808, "optim/total_tokens": 245891072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.4991655349731445, "created_at": "2025-01-15T06:51:40.962295+00:00"} {"global_step": 470, "acc_step": 0, "speed/wps": 12864.22939524457, "speed/FLOPS": 202050399955954.1, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4780162274837494, "optim/lr": 0.0002814, "optim/total_tokens": 246415360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.491846084594727, "created_at": "2025-01-15T06:51:51.159222+00:00"} {"global_step": 471, "acc_step": 0, "speed/wps": 12866.435135541013, "speed/FLOPS": 202085044138314.5, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3757362961769104, "optim/lr": 0.000282, "optim/total_tokens": 246939648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 5.563437461853027, "created_at": "2025-01-15T06:52:01.354431+00:00"} {"global_step": 472, "acc_step": 0, "speed/wps": 12867.277241679834, "speed/FLOPS": 202098270572399.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5593335032463074, "optim/lr": 0.00028260000000000004, "optim/total_tokens": 247463936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416629, "loss/out": 5.687925338745117, "created_at": "2025-01-15T06:52:11.546675+00:00"} {"global_step": 473, "acc_step": 0, "speed/wps": 12852.129639482717, "speed/FLOPS": 201860356664907.38, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.48211660981178284, "optim/lr": 0.0002832, "optim/total_tokens": 247988224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 5.512939929962158, "created_at": "2025-01-15T06:52:21.750640+00:00"} {"global_step": 474, "acc_step": 0, "speed/wps": 12859.431275498275, "speed/FLOPS": 201975038892031.03, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47969940304756165, "optim/lr": 0.0002838, "optim/total_tokens": 248512512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 5.474346160888672, "created_at": "2025-01-15T06:52:31.947788+00:00"} {"global_step": 475, "acc_step": 0, "speed/wps": 12858.98491862143, "speed/FLOPS": 201968028243921.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.5153180956840515, "optim/lr": 0.0002844, "optim/total_tokens": 249036800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313228, "loss/out": 5.401482582092285, "created_at": "2025-01-15T06:52:42.144434+00:00"} {"global_step": 476, "acc_step": 0, "speed/wps": 12859.48012936825, "speed/FLOPS": 201975806209191.25, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.4494660198688507, "optim/lr": 0.000285, "optim/total_tokens": 249561088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 5.382373809814453, "created_at": "2025-01-15T06:52:52.342642+00:00"} {"global_step": 477, "acc_step": 0, "speed/wps": 12863.055271080613, "speed/FLOPS": 202031958722543.47, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.671030580997467, "optim/lr": 0.0002856, "optim/total_tokens": 250085376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 5.458707332611084, "created_at": "2025-01-15T06:53:02.536214+00:00"} {"global_step": 478, "acc_step": 0, "speed/wps": 12857.89839043145, "speed/FLOPS": 201950962825651.75, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5773696303367615, "optim/lr": 0.0002862, "optim/total_tokens": 250609664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288973, "loss/out": 5.435952663421631, "created_at": "2025-01-15T06:53:12.733700+00:00"} {"global_step": 479, "acc_step": 0, "speed/wps": 12864.042626961698, "speed/FLOPS": 202047466503425.06, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6587324738502502, "optim/lr": 0.00028680000000000003, "optim/total_tokens": 251133952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 5.59232234954834, "created_at": "2025-01-15T06:53:22.930311+00:00"} {"global_step": 480, "acc_step": 0, "speed/wps": 12867.06629131195, "speed/FLOPS": 202094957306996.0, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6016582250595093, "optim/lr": 0.0002874, "optim/total_tokens": 251658240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 5.476851463317871, "created_at": "2025-01-15T06:53:33.123121+00:00"} {"global_step": 481, "acc_step": 0, "speed/wps": 12864.891495849659, "speed/FLOPS": 202060799155778.56, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5311970114707947, "optim/lr": 0.000288, "optim/total_tokens": 252182528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 5.45361328125, "created_at": "2025-01-15T06:53:43.315113+00:00"} {"global_step": 482, "acc_step": 0, "speed/wps": 12861.80760300032, "speed/FLOPS": 202012362380866.84, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5198078751564026, "optim/lr": 0.00028859999999999997, "optim/total_tokens": 252706816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 5.552944183349609, "created_at": "2025-01-15T06:53:53.512294+00:00"} {"global_step": 483, "acc_step": 0, "speed/wps": 12866.514373041662, "speed/FLOPS": 202086288672145.94, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4474731385707855, "optim/lr": 0.0002892, "optim/total_tokens": 253231104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 5.4378743171691895, "created_at": "2025-01-15T06:54:03.706392+00:00"} {"global_step": 484, "acc_step": 0, "speed/wps": 12865.982509015044, "speed/FLOPS": 202077935016750.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36226364970207214, "optim/lr": 0.0002898, "optim/total_tokens": 253755392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 5.318795680999756, "created_at": "2025-01-15T06:54:13.897441+00:00"} {"global_step": 485, "acc_step": 0, "speed/wps": 12864.855099160153, "speed/FLOPS": 202060227495755.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42873653769493103, "optim/lr": 0.0002904, "optim/total_tokens": 254279680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.451604843139648, "created_at": "2025-01-15T06:54:24.090881+00:00"} {"global_step": 486, "acc_step": 0, "speed/wps": 12863.948581106637, "speed/FLOPS": 202045989384038.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4968014359474182, "optim/lr": 0.000291, "optim/total_tokens": 254803968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 5.460745811462402, "created_at": "2025-01-15T06:54:34.285428+00:00"} {"global_step": 487, "acc_step": 0, "speed/wps": 12864.077451661857, "speed/FLOPS": 202048013473195.2, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5400617122650146, "optim/lr": 0.0002916, "optim/total_tokens": 255328256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 5.446353912353516, "created_at": "2025-01-15T06:54:44.477870+00:00"} {"global_step": 488, "acc_step": 0, "speed/wps": 12870.429334518969, "speed/FLOPS": 202147778521867.88, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5418710708618164, "optim/lr": 0.0002922, "optim/total_tokens": 255852544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415534, "loss/out": 5.418607711791992, "created_at": "2025-01-15T06:54:54.665531+00:00"} {"global_step": 489, "acc_step": 0, "speed/wps": 12866.128830607393, "speed/FLOPS": 202080233198424.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5522338151931763, "optim/lr": 0.0002928, "optim/total_tokens": 256376832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 5.49833869934082, "created_at": "2025-01-15T06:55:04.856631+00:00"} {"global_step": 490, "acc_step": 0, "speed/wps": 12864.86986243226, "speed/FLOPS": 202060459373230.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.47802048921585083, "optim/lr": 0.0002934, "optim/total_tokens": 256901120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 5.355588912963867, "created_at": "2025-01-15T06:55:15.048547+00:00"} {"global_step": 491, "acc_step": 0, "speed/wps": 12866.924669121383, "speed/FLOPS": 202092732935883.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.46517300605773926, "optim/lr": 0.000294, "optim/total_tokens": 257425408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313228, "loss/out": 5.399226665496826, "created_at": "2025-01-15T06:55:25.238997+00:00"} {"global_step": 492, "acc_step": 0, "speed/wps": 12867.957603095148, "speed/FLOPS": 202108956583341.84, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4774914085865021, "optim/lr": 0.0002946, "optim/total_tokens": 257949696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 5.440483093261719, "created_at": "2025-01-15T06:55:35.430133+00:00"} {"global_step": 493, "acc_step": 0, "speed/wps": 12862.41889202822, "speed/FLOPS": 202021963515049.16, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5241953730583191, "optim/lr": 0.0002952, "optim/total_tokens": 258473984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 5.492912292480469, "created_at": "2025-01-15T06:55:45.624826+00:00"} {"global_step": 494, "acc_step": 0, "speed/wps": 12864.996708724093, "speed/FLOPS": 202062451668549.2, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5270798206329346, "optim/lr": 0.0002958, "optim/total_tokens": 258998272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 5.414461135864258, "created_at": "2025-01-15T06:55:55.820289+00:00"} {"global_step": 495, "acc_step": 0, "speed/wps": 12869.17746311318, "speed/FLOPS": 202128116161188.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5135361552238464, "optim/lr": 0.0002964, "optim/total_tokens": 259522560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482329, "loss/out": 5.410820007324219, "created_at": "2025-01-15T06:56:06.012763+00:00"} {"global_step": 496, "acc_step": 0, "speed/wps": 12869.478076166633, "speed/FLOPS": 202132837702278.47, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.47768715023994446, "optim/lr": 0.000297, "optim/total_tokens": 260046848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 5.332475662231445, "created_at": "2025-01-15T06:56:16.201833+00:00"} {"global_step": 497, "acc_step": 0, "speed/wps": 12864.945793742878, "speed/FLOPS": 202061651978804.7, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44683411717414856, "optim/lr": 0.00029759999999999997, "optim/total_tokens": 260571136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 5.408856391906738, "created_at": "2025-01-15T06:56:26.396214+00:00"} {"global_step": 498, "acc_step": 0, "speed/wps": 12866.024048520772, "speed/FLOPS": 202078587451768.88, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.617771565914154, "optim/lr": 0.00029820000000000004, "optim/total_tokens": 261095424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 5.450464248657227, "created_at": "2025-01-15T06:56:36.587194+00:00"} {"global_step": 499, "acc_step": 0, "speed/wps": 12865.831129738228, "speed/FLOPS": 202075557397190.8, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5987701416015625, "optim/lr": 0.0002988, "optim/total_tokens": 261619712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 5.442814826965332, "created_at": "2025-01-15T06:56:46.778337+00:00"} {"global_step": 500, "acc_step": 0, "speed/wps": 12869.92935396854, "speed/FLOPS": 202139925640278.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6780834794044495, "optim/lr": 0.0002994, "optim/total_tokens": 262144000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 5.583305358886719, "created_at": "2025-01-15T06:56:56.969787+00:00"} {"global_step": 501, "acc_step": 0, "speed/wps": 12859.386100862532, "speed/FLOPS": 201974329362300.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5277129411697388, "optim/lr": 0.00030000000000000003, "optim/total_tokens": 262668288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311352, "loss/out": 5.446858882904053, "created_at": "2025-01-15T06:57:07.169306+00:00"} {"global_step": 502, "acc_step": 0, "speed/wps": 12867.343392287341, "speed/FLOPS": 202099309558591.16, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48137521743774414, "optim/lr": 0.0003006, "optim/total_tokens": 263192576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 5.418934345245361, "created_at": "2025-01-15T06:57:17.359414+00:00"} {"global_step": 503, "acc_step": 0, "speed/wps": 12865.87963940742, "speed/FLOPS": 202076319308204.75, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5098386406898499, "optim/lr": 0.0003012, "optim/total_tokens": 263716864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 5.496289253234863, "created_at": "2025-01-15T06:57:27.553451+00:00"} {"global_step": 504, "acc_step": 0, "speed/wps": 12868.296865966775, "speed/FLOPS": 202114285172936.78, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5190509557723999, "optim/lr": 0.0003018, "optim/total_tokens": 264241152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 5.365684509277344, "created_at": "2025-01-15T06:57:37.743858+00:00"} {"global_step": 505, "acc_step": 0, "speed/wps": 12862.152725886284, "speed/FLOPS": 202017783010039.47, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6707389950752258, "optim/lr": 0.00030240000000000003, "optim/total_tokens": 264765440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.480635643005371, "created_at": "2025-01-15T06:57:47.938036+00:00"} {"global_step": 506, "acc_step": 0, "speed/wps": 12864.92628398622, "speed/FLOPS": 202061345551267.2, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5762478709220886, "optim/lr": 0.00030300000000000005, "optim/total_tokens": 265289728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.324214935302734, "created_at": "2025-01-15T06:57:58.132640+00:00"} {"global_step": 507, "acc_step": 0, "speed/wps": 12869.987479365334, "speed/FLOPS": 202140838579507.4, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6964262127876282, "optim/lr": 0.0003036, "optim/total_tokens": 265814016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.44190788269043, "created_at": "2025-01-15T06:58:08.323570+00:00"} {"global_step": 508, "acc_step": 0, "speed/wps": 12864.77313503988, "speed/FLOPS": 202058940136616.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5679446458816528, "optim/lr": 0.0003042, "optim/total_tokens": 266338304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.435542106628418, "created_at": "2025-01-15T06:58:18.515448+00:00"} {"global_step": 509, "acc_step": 0, "speed/wps": 12869.911022475018, "speed/FLOPS": 202139637718982.22, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5435681343078613, "optim/lr": 0.0003048, "optim/total_tokens": 266862592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 5.394337177276611, "created_at": "2025-01-15T06:58:28.706956+00:00"} {"global_step": 510, "acc_step": 0, "speed/wps": 12865.547575905015, "speed/FLOPS": 202071103794596.4, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7881654500961304, "optim/lr": 0.0003054, "optim/total_tokens": 267386880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.344787120819092, "created_at": "2025-01-15T06:58:38.898492+00:00"} {"global_step": 511, "acc_step": 0, "speed/wps": 12866.060097788513, "speed/FLOPS": 202079153655055.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6049123406410217, "optim/lr": 0.000306, "optim/total_tokens": 267911168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 5.31032133102417, "created_at": "2025-01-15T06:58:49.090380+00:00"} {"global_step": 512, "acc_step": 0, "speed/wps": 12867.946667095906, "speed/FLOPS": 202108784818446.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5056230425834656, "optim/lr": 0.0003066, "optim/total_tokens": 268435456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396624, "loss/out": 5.388782024383545, "created_at": "2025-01-15T06:58:59.287884+00:00"} {"global_step": 513, "acc_step": 0, "speed/wps": 12865.946877725903, "speed/FLOPS": 202077375378391.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6635103225708008, "optim/lr": 0.00030720000000000004, "optim/total_tokens": 268959744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 5.352825164794922, "created_at": "2025-01-15T06:59:09.484377+00:00"} {"global_step": 514, "acc_step": 0, "speed/wps": 12865.422763959807, "speed/FLOPS": 202069143451487.3, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5636512637138367, "optim/lr": 0.0003078, "optim/total_tokens": 269484032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333121, "loss/out": 5.533177852630615, "created_at": "2025-01-15T06:59:19.676885+00:00"} {"global_step": 515, "acc_step": 0, "speed/wps": 12869.164259761394, "speed/FLOPS": 202127908784405.25, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48862046003341675, "optim/lr": 0.0003084, "optim/total_tokens": 270008320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.393943786621094, "created_at": "2025-01-15T06:59:29.867574+00:00"} {"global_step": 516, "acc_step": 0, "speed/wps": 12864.05117879095, "speed/FLOPS": 202047600821654.9, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5098924040794373, "optim/lr": 0.000309, "optim/total_tokens": 270532608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 5.427488803863525, "created_at": "2025-01-15T06:59:40.061972+00:00"} {"global_step": 517, "acc_step": 0, "speed/wps": 12865.023241827723, "speed/FLOPS": 202062868407402.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4407099783420563, "optim/lr": 0.0003096, "optim/total_tokens": 271056896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 5.437778472900391, "created_at": "2025-01-15T06:59:50.253783+00:00"} {"global_step": 518, "acc_step": 0, "speed/wps": 12862.899039520387, "speed/FLOPS": 202029504891205.44, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0209245681762695, "optim/lr": 0.0003102, "optim/total_tokens": 271581184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.413632392883301, "created_at": "2025-01-15T07:00:00.451331+00:00"} {"global_step": 519, "acc_step": 0, "speed/wps": 12867.782941114427, "speed/FLOPS": 202106213276923.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38458871841430664, "optim/lr": 0.0003108, "optim/total_tokens": 272105472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 5.5485520362854, "created_at": "2025-01-15T07:00:10.641106+00:00"} {"global_step": 520, "acc_step": 0, "speed/wps": 12863.573662430454, "speed/FLOPS": 202040100771037.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3673086166381836, "optim/lr": 0.00031140000000000003, "optim/total_tokens": 272629760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.328476428985596, "created_at": "2025-01-15T07:00:20.834127+00:00"} {"global_step": 521, "acc_step": 0, "speed/wps": 12865.895153075362, "speed/FLOPS": 202076562971677.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38873162865638733, "optim/lr": 0.000312, "optim/total_tokens": 273154048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 5.279761791229248, "created_at": "2025-01-15T07:00:31.025470+00:00"} {"global_step": 522, "acc_step": 0, "speed/wps": 12863.236014702077, "speed/FLOPS": 202034797549486.3, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4530465006828308, "optim/lr": 0.0003126, "optim/total_tokens": 273678336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 5.292891502380371, "created_at": "2025-01-15T07:00:41.219663+00:00"} {"global_step": 523, "acc_step": 0, "speed/wps": 12864.506262331106, "speed/FLOPS": 202054748534003.38, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5118036270141602, "optim/lr": 0.0003132, "optim/total_tokens": 274202624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 5.315817832946777, "created_at": "2025-01-15T07:00:51.412755+00:00"} {"global_step": 524, "acc_step": 0, "speed/wps": 12867.145118965102, "speed/FLOPS": 202096195403609.28, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46208637952804565, "optim/lr": 0.0003138, "optim/total_tokens": 274726912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 5.486350059509277, "created_at": "2025-01-15T07:01:01.603780+00:00"} {"global_step": 525, "acc_step": 0, "speed/wps": 12862.637558155233, "speed/FLOPS": 202025397967052.66, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47246190905570984, "optim/lr": 0.0003144, "optim/total_tokens": 275251200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 5.348771095275879, "created_at": "2025-01-15T07:01:11.802133+00:00"} {"global_step": 526, "acc_step": 0, "speed/wps": 12865.577852493134, "speed/FLOPS": 202071579330017.22, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.503632128238678, "optim/lr": 0.000315, "optim/total_tokens": 275775488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333025, "loss/out": 5.328308582305908, "created_at": "2025-01-15T07:01:22.000134+00:00"} {"global_step": 527, "acc_step": 0, "speed/wps": 12867.307881690822, "speed/FLOPS": 202098751815876.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.409555584192276, "optim/lr": 0.00031560000000000003, "optim/total_tokens": 276299776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348945, "loss/out": 5.38477087020874, "created_at": "2025-01-15T07:01:32.191030+00:00"} {"global_step": 528, "acc_step": 0, "speed/wps": 12867.355201354054, "speed/FLOPS": 202099495036211.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4632096290588379, "optim/lr": 0.0003162, "optim/total_tokens": 276824064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 5.299031734466553, "created_at": "2025-01-15T07:01:42.386569+00:00"} {"global_step": 529, "acc_step": 0, "speed/wps": 12864.849480329312, "speed/FLOPS": 202060139244296.16, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3748205602169037, "optim/lr": 0.0003168, "optim/total_tokens": 277348352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 5.547642707824707, "created_at": "2025-01-15T07:01:52.585904+00:00"} {"global_step": 530, "acc_step": 0, "speed/wps": 12866.004904817397, "speed/FLOPS": 202078286773601.16, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3871896266937256, "optim/lr": 0.0003174, "optim/total_tokens": 277872640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 5.369524955749512, "created_at": "2025-01-15T07:02:02.780137+00:00"} {"global_step": 531, "acc_step": 0, "speed/wps": 12870.398337135197, "speed/FLOPS": 202147291665360.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4523596167564392, "optim/lr": 0.000318, "optim/total_tokens": 278396928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343732, "loss/out": 5.454339027404785, "created_at": "2025-01-15T07:02:12.970161+00:00"} {"global_step": 532, "acc_step": 0, "speed/wps": 12865.530471806835, "speed/FLOPS": 202070835151231.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.470542848110199, "optim/lr": 0.0003186, "optim/total_tokens": 278921216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 5.373271465301514, "created_at": "2025-01-15T07:02:23.166120+00:00"} {"global_step": 533, "acc_step": 0, "speed/wps": 12867.56550472175, "speed/FLOPS": 202102798139588.0, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4663696587085724, "optim/lr": 0.0003192, "optim/total_tokens": 279445504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 5.369527339935303, "created_at": "2025-01-15T07:02:33.355876+00:00"} {"global_step": 534, "acc_step": 0, "speed/wps": 12867.41010238747, "speed/FLOPS": 202100357332382.94, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3664204478263855, "optim/lr": 0.0003198, "optim/total_tokens": 279969792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 5.282278537750244, "created_at": "2025-01-15T07:02:43.548671+00:00"} {"global_step": 535, "acc_step": 0, "speed/wps": 12867.626417041996, "speed/FLOPS": 202103754851279.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47191378474235535, "optim/lr": 0.00032040000000000004, "optim/total_tokens": 280494080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 5.320575714111328, "created_at": "2025-01-15T07:02:53.738389+00:00"} {"global_step": 536, "acc_step": 0, "speed/wps": 12864.25492098979, "speed/FLOPS": 202050800872858.78, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4303925037384033, "optim/lr": 0.000321, "optim/total_tokens": 281018368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.356904029846191, "created_at": "2025-01-15T07:03:03.930785+00:00"} {"global_step": 537, "acc_step": 0, "speed/wps": 12865.102007993286, "speed/FLOPS": 202064105538268.22, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4327523708343506, "optim/lr": 0.0003216, "optim/total_tokens": 281542656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 5.350139617919922, "created_at": "2025-01-15T07:03:14.123936+00:00"} {"global_step": 538, "acc_step": 0, "speed/wps": 12867.169303418708, "speed/FLOPS": 202096575253686.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41618359088897705, "optim/lr": 0.00032219999999999997, "optim/total_tokens": 282066944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 5.307023048400879, "created_at": "2025-01-15T07:03:24.316150+00:00"} {"global_step": 539, "acc_step": 0, "speed/wps": 12868.984500352019, "speed/FLOPS": 202125085415865.72, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5210062265396118, "optim/lr": 0.0003228, "optim/total_tokens": 282591232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 5.184450149536133, "created_at": "2025-01-15T07:03:34.509004+00:00"} {"global_step": 540, "acc_step": 0, "speed/wps": 12870.104796864762, "speed/FLOPS": 202142681212047.06, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4353039860725403, "optim/lr": 0.00032340000000000005, "optim/total_tokens": 283115520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 5.223653316497803, "created_at": "2025-01-15T07:03:44.701212+00:00"} {"global_step": 541, "acc_step": 0, "speed/wps": 12868.08129942368, "speed/FLOPS": 202110899404158.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.550046980381012, "optim/lr": 0.000324, "optim/total_tokens": 283639808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 5.289861679077148, "created_at": "2025-01-15T07:03:54.895483+00:00"} {"global_step": 542, "acc_step": 0, "speed/wps": 12872.388113823947, "speed/FLOPS": 202178543842497.38, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6050761342048645, "optim/lr": 0.00032460000000000003, "optim/total_tokens": 284164096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 5.30824613571167, "created_at": "2025-01-15T07:04:05.083507+00:00"} {"global_step": 543, "acc_step": 0, "speed/wps": 12866.496614713451, "speed/FLOPS": 202086009753198.88, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4512723386287689, "optim/lr": 0.0003252, "optim/total_tokens": 284688384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 5.218911170959473, "created_at": "2025-01-15T07:04:15.278898+00:00"} {"global_step": 544, "acc_step": 0, "speed/wps": 12864.879058624938, "speed/FLOPS": 202060603812073.78, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4612451195716858, "optim/lr": 0.0003258, "optim/total_tokens": 285212672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299645, "loss/out": 5.30799674987793, "created_at": "2025-01-15T07:04:25.472576+00:00"} {"global_step": 545, "acc_step": 0, "speed/wps": 12864.451096801355, "speed/FLOPS": 202053882083553.25, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.518020749092102, "optim/lr": 0.00032639999999999996, "optim/total_tokens": 285736960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.437898635864258, "created_at": "2025-01-15T07:04:35.665844+00:00"} {"global_step": 546, "acc_step": 0, "speed/wps": 12865.008290042277, "speed/FLOPS": 202062633569065.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4820213317871094, "optim/lr": 0.000327, "optim/total_tokens": 286261248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 5.259490013122559, "created_at": "2025-01-15T07:04:45.859427+00:00"} {"global_step": 547, "acc_step": 0, "speed/wps": 12862.618879474334, "speed/FLOPS": 202025104592701.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4787333309650421, "optim/lr": 0.00032760000000000005, "optim/total_tokens": 286785536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 5.3835601806640625, "created_at": "2025-01-15T07:04:56.054259+00:00"} {"global_step": 548, "acc_step": 0, "speed/wps": 12868.51540253499, "speed/FLOPS": 202117717590041.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.37850263714790344, "optim/lr": 0.0003282, "optim/total_tokens": 287309824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.300674915313721, "created_at": "2025-01-15T07:05:06.246638+00:00"} {"global_step": 549, "acc_step": 0, "speed/wps": 12865.264352105021, "speed/FLOPS": 202066655375627.22, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42018070816993713, "optim/lr": 0.0003288, "optim/total_tokens": 287834112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.277455806732178, "created_at": "2025-01-15T07:05:16.439308+00:00"} {"global_step": 550, "acc_step": 0, "speed/wps": 12865.241769536553, "speed/FLOPS": 202066300685357.72, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5379981398582458, "optim/lr": 0.0003294, "optim/total_tokens": 288358400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 5.3640594482421875, "created_at": "2025-01-15T07:05:26.635642+00:00"} {"global_step": 551, "acc_step": 0, "speed/wps": 12867.082780714127, "speed/FLOPS": 202095216295715.97, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49902990460395813, "optim/lr": 0.00033, "optim/total_tokens": 288882688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 5.324794769287109, "created_at": "2025-01-15T07:05:36.827851+00:00"} {"global_step": 552, "acc_step": 0, "speed/wps": 12868.45363226517, "speed/FLOPS": 202116747403072.53, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4181171655654907, "optim/lr": 0.0003306, "optim/total_tokens": 289406976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 5.234595775604248, "created_at": "2025-01-15T07:05:47.019052+00:00"} {"global_step": 553, "acc_step": 0, "speed/wps": 12861.76388916381, "speed/FLOPS": 202011675794995.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.001, "optim/grad_norm": 0.48716670274734497, "optim/lr": 0.00033120000000000003, "optim/total_tokens": 289931264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 5.3247270584106445, "created_at": "2025-01-15T07:05:57.219082+00:00"} {"global_step": 554, "acc_step": 0, "speed/wps": 12864.285307905902, "speed/FLOPS": 202051278141132.34, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6035025119781494, "optim/lr": 0.00033180000000000004, "optim/total_tokens": 290455552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 5.329646587371826, "created_at": "2025-01-15T07:06:07.412004+00:00"} {"global_step": 555, "acc_step": 0, "speed/wps": 12868.983955523627, "speed/FLOPS": 202125076858587.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7049612998962402, "optim/lr": 0.0003324, "optim/total_tokens": 290979840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.304428577423096, "created_at": "2025-01-15T07:06:17.609145+00:00"} {"global_step": 556, "acc_step": 0, "speed/wps": 12866.01442490951, "speed/FLOPS": 202078436299729.75, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45060351490974426, "optim/lr": 0.000333, "optim/total_tokens": 291504128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.242146015167236, "created_at": "2025-01-15T07:06:27.800628+00:00"} {"global_step": 557, "acc_step": 0, "speed/wps": 12862.969933480646, "speed/FLOPS": 202030618378269.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5535264611244202, "optim/lr": 0.0003336, "optim/total_tokens": 292028416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.353569507598877, "created_at": "2025-01-15T07:06:37.998203+00:00"} {"global_step": 558, "acc_step": 0, "speed/wps": 12866.18641943253, "speed/FLOPS": 202081137710059.03, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49685272574424744, "optim/lr": 0.0003342, "optim/total_tokens": 292552704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 5.362618923187256, "created_at": "2025-01-15T07:06:48.189714+00:00"} {"global_step": 559, "acc_step": 0, "speed/wps": 12867.914282763923, "speed/FLOPS": 202108276178011.56, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4060404896736145, "optim/lr": 0.0003348, "optim/total_tokens": 293076992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 5.256722927093506, "created_at": "2025-01-15T07:06:58.379439+00:00"} {"global_step": 560, "acc_step": 0, "speed/wps": 12868.319555744165, "speed/FLOPS": 202114641547069.72, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5399123430252075, "optim/lr": 0.0003354, "optim/total_tokens": 293601280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 5.335872173309326, "created_at": "2025-01-15T07:07:08.568684+00:00"} {"global_step": 561, "acc_step": 0, "speed/wps": 12867.24029504399, "speed/FLOPS": 202097690274713.84, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5774452686309814, "optim/lr": 0.00033600000000000004, "optim/total_tokens": 294125568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.160541534423828, "created_at": "2025-01-15T07:07:18.759210+00:00"} {"global_step": 562, "acc_step": 0, "speed/wps": 12865.0257396754, "speed/FLOPS": 202062907639532.2, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3868306577205658, "optim/lr": 0.0003366, "optim/total_tokens": 294649856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.33310604095459, "created_at": "2025-01-15T07:07:28.952975+00:00"} {"global_step": 563, "acc_step": 0, "speed/wps": 12866.986619990992, "speed/FLOPS": 202093705959420.47, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3798570930957794, "optim/lr": 0.0003372, "optim/total_tokens": 295174144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 5.272915363311768, "created_at": "2025-01-15T07:07:39.143178+00:00"} {"global_step": 564, "acc_step": 0, "speed/wps": 12863.075107873163, "speed/FLOPS": 202032270286629.06, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3815588355064392, "optim/lr": 0.0003378, "optim/total_tokens": 295698432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 5.207270622253418, "created_at": "2025-01-15T07:07:49.336495+00:00"} {"global_step": 565, "acc_step": 0, "speed/wps": 12866.996959114324, "speed/FLOPS": 202093868349559.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35659509897232056, "optim/lr": 0.0003384, "optim/total_tokens": 296222720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486334, "loss/out": 5.1748809814453125, "created_at": "2025-01-15T07:07:59.536671+00:00"} {"global_step": 566, "acc_step": 0, "speed/wps": 12868.01189392968, "speed/FLOPS": 202109809295501.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.373138964176178, "optim/lr": 0.000339, "optim/total_tokens": 296747008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 5.144243240356445, "created_at": "2025-01-15T07:08:09.730155+00:00"} {"global_step": 567, "acc_step": 0, "speed/wps": 12868.633646301638, "speed/FLOPS": 202119574770880.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3912786841392517, "optim/lr": 0.0003396, "optim/total_tokens": 297271296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 5.297511100769043, "created_at": "2025-01-15T07:08:19.924054+00:00"} {"global_step": 568, "acc_step": 0, "speed/wps": 12870.665206007327, "speed/FLOPS": 202151483207713.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46814990043640137, "optim/lr": 0.00034020000000000003, "optim/total_tokens": 297795584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 5.2311482429504395, "created_at": "2025-01-15T07:08:30.112526+00:00"} {"global_step": 569, "acc_step": 0, "speed/wps": 12863.893564983604, "speed/FLOPS": 202045125280226.47, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4859252870082855, "optim/lr": 0.00034080000000000004, "optim/total_tokens": 298319872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.280145645141602, "created_at": "2025-01-15T07:08:40.305292+00:00"} {"global_step": 570, "acc_step": 0, "speed/wps": 12868.607051236932, "speed/FLOPS": 202119157058843.8, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6470434069633484, "optim/lr": 0.0003414, "optim/total_tokens": 298844160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 5.276942729949951, "created_at": "2025-01-15T07:08:50.497083+00:00"} {"global_step": 571, "acc_step": 0, "speed/wps": 12868.695408886197, "speed/FLOPS": 202120544837141.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.649692952632904, "optim/lr": 0.000342, "optim/total_tokens": 299368448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 5.203901767730713, "created_at": "2025-01-15T07:09:00.688944+00:00"} {"global_step": 572, "acc_step": 0, "speed/wps": 12868.748483363448, "speed/FLOPS": 202121378444738.88, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5798019766807556, "optim/lr": 0.0003426, "optim/total_tokens": 299892736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310540, "loss/out": 5.240136623382568, "created_at": "2025-01-15T07:09:10.877690+00:00"} {"global_step": 573, "acc_step": 0, "speed/wps": 12864.510480344634, "speed/FLOPS": 202054814783701.97, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.502007782459259, "optim/lr": 0.0003432, "optim/total_tokens": 300417024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 5.20894193649292, "created_at": "2025-01-15T07:09:21.071146+00:00"} {"global_step": 574, "acc_step": 0, "speed/wps": 12867.33059033295, "speed/FLOPS": 202099108486305.7, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5805234909057617, "optim/lr": 0.0003438, "optim/total_tokens": 300941312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.195693016052246, "created_at": "2025-01-15T07:09:31.261372+00:00"} {"global_step": 575, "acc_step": 0, "speed/wps": 12865.117094063771, "speed/FLOPS": 202064342485735.62, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42427292466163635, "optim/lr": 0.0003444, "optim/total_tokens": 301465600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.145325183868408, "created_at": "2025-01-15T07:09:41.454216+00:00"} {"global_step": 576, "acc_step": 0, "speed/wps": 12866.600945740716, "speed/FLOPS": 202087648415347.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4298679530620575, "optim/lr": 0.00034500000000000004, "optim/total_tokens": 301989888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.173248767852783, "created_at": "2025-01-15T07:09:51.652423+00:00"} {"global_step": 577, "acc_step": 0, "speed/wps": 12864.42085491371, "speed/FLOPS": 202053407093151.06, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4003359079360962, "optim/lr": 0.0003456, "optim/total_tokens": 302514176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 5.2597737312316895, "created_at": "2025-01-15T07:10:01.844773+00:00"} {"global_step": 578, "acc_step": 0, "speed/wps": 12867.698299615882, "speed/FLOPS": 202104883865879.66, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4037249982357025, "optim/lr": 0.0003462, "optim/total_tokens": 303038464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 5.154718399047852, "created_at": "2025-01-15T07:10:12.037685+00:00"} {"global_step": 579, "acc_step": 0, "speed/wps": 12868.049142667493, "speed/FLOPS": 202110394338114.06, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4222078323364258, "optim/lr": 0.00034679999999999997, "optim/total_tokens": 303562752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 5.300076484680176, "created_at": "2025-01-15T07:10:22.229254+00:00"} {"global_step": 580, "acc_step": 0, "speed/wps": 12864.934448912141, "speed/FLOPS": 202061473792648.53, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41323214769363403, "optim/lr": 0.0003474, "optim/total_tokens": 304087040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 5.181986331939697, "created_at": "2025-01-15T07:10:32.424433+00:00"} {"global_step": 581, "acc_step": 0, "speed/wps": 12868.427359165407, "speed/FLOPS": 202116334747937.75, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5000501275062561, "optim/lr": 0.000348, "optim/total_tokens": 304611328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 5.2074432373046875, "created_at": "2025-01-15T07:10:42.618829+00:00"} {"global_step": 582, "acc_step": 0, "speed/wps": 12865.284639085758, "speed/FLOPS": 202066974010536.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5670238137245178, "optim/lr": 0.0003486, "optim/total_tokens": 305135616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 5.319972515106201, "created_at": "2025-01-15T07:10:52.811359+00:00"} {"global_step": 583, "acc_step": 0, "speed/wps": 12865.278109611121, "speed/FLOPS": 202066871456165.3, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5430800914764404, "optim/lr": 0.00034920000000000003, "optim/total_tokens": 305659904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 5.194026947021484, "created_at": "2025-01-15T07:11:03.002894+00:00"} {"global_step": 584, "acc_step": 0, "speed/wps": 12859.876811437818, "speed/FLOPS": 201982036646192.12, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5066615343093872, "optim/lr": 0.0003498, "optim/total_tokens": 306184192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 5.1895647048950195, "created_at": "2025-01-15T07:11:13.200713+00:00"} {"global_step": 585, "acc_step": 0, "speed/wps": 12867.639852536278, "speed/FLOPS": 202103965874179.78, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5634413957595825, "optim/lr": 0.0003504, "optim/total_tokens": 306708480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 5.149653434753418, "created_at": "2025-01-15T07:11:23.394315+00:00"} {"global_step": 586, "acc_step": 0, "speed/wps": 12867.365128139594, "speed/FLOPS": 202099650950018.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.562688410282135, "optim/lr": 0.000351, "optim/total_tokens": 307232768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 5.222074031829834, "created_at": "2025-01-15T07:11:33.587764+00:00"} {"global_step": 587, "acc_step": 0, "speed/wps": 12851.23418679203, "speed/FLOPS": 201846292349919.2, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5590258240699768, "optim/lr": 0.0003516, "optim/total_tokens": 307757056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 5.2067742347717285, "created_at": "2025-01-15T07:11:43.800914+00:00"} {"global_step": 588, "acc_step": 0, "speed/wps": 12863.392321463825, "speed/FLOPS": 202037252561967.56, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5476362705230713, "optim/lr": 0.0003522, "optim/total_tokens": 308281344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 5.175442695617676, "created_at": "2025-01-15T07:11:53.994765+00:00"} {"global_step": 589, "acc_step": 0, "speed/wps": 12859.457964249297, "speed/FLOPS": 201975458075540.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5440067648887634, "optim/lr": 0.0003528, "optim/total_tokens": 308805632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 5.213016510009766, "created_at": "2025-01-15T07:12:04.191309+00:00"} {"global_step": 590, "acc_step": 0, "speed/wps": 12861.707536484104, "speed/FLOPS": 202010790698723.88, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.558741569519043, "optim/lr": 0.0003534, "optim/total_tokens": 309329920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359993, "loss/out": 5.301230430603027, "created_at": "2025-01-15T07:12:14.386239+00:00"} {"global_step": 591, "acc_step": 0, "speed/wps": 12859.161991328958, "speed/FLOPS": 201970809414116.7, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.680176317691803, "optim/lr": 0.000354, "optim/total_tokens": 309854208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 5.2545976638793945, "created_at": "2025-01-15T07:12:24.584101+00:00"} {"global_step": 592, "acc_step": 0, "speed/wps": 12864.190809324407, "speed/FLOPS": 202049793911055.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6855819225311279, "optim/lr": 0.0003546, "optim/total_tokens": 310378496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.194292068481445, "created_at": "2025-01-15T07:12:34.782358+00:00"} {"global_step": 593, "acc_step": 0, "speed/wps": 12862.192386912337, "speed/FLOPS": 202018405940953.56, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4226595461368561, "optim/lr": 0.0003552, "optim/total_tokens": 310902784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 5.15634822845459, "created_at": "2025-01-15T07:12:44.976477+00:00"} {"global_step": 594, "acc_step": 0, "speed/wps": 12861.163666766231, "speed/FLOPS": 202002248477449.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.555421769618988, "optim/lr": 0.0003558, "optim/total_tokens": 311427072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 5.173102378845215, "created_at": "2025-01-15T07:12:55.171755+00:00"} {"global_step": 595, "acc_step": 0, "speed/wps": 12863.419185380966, "speed/FLOPS": 202037674496701.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35246917605400085, "optim/lr": 0.0003564, "optim/total_tokens": 311951360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 5.249251842498779, "created_at": "2025-01-15T07:13:05.370088+00:00"} {"global_step": 596, "acc_step": 0, "speed/wps": 12865.327753542511, "speed/FLOPS": 202067651182325.72, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35378342866897583, "optim/lr": 0.000357, "optim/total_tokens": 312475648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 5.272199630737305, "created_at": "2025-01-15T07:13:15.562093+00:00"} {"global_step": 597, "acc_step": 0, "speed/wps": 12862.335610626733, "speed/FLOPS": 202020655466198.16, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3919176161289215, "optim/lr": 0.0003576, "optim/total_tokens": 312999936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.071545124053955, "created_at": "2025-01-15T07:13:25.759867+00:00"} {"global_step": 598, "acc_step": 0, "speed/wps": 12863.704036337855, "speed/FLOPS": 202042148472405.3, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44759511947631836, "optim/lr": 0.00035820000000000003, "optim/total_tokens": 313524224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.137224197387695, "created_at": "2025-01-15T07:13:35.952623+00:00"} {"global_step": 599, "acc_step": 0, "speed/wps": 12861.713165500201, "speed/FLOPS": 202010879110156.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.39846116304397583, "optim/lr": 0.0003588, "optim/total_tokens": 314048512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 5.071236610412598, "created_at": "2025-01-15T07:13:46.147298+00:00"} {"global_step": 600, "acc_step": 0, "speed/wps": 12863.703126292132, "speed/FLOPS": 202042134178886.72, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5652667880058289, "optim/lr": 0.0003594, "optim/total_tokens": 314572800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.073694705963135, "created_at": "2025-01-15T07:13:56.340487+00:00"} {"global_step": 601, "acc_step": 0, "speed/wps": 12869.204698347934, "speed/FLOPS": 202128543927975.3, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.613560140132904, "optim/lr": 0.00035999999999999997, "optim/total_tokens": 315097088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.17837381362915, "created_at": "2025-01-15T07:14:06.531326+00:00"} {"global_step": 602, "acc_step": 0, "speed/wps": 12861.36405593421, "speed/FLOPS": 202005395864696.03, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5470243096351624, "optim/lr": 0.00036060000000000004, "optim/total_tokens": 315621376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 5.242631912231445, "created_at": "2025-01-15T07:14:16.729506+00:00"} {"global_step": 603, "acc_step": 0, "speed/wps": 12864.430771292726, "speed/FLOPS": 202053562843510.03, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6185039281845093, "optim/lr": 0.0003612, "optim/total_tokens": 316145664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.149819374084473, "created_at": "2025-01-15T07:14:26.925054+00:00"} {"global_step": 604, "acc_step": 0, "speed/wps": 12862.228313618405, "speed/FLOPS": 202018970219240.84, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6036338210105896, "optim/lr": 0.0003618, "optim/total_tokens": 316669952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 5.041830062866211, "created_at": "2025-01-15T07:14:37.119216+00:00"} {"global_step": 605, "acc_step": 0, "speed/wps": 12863.279831425612, "speed/FLOPS": 202035485751339.9, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40565505623817444, "optim/lr": 0.0003624, "optim/total_tokens": 317194240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 5.217276573181152, "created_at": "2025-01-15T07:14:47.312874+00:00"} {"global_step": 606, "acc_step": 0, "speed/wps": 12865.509030296851, "speed/FLOPS": 202070498382852.97, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3742862641811371, "optim/lr": 0.000363, "optim/total_tokens": 317718528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478609, "loss/out": 5.093981742858887, "created_at": "2025-01-15T07:14:57.504229+00:00"} {"global_step": 607, "acc_step": 0, "speed/wps": 12865.50081877375, "speed/FLOPS": 202070369409598.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4049222469329834, "optim/lr": 0.0003636, "optim/total_tokens": 318242816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.21997594833374, "created_at": "2025-01-15T07:15:07.696477+00:00"} {"global_step": 608, "acc_step": 0, "speed/wps": 12868.828941363963, "speed/FLOPS": 202122642148197.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38346728682518005, "optim/lr": 0.0003642, "optim/total_tokens": 318767104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 5.127585411071777, "created_at": "2025-01-15T07:15:17.893132+00:00"} {"global_step": 609, "acc_step": 0, "speed/wps": 12863.107432582634, "speed/FLOPS": 202032777990610.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5132274031639099, "optim/lr": 0.00036480000000000003, "optim/total_tokens": 319291392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391480, "loss/out": 5.085460662841797, "created_at": "2025-01-15T07:15:28.088820+00:00"} {"global_step": 610, "acc_step": 0, "speed/wps": 12863.278099631227, "speed/FLOPS": 202035458551129.34, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.558264970779419, "optim/lr": 0.00036540000000000005, "optim/total_tokens": 319815680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 5.177809715270996, "created_at": "2025-01-15T07:15:38.283491+00:00"} {"global_step": 611, "acc_step": 0, "speed/wps": 12865.610489692715, "speed/FLOPS": 202072091942084.84, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5446237921714783, "optim/lr": 0.000366, "optim/total_tokens": 320339968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 5.231513023376465, "created_at": "2025-01-15T07:15:48.474710+00:00"} {"global_step": 612, "acc_step": 0, "speed/wps": 12867.562143531413, "speed/FLOPS": 202102745347475.0, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7062504291534424, "optim/lr": 0.0003666, "optim/total_tokens": 320864256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 5.162709712982178, "created_at": "2025-01-15T07:15:58.664400+00:00"} {"global_step": 613, "acc_step": 0, "speed/wps": 12865.746123435325, "speed/FLOPS": 202074222256392.8, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5628079175949097, "optim/lr": 0.0003672, "optim/total_tokens": 321388544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.125505447387695, "created_at": "2025-01-15T07:16:08.857558+00:00"} {"global_step": 614, "acc_step": 0, "speed/wps": 12868.730684743705, "speed/FLOPS": 202121098892957.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.461101770401001, "optim/lr": 0.0003678, "optim/total_tokens": 321912832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.108926296234131, "created_at": "2025-01-15T07:16:19.048237+00:00"} {"global_step": 615, "acc_step": 0, "speed/wps": 12867.505418856399, "speed/FLOPS": 202101854408506.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3974452614784241, "optim/lr": 0.0003684, "optim/total_tokens": 322437120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 5.1466474533081055, "created_at": "2025-01-15T07:16:29.239901+00:00"} {"global_step": 616, "acc_step": 0, "speed/wps": 12862.630250786417, "speed/FLOPS": 202025283194784.0, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4693358242511749, "optim/lr": 0.000369, "optim/total_tokens": 322961408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.077458381652832, "created_at": "2025-01-15T07:16:39.435807+00:00"} {"global_step": 617, "acc_step": 0, "speed/wps": 12864.198130225066, "speed/FLOPS": 202049908895860.12, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5145597457885742, "optim/lr": 0.00036960000000000004, "optim/total_tokens": 323485696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464770, "loss/out": 5.047416687011719, "created_at": "2025-01-15T07:16:49.633075+00:00"} {"global_step": 618, "acc_step": 0, "speed/wps": 12865.751765396162, "speed/FLOPS": 202074310871140.62, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4237929582595825, "optim/lr": 0.0003702, "optim/total_tokens": 324009984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 5.1953630447387695, "created_at": "2025-01-15T07:16:59.824387+00:00"} {"global_step": 619, "acc_step": 0, "speed/wps": 12864.281238200314, "speed/FLOPS": 202051214220813.72, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.343557745218277, "optim/lr": 0.0003708, "optim/total_tokens": 324534272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462696, "loss/out": 5.11590576171875, "created_at": "2025-01-15T07:17:10.017136+00:00"} {"global_step": 620, "acc_step": 0, "speed/wps": 12865.015363735401, "speed/FLOPS": 202062744671136.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3233740031719208, "optim/lr": 0.0003714, "optim/total_tokens": 325058560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388103, "loss/out": 5.122259140014648, "created_at": "2025-01-15T07:17:20.214998+00:00"} {"global_step": 621, "acc_step": 0, "speed/wps": 12863.661701711551, "speed/FLOPS": 202041483548925.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2980579733848572, "optim/lr": 0.000372, "optim/total_tokens": 325582848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 5.072112083435059, "created_at": "2025-01-15T07:17:30.410262+00:00"} {"global_step": 622, "acc_step": 0, "speed/wps": 12868.812523971757, "speed/FLOPS": 202122384290493.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3159002661705017, "optim/lr": 0.0003726, "optim/total_tokens": 326107136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392997, "loss/out": 5.056568145751953, "created_at": "2025-01-15T07:17:40.600963+00:00"} {"global_step": 623, "acc_step": 0, "speed/wps": 12868.395094783153, "speed/FLOPS": 202115827991478.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3476859927177429, "optim/lr": 0.0003732, "optim/total_tokens": 326631424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 5.134458541870117, "created_at": "2025-01-15T07:17:50.790410+00:00"} {"global_step": 624, "acc_step": 0, "speed/wps": 12868.330021730988, "speed/FLOPS": 202114805929774.56, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37817734479904175, "optim/lr": 0.00037380000000000003, "optim/total_tokens": 327155712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293783, "loss/out": 5.059591293334961, "created_at": "2025-01-15T07:18:00.980872+00:00"} {"global_step": 625, "acc_step": 0, "speed/wps": 12865.516772566058, "speed/FLOPS": 202070619985829.84, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44134268164634705, "optim/lr": 0.0003744, "optim/total_tokens": 327680000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 5.036123275756836, "created_at": "2025-01-15T07:18:11.173541+00:00"} {"global_step": 626, "acc_step": 0, "speed/wps": 12867.832166711656, "speed/FLOPS": 202106986432571.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47278669476509094, "optim/lr": 0.000375, "optim/total_tokens": 328204288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.056946277618408, "created_at": "2025-01-15T07:18:21.364692+00:00"} {"global_step": 627, "acc_step": 0, "speed/wps": 12866.53222081669, "speed/FLOPS": 202086568995978.16, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4569271206855774, "optim/lr": 0.0003756, "optim/total_tokens": 328728576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469813, "loss/out": 5.104144096374512, "created_at": "2025-01-15T07:18:31.556403+00:00"} {"global_step": 628, "acc_step": 0, "speed/wps": 12866.821499773023, "speed/FLOPS": 202091112519498.3, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4170956015586853, "optim/lr": 0.00037620000000000004, "optim/total_tokens": 329252864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.0722808837890625, "created_at": "2025-01-15T07:18:41.752563+00:00"} {"global_step": 629, "acc_step": 0, "speed/wps": 12864.196398183409, "speed/FLOPS": 202049881691765.8, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6359784007072449, "optim/lr": 0.0003768, "optim/total_tokens": 329777152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.1126627922058105, "created_at": "2025-01-15T07:18:51.946048+00:00"} {"global_step": 630, "acc_step": 0, "speed/wps": 12860.544137891939, "speed/FLOPS": 201992517925156.72, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6366978883743286, "optim/lr": 0.0003774, "optim/total_tokens": 330301440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 5.161751747131348, "created_at": "2025-01-15T07:19:02.143757+00:00"} {"global_step": 631, "acc_step": 0, "speed/wps": 12865.124304295956, "speed/FLOPS": 202064455732340.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5636841058731079, "optim/lr": 0.000378, "optim/total_tokens": 330825728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 5.176318168640137, "created_at": "2025-01-15T07:19:12.339428+00:00"} {"global_step": 632, "acc_step": 0, "speed/wps": 12867.574063588516, "speed/FLOPS": 202102932568351.7, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.611628532409668, "optim/lr": 0.00037860000000000004, "optim/total_tokens": 331350016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 5.116231918334961, "created_at": "2025-01-15T07:19:22.530582+00:00"} {"global_step": 633, "acc_step": 0, "speed/wps": 12867.006171288387, "speed/FLOPS": 202094013039412.34, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.697178065776825, "optim/lr": 0.00037920000000000006, "optim/total_tokens": 331874304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 5.195810794830322, "created_at": "2025-01-15T07:19:32.723635+00:00"} {"global_step": 634, "acc_step": 0, "speed/wps": 12863.005395449352, "speed/FLOPS": 202031175357218.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6957719922065735, "optim/lr": 0.00037979999999999996, "optim/total_tokens": 332398592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 5.276994228363037, "created_at": "2025-01-15T07:19:42.917036+00:00"} {"global_step": 635, "acc_step": 0, "speed/wps": 12859.944553417856, "speed/FLOPS": 201983100627075.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.6581951379776001, "optim/lr": 0.0003804, "optim/total_tokens": 332922880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472193, "loss/out": 5.044441223144531, "created_at": "2025-01-15T07:19:53.115680+00:00"} {"global_step": 636, "acc_step": 0, "speed/wps": 12865.524174950951, "speed/FLOPS": 202070736250456.56, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8323049545288086, "optim/lr": 0.000381, "optim/total_tokens": 333447168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.071212291717529, "created_at": "2025-01-15T07:20:03.307180+00:00"} {"global_step": 637, "acc_step": 0, "speed/wps": 12867.559244697335, "speed/FLOPS": 202102699817302.4, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.01610267162323, "optim/lr": 0.0003816, "optim/total_tokens": 333971456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415534, "loss/out": 5.1091132164001465, "created_at": "2025-01-15T07:20:13.496900+00:00"} {"global_step": 638, "acc_step": 0, "speed/wps": 12867.674354355295, "speed/FLOPS": 202104507772657.9, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6655789017677307, "optim/lr": 0.0003822000000000001, "optim/total_tokens": 334495744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 281309, "loss/out": 5.32569694519043, "created_at": "2025-01-15T07:20:23.690430+00:00"} {"global_step": 639, "acc_step": 0, "speed/wps": 12871.78450373572, "speed/FLOPS": 202169063316614.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44600149989128113, "optim/lr": 0.0003828, "optim/total_tokens": 335020032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.102102279663086, "created_at": "2025-01-15T07:20:33.877674+00:00"} {"global_step": 640, "acc_step": 0, "speed/wps": 12861.394185897754, "speed/FLOPS": 202005869097176.4, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.7617796063423157, "optim/lr": 0.0003834, "optim/total_tokens": 335544320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 5.176878929138184, "created_at": "2025-01-15T07:20:44.073996+00:00"} {"global_step": 641, "acc_step": 0, "speed/wps": 12863.457796065433, "speed/FLOPS": 202038280930557.34, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6787853240966797, "optim/lr": 0.000384, "optim/total_tokens": 336068608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.167112350463867, "created_at": "2025-01-15T07:20:54.267443+00:00"} {"global_step": 642, "acc_step": 0, "speed/wps": 12865.677913378911, "speed/FLOPS": 202073150923726.4, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5166961550712585, "optim/lr": 0.0003846, "optim/total_tokens": 336592896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.189645290374756, "created_at": "2025-01-15T07:21:04.463212+00:00"} {"global_step": 643, "acc_step": 0, "speed/wps": 12870.510602138524, "speed/FLOPS": 202149054941506.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7334785461425781, "optim/lr": 0.0003852, "optim/total_tokens": 337117184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 5.113225936889648, "created_at": "2025-01-15T07:21:14.654695+00:00"} {"global_step": 644, "acc_step": 0, "speed/wps": 12872.519675480731, "speed/FLOPS": 202180610199101.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4740726053714752, "optim/lr": 0.0003858, "optim/total_tokens": 337641472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 5.190892696380615, "created_at": "2025-01-15T07:21:24.843491+00:00"} {"global_step": 645, "acc_step": 0, "speed/wps": 12864.53304741075, "speed/FLOPS": 202055169230485.75, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.5290024280548096, "optim/lr": 0.0003864, "optim/total_tokens": 338165760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379737, "loss/out": 5.233546257019043, "created_at": "2025-01-15T07:21:35.035554+00:00"} {"global_step": 646, "acc_step": 0, "speed/wps": 12869.14772972093, "speed/FLOPS": 202127649157405.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5043261051177979, "optim/lr": 0.00038700000000000003, "optim/total_tokens": 338690048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 5.092872619628906, "created_at": "2025-01-15T07:21:45.224948+00:00"} {"global_step": 647, "acc_step": 0, "speed/wps": 12866.914032495994, "speed/FLOPS": 202092565873065.84, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.44716930389404297, "optim/lr": 0.00038760000000000004, "optim/total_tokens": 339214336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 5.210513114929199, "created_at": "2025-01-15T07:21:55.417222+00:00"} {"global_step": 648, "acc_step": 0, "speed/wps": 12870.31719569314, "speed/FLOPS": 202146017227513.3, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33628109097480774, "optim/lr": 0.00038819999999999995, "optim/total_tokens": 339738624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.032919883728027, "created_at": "2025-01-15T07:22:05.604874+00:00"} {"global_step": 649, "acc_step": 0, "speed/wps": 12863.613055918056, "speed/FLOPS": 202040719499892.2, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.32078978419303894, "optim/lr": 0.00038879999999999996, "optim/total_tokens": 340262912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.1396074295043945, "created_at": "2025-01-15T07:22:15.800301+00:00"} {"global_step": 650, "acc_step": 0, "speed/wps": 12867.67418846941, "speed/FLOPS": 202104505167192.12, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33055171370506287, "optim/lr": 0.0003894, "optim/total_tokens": 340787200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 5.09037971496582, "created_at": "2025-01-15T07:22:25.990754+00:00"} {"global_step": 651, "acc_step": 0, "speed/wps": 12865.129505876075, "speed/FLOPS": 202064537430303.47, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2877998650074005, "optim/lr": 0.00039000000000000005, "optim/total_tokens": 341311488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 5.123068809509277, "created_at": "2025-01-15T07:22:36.182411+00:00"} {"global_step": 652, "acc_step": 0, "speed/wps": 12857.249424827845, "speed/FLOPS": 201940769929074.94, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.3151896893978119, "optim/lr": 0.00039060000000000006, "optim/total_tokens": 341835776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 4.997811317443848, "created_at": "2025-01-15T07:22:46.383568+00:00"} {"global_step": 653, "acc_step": 0, "speed/wps": 12869.627532387898, "speed/FLOPS": 202135185117607.8, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30916735529899597, "optim/lr": 0.00039119999999999997, "optim/total_tokens": 342360064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.9576416015625, "created_at": "2025-01-15T07:22:56.571709+00:00"} {"global_step": 654, "acc_step": 0, "speed/wps": 12869.149059465619, "speed/FLOPS": 202127670042872.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3840949237346649, "optim/lr": 0.0003918, "optim/total_tokens": 342884352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 5.055215835571289, "created_at": "2025-01-15T07:23:06.763673+00:00"} {"global_step": 655, "acc_step": 0, "speed/wps": 12863.357070164502, "speed/FLOPS": 202036698891871.34, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5188773274421692, "optim/lr": 0.0003924, "optim/total_tokens": 343408640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 5.037200927734375, "created_at": "2025-01-15T07:23:16.958722+00:00"} {"global_step": 656, "acc_step": 0, "speed/wps": 12867.781059866402, "speed/FLOPS": 202106183729338.56, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4890947937965393, "optim/lr": 0.000393, "optim/total_tokens": 343932928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.108858108520508, "created_at": "2025-01-15T07:23:27.152798+00:00"} {"global_step": 657, "acc_step": 0, "speed/wps": 12868.697390419615, "speed/FLOPS": 202120575959846.9, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5122569799423218, "optim/lr": 0.0003936, "optim/total_tokens": 344457216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 4.994830131530762, "created_at": "2025-01-15T07:23:37.345997+00:00"} {"global_step": 658, "acc_step": 0, "speed/wps": 12862.488092609487, "speed/FLOPS": 202023050405269.53, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4808679521083832, "optim/lr": 0.0003942, "optim/total_tokens": 344981504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 5.056944370269775, "created_at": "2025-01-15T07:23:47.543715+00:00"} {"global_step": 659, "acc_step": 0, "speed/wps": 12867.422632740208, "speed/FLOPS": 202100554138791.22, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4816454350948334, "optim/lr": 0.0003948, "optim/total_tokens": 345505792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 5.052321910858154, "created_at": "2025-01-15T07:23:57.743393+00:00"} {"global_step": 660, "acc_step": 0, "speed/wps": 12864.77516122823, "speed/FLOPS": 202058971960688.5, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.45983240008354187, "optim/lr": 0.0003954, "optim/total_tokens": 346030080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 4.959825038909912, "created_at": "2025-01-15T07:24:07.937480+00:00"} {"global_step": 661, "acc_step": 0, "speed/wps": 12867.58054127587, "speed/FLOPS": 202103034309332.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5834612846374512, "optim/lr": 0.00039600000000000003, "optim/total_tokens": 346554368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 5.020351886749268, "created_at": "2025-01-15T07:24:18.127958+00:00"} {"global_step": 662, "acc_step": 0, "speed/wps": 12869.730088869015, "speed/FLOPS": 202136795908071.12, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42467328906059265, "optim/lr": 0.00039660000000000004, "optim/total_tokens": 347078656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 5.065800666809082, "created_at": "2025-01-15T07:24:28.320730+00:00"} {"global_step": 663, "acc_step": 0, "speed/wps": 12867.438338361855, "speed/FLOPS": 202100800817160.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3394508957862854, "optim/lr": 0.00039719999999999995, "optim/total_tokens": 347602944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 4.995198726654053, "created_at": "2025-01-15T07:24:38.510664+00:00"} {"global_step": 664, "acc_step": 0, "speed/wps": 12865.575779007453, "speed/FLOPS": 202071546763075.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3117404878139496, "optim/lr": 0.0003978, "optim/total_tokens": 348127232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 5.12399435043335, "created_at": "2025-01-15T07:24:48.703480+00:00"} {"global_step": 665, "acc_step": 0, "speed/wps": 12869.373694944516, "speed/FLOPS": 202131198251750.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38433533906936646, "optim/lr": 0.00039840000000000003, "optim/total_tokens": 348651520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 5.0448384284973145, "created_at": "2025-01-15T07:24:58.894900+00:00"} {"global_step": 666, "acc_step": 0, "speed/wps": 12867.932101479735, "speed/FLOPS": 202108556045428.9, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4788452088832855, "optim/lr": 0.00039900000000000005, "optim/total_tokens": 349175808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 5.155755043029785, "created_at": "2025-01-15T07:25:09.085473+00:00"} {"global_step": 667, "acc_step": 0, "speed/wps": 12867.239137450473, "speed/FLOPS": 202097672093116.97, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3718571066856384, "optim/lr": 0.00039960000000000006, "optim/total_tokens": 349700096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 5.047600269317627, "created_at": "2025-01-15T07:25:19.276395+00:00"} {"global_step": 668, "acc_step": 0, "speed/wps": 12870.842690004141, "speed/FLOPS": 202154270837772.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35984349250793457, "optim/lr": 0.00040019999999999997, "optim/total_tokens": 350224384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 4.817962646484375, "created_at": "2025-01-15T07:25:29.464388+00:00"} {"global_step": 669, "acc_step": 0, "speed/wps": 12863.691535566017, "speed/FLOPS": 202041952130605.75, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.4864009916782379, "optim/lr": 0.0004008, "optim/total_tokens": 350748672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.973019123077393, "created_at": "2025-01-15T07:25:39.658665+00:00"} {"global_step": 670, "acc_step": 0, "speed/wps": 12870.4797667905, "speed/FLOPS": 202148570629993.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3656120002269745, "optim/lr": 0.0004014, "optim/total_tokens": 351272960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 4.876772880554199, "created_at": "2025-01-15T07:25:49.846732+00:00"} {"global_step": 671, "acc_step": 0, "speed/wps": 12869.61607220515, "speed/FLOPS": 202135005119689.8, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3670708239078522, "optim/lr": 0.000402, "optim/total_tokens": 351797248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 4.840048313140869, "created_at": "2025-01-15T07:26:00.037186+00:00"} {"global_step": 672, "acc_step": 0, "speed/wps": 12866.357668103434, "speed/FLOPS": 202083827405755.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.38858917355537415, "optim/lr": 0.00040260000000000003, "optim/total_tokens": 352321536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.960776329040527, "created_at": "2025-01-15T07:26:10.228832+00:00"} {"global_step": 673, "acc_step": 0, "speed/wps": 12861.70733196367, "speed/FLOPS": 202010787486449.4, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38552361726760864, "optim/lr": 0.0004032, "optim/total_tokens": 352845824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478894, "loss/out": 4.838149547576904, "created_at": "2025-01-15T07:26:20.424845+00:00"} {"global_step": 674, "acc_step": 0, "speed/wps": 12871.341298316345, "speed/FLOPS": 202162102166475.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3624381124973297, "optim/lr": 0.0004038, "optim/total_tokens": 353370112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 4.816775321960449, "created_at": "2025-01-15T07:26:30.611662+00:00"} {"global_step": 675, "acc_step": 0, "speed/wps": 12868.85924163518, "speed/FLOPS": 202123118055594.25, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4183555841445923, "optim/lr": 0.0004044, "optim/total_tokens": 353894400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.986179351806641, "created_at": "2025-01-15T07:26:40.806297+00:00"} {"global_step": 676, "acc_step": 0, "speed/wps": 12869.006350009044, "speed/FLOPS": 202125428594753.84, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4191886782646179, "optim/lr": 0.00040500000000000003, "optim/total_tokens": 354418688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 4.884645462036133, "created_at": "2025-01-15T07:26:50.997331+00:00"} {"global_step": 677, "acc_step": 0, "speed/wps": 12864.862233835855, "speed/FLOPS": 202060339555641.56, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42788904905319214, "optim/lr": 0.0004056, "optim/total_tokens": 354942976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 4.953722953796387, "created_at": "2025-01-15T07:27:01.189751+00:00"} {"global_step": 678, "acc_step": 0, "speed/wps": 12867.6435561258, "speed/FLOPS": 202104024044142.47, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49847033619880676, "optim/lr": 0.0004062, "optim/total_tokens": 355467264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.932706832885742, "created_at": "2025-01-15T07:27:11.382202+00:00"} {"global_step": 679, "acc_step": 0, "speed/wps": 12865.811012752576, "speed/FLOPS": 202075241432287.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5182651877403259, "optim/lr": 0.0004068, "optim/total_tokens": 355991552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 4.970700263977051, "created_at": "2025-01-15T07:27:21.577747+00:00"} {"global_step": 680, "acc_step": 0, "speed/wps": 12867.310781588074, "speed/FLOPS": 202098797362747.34, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4229259490966797, "optim/lr": 0.00040740000000000004, "optim/total_tokens": 356515840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 5.08233642578125, "created_at": "2025-01-15T07:27:31.772337+00:00"} {"global_step": 681, "acc_step": 0, "speed/wps": 12866.816485038815, "speed/FLOPS": 202091033756206.6, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4694736897945404, "optim/lr": 0.00040800000000000005, "optim/total_tokens": 357040128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.892672538757324, "created_at": "2025-01-15T07:27:41.966432+00:00"} {"global_step": 682, "acc_step": 0, "speed/wps": 12864.651740122445, "speed/FLOPS": 202057033462623.78, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4387277066707611, "optim/lr": 0.00040859999999999996, "optim/total_tokens": 357564416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 4.926109313964844, "created_at": "2025-01-15T07:27:52.159866+00:00"} {"global_step": 683, "acc_step": 0, "speed/wps": 12863.485493848424, "speed/FLOPS": 202038715962299.97, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.46165168285369873, "optim/lr": 0.00040919999999999997, "optim/total_tokens": 358088704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.047647476196289, "created_at": "2025-01-15T07:28:02.355720+00:00"} {"global_step": 684, "acc_step": 0, "speed/wps": 12869.595966918454, "speed/FLOPS": 202134689338534.7, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.408572256565094, "optim/lr": 0.0004098, "optim/total_tokens": 358612992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342633, "loss/out": 4.889023780822754, "created_at": "2025-01-15T07:28:12.544801+00:00"} {"global_step": 685, "acc_step": 0, "speed/wps": 12869.430173662711, "speed/FLOPS": 202132085327629.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3464217782020569, "optim/lr": 0.0004104, "optim/total_tokens": 359137280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 4.868029594421387, "created_at": "2025-01-15T07:28:22.734575+00:00"} {"global_step": 686, "acc_step": 0, "speed/wps": 12869.634099230096, "speed/FLOPS": 202135288258888.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39760223031044006, "optim/lr": 0.000411, "optim/total_tokens": 359661568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 4.757624626159668, "created_at": "2025-01-15T07:28:32.922696+00:00"} {"global_step": 687, "acc_step": 0, "speed/wps": 12868.670439734726, "speed/FLOPS": 202120152662306.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43705078959465027, "optim/lr": 0.0004116, "optim/total_tokens": 360185856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 4.77957820892334, "created_at": "2025-01-15T07:28:43.115502+00:00"} {"global_step": 688, "acc_step": 0, "speed/wps": 12867.356595432399, "speed/FLOPS": 202099516932127.28, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41798242926597595, "optim/lr": 0.0004122, "optim/total_tokens": 360710144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 4.904242038726807, "created_at": "2025-01-15T07:28:53.305368+00:00"} {"global_step": 689, "acc_step": 0, "speed/wps": 12860.240548261749, "speed/FLOPS": 201987749632838.72, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5736222863197327, "optim/lr": 0.0004128, "optim/total_tokens": 361234432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.900503635406494, "created_at": "2025-01-15T07:29:03.502967+00:00"} {"global_step": 690, "acc_step": 0, "speed/wps": 12870.084334616238, "speed/FLOPS": 202142359824315.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4888719320297241, "optim/lr": 0.0004134, "optim/total_tokens": 361758720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 4.866453170776367, "created_at": "2025-01-15T07:29:13.695286+00:00"} {"global_step": 691, "acc_step": 0, "speed/wps": 12869.409042775927, "speed/FLOPS": 202131753438015.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5252487659454346, "optim/lr": 0.00041400000000000003, "optim/total_tokens": 362283008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.857228755950928, "created_at": "2025-01-15T07:29:23.883783+00:00"} {"global_step": 692, "acc_step": 0, "speed/wps": 12867.022253510224, "speed/FLOPS": 202094265632805.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.44017964601516724, "optim/lr": 0.0004146, "optim/total_tokens": 362807296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351470, "loss/out": 4.931666851043701, "created_at": "2025-01-15T07:29:34.075140+00:00"} {"global_step": 693, "acc_step": 0, "speed/wps": 12872.20278154127, "speed/FLOPS": 202175632944324.7, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38603445887565613, "optim/lr": 0.0004152, "optim/total_tokens": 363331584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293250, "loss/out": 4.841200828552246, "created_at": "2025-01-15T07:29:44.263376+00:00"} {"global_step": 694, "acc_step": 0, "speed/wps": 12867.707812802173, "speed/FLOPS": 202105033283542.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3292747735977173, "optim/lr": 0.0004158, "optim/total_tokens": 363855872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 4.87709903717041, "created_at": "2025-01-15T07:29:54.453965+00:00"} {"global_step": 695, "acc_step": 0, "speed/wps": 12869.665116469947, "speed/FLOPS": 202135775427262.3, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3338552415370941, "optim/lr": 0.00041640000000000004, "optim/total_tokens": 364380160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.935416221618652, "created_at": "2025-01-15T07:30:04.642966+00:00"} {"global_step": 696, "acc_step": 0, "speed/wps": 12869.440466110198, "speed/FLOPS": 202132246984660.53, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3659060001373291, "optim/lr": 0.00041700000000000005, "optim/total_tokens": 364904448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.813509464263916, "created_at": "2025-01-15T07:30:14.831203+00:00"} {"global_step": 697, "acc_step": 0, "speed/wps": 12865.854891714795, "speed/FLOPS": 202075930611685.16, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3887545168399811, "optim/lr": 0.00041759999999999996, "optim/total_tokens": 365428736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 4.774846076965332, "created_at": "2025-01-15T07:30:25.022341+00:00"} {"global_step": 698, "acc_step": 0, "speed/wps": 12871.57807201325, "speed/FLOPS": 202165821022746.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30842381715774536, "optim/lr": 0.0004182, "optim/total_tokens": 365953024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 4.707123756408691, "created_at": "2025-01-15T07:30:35.208896+00:00"} {"global_step": 699, "acc_step": 0, "speed/wps": 12869.049948345419, "speed/FLOPS": 202126113366537.03, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.44753846526145935, "optim/lr": 0.0004188, "optim/total_tokens": 366477312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 4.886107444763184, "created_at": "2025-01-15T07:30:45.402124+00:00"} {"global_step": 700, "acc_step": 0, "speed/wps": 12868.784696164032, "speed/FLOPS": 202121947216533.66, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49023598432540894, "optim/lr": 0.00041940000000000006, "optim/total_tokens": 367001600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 4.738515377044678, "created_at": "2025-01-15T07:30:55.590835+00:00"} {"global_step": 701, "acc_step": 0, "speed/wps": 12871.927874047586, "speed/FLOPS": 202171315144373.88, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5609292984008789, "optim/lr": 0.00042000000000000007, "optim/total_tokens": 367525888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487669, "loss/out": 4.832773685455322, "created_at": "2025-01-15T07:31:05.778060+00:00"} {"global_step": 702, "acc_step": 0, "speed/wps": 12871.612189992247, "speed/FLOPS": 202166356892489.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6131625175476074, "optim/lr": 0.0004206, "optim/total_tokens": 368050176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364255, "loss/out": 4.8597869873046875, "created_at": "2025-01-15T07:31:15.964773+00:00"} {"global_step": 703, "acc_step": 0, "speed/wps": 12867.313641488025, "speed/FLOPS": 202098842281406.0, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8134759068489075, "optim/lr": 0.0004212, "optim/total_tokens": 368574464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 4.879924774169922, "created_at": "2025-01-15T07:31:26.155706+00:00"} {"global_step": 704, "acc_step": 0, "speed/wps": 12871.207369969765, "speed/FLOPS": 202159998637755.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7834075689315796, "optim/lr": 0.0004218, "optim/total_tokens": 369098752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.8967390060424805, "created_at": "2025-01-15T07:31:36.343159+00:00"} {"global_step": 705, "acc_step": 0, "speed/wps": 12867.0624951204, "speed/FLOPS": 202094897682591.2, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6755993366241455, "optim/lr": 0.0004224, "optim/total_tokens": 369623040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.873385429382324, "created_at": "2025-01-15T07:31:46.533785+00:00"} {"global_step": 706, "acc_step": 0, "speed/wps": 12869.894980107223, "speed/FLOPS": 202139385751551.47, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9511037468910217, "optim/lr": 0.000423, "optim/total_tokens": 370147328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.089483261108398, "created_at": "2025-01-15T07:31:56.730110+00:00"} {"global_step": 707, "acc_step": 0, "speed/wps": 12871.972657690498, "speed/FLOPS": 202172018533025.0, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6751437187194824, "optim/lr": 0.0004236, "optim/total_tokens": 370671616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 5.031746864318848, "created_at": "2025-01-15T07:32:06.918538+00:00"} {"global_step": 708, "acc_step": 0, "speed/wps": 12872.39245945295, "speed/FLOPS": 202178612096572.4, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7524236440658569, "optim/lr": 0.0004242, "optim/total_tokens": 371195904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 5.024625301361084, "created_at": "2025-01-15T07:32:17.108129+00:00"} {"global_step": 709, "acc_step": 0, "speed/wps": 12870.183081513516, "speed/FLOPS": 202143910780030.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7373316287994385, "optim/lr": 0.0004248, "optim/total_tokens": 371720192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.035239219665527, "created_at": "2025-01-15T07:32:27.299728+00:00"} {"global_step": 710, "acc_step": 0, "speed/wps": 12867.095689226897, "speed/FLOPS": 202095419041647.2, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5481840372085571, "optim/lr": 0.00042540000000000004, "optim/total_tokens": 372244480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 4.932871341705322, "created_at": "2025-01-15T07:32:37.492085+00:00"} {"global_step": 711, "acc_step": 0, "speed/wps": 12869.702718424513, "speed/FLOPS": 202136366017629.28, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4797435700893402, "optim/lr": 0.00042599999999999995, "optim/total_tokens": 372768768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292143, "loss/out": 4.9404497146606445, "created_at": "2025-01-15T07:32:47.683130+00:00"} {"global_step": 712, "acc_step": 0, "speed/wps": 12863.58584434544, "speed/FLOPS": 202040292104751.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5687314867973328, "optim/lr": 0.00042659999999999996, "optim/total_tokens": 373293056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294596, "loss/out": 4.855950832366943, "created_at": "2025-01-15T07:32:57.883701+00:00"} {"global_step": 713, "acc_step": 0, "speed/wps": 12869.695691357938, "speed/FLOPS": 202136255647892.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.515239417552948, "optim/lr": 0.00042720000000000003, "optim/total_tokens": 373817344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 4.925111770629883, "created_at": "2025-01-15T07:33:08.073383+00:00"} {"global_step": 714, "acc_step": 0, "speed/wps": 12868.720578171346, "speed/FLOPS": 202120940155351.16, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5032181739807129, "optim/lr": 0.00042780000000000004, "optim/total_tokens": 374341632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.925655364990234, "created_at": "2025-01-15T07:33:18.262189+00:00"} {"global_step": 715, "acc_step": 0, "speed/wps": 12866.889223182621, "speed/FLOPS": 202092176208707.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4680008590221405, "optim/lr": 0.00042840000000000006, "optim/total_tokens": 374865920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.839699745178223, "created_at": "2025-01-15T07:33:28.454866+00:00"} {"global_step": 716, "acc_step": 0, "speed/wps": 12866.901954853867, "speed/FLOPS": 202092376177099.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6219116449356079, "optim/lr": 0.00042899999999999997, "optim/total_tokens": 375390208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465291, "loss/out": 4.888193130493164, "created_at": "2025-01-15T07:33:38.645238+00:00"} {"global_step": 717, "acc_step": 0, "speed/wps": 12871.616962427026, "speed/FLOPS": 202166431850135.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.39460352063179016, "optim/lr": 0.0004296, "optim/total_tokens": 375914496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 4.952549934387207, "created_at": "2025-01-15T07:33:48.831729+00:00"} {"global_step": 718, "acc_step": 0, "speed/wps": 12868.139004561464, "speed/FLOPS": 202111805742642.1, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3428117632865906, "optim/lr": 0.0004302, "optim/total_tokens": 376438784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 5.001911163330078, "created_at": "2025-01-15T07:33:59.020970+00:00"} {"global_step": 719, "acc_step": 0, "speed/wps": 12869.401791259043, "speed/FLOPS": 202131639542978.3, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38218745589256287, "optim/lr": 0.0004308, "optim/total_tokens": 376963072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.839639663696289, "created_at": "2025-01-15T07:34:09.214245+00:00"} {"global_step": 720, "acc_step": 0, "speed/wps": 12865.840419045748, "speed/FLOPS": 202075703298530.25, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4293595850467682, "optim/lr": 0.0004314, "optim/total_tokens": 377487360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.999822616577148, "created_at": "2025-01-15T07:34:19.407103+00:00"} {"global_step": 721, "acc_step": 0, "speed/wps": 12869.229734295604, "speed/FLOPS": 202128937151936.47, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3777157962322235, "optim/lr": 0.000432, "optim/total_tokens": 378011648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 4.830814361572266, "created_at": "2025-01-15T07:34:29.599408+00:00"} {"global_step": 722, "acc_step": 0, "speed/wps": 12869.054481177263, "speed/FLOPS": 202126184560889.9, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3085128366947174, "optim/lr": 0.0004326, "optim/total_tokens": 378535936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 4.7710347175598145, "created_at": "2025-01-15T07:34:39.791540+00:00"} {"global_step": 723, "acc_step": 0, "speed/wps": 12867.75281885835, "speed/FLOPS": 202105740165499.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3419358730316162, "optim/lr": 0.0004332, "optim/total_tokens": 379060224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463795, "loss/out": 4.8750810623168945, "created_at": "2025-01-15T07:34:49.984149+00:00"} {"global_step": 724, "acc_step": 0, "speed/wps": 12870.320368829236, "speed/FLOPS": 202146067065976.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3175961375236511, "optim/lr": 0.00043380000000000003, "optim/total_tokens": 379584512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 4.7319254875183105, "created_at": "2025-01-15T07:35:00.171644+00:00"} {"global_step": 725, "acc_step": 0, "speed/wps": 12867.003940884671, "speed/FLOPS": 202093978007857.1, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3108435571193695, "optim/lr": 0.00043440000000000004, "optim/total_tokens": 380108800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 4.793890953063965, "created_at": "2025-01-15T07:35:10.364972+00:00"} {"global_step": 726, "acc_step": 0, "speed/wps": 12845.35941066859, "speed/FLOPS": 201754020918112.22, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2922436594963074, "optim/lr": 0.000435, "optim/total_tokens": 380633088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 4.673841953277588, "created_at": "2025-01-15T07:35:20.575769+00:00"} {"global_step": 727, "acc_step": 0, "speed/wps": 12856.872137586117, "speed/FLOPS": 201934844114496.6, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2625467777252197, "optim/lr": 0.0004356, "optim/total_tokens": 381157376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283675, "loss/out": 4.665916442871094, "created_at": "2025-01-15T07:35:30.775357+00:00"} {"global_step": 728, "acc_step": 0, "speed/wps": 12867.757615491413, "speed/FLOPS": 202105815503213.44, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2700667977333069, "optim/lr": 0.00043620000000000003, "optim/total_tokens": 381681664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312171, "loss/out": 4.810492515563965, "created_at": "2025-01-15T07:35:40.964943+00:00"} {"global_step": 729, "acc_step": 0, "speed/wps": 12859.617267165662, "speed/FLOPS": 201977960146746.7, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2714661657810211, "optim/lr": 0.00043680000000000005, "optim/total_tokens": 382205952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306261, "loss/out": 4.74385929107666, "created_at": "2025-01-15T07:35:51.160993+00:00"} {"global_step": 730, "acc_step": 0, "speed/wps": 12864.9585215684, "speed/FLOPS": 202061851886794.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3252193331718445, "optim/lr": 0.00043740000000000006, "optim/total_tokens": 382730240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486618, "loss/out": 4.676113128662109, "created_at": "2025-01-15T07:36:01.355707+00:00"} {"global_step": 731, "acc_step": 0, "speed/wps": 12866.374194505943, "speed/FLOPS": 202084086975616.97, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.47168195247650146, "optim/lr": 0.00043799999999999997, "optim/total_tokens": 383254528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 4.875485420227051, "created_at": "2025-01-15T07:36:11.548371+00:00"} {"global_step": 732, "acc_step": 0, "speed/wps": 12867.227052139531, "speed/FLOPS": 202097482276701.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49364781379699707, "optim/lr": 0.0004386, "optim/total_tokens": 383778816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 4.742318153381348, "created_at": "2025-01-15T07:36:21.751537+00:00"} {"global_step": 733, "acc_step": 0, "speed/wps": 12868.672454207444, "speed/FLOPS": 202120184302368.84, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7009732127189636, "optim/lr": 0.0004392, "optim/total_tokens": 384303104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.8917436599731445, "created_at": "2025-01-15T07:36:31.943894+00:00"} {"global_step": 734, "acc_step": 0, "speed/wps": 12866.845527732703, "speed/FLOPS": 202091489911622.97, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6958569884300232, "optim/lr": 0.0004398, "optim/total_tokens": 384827392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.725070476531982, "created_at": "2025-01-15T07:36:42.135144+00:00"} {"global_step": 735, "acc_step": 0, "speed/wps": 12867.22655804573, "speed/FLOPS": 202097474516279.56, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7081826329231262, "optim/lr": 0.0004404, "optim/total_tokens": 385351680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.952656269073486, "created_at": "2025-01-15T07:36:52.325393+00:00"} {"global_step": 736, "acc_step": 0, "speed/wps": 12867.917532366917, "speed/FLOPS": 202108327217492.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4930509328842163, "optim/lr": 0.000441, "optim/total_tokens": 385875968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 4.6684250831604, "created_at": "2025-01-15T07:37:02.549113+00:00"} {"global_step": 737, "acc_step": 0, "speed/wps": 12870.544913375266, "speed/FLOPS": 202149593846628.0, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5175954103469849, "optim/lr": 0.0004416, "optim/total_tokens": 386400256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.868345260620117, "created_at": "2025-01-15T07:37:12.741947+00:00"} {"global_step": 738, "acc_step": 0, "speed/wps": 12866.443398836465, "speed/FLOPS": 202085173924724.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45432448387145996, "optim/lr": 0.0004422, "optim/total_tokens": 386924544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.803179740905762, "created_at": "2025-01-15T07:37:22.935091+00:00"} {"global_step": 739, "acc_step": 0, "speed/wps": 12867.695948969249, "speed/FLOPS": 202104846945744.16, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4030098021030426, "optim/lr": 0.00044280000000000003, "optim/total_tokens": 387448832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.739442348480225, "created_at": "2025-01-15T07:37:33.128525+00:00"} {"global_step": 740, "acc_step": 0, "speed/wps": 12867.466597989314, "speed/FLOPS": 202101244673443.0, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3727189600467682, "optim/lr": 0.0004434, "optim/total_tokens": 387973120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.805729866027832, "created_at": "2025-01-15T07:37:43.318383+00:00"} {"global_step": 741, "acc_step": 0, "speed/wps": 12867.010723855334, "speed/FLOPS": 202094084543732.12, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34624814987182617, "optim/lr": 0.000444, "optim/total_tokens": 388497408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 4.708446502685547, "created_at": "2025-01-15T07:37:53.511080+00:00"} {"global_step": 742, "acc_step": 0, "speed/wps": 12870.357492048455, "speed/FLOPS": 202146650137146.53, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30691248178482056, "optim/lr": 0.0004446, "optim/total_tokens": 389021696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 4.842005729675293, "created_at": "2025-01-15T07:38:03.698639+00:00"} {"global_step": 743, "acc_step": 0, "speed/wps": 12862.937109819655, "speed/FLOPS": 202030102837569.53, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.3413686454296112, "optim/lr": 0.00044520000000000003, "optim/total_tokens": 389545984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 4.742314338684082, "created_at": "2025-01-15T07:38:13.896809+00:00"} {"global_step": 744, "acc_step": 0, "speed/wps": 12869.38169840993, "speed/FLOPS": 202131323957172.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4153463542461395, "optim/lr": 0.00044580000000000005, "optim/total_tokens": 390070272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.695611953735352, "created_at": "2025-01-15T07:38:24.088047+00:00"} {"global_step": 745, "acc_step": 0, "speed/wps": 12870.692489913232, "speed/FLOPS": 202151911738948.16, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5191706418991089, "optim/lr": 0.00044639999999999995, "optim/total_tokens": 390594560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 4.657615661621094, "created_at": "2025-01-15T07:38:34.278555+00:00"} {"global_step": 746, "acc_step": 0, "speed/wps": 12866.550642670816, "speed/FLOPS": 202086858336511.5, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37549343705177307, "optim/lr": 0.00044699999999999997, "optim/total_tokens": 391118848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.73444938659668, "created_at": "2025-01-15T07:38:44.473690+00:00"} {"global_step": 747, "acc_step": 0, "speed/wps": 12869.807609868, "speed/FLOPS": 202138013481885.2, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37080228328704834, "optim/lr": 0.0004476, "optim/total_tokens": 391643136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507176, "loss/out": 4.761996269226074, "created_at": "2025-01-15T07:38:54.663521+00:00"} {"global_step": 748, "acc_step": 0, "speed/wps": 12870.395553545426, "speed/FLOPS": 202147247945258.44, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3272250294685364, "optim/lr": 0.0004482, "optim/total_tokens": 392167424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 4.649105548858643, "created_at": "2025-01-15T07:39:04.852384+00:00"} {"global_step": 749, "acc_step": 0, "speed/wps": 12869.96284297396, "speed/FLOPS": 202140451631126.94, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.360489159822464, "optim/lr": 0.00044880000000000007, "optim/total_tokens": 392691712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 4.69715690612793, "created_at": "2025-01-15T07:39:15.040824+00:00"} {"global_step": 750, "acc_step": 0, "speed/wps": 12867.040539131167, "speed/FLOPS": 202094552833609.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42260926961898804, "optim/lr": 0.0004494, "optim/total_tokens": 393216000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 4.756678581237793, "created_at": "2025-01-15T07:39:25.231002+00:00"} {"global_step": 751, "acc_step": 0, "speed/wps": 12865.160938891977, "speed/FLOPS": 202065031129011.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4264943599700928, "optim/lr": 0.00045, "optim/total_tokens": 393740288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 4.64656925201416, "created_at": "2025-01-15T07:39:35.422559+00:00"} {"global_step": 752, "acc_step": 0, "speed/wps": 12869.030730915452, "speed/FLOPS": 202125811530391.88, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37384459376335144, "optim/lr": 0.0004506, "optim/total_tokens": 394264576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 4.675127983093262, "created_at": "2025-01-15T07:39:45.612616+00:00"} {"global_step": 753, "acc_step": 0, "speed/wps": 12868.924396526902, "speed/FLOPS": 202124141402700.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41895052790641785, "optim/lr": 0.0004512, "optim/total_tokens": 394788864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 4.630973815917969, "created_at": "2025-01-15T07:39:55.801074+00:00"} {"global_step": 754, "acc_step": 0, "speed/wps": 12863.026915612883, "speed/FLOPS": 202031513360958.25, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44381609559059143, "optim/lr": 0.00045180000000000003, "optim/total_tokens": 395313152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.691001892089844, "created_at": "2025-01-15T07:40:06.000790+00:00"} {"global_step": 755, "acc_step": 0, "speed/wps": 12866.350912875118, "speed/FLOPS": 202083721305612.53, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3699035346508026, "optim/lr": 0.0004524, "optim/total_tokens": 395837440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 4.688871383666992, "created_at": "2025-01-15T07:40:16.194050+00:00"} {"global_step": 756, "acc_step": 0, "speed/wps": 12866.235143502798, "speed/FLOPS": 202081902988536.47, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4212709367275238, "optim/lr": 0.000453, "optim/total_tokens": 396361728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 4.639625072479248, "created_at": "2025-01-15T07:40:26.400646+00:00"} {"global_step": 757, "acc_step": 0, "speed/wps": 12868.239098230033, "speed/FLOPS": 202113377851250.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.39206019043922424, "optim/lr": 0.0004536, "optim/total_tokens": 396886016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.7042059898376465, "created_at": "2025-01-15T07:40:36.595190+00:00"} {"global_step": 758, "acc_step": 0, "speed/wps": 12864.877606284172, "speed/FLOPS": 202060581001066.3, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30002328753471375, "optim/lr": 0.00045420000000000004, "optim/total_tokens": 397410304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 4.66917610168457, "created_at": "2025-01-15T07:40:46.788348+00:00"} {"global_step": 759, "acc_step": 0, "speed/wps": 12865.199358929844, "speed/FLOPS": 202065634568501.06, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4479452669620514, "optim/lr": 0.00045480000000000005, "optim/total_tokens": 397934592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.619910717010498, "created_at": "2025-01-15T07:40:56.979958+00:00"} {"global_step": 760, "acc_step": 0, "speed/wps": 12868.844440964485, "speed/FLOPS": 202122885590722.72, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3981997072696686, "optim/lr": 0.00045539999999999996, "optim/total_tokens": 398458880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.608669281005859, "created_at": "2025-01-15T07:41:07.172172+00:00"} {"global_step": 761, "acc_step": 0, "speed/wps": 12869.913539874638, "speed/FLOPS": 202139677258202.53, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3227808475494385, "optim/lr": 0.000456, "optim/total_tokens": 398983168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 4.606617450714111, "created_at": "2025-01-15T07:41:17.360337+00:00"} {"global_step": 762, "acc_step": 0, "speed/wps": 12870.20869209321, "speed/FLOPS": 202144313029376.88, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3048328459262848, "optim/lr": 0.00045660000000000004, "optim/total_tokens": 399507456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 4.605299472808838, "created_at": "2025-01-15T07:41:27.549696+00:00"} {"global_step": 763, "acc_step": 0, "speed/wps": 12870.199565958592, "speed/FLOPS": 202144169690892.06, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2493465542793274, "optim/lr": 0.00045720000000000005, "optim/total_tokens": 400031744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.6453351974487305, "created_at": "2025-01-15T07:41:37.740361+00:00"} {"global_step": 764, "acc_step": 0, "speed/wps": 12868.708373585625, "speed/FLOPS": 202120748465561.56, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24757426977157593, "optim/lr": 0.00045780000000000007, "optim/total_tokens": 400556032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350002, "loss/out": 4.580617427825928, "created_at": "2025-01-15T07:41:47.929614+00:00"} {"global_step": 765, "acc_step": 0, "speed/wps": 12865.364223470891, "speed/FLOPS": 202068223992665.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.268549382686615, "optim/lr": 0.0004584, "optim/total_tokens": 401080320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 4.480502128601074, "created_at": "2025-01-15T07:41:58.121379+00:00"} {"global_step": 766, "acc_step": 0, "speed/wps": 12869.267583184894, "speed/FLOPS": 202129531620752.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36160269379615784, "optim/lr": 0.000459, "optim/total_tokens": 401604608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 4.521074295043945, "created_at": "2025-01-15T07:42:08.311014+00:00"} {"global_step": 767, "acc_step": 0, "speed/wps": 12866.733013756275, "speed/FLOPS": 202089722725012.75, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45124903321266174, "optim/lr": 0.0004596, "optim/total_tokens": 402128896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 4.614980697631836, "created_at": "2025-01-15T07:42:18.501605+00:00"} {"global_step": 768, "acc_step": 0, "speed/wps": 12867.764333372035, "speed/FLOPS": 202105921016759.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42540737986564636, "optim/lr": 0.0004602, "optim/total_tokens": 402653184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.657537937164307, "created_at": "2025-01-15T07:42:28.691691+00:00"} {"global_step": 769, "acc_step": 0, "speed/wps": 12866.466578466088, "speed/FLOPS": 202085537992659.9, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43860435485839844, "optim/lr": 0.0004608, "optim/total_tokens": 403177472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 4.57326602935791, "created_at": "2025-01-15T07:42:38.882579+00:00"} {"global_step": 770, "acc_step": 0, "speed/wps": 12866.902433630134, "speed/FLOPS": 202092383696939.0, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4129566550254822, "optim/lr": 0.0004614, "optim/total_tokens": 403701760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.743912696838379, "created_at": "2025-01-15T07:42:49.076181+00:00"} {"global_step": 771, "acc_step": 0, "speed/wps": 12868.324116300242, "speed/FLOPS": 202114713176869.78, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38032665848731995, "optim/lr": 0.000462, "optim/total_tokens": 404226048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 4.591153144836426, "created_at": "2025-01-15T07:42:59.268075+00:00"} {"global_step": 772, "acc_step": 0, "speed/wps": 12867.164302530218, "speed/FLOPS": 202096496707860.4, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4419194459915161, "optim/lr": 0.0004626, "optim/total_tokens": 404750336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 4.638462066650391, "created_at": "2025-01-15T07:43:09.461568+00:00"} {"global_step": 773, "acc_step": 0, "speed/wps": 12869.316713053337, "speed/FLOPS": 202130303272848.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3894628584384918, "optim/lr": 0.00046320000000000004, "optim/total_tokens": 405274624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 4.5951995849609375, "created_at": "2025-01-15T07:43:19.651684+00:00"} {"global_step": 774, "acc_step": 0, "speed/wps": 12866.511081801138, "speed/FLOPS": 202086236978690.88, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29446858167648315, "optim/lr": 0.0004638, "optim/total_tokens": 405798912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.651383876800537, "created_at": "2025-01-15T07:43:29.845469+00:00"} {"global_step": 775, "acc_step": 0, "speed/wps": 12865.626069750488, "speed/FLOPS": 202072336648301.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33131149411201477, "optim/lr": 0.0004644, "optim/total_tokens": 406323200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 4.697446346282959, "created_at": "2025-01-15T07:43:40.040567+00:00"} {"global_step": 776, "acc_step": 0, "speed/wps": 12867.289860005716, "speed/FLOPS": 202098468760547.16, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3176417350769043, "optim/lr": 0.000465, "optim/total_tokens": 406847488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 4.660459518432617, "created_at": "2025-01-15T07:43:50.233614+00:00"} {"global_step": 777, "acc_step": 0, "speed/wps": 12868.52041388696, "speed/FLOPS": 202117796300210.3, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31038716435432434, "optim/lr": 0.00046560000000000004, "optim/total_tokens": 407371776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 4.493017673492432, "created_at": "2025-01-15T07:44:00.425270+00:00"} {"global_step": 778, "acc_step": 0, "speed/wps": 12865.213160986677, "speed/FLOPS": 202065851348769.66, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2669067084789276, "optim/lr": 0.00046620000000000006, "optim/total_tokens": 407896064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 4.557119369506836, "created_at": "2025-01-15T07:44:10.623518+00:00"} {"global_step": 779, "acc_step": 0, "speed/wps": 12860.917014852326, "speed/FLOPS": 201998374470206.25, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.28786587715148926, "optim/lr": 0.00046679999999999996, "optim/total_tokens": 408420352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361335, "loss/out": 4.557666778564453, "created_at": "2025-01-15T07:44:20.820466+00:00"} {"global_step": 780, "acc_step": 0, "speed/wps": 12866.459072660913, "speed/FLOPS": 202085420103675.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2516863942146301, "optim/lr": 0.0004674, "optim/total_tokens": 408944640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 4.741273880004883, "created_at": "2025-01-15T07:44:31.011267+00:00"} {"global_step": 781, "acc_step": 0, "speed/wps": 12866.46736186516, "speed/FLOPS": 202085550297018.84, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.305488646030426, "optim/lr": 0.000468, "optim/total_tokens": 409468928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 4.563762664794922, "created_at": "2025-01-15T07:44:41.205097+00:00"} {"global_step": 782, "acc_step": 0, "speed/wps": 12860.971151096608, "speed/FLOPS": 201999224754313.7, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3019106090068817, "optim/lr": 0.0004686, "optim/total_tokens": 409993216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 4.565082550048828, "created_at": "2025-01-15T07:44:51.400134+00:00"} {"global_step": 783, "acc_step": 0, "speed/wps": 12865.495094706197, "speed/FLOPS": 202070279505252.25, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.33005550503730774, "optim/lr": 0.0004692, "optim/total_tokens": 410517504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 4.486059665679932, "created_at": "2025-01-15T07:45:01.592533+00:00"} {"global_step": 784, "acc_step": 0, "speed/wps": 12866.297157131183, "speed/FLOPS": 202082876997785.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37735864520072937, "optim/lr": 0.0004698, "optim/total_tokens": 411041792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 4.617527961730957, "created_at": "2025-01-15T07:45:11.784492+00:00"} {"global_step": 785, "acc_step": 0, "speed/wps": 12858.602629271989, "speed/FLOPS": 201962023864367.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3520631492137909, "optim/lr": 0.0004704, "optim/total_tokens": 411566080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 4.5846052169799805, "created_at": "2025-01-15T07:45:21.983766+00:00"} {"global_step": 786, "acc_step": 0, "speed/wps": 12859.263568008913, "speed/FLOPS": 201972404815456.38, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3515315055847168, "optim/lr": 0.000471, "optim/total_tokens": 412090368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.551861763000488, "created_at": "2025-01-15T07:45:32.182054+00:00"} {"global_step": 787, "acc_step": 0, "speed/wps": 12865.704686871855, "speed/FLOPS": 202073571438223.72, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3438831865787506, "optim/lr": 0.0004716, "optim/total_tokens": 412614656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 4.509279251098633, "created_at": "2025-01-15T07:45:42.373587+00:00"} {"global_step": 788, "acc_step": 0, "speed/wps": 12862.100978841805, "speed/FLOPS": 202016970251598.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39074960350990295, "optim/lr": 0.00047220000000000004, "optim/total_tokens": 413138944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 4.579800605773926, "created_at": "2025-01-15T07:45:52.569500+00:00"} {"global_step": 789, "acc_step": 0, "speed/wps": 12865.390386449819, "speed/FLOPS": 202068634918201.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37245890498161316, "optim/lr": 0.0004728, "optim/total_tokens": 413663232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 4.663834571838379, "created_at": "2025-01-15T07:46:02.763741+00:00"} {"global_step": 790, "acc_step": 0, "speed/wps": 12866.210834351796, "speed/FLOPS": 202081521179915.72, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31296810507774353, "optim/lr": 0.0004734, "optim/total_tokens": 414187520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 4.548070907592773, "created_at": "2025-01-15T07:46:12.962877+00:00"} {"global_step": 791, "acc_step": 0, "speed/wps": 12866.377632708753, "speed/FLOPS": 202084140977316.7, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35030990839004517, "optim/lr": 0.00047400000000000003, "optim/total_tokens": 414711808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 4.5391435623168945, "created_at": "2025-01-15T07:46:23.153620+00:00"} {"global_step": 792, "acc_step": 0, "speed/wps": 12864.534926533266, "speed/FLOPS": 202055198744687.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3608267605304718, "optim/lr": 0.00047460000000000004, "optim/total_tokens": 415236096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 4.581987380981445, "created_at": "2025-01-15T07:46:33.348519+00:00"} {"global_step": 793, "acc_step": 0, "speed/wps": 12865.644503123407, "speed/FLOPS": 202072626169753.22, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34748202562332153, "optim/lr": 0.00047520000000000006, "optim/total_tokens": 415760384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.543238639831543, "created_at": "2025-01-15T07:46:43.547803+00:00"} {"global_step": 794, "acc_step": 0, "speed/wps": 12865.77224209777, "speed/FLOPS": 202074632485877.3, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3604676127433777, "optim/lr": 0.00047579999999999996, "optim/total_tokens": 416284672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.555464267730713, "created_at": "2025-01-15T07:46:53.741764+00:00"} {"global_step": 795, "acc_step": 0, "speed/wps": 12866.976210302195, "speed/FLOPS": 202093542460953.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2681330442428589, "optim/lr": 0.0004764, "optim/total_tokens": 416808960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505844, "loss/out": 4.597837448120117, "created_at": "2025-01-15T07:47:03.932985+00:00"} {"global_step": 796, "acc_step": 0, "speed/wps": 12864.935910673701, "speed/FLOPS": 202061496751622.5, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3362072706222534, "optim/lr": 0.000477, "optim/total_tokens": 417333248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 4.599104404449463, "created_at": "2025-01-15T07:47:14.124729+00:00"} {"global_step": 797, "acc_step": 0, "speed/wps": 12867.628177066621, "speed/FLOPS": 202103782494885.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3475836217403412, "optim/lr": 0.0004776, "optim/total_tokens": 417857536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 4.501576900482178, "created_at": "2025-01-15T07:47:24.314420+00:00"} {"global_step": 798, "acc_step": 0, "speed/wps": 12867.795632199744, "speed/FLOPS": 202106412607858.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.283828467130661, "optim/lr": 0.00047819999999999997, "optim/total_tokens": 418381824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 4.558442115783691, "created_at": "2025-01-15T07:47:34.505879+00:00"} {"global_step": 799, "acc_step": 0, "speed/wps": 12873.436338400656, "speed/FLOPS": 202195007649886.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33140048384666443, "optim/lr": 0.0004788, "optim/total_tokens": 418906112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.413414478302002, "created_at": "2025-01-15T07:47:44.690957+00:00"} {"global_step": 800, "acc_step": 0, "speed/wps": 12873.866736544684, "speed/FLOPS": 202201767644167.1, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36107978224754333, "optim/lr": 0.0004794, "optim/total_tokens": 419430400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 4.626323699951172, "created_at": "2025-01-15T07:47:54.878594+00:00"} {"global_step": 801, "acc_step": 0, "speed/wps": 12872.89419741128, "speed/FLOPS": 202186492580668.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3823297917842865, "optim/lr": 0.00048, "optim/total_tokens": 419954688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.436769485473633, "created_at": "2025-01-15T07:48:05.065927+00:00"} {"global_step": 802, "acc_step": 0, "speed/wps": 12873.79859344561, "speed/FLOPS": 202200697363157.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48729559779167175, "optim/lr": 0.0004806, "optim/total_tokens": 420478976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.568354606628418, "created_at": "2025-01-15T07:48:15.253384+00:00"} {"global_step": 803, "acc_step": 0, "speed/wps": 12876.063793256064, "speed/FLOPS": 202236275438891.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43278470635414124, "optim/lr": 0.0004812, "optim/total_tokens": 421003264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.548683166503906, "created_at": "2025-01-15T07:48:25.438140+00:00"} {"global_step": 804, "acc_step": 0, "speed/wps": 12878.710590159575, "speed/FLOPS": 202277847021341.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36623653769493103, "optim/lr": 0.0004818, "optim/total_tokens": 421527552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413950, "loss/out": 4.444465637207031, "created_at": "2025-01-15T07:48:35.622855+00:00"} {"global_step": 805, "acc_step": 0, "speed/wps": 12874.883680037563, "speed/FLOPS": 202217740139149.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3995569944381714, "optim/lr": 0.0004824, "optim/total_tokens": 422051840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 4.6548309326171875, "created_at": "2025-01-15T07:48:45.808736+00:00"} {"global_step": 806, "acc_step": 0, "speed/wps": 12881.337077812947, "speed/FLOPS": 202319099619109.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40395882725715637, "optim/lr": 0.00048300000000000003, "optim/total_tokens": 422576128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.478141784667969, "created_at": "2025-01-15T07:48:55.993459+00:00"} {"global_step": 807, "acc_step": 0, "speed/wps": 12879.540230677865, "speed/FLOPS": 202290877665726.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6249206066131592, "optim/lr": 0.00048360000000000005, "optim/total_tokens": 423100416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.5872344970703125, "created_at": "2025-01-15T07:49:06.179086+00:00"} {"global_step": 808, "acc_step": 0, "speed/wps": 12878.28654745882, "speed/FLOPS": 202271186848030.25, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3806302547454834, "optim/lr": 0.00048419999999999995, "optim/total_tokens": 423624704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 4.472072601318359, "created_at": "2025-01-15T07:49:16.363411+00:00"} {"global_step": 809, "acc_step": 0, "speed/wps": 12882.710106468237, "speed/FLOPS": 202340664920880.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6666455864906311, "optim/lr": 0.00048479999999999997, "optim/total_tokens": 424148992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.586348056793213, "created_at": "2025-01-15T07:49:26.544960+00:00"} {"global_step": 810, "acc_step": 0, "speed/wps": 12880.53392551777, "speed/FLOPS": 202306485008667.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0470142364501953, "optim/lr": 0.00048540000000000003, "optim/total_tokens": 424673280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464317, "loss/out": 4.784092426300049, "created_at": "2025-01-15T07:49:36.731035+00:00"} {"global_step": 811, "acc_step": 0, "speed/wps": 12878.893321442816, "speed/FLOPS": 202280717067243.72, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7095128297805786, "optim/lr": 0.00048600000000000005, "optim/total_tokens": 425197568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277225, "loss/out": 4.779707908630371, "created_at": "2025-01-15T07:49:46.911850+00:00"} {"global_step": 812, "acc_step": 0, "speed/wps": 12876.762697636805, "speed/FLOPS": 202247252692586.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7621821165084839, "optim/lr": 0.00048660000000000006, "optim/total_tokens": 425721856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 4.809318542480469, "created_at": "2025-01-15T07:49:57.096131+00:00"} {"global_step": 813, "acc_step": 0, "speed/wps": 12882.626520824726, "speed/FLOPS": 202339352093490.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9198881983757019, "optim/lr": 0.00048719999999999997, "optim/total_tokens": 426246144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 4.799981117248535, "created_at": "2025-01-15T07:50:07.274112+00:00"} {"global_step": 814, "acc_step": 0, "speed/wps": 12879.26968997729, "speed/FLOPS": 202286628452262.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5508379936218262, "optim/lr": 0.0004878, "optim/total_tokens": 426770432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.854776382446289, "created_at": "2025-01-15T07:50:17.459711+00:00"} {"global_step": 815, "acc_step": 0, "speed/wps": 12885.149816638148, "speed/FLOPS": 202378983921611.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.608350396156311, "optim/lr": 0.0004884, "optim/total_tokens": 427294720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.772674560546875, "created_at": "2025-01-15T07:50:27.638572+00:00"} {"global_step": 816, "acc_step": 0, "speed/wps": 12882.252934640528, "speed/FLOPS": 202333484409106.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6409499645233154, "optim/lr": 0.0004890000000000001, "optim/total_tokens": 427819008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.884284973144531, "created_at": "2025-01-15T07:50:37.818566+00:00"} {"global_step": 817, "acc_step": 0, "speed/wps": 12885.332134933944, "speed/FLOPS": 202381847480978.8, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7665430307388306, "optim/lr": 0.0004896000000000001, "optim/total_tokens": 428343296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 4.688636779785156, "created_at": "2025-01-15T07:50:47.994334+00:00"} {"global_step": 818, "acc_step": 0, "speed/wps": 12880.847017563812, "speed/FLOPS": 202311402549483.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6377753615379333, "optim/lr": 0.0004902, "optim/total_tokens": 428867584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.845278739929199, "created_at": "2025-01-15T07:50:58.173603+00:00"} {"global_step": 819, "acc_step": 0, "speed/wps": 12883.583362467078, "speed/FLOPS": 202354380606321.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6808437705039978, "optim/lr": 0.0004908, "optim/total_tokens": 429391872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.683484077453613, "created_at": "2025-01-15T07:51:08.351530+00:00"} {"global_step": 820, "acc_step": 0, "speed/wps": 12880.077073154647, "speed/FLOPS": 202299309514523.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5507248640060425, "optim/lr": 0.0004914, "optim/total_tokens": 429916160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 4.72309684753418, "created_at": "2025-01-15T07:51:18.534362+00:00"} {"global_step": 821, "acc_step": 0, "speed/wps": 12877.16904895514, "speed/FLOPS": 202253634998426.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4532642662525177, "optim/lr": 0.000492, "optim/total_tokens": 430440448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.719575881958008, "created_at": "2025-01-15T07:51:28.743889+00:00"} {"global_step": 822, "acc_step": 0, "speed/wps": 12882.434637969416, "speed/FLOPS": 202336338309572.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.001, "optim/grad_norm": 0.35710400342941284, "optim/lr": 0.0004926, "optim/total_tokens": 430964736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 4.588325500488281, "created_at": "2025-01-15T07:51:38.921969+00:00"} {"global_step": 823, "acc_step": 0, "speed/wps": 12885.563306544154, "speed/FLOPS": 202385478348780.25, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4265560507774353, "optim/lr": 0.0004932, "optim/total_tokens": 431489024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 4.735571384429932, "created_at": "2025-01-15T07:51:49.097679+00:00"} {"global_step": 824, "acc_step": 0, "speed/wps": 12879.798864758386, "speed/FLOPS": 202294939869361.66, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.001, "optim/grad_norm": 0.43681371212005615, "optim/lr": 0.0004938, "optim/total_tokens": 432013312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.677741050720215, "created_at": "2025-01-15T07:51:59.280371+00:00"} {"global_step": 825, "acc_step": 0, "speed/wps": 12885.54310302978, "speed/FLOPS": 202385161024824.47, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39183762669563293, "optim/lr": 0.0004944, "optim/total_tokens": 432537600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.571809768676758, "created_at": "2025-01-15T07:52:09.456113+00:00"} {"global_step": 826, "acc_step": 0, "speed/wps": 12884.770388867682, "speed/FLOPS": 202373024487088.2, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3822130858898163, "optim/lr": 0.000495, "optim/total_tokens": 433061888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.650404453277588, "created_at": "2025-01-15T07:52:19.632921+00:00"} {"global_step": 827, "acc_step": 0, "speed/wps": 12882.879784615467, "speed/FLOPS": 202343329949344.66, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3006029725074768, "optim/lr": 0.0004956, "optim/total_tokens": 433586176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.639608860015869, "created_at": "2025-01-15T07:52:29.811409+00:00"} {"global_step": 828, "acc_step": 0, "speed/wps": 12881.545366037099, "speed/FLOPS": 202322371071888.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30920302867889404, "optim/lr": 0.0004962, "optim/total_tokens": 434110464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379498, "loss/out": 4.611130237579346, "created_at": "2025-01-15T07:52:39.992085+00:00"} {"global_step": 829, "acc_step": 0, "speed/wps": 12882.913173593528, "speed/FLOPS": 202343854369126.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3146301209926605, "optim/lr": 0.0004968, "optim/total_tokens": 434634752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 4.495153427124023, "created_at": "2025-01-15T07:52:50.172353+00:00"} {"global_step": 830, "acc_step": 0, "speed/wps": 12877.883109225946, "speed/FLOPS": 202264850296200.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2734779715538025, "optim/lr": 0.0004974000000000001, "optim/total_tokens": 435159040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.452763080596924, "created_at": "2025-01-15T07:53:00.358128+00:00"} {"global_step": 831, "acc_step": 0, "speed/wps": 12882.162276509518, "speed/FLOPS": 202332060498581.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.28278398513793945, "optim/lr": 0.0004980000000000001, "optim/total_tokens": 435683328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 4.472281455993652, "created_at": "2025-01-15T07:53:10.536361+00:00"} {"global_step": 832, "acc_step": 0, "speed/wps": 12883.686577881685, "speed/FLOPS": 202356001746240.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.298869252204895, "optim/lr": 0.0004986, "optim/total_tokens": 436207616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 4.519031047821045, "created_at": "2025-01-15T07:53:20.714990+00:00"} {"global_step": 833, "acc_step": 0, "speed/wps": 12882.544050625162, "speed/FLOPS": 202338056785680.06, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2809235751628876, "optim/lr": 0.0004992, "optim/total_tokens": 436731904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357619, "loss/out": 4.583465576171875, "created_at": "2025-01-15T07:53:30.892835+00:00"} {"global_step": 834, "acc_step": 0, "speed/wps": 12882.203691802217, "speed/FLOPS": 202332710982663.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3062553405761719, "optim/lr": 0.0004998, "optim/total_tokens": 437256192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 4.497365474700928, "created_at": "2025-01-15T07:53:41.073487+00:00"} {"global_step": 835, "acc_step": 0, "speed/wps": 12881.91320079496, "speed/FLOPS": 202328148422218.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3086269497871399, "optim/lr": 0.0005004, "optim/total_tokens": 437780480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.485457897186279, "created_at": "2025-01-15T07:53:51.255501+00:00"} {"global_step": 836, "acc_step": 0, "speed/wps": 12884.681567337533, "speed/FLOPS": 202371629422903.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29386699199676514, "optim/lr": 0.000501, "optim/total_tokens": 438304768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 4.486127853393555, "created_at": "2025-01-15T07:54:01.433753+00:00"} {"global_step": 837, "acc_step": 0, "speed/wps": 12887.375888577224, "speed/FLOPS": 202413947440358.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.29992175102233887, "optim/lr": 0.0005015999999999999, "optim/total_tokens": 438829056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 4.492403030395508, "created_at": "2025-01-15T07:54:11.637150+00:00"} {"global_step": 838, "acc_step": 0, "speed/wps": 12880.96129933343, "speed/FLOPS": 202313197501715.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3021200895309448, "optim/lr": 0.0005022, "optim/total_tokens": 439353344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 4.530375003814697, "created_at": "2025-01-15T07:54:21.818503+00:00"} {"global_step": 839, "acc_step": 0, "speed/wps": 12881.680561330268, "speed/FLOPS": 202324494499745.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30231472849845886, "optim/lr": 0.0005028, "optim/total_tokens": 439877632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.477451801300049, "created_at": "2025-01-15T07:54:32.005675+00:00"} {"global_step": 840, "acc_step": 0, "speed/wps": 12879.017012427686, "speed/FLOPS": 202282659804130.34, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.25921133160591125, "optim/lr": 0.0005034, "optim/total_tokens": 440401920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.382424354553223, "created_at": "2025-01-15T07:54:42.186373+00:00"} {"global_step": 841, "acc_step": 0, "speed/wps": 12882.830357735891, "speed/FLOPS": 202342553632281.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31426161527633667, "optim/lr": 0.000504, "optim/total_tokens": 440926208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 4.479162216186523, "created_at": "2025-01-15T07:54:52.369763+00:00"} {"global_step": 842, "acc_step": 0, "speed/wps": 12882.087068631923, "speed/FLOPS": 202330879255517.2, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2445567548274994, "optim/lr": 0.0005046, "optim/total_tokens": 441450496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 4.462549209594727, "created_at": "2025-01-15T07:55:02.549885+00:00"} {"global_step": 843, "acc_step": 0, "speed/wps": 12887.735170916048, "speed/FLOPS": 202419590463195.25, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2579631507396698, "optim/lr": 0.0005052, "optim/total_tokens": 441974784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 4.4226531982421875, "created_at": "2025-01-15T07:55:12.734102+00:00"} {"global_step": 844, "acc_step": 0, "speed/wps": 12883.274615375607, "speed/FLOPS": 202349531308986.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2567557692527771, "optim/lr": 0.0005058, "optim/total_tokens": 442499072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500101, "loss/out": 4.391462326049805, "created_at": "2025-01-15T07:55:22.911477+00:00"} {"global_step": 845, "acc_step": 0, "speed/wps": 12889.004053084678, "speed/FLOPS": 202439520001280.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25967949628829956, "optim/lr": 0.0005064000000000001, "optim/total_tokens": 443023360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 4.387927055358887, "created_at": "2025-01-15T07:55:33.087855+00:00"} {"global_step": 846, "acc_step": 0, "speed/wps": 12882.461035511347, "speed/FLOPS": 202336752919242.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2848162055015564, "optim/lr": 0.0005070000000000001, "optim/total_tokens": 443547648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 4.4912333488464355, "created_at": "2025-01-15T07:55:43.266613+00:00"} {"global_step": 847, "acc_step": 0, "speed/wps": 12884.123116913637, "speed/FLOPS": 202362858191605.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3037416636943817, "optim/lr": 0.0005076, "optim/total_tokens": 444071936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 4.535124778747559, "created_at": "2025-01-15T07:55:53.443236+00:00"} {"global_step": 848, "acc_step": 0, "speed/wps": 12883.567185746695, "speed/FLOPS": 202354126528698.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.354857861995697, "optim/lr": 0.0005082, "optim/total_tokens": 444596224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 4.42104959487915, "created_at": "2025-01-15T07:56:03.627823+00:00"} {"global_step": 849, "acc_step": 0, "speed/wps": 12885.889247241787, "speed/FLOPS": 202390597695326.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5736861824989319, "optim/lr": 0.0005088, "optim/total_tokens": 445120512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 4.456495761871338, "created_at": "2025-01-15T07:56:13.805317+00:00"} {"global_step": 850, "acc_step": 0, "speed/wps": 12883.279286782874, "speed/FLOPS": 202349604679857.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43166017532348633, "optim/lr": 0.0005094, "optim/total_tokens": 445644800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308039, "loss/out": 4.524689674377441, "created_at": "2025-01-15T07:56:23.986118+00:00"} {"global_step": 851, "acc_step": 0, "speed/wps": 12886.309346059606, "speed/FLOPS": 202397195924536.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8932659029960632, "optim/lr": 0.00051, "optim/total_tokens": 446169088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.5841965675354, "created_at": "2025-01-15T07:56:34.161361+00:00"} {"global_step": 852, "acc_step": 0, "speed/wps": 12886.685350509268, "speed/FLOPS": 202403101591102.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1006758213043213, "optim/lr": 0.0005105999999999999, "optim/total_tokens": 446693376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.738288879394531, "created_at": "2025-01-15T07:56:44.336017+00:00"} {"global_step": 853, "acc_step": 0, "speed/wps": 12881.434411688864, "speed/FLOPS": 202320628381381.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.772651195526123, "optim/lr": 0.0005112, "optim/total_tokens": 447217664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.576052665710449, "created_at": "2025-01-15T07:56:54.518061+00:00"} {"global_step": 854, "acc_step": 0, "speed/wps": 12884.421728602674, "speed/FLOPS": 202367548298516.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.835858941078186, "optim/lr": 0.0005118, "optim/total_tokens": 447741952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 4.661455154418945, "created_at": "2025-01-15T07:57:04.696155+00:00"} {"global_step": 855, "acc_step": 0, "speed/wps": 12886.50978566272, "speed/FLOPS": 202400344103935.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8892427682876587, "optim/lr": 0.0005124, "optim/total_tokens": 448266240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366026, "loss/out": 4.726611137390137, "created_at": "2025-01-15T07:57:14.871951+00:00"} {"global_step": 856, "acc_step": 0, "speed/wps": 12888.136966524544, "speed/FLOPS": 202425901215351.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7556094527244568, "optim/lr": 0.000513, "optim/total_tokens": 448790528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 4.712546348571777, "created_at": "2025-01-15T07:57:25.048743+00:00"} {"global_step": 857, "acc_step": 0, "speed/wps": 12883.90953325621, "speed/FLOPS": 202359503566770.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8359956741333008, "optim/lr": 0.0005136, "optim/total_tokens": 449314816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 4.7600202560424805, "created_at": "2025-01-15T07:57:35.227896+00:00"} {"global_step": 858, "acc_step": 0, "speed/wps": 12884.769664578636, "speed/FLOPS": 202373013111133.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7011427283287048, "optim/lr": 0.0005142, "optim/total_tokens": 449839104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.576641082763672, "created_at": "2025-01-15T07:57:45.404423+00:00"} {"global_step": 859, "acc_step": 0, "speed/wps": 12885.453730520807, "speed/FLOPS": 202383757306760.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5721981525421143, "optim/lr": 0.0005148, "optim/total_tokens": 450363392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367469, "loss/out": 4.621487617492676, "created_at": "2025-01-15T07:57:55.581556+00:00"} {"global_step": 860, "acc_step": 0, "speed/wps": 12885.427357284829, "speed/FLOPS": 202383343078848.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5382391214370728, "optim/lr": 0.0005154000000000001, "optim/total_tokens": 450887680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 4.674420356750488, "created_at": "2025-01-15T07:58:05.758612+00:00"} {"global_step": 861, "acc_step": 0, "speed/wps": 12887.097802586226, "speed/FLOPS": 202409579717739.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4024108648300171, "optim/lr": 0.000516, "optim/total_tokens": 451411968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 4.586752414703369, "created_at": "2025-01-15T07:58:15.935393+00:00"} {"global_step": 862, "acc_step": 0, "speed/wps": 12887.938195856614, "speed/FLOPS": 202422779248872.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40559399127960205, "optim/lr": 0.0005166, "optim/total_tokens": 451936256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 4.704413414001465, "created_at": "2025-01-15T07:58:26.115167+00:00"} {"global_step": 863, "acc_step": 0, "speed/wps": 12886.158722234972, "speed/FLOPS": 202394830170391.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3909004032611847, "optim/lr": 0.0005172, "optim/total_tokens": 452460544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 4.657512664794922, "created_at": "2025-01-15T07:58:36.292290+00:00"} {"global_step": 864, "acc_step": 0, "speed/wps": 12886.573278677886, "speed/FLOPS": 202401341348987.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.49461808800697327, "optim/lr": 0.0005178, "optim/total_tokens": 452984832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 4.6308207511901855, "created_at": "2025-01-15T07:58:46.467412+00:00"} {"global_step": 865, "acc_step": 0, "speed/wps": 12885.727095971952, "speed/FLOPS": 202388050886814.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5521669983863831, "optim/lr": 0.0005184, "optim/total_tokens": 453509120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 4.741760730743408, "created_at": "2025-01-15T07:58:56.646726+00:00"} {"global_step": 866, "acc_step": 0, "speed/wps": 12886.301073745863, "speed/FLOPS": 202397065996481.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7794849872589111, "optim/lr": 0.0005189999999999999, "optim/total_tokens": 454033408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.670150279998779, "created_at": "2025-01-15T07:59:06.822649+00:00"} {"global_step": 867, "acc_step": 0, "speed/wps": 12883.850103761812, "speed/FLOPS": 202358570144896.25, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5293033123016357, "optim/lr": 0.0005195999999999999, "optim/total_tokens": 454557696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.6536760330200195, "created_at": "2025-01-15T07:59:17.002819+00:00"} {"global_step": 868, "acc_step": 0, "speed/wps": 12886.30799270788, "speed/FLOPS": 202397174668287.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5897225737571716, "optim/lr": 0.0005202, "optim/total_tokens": 455081984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.545059680938721, "created_at": "2025-01-15T07:59:27.180338+00:00"} {"global_step": 869, "acc_step": 0, "speed/wps": 12884.919398325492, "speed/FLOPS": 202375364885383.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5696394443511963, "optim/lr": 0.0005208, "optim/total_tokens": 455606272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342387, "loss/out": 4.655447006225586, "created_at": "2025-01-15T07:59:37.359761+00:00"} {"global_step": 870, "acc_step": 0, "speed/wps": 12885.263466943095, "speed/FLOPS": 202380768955822.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4155213534832001, "optim/lr": 0.0005214, "optim/total_tokens": 456130560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 4.523919105529785, "created_at": "2025-01-15T07:59:47.535577+00:00"} {"global_step": 871, "acc_step": 0, "speed/wps": 12887.267006293992, "speed/FLOPS": 202412237294480.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3897581100463867, "optim/lr": 0.000522, "optim/total_tokens": 456654848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.39534330368042, "created_at": "2025-01-15T07:59:57.713636+00:00"} {"global_step": 872, "acc_step": 0, "speed/wps": 12888.212845809085, "speed/FLOPS": 202427093003784.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.31421858072280884, "optim/lr": 0.0005226, "optim/total_tokens": 457179136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 4.540709495544434, "created_at": "2025-01-15T08:00:07.887189+00:00"} {"global_step": 873, "acc_step": 0, "speed/wps": 12880.991802951661, "speed/FLOPS": 202313676602955.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28959518671035767, "optim/lr": 0.0005232, "optim/total_tokens": 457703424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 4.572047233581543, "created_at": "2025-01-15T08:00:18.070500+00:00"} {"global_step": 874, "acc_step": 0, "speed/wps": 12883.969640318812, "speed/FLOPS": 202360447630784.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2588850259780884, "optim/lr": 0.0005238, "optim/total_tokens": 458227712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.406301021575928, "created_at": "2025-01-15T08:00:28.247649+00:00"} {"global_step": 875, "acc_step": 0, "speed/wps": 12887.488573871147, "speed/FLOPS": 202415717317745.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2661459445953369, "optim/lr": 0.0005244000000000001, "optim/total_tokens": 458752000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 4.473996639251709, "created_at": "2025-01-15T08:00:38.422932+00:00"} {"global_step": 876, "acc_step": 0, "speed/wps": 12881.366569270253, "speed/FLOPS": 202319562822972.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2562490999698639, "optim/lr": 0.000525, "optim/total_tokens": 459276288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 4.426822185516357, "created_at": "2025-01-15T08:00:48.604061+00:00"} {"global_step": 877, "acc_step": 0, "speed/wps": 12873.080820110104, "speed/FLOPS": 202189423746599.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2543988823890686, "optim/lr": 0.0005256, "optim/total_tokens": 459800576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470104, "loss/out": 4.57494592666626, "created_at": "2025-01-15T08:00:58.793985+00:00"} {"global_step": 878, "acc_step": 0, "speed/wps": 12875.026733065788, "speed/FLOPS": 202219986983533.28, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.21919934451580048, "optim/lr": 0.0005262, "optim/total_tokens": 460324864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389901, "loss/out": 4.47795295715332, "created_at": "2025-01-15T08:01:08.978585+00:00"} {"global_step": 879, "acc_step": 0, "speed/wps": 12882.922664477215, "speed/FLOPS": 202344003436496.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21915286779403687, "optim/lr": 0.0005268, "optim/total_tokens": 460849152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 4.49525785446167, "created_at": "2025-01-15T08:01:19.158687+00:00"} {"global_step": 880, "acc_step": 0, "speed/wps": 12884.668263736097, "speed/FLOPS": 202371420471562.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23366932570934296, "optim/lr": 0.0005274, "optim/total_tokens": 461373440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 4.409564971923828, "created_at": "2025-01-15T08:01:29.335481+00:00"} {"global_step": 881, "acc_step": 0, "speed/wps": 12884.338527690014, "speed/FLOPS": 202366241513853.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2836194336414337, "optim/lr": 0.0005279999999999999, "optim/total_tokens": 461897728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 4.417449951171875, "created_at": "2025-01-15T08:01:39.512306+00:00"} {"global_step": 882, "acc_step": 0, "speed/wps": 12887.396736276036, "speed/FLOPS": 202414274882115.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2677065432071686, "optim/lr": 0.0005286, "optim/total_tokens": 462422016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 4.497450828552246, "created_at": "2025-01-15T08:01:49.686332+00:00"} {"global_step": 883, "acc_step": 0, "speed/wps": 12887.764793171986, "speed/FLOPS": 202420055721430.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2195439338684082, "optim/lr": 0.0005292, "optim/total_tokens": 462946304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 4.469263553619385, "created_at": "2025-01-15T08:01:59.863016+00:00"} {"global_step": 884, "acc_step": 0, "speed/wps": 12886.345854806055, "speed/FLOPS": 202397769344567.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2583017349243164, "optim/lr": 0.0005298000000000001, "optim/total_tokens": 463470592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.408348083496094, "created_at": "2025-01-15T08:02:10.042457+00:00"} {"global_step": 885, "acc_step": 0, "speed/wps": 12884.241085047537, "speed/FLOPS": 202364711043253.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2728859782218933, "optim/lr": 0.0005304000000000001, "optim/total_tokens": 463994880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 4.390022277832031, "created_at": "2025-01-15T08:02:20.219193+00:00"} {"global_step": 886, "acc_step": 0, "speed/wps": 12886.983077955001, "speed/FLOPS": 202407777809757.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2850431799888611, "optim/lr": 0.000531, "optim/total_tokens": 464519168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.333884239196777, "created_at": "2025-01-15T08:02:30.394636+00:00"} {"global_step": 887, "acc_step": 0, "speed/wps": 12886.81748187807, "speed/FLOPS": 202405176895816.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2850317358970642, "optim/lr": 0.0005316, "optim/total_tokens": 465043456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.493710517883301, "created_at": "2025-01-15T08:02:40.571263+00:00"} {"global_step": 888, "acc_step": 0, "speed/wps": 12885.502746920693, "speed/FLOPS": 202384527176676.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.26298344135284424, "optim/lr": 0.0005322, "optim/total_tokens": 465567744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352941, "loss/out": 4.4513654708862305, "created_at": "2025-01-15T08:02:50.747096+00:00"} {"global_step": 889, "acc_step": 0, "speed/wps": 12883.570831284322, "speed/FLOPS": 202354183786876.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36824217438697815, "optim/lr": 0.0005328, "optim/total_tokens": 466092032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.459596633911133, "created_at": "2025-01-15T08:03:00.925573+00:00"} {"global_step": 890, "acc_step": 0, "speed/wps": 12881.930657406234, "speed/FLOPS": 202328422602286.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3275637924671173, "optim/lr": 0.0005334000000000001, "optim/total_tokens": 466616320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.367779731750488, "created_at": "2025-01-15T08:03:11.104003+00:00"} {"global_step": 891, "acc_step": 0, "speed/wps": 12883.344283498222, "speed/FLOPS": 202350625542586.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4013930857181549, "optim/lr": 0.000534, "optim/total_tokens": 467140608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.537389755249023, "created_at": "2025-01-15T08:03:21.284526+00:00"} {"global_step": 892, "acc_step": 0, "speed/wps": 12882.199710979552, "speed/FLOPS": 202332648458373.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.32731080055236816, "optim/lr": 0.0005346, "optim/total_tokens": 467664896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365889, "loss/out": 4.479383945465088, "created_at": "2025-01-15T08:03:31.462777+00:00"} {"global_step": 893, "acc_step": 0, "speed/wps": 12887.043669194494, "speed/FLOPS": 202408729478435.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3311581015586853, "optim/lr": 0.0005352, "optim/total_tokens": 468189184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 4.375881671905518, "created_at": "2025-01-15T08:03:41.637801+00:00"} {"global_step": 894, "acc_step": 0, "speed/wps": 12881.244204302542, "speed/FLOPS": 202317640913007.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3160264790058136, "optim/lr": 0.0005358, "optim/total_tokens": 468713472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 4.386460781097412, "created_at": "2025-01-15T08:03:51.816994+00:00"} {"global_step": 895, "acc_step": 0, "speed/wps": 12884.569867717886, "speed/FLOPS": 202369875026885.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2605012059211731, "optim/lr": 0.0005363999999999999, "optim/total_tokens": 469237760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.457608699798584, "created_at": "2025-01-15T08:04:01.997192+00:00"} {"global_step": 896, "acc_step": 0, "speed/wps": 12875.01911950931, "speed/FLOPS": 202219867402166.7, "speed/curr_iter_time": 1.2909, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29727545380592346, "optim/lr": 0.000537, "optim/total_tokens": 469762048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.4391069412231445, "created_at": "2025-01-15T08:04:12.181340+00:00"} {"global_step": 897, "acc_step": 0, "speed/wps": 12879.578487947316, "speed/FLOPS": 202291478548714.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3180823028087616, "optim/lr": 0.0005376000000000001, "optim/total_tokens": 470286336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 278812, "loss/out": 4.457326412200928, "created_at": "2025-01-15T08:04:22.361934+00:00"} {"global_step": 898, "acc_step": 0, "speed/wps": 12863.601440663484, "speed/FLOPS": 202040537066358.1, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2633056938648224, "optim/lr": 0.0005382000000000001, "optim/total_tokens": 470810624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 4.516852855682373, "created_at": "2025-01-15T08:04:32.558107+00:00"} {"global_step": 899, "acc_step": 0, "speed/wps": 12880.65176781227, "speed/FLOPS": 202308335883834.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2380695641040802, "optim/lr": 0.0005388000000000001, "optim/total_tokens": 471334912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 4.491059303283691, "created_at": "2025-01-15T08:04:42.740762+00:00"} {"global_step": 900, "acc_step": 0, "speed/wps": 12884.06340826763, "speed/FLOPS": 202361920385271.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2895808517932892, "optim/lr": 0.0005394, "optim/total_tokens": 471859200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 4.395750999450684, "created_at": "2025-01-15T08:04:52.917776+00:00"} {"global_step": 901, "acc_step": 0, "speed/wps": 12882.447274249107, "speed/FLOPS": 202336536779709.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.238485649228096, "optim/lr": 0.00054, "optim/total_tokens": 472383488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.338047027587891, "created_at": "2025-01-15T08:05:03.098876+00:00"} {"global_step": 902, "acc_step": 0, "speed/wps": 12882.84007727603, "speed/FLOPS": 202342706291015.22, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27113527059555054, "optim/lr": 0.0005406, "optim/total_tokens": 472907776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 4.3674845695495605, "created_at": "2025-01-15T08:05:13.277841+00:00"} {"global_step": 903, "acc_step": 0, "speed/wps": 12884.42055022226, "speed/FLOPS": 202367529790433.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29508474469184875, "optim/lr": 0.0005412, "optim/total_tokens": 473432064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 4.472416400909424, "created_at": "2025-01-15T08:05:23.454547+00:00"} {"global_step": 904, "acc_step": 0, "speed/wps": 12881.741739415258, "speed/FLOPS": 202325455385637.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3044950067996979, "optim/lr": 0.0005418, "optim/total_tokens": 473956352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 4.418232440948486, "created_at": "2025-01-15T08:05:33.635023+00:00"} {"global_step": 905, "acc_step": 0, "speed/wps": 12882.230472803478, "speed/FLOPS": 202333131615089.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.28871825337409973, "optim/lr": 0.0005424, "optim/total_tokens": 474480640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.317686080932617, "created_at": "2025-01-15T08:05:43.814951+00:00"} {"global_step": 906, "acc_step": 0, "speed/wps": 12882.473826318475, "speed/FLOPS": 202336953816445.25, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3593020439147949, "optim/lr": 0.000543, "optim/total_tokens": 475004928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.42323637008667, "created_at": "2025-01-15T08:05:53.996884+00:00"} {"global_step": 907, "acc_step": 0, "speed/wps": 12884.134197196961, "speed/FLOPS": 202363032222680.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2734982371330261, "optim/lr": 0.0005436, "optim/total_tokens": 475529216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.402863025665283, "created_at": "2025-01-15T08:06:04.176032+00:00"} {"global_step": 908, "acc_step": 0, "speed/wps": 12884.760595637326, "speed/FLOPS": 202372870670948.16, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2881162762641907, "optim/lr": 0.0005442, "optim/total_tokens": 476053504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.338925361633301, "created_at": "2025-01-15T08:06:14.352257+00:00"} {"global_step": 909, "acc_step": 0, "speed/wps": 12881.579410718883, "speed/FLOPS": 202322905790397.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.31946074962615967, "optim/lr": 0.0005448, "optim/total_tokens": 476577792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288153, "loss/out": 4.337482929229736, "created_at": "2025-01-15T08:06:24.532709+00:00"} {"global_step": 910, "acc_step": 0, "speed/wps": 12886.813063964146, "speed/FLOPS": 202405107506407.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2947985827922821, "optim/lr": 0.0005454, "optim/total_tokens": 477102080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 4.3747968673706055, "created_at": "2025-01-15T08:06:34.707409+00:00"} {"global_step": 911, "acc_step": 0, "speed/wps": 12885.58289310073, "speed/FLOPS": 202385785982566.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.381991446018219, "optim/lr": 0.000546, "optim/total_tokens": 477626368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.397697448730469, "created_at": "2025-01-15T08:06:44.887673+00:00"} {"global_step": 912, "acc_step": 0, "speed/wps": 12881.624044600974, "speed/FLOPS": 202323606826849.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4249407947063446, "optim/lr": 0.0005466000000000001, "optim/total_tokens": 478150656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.412092685699463, "created_at": "2025-01-15T08:06:55.067864+00:00"} {"global_step": 913, "acc_step": 0, "speed/wps": 12884.060044350266, "speed/FLOPS": 202361867550326.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5035578608512878, "optim/lr": 0.0005472000000000001, "optim/total_tokens": 478674944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 4.356297016143799, "created_at": "2025-01-15T08:07:05.245818+00:00"} {"global_step": 914, "acc_step": 0, "speed/wps": 12880.96227666363, "speed/FLOPS": 202313212852029.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5945838689804077, "optim/lr": 0.0005478000000000001, "optim/total_tokens": 479199232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 4.2993974685668945, "created_at": "2025-01-15T08:07:15.430061+00:00"} {"global_step": 915, "acc_step": 0, "speed/wps": 12881.893315290352, "speed/FLOPS": 202327836093042.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4993235766887665, "optim/lr": 0.0005484, "optim/total_tokens": 479723520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 4.451369285583496, "created_at": "2025-01-15T08:07:25.610577+00:00"} {"global_step": 916, "acc_step": 0, "speed/wps": 12880.951687552302, "speed/FLOPS": 202313046535484.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.4527733325958252, "optim/lr": 0.000549, "optim/total_tokens": 480247808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.387568473815918, "created_at": "2025-01-15T08:07:35.795187+00:00"} {"global_step": 917, "acc_step": 0, "speed/wps": 12886.366609498828, "speed/FLOPS": 202398095325537.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3797396719455719, "optim/lr": 0.0005496, "optim/total_tokens": 480772096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 4.39445686340332, "created_at": "2025-01-15T08:07:45.970283+00:00"} {"global_step": 918, "acc_step": 0, "speed/wps": 12881.688339582115, "speed/FLOPS": 202324616667878.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.39247071743011475, "optim/lr": 0.0005502, "optim/total_tokens": 481296384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 4.5220232009887695, "created_at": "2025-01-15T08:07:56.154035+00:00"} {"global_step": 919, "acc_step": 0, "speed/wps": 12887.249005361948, "speed/FLOPS": 202411954565106.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3606473207473755, "optim/lr": 0.0005508, "optim/total_tokens": 481820672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 4.385117530822754, "created_at": "2025-01-15T08:08:06.328697+00:00"} {"global_step": 920, "acc_step": 0, "speed/wps": 12885.591296627148, "speed/FLOPS": 202385917971496.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3387679159641266, "optim/lr": 0.0005514, "optim/total_tokens": 482344960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.402121067047119, "created_at": "2025-01-15T08:08:16.506799+00:00"} {"global_step": 921, "acc_step": 0, "speed/wps": 12885.717159711552, "speed/FLOPS": 202387894824190.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3217432498931885, "optim/lr": 0.000552, "optim/total_tokens": 482869248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 4.467331409454346, "created_at": "2025-01-15T08:08:26.682513+00:00"} {"global_step": 922, "acc_step": 0, "speed/wps": 12886.782499729146, "speed/FLOPS": 202404627453097.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2801218330860138, "optim/lr": 0.0005526, "optim/total_tokens": 483393536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 4.232789993286133, "created_at": "2025-01-15T08:08:36.858190+00:00"} {"global_step": 923, "acc_step": 0, "speed/wps": 12882.850099903819, "speed/FLOPS": 202342863710157.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29480481147766113, "optim/lr": 0.0005532, "optim/total_tokens": 483917824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 4.378745079040527, "created_at": "2025-01-15T08:08:47.040674+00:00"} {"global_step": 924, "acc_step": 0, "speed/wps": 12888.240206523104, "speed/FLOPS": 202427522741395.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3104386329650879, "optim/lr": 0.0005538, "optim/total_tokens": 484442112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 4.363844871520996, "created_at": "2025-01-15T08:08:57.215302+00:00"} {"global_step": 925, "acc_step": 0, "speed/wps": 12884.397989999765, "speed/FLOPS": 202367175451137.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37154990434646606, "optim/lr": 0.0005544, "optim/total_tokens": 484966400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 4.408888816833496, "created_at": "2025-01-15T08:09:07.392096+00:00"} {"global_step": 926, "acc_step": 0, "speed/wps": 12881.216513697487, "speed/FLOPS": 202317205994004.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31059086322784424, "optim/lr": 0.000555, "optim/total_tokens": 485490688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 4.3970255851745605, "created_at": "2025-01-15T08:09:17.571233+00:00"} {"global_step": 927, "acc_step": 0, "speed/wps": 12882.326689350868, "speed/FLOPS": 202334642828181.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.25752779841423035, "optim/lr": 0.0005556000000000001, "optim/total_tokens": 486014976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.335503101348877, "created_at": "2025-01-15T08:09:27.749640+00:00"} {"global_step": 928, "acc_step": 0, "speed/wps": 12886.703923335992, "speed/FLOPS": 202403393302867.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2502647340297699, "optim/lr": 0.0005562000000000001, "optim/total_tokens": 486539264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369865, "loss/out": 4.357514381408691, "created_at": "2025-01-15T08:09:37.925108+00:00"} {"global_step": 929, "acc_step": 0, "speed/wps": 12882.4122379959, "speed/FLOPS": 202335986487208.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2606560289859772, "optim/lr": 0.0005568, "optim/total_tokens": 487063552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.30336856842041, "created_at": "2025-01-15T08:09:48.105314+00:00"} {"global_step": 930, "acc_step": 0, "speed/wps": 12886.614243273229, "speed/FLOPS": 202401984754248.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2774963676929474, "optim/lr": 0.0005574, "optim/total_tokens": 487587840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.356391906738281, "created_at": "2025-01-15T08:09:58.280011+00:00"} {"global_step": 931, "acc_step": 0, "speed/wps": 12878.80045488281, "speed/FLOPS": 202279258470306.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2645528316497803, "optim/lr": 0.000558, "optim/total_tokens": 488112128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 4.260987281799316, "created_at": "2025-01-15T08:10:08.464796+00:00"} {"global_step": 932, "acc_step": 0, "speed/wps": 12884.173838204828, "speed/FLOPS": 202363654839181.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24927270412445068, "optim/lr": 0.0005586, "optim/total_tokens": 488636416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413748, "loss/out": 4.38124942779541, "created_at": "2025-01-15T08:10:18.641423+00:00"} {"global_step": 933, "acc_step": 0, "speed/wps": 12884.030265737638, "speed/FLOPS": 202361399836295.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2723354697227478, "optim/lr": 0.0005592, "optim/total_tokens": 489160704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 4.280298709869385, "created_at": "2025-01-15T08:10:28.827841+00:00"} {"global_step": 934, "acc_step": 0, "speed/wps": 12889.909912634314, "speed/FLOPS": 202453747770288.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2621883451938629, "optim/lr": 0.0005598, "optim/total_tokens": 489684992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.21574592590332, "created_at": "2025-01-15T08:10:39.002437+00:00"} {"global_step": 935, "acc_step": 0, "speed/wps": 12883.450692386741, "speed/FLOPS": 202352296840402.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.256815642118454, "optim/lr": 0.0005604, "optim/total_tokens": 490209280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 4.232883930206299, "created_at": "2025-01-15T08:10:49.180658+00:00"} {"global_step": 936, "acc_step": 0, "speed/wps": 12886.068261612674, "speed/FLOPS": 202393409362012.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3135489821434021, "optim/lr": 0.000561, "optim/total_tokens": 490733568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 4.406730651855469, "created_at": "2025-01-15T08:10:59.358569+00:00"} {"global_step": 937, "acc_step": 0, "speed/wps": 12885.41833696685, "speed/FLOPS": 202383201402359.62, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30222949385643005, "optim/lr": 0.0005616, "optim/total_tokens": 491257856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 4.256891250610352, "created_at": "2025-01-15T08:11:09.537697+00:00"} {"global_step": 938, "acc_step": 0, "speed/wps": 12881.712425457785, "speed/FLOPS": 202324994969653.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30989983677864075, "optim/lr": 0.0005622, "optim/total_tokens": 491782144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.249396324157715, "created_at": "2025-01-15T08:11:19.716255+00:00"} {"global_step": 939, "acc_step": 0, "speed/wps": 12876.880599357753, "speed/FLOPS": 202249104501127.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27306047081947327, "optim/lr": 0.0005628, "optim/total_tokens": 492306432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.287375450134277, "created_at": "2025-01-15T08:11:29.899066+00:00"} {"global_step": 940, "acc_step": 0, "speed/wps": 12884.32175340457, "speed/FLOPS": 202365978050650.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2519710063934326, "optim/lr": 0.0005634, "optim/total_tokens": 492830720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 4.2941365242004395, "created_at": "2025-01-15T08:11:40.076700+00:00"} {"global_step": 941, "acc_step": 0, "speed/wps": 12886.022519721488, "speed/FLOPS": 202392690922755.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2302069216966629, "optim/lr": 0.000564, "optim/total_tokens": 493355008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 4.3867621421813965, "created_at": "2025-01-15T08:11:50.255700+00:00"} {"global_step": 942, "acc_step": 0, "speed/wps": 12882.916245628778, "speed/FLOPS": 202343902619661.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3277542293071747, "optim/lr": 0.0005646000000000001, "optim/total_tokens": 493879296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 4.477817058563232, "created_at": "2025-01-15T08:12:00.433607+00:00"} {"global_step": 943, "acc_step": 0, "speed/wps": 12882.687440207163, "speed/FLOPS": 202340308916103.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3638428747653961, "optim/lr": 0.0005652000000000001, "optim/total_tokens": 494403584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.3340559005737305, "created_at": "2025-01-15T08:12:10.613434+00:00"} {"global_step": 944, "acc_step": 0, "speed/wps": 12880.856279053241, "speed/FLOPS": 202311548013901.88, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.287923127412796, "optim/lr": 0.0005658, "optim/total_tokens": 494927872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.360566139221191, "created_at": "2025-01-15T08:12:20.794093+00:00"} {"global_step": 945, "acc_step": 0, "speed/wps": 12886.105108795151, "speed/FLOPS": 202393988097646.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30308395624160767, "optim/lr": 0.0005664, "optim/total_tokens": 495452160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 4.263539791107178, "created_at": "2025-01-15T08:12:30.970705+00:00"} {"global_step": 946, "acc_step": 0, "speed/wps": 12880.360014728343, "speed/FLOPS": 202303753500740.72, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30777910351753235, "optim/lr": 0.000567, "optim/total_tokens": 495976448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.320089340209961, "created_at": "2025-01-15T08:12:41.152854+00:00"} {"global_step": 947, "acc_step": 0, "speed/wps": 12882.648659753162, "speed/FLOPS": 202339699815783.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3132437765598297, "optim/lr": 0.0005676, "optim/total_tokens": 496500736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 4.285765647888184, "created_at": "2025-01-15T08:12:51.330537+00:00"} {"global_step": 948, "acc_step": 0, "speed/wps": 12881.342255962623, "speed/FLOPS": 202319180949065.97, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2939061224460602, "optim/lr": 0.0005682, "optim/total_tokens": 497025024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.326603889465332, "created_at": "2025-01-15T08:13:01.512438+00:00"} {"global_step": 949, "acc_step": 0, "speed/wps": 12883.51604464987, "speed/FLOPS": 202353323287497.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.288195937871933, "optim/lr": 0.0005688, "optim/total_tokens": 497549312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.248201370239258, "created_at": "2025-01-15T08:13:11.697783+00:00"} {"global_step": 950, "acc_step": 0, "speed/wps": 12882.89511052761, "speed/FLOPS": 202343570663854.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23819954693317413, "optim/lr": 0.0005694, "optim/total_tokens": 498073600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.1933488845825195, "created_at": "2025-01-15T08:13:21.878981+00:00"} {"global_step": 951, "acc_step": 0, "speed/wps": 12881.51390006838, "speed/FLOPS": 202321876855610.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26494133472442627, "optim/lr": 0.00057, "optim/total_tokens": 498597888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.274994850158691, "created_at": "2025-01-15T08:13:32.057579+00:00"} {"global_step": 952, "acc_step": 0, "speed/wps": 12882.604841871065, "speed/FLOPS": 202339011595732.9, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25873327255249023, "optim/lr": 0.0005706, "optim/total_tokens": 499122176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349184, "loss/out": 4.277529716491699, "created_at": "2025-01-15T08:13:42.237764+00:00"} {"global_step": 953, "acc_step": 0, "speed/wps": 12878.185439535957, "speed/FLOPS": 202269598809164.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2638855278491974, "optim/lr": 0.0005712, "optim/total_tokens": 499646464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 4.196564197540283, "created_at": "2025-01-15T08:13:52.423514+00:00"} {"global_step": 954, "acc_step": 0, "speed/wps": 12884.402735354235, "speed/FLOPS": 202367249983450.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23157022893428802, "optim/lr": 0.0005718, "optim/total_tokens": 500170752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 4.269391059875488, "created_at": "2025-01-15T08:14:02.600251+00:00"} {"global_step": 955, "acc_step": 0, "speed/wps": 12880.355728552959, "speed/FLOPS": 202303686180466.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.3430815041065216, "optim/lr": 0.0005724, "optim/total_tokens": 500695040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.315501689910889, "created_at": "2025-01-15T08:14:12.780546+00:00"} {"global_step": 956, "acc_step": 0, "speed/wps": 12879.870763165007, "speed/FLOPS": 202296069132636.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38319531083106995, "optim/lr": 0.000573, "optim/total_tokens": 501219328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 4.277970790863037, "created_at": "2025-01-15T08:14:22.960530+00:00"} {"global_step": 957, "acc_step": 0, "speed/wps": 12882.14890854048, "speed/FLOPS": 202331850536258.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3599289655685425, "optim/lr": 0.0005736000000000001, "optim/total_tokens": 501743616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 4.299346923828125, "created_at": "2025-01-15T08:14:33.146902+00:00"} {"global_step": 958, "acc_step": 0, "speed/wps": 12881.55268666015, "speed/FLOPS": 202322486052332.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3364027738571167, "optim/lr": 0.0005742, "optim/total_tokens": 502267904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 4.375456809997559, "created_at": "2025-01-15T08:14:43.326516+00:00"} {"global_step": 959, "acc_step": 0, "speed/wps": 12885.143714087093, "speed/FLOPS": 202378888072661.66, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30365896224975586, "optim/lr": 0.0005748, "optim/total_tokens": 502792192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 4.361911296844482, "created_at": "2025-01-15T08:14:53.502691+00:00"} {"global_step": 960, "acc_step": 0, "speed/wps": 12881.412673080693, "speed/FLOPS": 202320286946668.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4079592525959015, "optim/lr": 0.0005754, "optim/total_tokens": 503316480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.207741737365723, "created_at": "2025-01-15T08:15:03.684094+00:00"} {"global_step": 961, "acc_step": 0, "speed/wps": 12884.141697689112, "speed/FLOPS": 202363150028216.53, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3528548777103424, "optim/lr": 0.000576, "optim/total_tokens": 503840768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.316590309143066, "created_at": "2025-01-15T08:15:13.864583+00:00"} {"global_step": 962, "acc_step": 0, "speed/wps": 12879.644129785394, "speed/FLOPS": 202292509543982.7, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28925931453704834, "optim/lr": 0.0005766, "optim/total_tokens": 504365056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.242393970489502, "created_at": "2025-01-15T08:15:24.044882+00:00"} {"global_step": 963, "acc_step": 0, "speed/wps": 12882.93344905325, "speed/FLOPS": 202344172823082.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3810250759124756, "optim/lr": 0.0005771999999999999, "optim/total_tokens": 504889344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 4.285499572753906, "created_at": "2025-01-15T08:15:34.229127+00:00"} {"global_step": 964, "acc_step": 0, "speed/wps": 12881.737475901607, "speed/FLOPS": 202325388421296.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2698054313659668, "optim/lr": 0.0005778, "optim/total_tokens": 505413632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 4.343491077423096, "created_at": "2025-01-15T08:15:44.407723+00:00"} {"global_step": 965, "acc_step": 0, "speed/wps": 12882.248432597722, "speed/FLOPS": 202333413698337.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27410170435905457, "optim/lr": 0.0005784, "optim/total_tokens": 505937920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 4.358621597290039, "created_at": "2025-01-15T08:15:54.588270+00:00"} {"global_step": 966, "acc_step": 0, "speed/wps": 12880.609992334827, "speed/FLOPS": 202307679742555.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3125874996185303, "optim/lr": 0.000579, "optim/total_tokens": 506462208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.271825790405273, "created_at": "2025-01-15T08:16:04.769314+00:00"} {"global_step": 967, "acc_step": 0, "speed/wps": 12882.697764484026, "speed/FLOPS": 202340471073058.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3407086730003357, "optim/lr": 0.0005796, "optim/total_tokens": 506986496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.182976245880127, "created_at": "2025-01-15T08:16:14.948702+00:00"} {"global_step": 968, "acc_step": 0, "speed/wps": 12877.533921373415, "speed/FLOPS": 202259365821142.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.3298192024230957, "optim/lr": 0.0005802, "optim/total_tokens": 507510784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464770, "loss/out": 4.184420585632324, "created_at": "2025-01-15T08:16:25.133804+00:00"} {"global_step": 969, "acc_step": 0, "speed/wps": 12881.847561765957, "speed/FLOPS": 202327117471070.28, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3626771569252014, "optim/lr": 0.0005808, "optim/total_tokens": 508035072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 4.3506999015808105, "created_at": "2025-01-15T08:16:35.314338+00:00"} {"global_step": 970, "acc_step": 0, "speed/wps": 12877.361801251416, "speed/FLOPS": 202256662438108.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.2835729718208313, "optim/lr": 0.0005814, "optim/total_tokens": 508559360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506128, "loss/out": 4.146026611328125, "created_at": "2025-01-15T08:16:45.496429+00:00"} {"global_step": 971, "acc_step": 0, "speed/wps": 12874.622813584157, "speed/FLOPS": 202213642873030.75, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.34348955750465393, "optim/lr": 0.000582, "optim/total_tokens": 509083648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.122866630554199, "created_at": "2025-01-15T08:16:55.683180+00:00"} {"global_step": 972, "acc_step": 0, "speed/wps": 12873.580068168247, "speed/FLOPS": 202197265123391.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3391687273979187, "optim/lr": 0.0005826000000000001, "optim/total_tokens": 509607936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.23919677734375, "created_at": "2025-01-15T08:17:05.870278+00:00"} {"global_step": 973, "acc_step": 0, "speed/wps": 12878.273073224193, "speed/FLOPS": 202270975216659.88, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29359331727027893, "optim/lr": 0.0005832, "optim/total_tokens": 510132224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 4.28210973739624, "created_at": "2025-01-15T08:17:16.057340+00:00"} {"global_step": 974, "acc_step": 0, "speed/wps": 12882.773792001473, "speed/FLOPS": 202341665189692.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2730683982372284, "optim/lr": 0.0005838, "optim/total_tokens": 510656512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.101686477661133, "created_at": "2025-01-15T08:17:26.236940+00:00"} {"global_step": 975, "acc_step": 0, "speed/wps": 12878.41460241093, "speed/FLOPS": 202273198127019.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33390116691589355, "optim/lr": 0.0005844, "optim/total_tokens": 511180800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334895, "loss/out": 4.185139179229736, "created_at": "2025-01-15T08:17:36.423945+00:00"} {"global_step": 976, "acc_step": 0, "speed/wps": 12882.41516710993, "speed/FLOPS": 202336032492968.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.30399322509765625, "optim/lr": 0.000585, "optim/total_tokens": 511705088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 4.175114154815674, "created_at": "2025-01-15T08:17:46.605421+00:00"} {"global_step": 977, "acc_step": 0, "speed/wps": 12876.02439533288, "speed/FLOPS": 202235656640369.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30100512504577637, "optim/lr": 0.0005856, "optim/total_tokens": 512229376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.245782852172852, "created_at": "2025-01-15T08:17:56.792448+00:00"} {"global_step": 978, "acc_step": 0, "speed/wps": 12880.800031271707, "speed/FLOPS": 202310664565200.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2821120321750641, "optim/lr": 0.0005861999999999999, "optim/total_tokens": 512753664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490582, "loss/out": 4.134413719177246, "created_at": "2025-01-15T08:18:06.980467+00:00"} {"global_step": 979, "acc_step": 0, "speed/wps": 12876.165571162166, "speed/FLOPS": 202237874000764.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28592655062675476, "optim/lr": 0.0005868, "optim/total_tokens": 513277952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.31035041809082, "created_at": "2025-01-15T08:18:17.163561+00:00"} {"global_step": 980, "acc_step": 0, "speed/wps": 12879.676566112592, "speed/FLOPS": 202293019001073.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3043511211872101, "optim/lr": 0.0005874, "optim/total_tokens": 513802240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.14407205581665, "created_at": "2025-01-15T08:18:27.346471+00:00"} {"global_step": 981, "acc_step": 0, "speed/wps": 12881.452448257618, "speed/FLOPS": 202320911670478.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2989449203014374, "optim/lr": 0.000588, "optim/total_tokens": 514326528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 4.238329887390137, "created_at": "2025-01-15T08:18:37.526976+00:00"} {"global_step": 982, "acc_step": 0, "speed/wps": 12876.587216907694, "speed/FLOPS": 202244496526599.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2708078622817993, "optim/lr": 0.0005886000000000001, "optim/total_tokens": 514850816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 278077, "loss/out": 4.3259453773498535, "created_at": "2025-01-15T08:18:47.713957+00:00"} {"global_step": 983, "acc_step": 0, "speed/wps": 12881.919847389703, "speed/FLOPS": 202328252816122.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34560301899909973, "optim/lr": 0.0005892, "optim/total_tokens": 515375104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 4.281136512756348, "created_at": "2025-01-15T08:18:57.893379+00:00"} {"global_step": 984, "acc_step": 0, "speed/wps": 12879.666497738446, "speed/FLOPS": 202292860863422.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.44568294286727905, "optim/lr": 0.0005898, "optim/total_tokens": 515899392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.192263126373291, "created_at": "2025-01-15T08:19:08.077500+00:00"} {"global_step": 985, "acc_step": 0, "speed/wps": 12880.224591354496, "speed/FLOPS": 202301626490563.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4119245409965515, "optim/lr": 0.0005904, "optim/total_tokens": 516423680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.207711696624756, "created_at": "2025-01-15T08:19:18.259765+00:00"} {"global_step": 986, "acc_step": 0, "speed/wps": 12879.478127041782, "speed/FLOPS": 202289902242782.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30123576521873474, "optim/lr": 0.000591, "optim/total_tokens": 516947968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 4.041093349456787, "created_at": "2025-01-15T08:19:28.447436+00:00"} {"global_step": 987, "acc_step": 0, "speed/wps": 12884.271932111662, "speed/FLOPS": 202365195538783.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35892313718795776, "optim/lr": 0.0005916, "optim/total_tokens": 517472256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.156414985656738, "created_at": "2025-01-15T08:19:38.629972+00:00"} {"global_step": 988, "acc_step": 0, "speed/wps": 12884.308009398452, "speed/FLOPS": 202365762182147.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25174465775489807, "optim/lr": 0.0005922, "optim/total_tokens": 517996544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302164, "loss/out": 4.146512508392334, "created_at": "2025-01-15T08:19:48.806713+00:00"} {"global_step": 989, "acc_step": 0, "speed/wps": 12879.187784871214, "speed/FLOPS": 202285342020021.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23644958436489105, "optim/lr": 0.0005928, "optim/total_tokens": 518520832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.0967512130737305, "created_at": "2025-01-15T08:19:58.987673+00:00"} {"global_step": 990, "acc_step": 0, "speed/wps": 12883.4552589566, "speed/FLOPS": 202352368564656.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23571418225765228, "optim/lr": 0.0005934, "optim/total_tokens": 519045120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.151639461517334, "created_at": "2025-01-15T08:20:09.167044+00:00"} {"global_step": 991, "acc_step": 0, "speed/wps": 12880.890226958567, "speed/FLOPS": 202312081212404.3, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24309684336185455, "optim/lr": 0.000594, "optim/total_tokens": 519569408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.2029571533203125, "created_at": "2025-01-15T08:20:19.349947+00:00"} {"global_step": 992, "acc_step": 0, "speed/wps": 12870.783517534412, "speed/FLOPS": 202153341452823.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31011077761650085, "optim/lr": 0.0005945999999999999, "optim/total_tokens": 520093696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.159106254577637, "created_at": "2025-01-15T08:20:29.537280+00:00"} {"global_step": 993, "acc_step": 0, "speed/wps": 12871.391191525716, "speed/FLOPS": 202162885807888.84, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30914002656936646, "optim/lr": 0.0005951999999999999, "optim/total_tokens": 520617984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.084331035614014, "created_at": "2025-01-15T08:20:39.727534+00:00"} {"global_step": 994, "acc_step": 0, "speed/wps": 12874.081865738786, "speed/FLOPS": 202205146543778.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.24433322250843048, "optim/lr": 0.0005958000000000001, "optim/total_tokens": 521142272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400828, "loss/out": 4.165831565856934, "created_at": "2025-01-15T08:20:49.915674+00:00"} {"global_step": 995, "acc_step": 0, "speed/wps": 12882.45894478812, "speed/FLOPS": 202336720081561.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3100040853023529, "optim/lr": 0.0005964000000000001, "optim/total_tokens": 521666560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 4.177938461303711, "created_at": "2025-01-15T08:21:00.094651+00:00"} {"global_step": 996, "acc_step": 0, "speed/wps": 12882.502430320086, "speed/FLOPS": 202337403081596.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27289050817489624, "optim/lr": 0.0005970000000000001, "optim/total_tokens": 522190848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 4.216307640075684, "created_at": "2025-01-15T08:21:10.277375+00:00"} {"global_step": 997, "acc_step": 0, "speed/wps": 12881.173899849102, "speed/FLOPS": 202316536684957.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33524203300476074, "optim/lr": 0.0005976, "optim/total_tokens": 522715136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 4.274397373199463, "created_at": "2025-01-15T08:21:20.461966+00:00"} {"global_step": 998, "acc_step": 0, "speed/wps": 12884.728145493913, "speed/FLOPS": 202372360996854.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3022649884223938, "optim/lr": 0.0005982, "optim/total_tokens": 523239424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475383, "loss/out": 4.140388488769531, "created_at": "2025-01-15T08:21:30.641266+00:00"} {"global_step": 999, "acc_step": 0, "speed/wps": 12881.795555906103, "speed/FLOPS": 202326300647577.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2592224180698395, "optim/lr": 0.0005988, "optim/total_tokens": 523763712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369048, "loss/out": 4.240930080413818, "created_at": "2025-01-15T08:21:40.823306+00:00"} {"global_step": 1000, "acc_step": 0, "speed/wps": 12878.542900320404, "speed/FLOPS": 202275213221987.16, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3319644629955292, "optim/lr": 0.0005994, "optim/total_tokens": 524288000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.096593856811523, "created_at": "2025-01-15T08:21:51.007641+00:00"} {"global_step": 1001, "acc_step": 0, "speed/wps": 12102.750758139826, "speed/FLOPS": 190090331578923.75, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.6614, "optim/grad_norm": 0.2921419143676758, "optim/lr": 0.0006000000000000001, "optim/total_tokens": 524812288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.116393566131592, "created_at": "2025-01-15T08:22:01.844943+00:00"} {"global_step": 1002, "acc_step": 0, "speed/wps": 12892.14439082801, "speed/FLOPS": 202488843320815.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2290591448545456, "optim/lr": 0.0006006, "optim/total_tokens": 525336576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 4.235013008117676, "created_at": "2025-01-15T08:22:12.015274+00:00"} {"global_step": 1003, "acc_step": 0, "speed/wps": 12883.48388146935, "speed/FLOPS": 202352818120550.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27964094281196594, "optim/lr": 0.0006012, "optim/total_tokens": 525860864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 4.125789165496826, "created_at": "2025-01-15T08:22:22.194575+00:00"} {"global_step": 1004, "acc_step": 0, "speed/wps": 12879.314969063425, "speed/FLOPS": 202287339622529.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3073536157608032, "optim/lr": 0.0006018, "optim/total_tokens": 526385152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 4.206882953643799, "created_at": "2025-01-15T08:22:32.375987+00:00"} {"global_step": 1005, "acc_step": 0, "speed/wps": 12878.18178526122, "speed/FLOPS": 202269541413758.6, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34116435050964355, "optim/lr": 0.0006024, "optim/total_tokens": 526909440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.121388912200928, "created_at": "2025-01-15T08:22:42.563706+00:00"} {"global_step": 1006, "acc_step": 0, "speed/wps": 12875.161817476273, "speed/FLOPS": 202222108669825.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.32519441843032837, "optim/lr": 0.000603, "optim/total_tokens": 527433728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471773, "loss/out": 4.156584739685059, "created_at": "2025-01-15T08:22:52.748450+00:00"} {"global_step": 1007, "acc_step": 0, "speed/wps": 12876.697822685652, "speed/FLOPS": 202246233742331.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.35135725140571594, "optim/lr": 0.0006036, "optim/total_tokens": 527958016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 4.288397789001465, "created_at": "2025-01-15T08:23:02.933120+00:00"} {"global_step": 1008, "acc_step": 0, "speed/wps": 12873.680318333938, "speed/FLOPS": 202198839690001.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2909160554409027, "optim/lr": 0.0006042, "optim/total_tokens": 528482304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 4.253291130065918, "created_at": "2025-01-15T08:23:13.120937+00:00"} {"global_step": 1009, "acc_step": 0, "speed/wps": 12880.031789888544, "speed/FLOPS": 202298598278603.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30360013246536255, "optim/lr": 0.0006048000000000001, "optim/total_tokens": 529006592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 4.1906843185424805, "created_at": "2025-01-15T08:23:23.304601+00:00"} {"global_step": 1010, "acc_step": 0, "speed/wps": 12880.351225478482, "speed/FLOPS": 202303615453493.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2699987590312958, "optim/lr": 0.0006054000000000001, "optim/total_tokens": 529530880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 4.158416748046875, "created_at": "2025-01-15T08:23:33.486701+00:00"} {"global_step": 1011, "acc_step": 0, "speed/wps": 12883.743762579927, "speed/FLOPS": 202356899910506.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27448514103889465, "optim/lr": 0.0006060000000000001, "optim/total_tokens": 530055168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.172125339508057, "created_at": "2025-01-15T08:23:43.663575+00:00"} {"global_step": 1012, "acc_step": 0, "speed/wps": 12879.302403752146, "speed/FLOPS": 202287142267049.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2681412994861603, "optim/lr": 0.0006066, "optim/total_tokens": 530579456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416139, "loss/out": 4.118644714355469, "created_at": "2025-01-15T08:23:53.844101+00:00"} {"global_step": 1013, "acc_step": 0, "speed/wps": 12880.023599894097, "speed/FLOPS": 202298469643486.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24713118374347687, "optim/lr": 0.0006072, "optim/total_tokens": 531103744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 4.168450355529785, "created_at": "2025-01-15T08:24:04.024991+00:00"} {"global_step": 1014, "acc_step": 0, "speed/wps": 12877.530461887685, "speed/FLOPS": 202259311485165.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2562180757522583, "optim/lr": 0.0006078, "optim/total_tokens": 531628032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.1656270027160645, "created_at": "2025-01-15T08:24:14.207194+00:00"} {"global_step": 1015, "acc_step": 0, "speed/wps": 12876.282976583538, "speed/FLOPS": 202239718014239.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30046266317367554, "optim/lr": 0.0006084, "optim/total_tokens": 532152320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 4.142251014709473, "created_at": "2025-01-15T08:24:24.399902+00:00"} {"global_step": 1016, "acc_step": 0, "speed/wps": 12875.48448289192, "speed/FLOPS": 202227176573566.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29014211893081665, "optim/lr": 0.0006090000000000001, "optim/total_tokens": 532676608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 4.117561340332031, "created_at": "2025-01-15T08:24:34.583334+00:00"} {"global_step": 1017, "acc_step": 0, "speed/wps": 12879.549095434379, "speed/FLOPS": 202291016898909.12, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3643970787525177, "optim/lr": 0.0006096, "optim/total_tokens": 533200896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.189563751220703, "created_at": "2025-01-15T08:24:44.767164+00:00"} {"global_step": 1018, "acc_step": 0, "speed/wps": 12881.007539357233, "speed/FLOPS": 202313923764829.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2871223986148834, "optim/lr": 0.0006102, "optim/total_tokens": 533725184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396093, "loss/out": 4.373337745666504, "created_at": "2025-01-15T08:24:54.947167+00:00"} {"global_step": 1019, "acc_step": 0, "speed/wps": 12879.695649154777, "speed/FLOPS": 202293318726473.88, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.28226712346076965, "optim/lr": 0.0006108, "optim/total_tokens": 534249472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 4.132757663726807, "created_at": "2025-01-15T08:25:05.129518+00:00"} {"global_step": 1020, "acc_step": 0, "speed/wps": 12876.07028302799, "speed/FLOPS": 202236377369677.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2944662272930145, "optim/lr": 0.0006114, "optim/total_tokens": 534773760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 4.188998699188232, "created_at": "2025-01-15T08:25:15.317129+00:00"} {"global_step": 1021, "acc_step": 0, "speed/wps": 12878.587145472775, "speed/FLOPS": 202275908152904.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.32520103454589844, "optim/lr": 0.000612, "optim/total_tokens": 535298048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 4.273441791534424, "created_at": "2025-01-15T08:25:25.500393+00:00"} {"global_step": 1022, "acc_step": 0, "speed/wps": 12882.123435543703, "speed/FLOPS": 202331450447840.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28848910331726074, "optim/lr": 0.0006126, "optim/total_tokens": 535822336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 4.342311859130859, "created_at": "2025-01-15T08:25:35.678807+00:00"} {"global_step": 1023, "acc_step": 0, "speed/wps": 12878.696188727123, "speed/FLOPS": 202277620827055.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3093132972717285, "optim/lr": 0.0006132, "optim/total_tokens": 536346624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.0448102951049805, "created_at": "2025-01-15T08:25:45.862428+00:00"} {"global_step": 1024, "acc_step": 0, "speed/wps": 12878.494420031328, "speed/FLOPS": 202274451772428.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3158286213874817, "optim/lr": 0.0006138000000000001, "optim/total_tokens": 536870912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 4.104498863220215, "created_at": "2025-01-15T08:25:56.045808+00:00"} {"global_step": 1025, "acc_step": 0, "speed/wps": 12881.185542143778, "speed/FLOPS": 202316719543193.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.262698769569397, "optim/lr": 0.0006144000000000001, "optim/total_tokens": 537395200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 4.023266792297363, "created_at": "2025-01-15T08:26:06.224723+00:00"} {"global_step": 1026, "acc_step": 0, "speed/wps": 12880.698727087389, "speed/FLOPS": 202309073443779.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2568911015987396, "optim/lr": 0.000615, "optim/total_tokens": 537919488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.21730899810791, "created_at": "2025-01-15T08:26:16.406174+00:00"} {"global_step": 1027, "acc_step": 0, "speed/wps": 12881.289226106272, "speed/FLOPS": 202318348042301.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22575753927230835, "optim/lr": 0.0006156, "optim/total_tokens": 538443776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.135952949523926, "created_at": "2025-01-15T08:26:26.594146+00:00"} {"global_step": 1028, "acc_step": 0, "speed/wps": 12881.058435362569, "speed/FLOPS": 202314723156531.75, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2509186863899231, "optim/lr": 0.0006162, "optim/total_tokens": 538968064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.006354331970215, "created_at": "2025-01-15T08:26:36.773261+00:00"} {"global_step": 1029, "acc_step": 0, "speed/wps": 12876.84963569111, "speed/FLOPS": 202248618174194.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23836162686347961, "optim/lr": 0.0006168, "optim/total_tokens": 539492352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372460, "loss/out": 4.185145854949951, "created_at": "2025-01-15T08:26:46.965077+00:00"} {"global_step": 1030, "acc_step": 0, "speed/wps": 12882.022822730903, "speed/FLOPS": 202329870185358.56, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.277936190366745, "optim/lr": 0.0006174, "optim/total_tokens": 540016640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 4.122560024261475, "created_at": "2025-01-15T08:26:57.143378+00:00"} {"global_step": 1031, "acc_step": 0, "speed/wps": 12878.559520500605, "speed/FLOPS": 202275474264755.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.26880574226379395, "optim/lr": 0.000618, "optim/total_tokens": 540540928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 4.111886978149414, "created_at": "2025-01-15T08:27:07.328169+00:00"} {"global_step": 1032, "acc_step": 0, "speed/wps": 12875.373627039195, "speed/FLOPS": 202225435430067.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21930649876594543, "optim/lr": 0.0006186, "optim/total_tokens": 541065216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.128634929656982, "created_at": "2025-01-15T08:27:17.514347+00:00"} {"global_step": 1033, "acc_step": 0, "speed/wps": 12878.539654776501, "speed/FLOPS": 202275162246260.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21259663999080658, "optim/lr": 0.0006192, "optim/total_tokens": 541589504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 4.149994850158691, "created_at": "2025-01-15T08:27:27.699102+00:00"} {"global_step": 1034, "acc_step": 0, "speed/wps": 12880.686869938312, "speed/FLOPS": 202308887210959.4, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20118306577205658, "optim/lr": 0.0006198, "optim/total_tokens": 542113792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.189447402954102, "created_at": "2025-01-15T08:27:37.878469+00:00"} {"global_step": 1035, "acc_step": 0, "speed/wps": 12875.187374846415, "speed/FLOPS": 202222510083442.97, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2122374325990677, "optim/lr": 0.0006204, "optim/total_tokens": 542638080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 4.104849815368652, "created_at": "2025-01-15T08:27:48.064814+00:00"} {"global_step": 1036, "acc_step": 0, "speed/wps": 12875.659961837187, "speed/FLOPS": 202229932711535.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2204081118106842, "optim/lr": 0.000621, "optim/total_tokens": 543162368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 4.125048637390137, "created_at": "2025-01-15T08:27:58.249955+00:00"} {"global_step": 1037, "acc_step": 0, "speed/wps": 12882.928473640804, "speed/FLOPS": 202344094677393.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22772814333438873, "optim/lr": 0.0006216, "optim/total_tokens": 543686656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 4.114144325256348, "created_at": "2025-01-15T08:28:08.427748+00:00"} {"global_step": 1038, "acc_step": 0, "speed/wps": 12880.05137024132, "speed/FLOPS": 202298905814950.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2551986873149872, "optim/lr": 0.0006222, "optim/total_tokens": 544210944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.072171211242676, "created_at": "2025-01-15T08:28:18.609587+00:00"} {"global_step": 1039, "acc_step": 0, "speed/wps": 12879.452958090204, "speed/FLOPS": 202289506929812.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49237796664237976, "optim/lr": 0.0006228000000000001, "optim/total_tokens": 544735232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 4.115813255310059, "created_at": "2025-01-15T08:28:28.793511+00:00"} {"global_step": 1040, "acc_step": 0, "speed/wps": 12881.51224942978, "speed/FLOPS": 202321850930063.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5306646227836609, "optim/lr": 0.0006234000000000001, "optim/total_tokens": 545259520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 4.176412105560303, "created_at": "2025-01-15T08:28:38.974431+00:00"} {"global_step": 1041, "acc_step": 0, "speed/wps": 12882.653821268828, "speed/FLOPS": 202339780884479.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5138390064239502, "optim/lr": 0.000624, "optim/total_tokens": 545783808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 4.2451171875, "created_at": "2025-01-15T08:28:49.154429+00:00"} {"global_step": 1042, "acc_step": 0, "speed/wps": 12881.759370064581, "speed/FLOPS": 202325732299212.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5365040302276611, "optim/lr": 0.0006246, "optim/total_tokens": 546308096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 4.074606895446777, "created_at": "2025-01-15T08:28:59.333104+00:00"} {"global_step": 1043, "acc_step": 0, "speed/wps": 12881.241525664282, "speed/FLOPS": 202317598841312.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.2190349102020264, "optim/lr": 0.0006252, "optim/total_tokens": 546832384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 4.332779407501221, "created_at": "2025-01-15T08:29:09.512148+00:00"} {"global_step": 1044, "acc_step": 0, "speed/wps": 12878.632900587989, "speed/FLOPS": 202276626799862.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0340890884399414, "optim/lr": 0.0006258, "optim/total_tokens": 547356672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.4052653312683105, "created_at": "2025-01-15T08:29:19.695906+00:00"} {"global_step": 1045, "acc_step": 0, "speed/wps": 12884.683540821448, "speed/FLOPS": 202371660419180.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8301153779029846, "optim/lr": 0.0006264, "optim/total_tokens": 547880960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.398211479187012, "created_at": "2025-01-15T08:29:29.874320+00:00"} {"global_step": 1046, "acc_step": 0, "speed/wps": 12881.574337312988, "speed/FLOPS": 202322826105586.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.671432614326477, "optim/lr": 0.000627, "optim/total_tokens": 548405248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 4.4859232902526855, "created_at": "2025-01-15T08:29:40.053417+00:00"} {"global_step": 1047, "acc_step": 0, "speed/wps": 12882.407113228744, "speed/FLOPS": 202335905995697.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5980238914489746, "optim/lr": 0.0006276, "optim/total_tokens": 548929536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406724, "loss/out": 4.35617208480835, "created_at": "2025-01-15T08:29:50.233938+00:00"} {"global_step": 1048, "acc_step": 0, "speed/wps": 12883.927537834461, "speed/FLOPS": 202359786353412.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5106629133224487, "optim/lr": 0.0006282, "optim/total_tokens": 549453824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.459864616394043, "created_at": "2025-01-15T08:30:00.411632+00:00"} {"global_step": 1049, "acc_step": 0, "speed/wps": 12883.852549952817, "speed/FLOPS": 202358608565687.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6646761894226074, "optim/lr": 0.0006288, "optim/total_tokens": 549978112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 4.272031307220459, "created_at": "2025-01-15T08:30:10.592498+00:00"} {"global_step": 1050, "acc_step": 0, "speed/wps": 12883.51358207513, "speed/FLOPS": 202353284609377.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0481029748916626, "optim/lr": 0.0006294, "optim/total_tokens": 550502400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.388304233551025, "created_at": "2025-01-15T08:30:20.772847+00:00"} {"global_step": 1051, "acc_step": 0, "speed/wps": 12886.772609040647, "speed/FLOPS": 202404472106243.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.057213544845581, "optim/lr": 0.00063, "optim/total_tokens": 551026688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.582295894622803, "created_at": "2025-01-15T08:30:30.953882+00:00"} {"global_step": 1052, "acc_step": 0, "speed/wps": 12886.600874322596, "speed/FLOPS": 202401774776508.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0172624588012695, "optim/lr": 0.0006306, "optim/total_tokens": 551550976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 4.582754611968994, "created_at": "2025-01-15T08:30:41.128650+00:00"} {"global_step": 1053, "acc_step": 0, "speed/wps": 12879.476518177788, "speed/FLOPS": 202289876973362.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7509729266166687, "optim/lr": 0.0006312000000000001, "optim/total_tokens": 552075264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332063, "loss/out": 4.511870861053467, "created_at": "2025-01-15T08:30:51.309367+00:00"} {"global_step": 1054, "acc_step": 0, "speed/wps": 12883.839913275038, "speed/FLOPS": 202358410089298.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9645535349845886, "optim/lr": 0.0006318000000000001, "optim/total_tokens": 552599552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 4.51434326171875, "created_at": "2025-01-15T08:31:01.488437+00:00"} {"global_step": 1055, "acc_step": 0, "speed/wps": 12882.572315322828, "speed/FLOPS": 202338500721596.7, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.5968338847160339, "optim/lr": 0.0006324, "optim/total_tokens": 553123840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 4.432780742645264, "created_at": "2025-01-15T08:31:11.669788+00:00"} {"global_step": 1056, "acc_step": 0, "speed/wps": 12881.227315459777, "speed/FLOPS": 202317375650523.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5308127403259277, "optim/lr": 0.000633, "optim/total_tokens": 553648128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464555, "loss/out": 4.331598281860352, "created_at": "2025-01-15T08:31:21.851788+00:00"} {"global_step": 1057, "acc_step": 0, "speed/wps": 12883.828075110589, "speed/FLOPS": 202358224154658.25, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48040416836738586, "optim/lr": 0.0006336, "optim/total_tokens": 554172416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 4.327575206756592, "created_at": "2025-01-15T08:31:32.028626+00:00"} {"global_step": 1058, "acc_step": 0, "speed/wps": 12875.421706902918, "speed/FLOPS": 202226190590396.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.368143230676651, "optim/lr": 0.0006342, "optim/total_tokens": 554696704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470104, "loss/out": 4.28769588470459, "created_at": "2025-01-15T08:31:42.214545+00:00"} {"global_step": 1059, "acc_step": 0, "speed/wps": 12876.600245816637, "speed/FLOPS": 202244701163518.47, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.371880441904068, "optim/lr": 0.0006348, "optim/total_tokens": 555220992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.180294990539551, "created_at": "2025-01-15T08:31:52.397546+00:00"} {"global_step": 1060, "acc_step": 0, "speed/wps": 12880.230577261615, "speed/FLOPS": 202301720507460.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.39618679881095886, "optim/lr": 0.0006353999999999999, "optim/total_tokens": 555745280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.315077781677246, "created_at": "2025-01-15T08:32:02.580609+00:00"} {"global_step": 1061, "acc_step": 0, "speed/wps": 12880.248303972327, "speed/FLOPS": 202301998929810.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3726499080657959, "optim/lr": 0.000636, "optim/total_tokens": 556269568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474863, "loss/out": 4.163424015045166, "created_at": "2025-01-15T08:32:12.764894+00:00"} {"global_step": 1062, "acc_step": 0, "speed/wps": 12882.387232095165, "speed/FLOPS": 202335593735175.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6291183829307556, "optim/lr": 0.0006366, "optim/total_tokens": 556793856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 4.29205322265625, "created_at": "2025-01-15T08:32:22.944322+00:00"} {"global_step": 1063, "acc_step": 0, "speed/wps": 12878.681459701289, "speed/FLOPS": 202277389487464.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5217145085334778, "optim/lr": 0.0006372, "optim/total_tokens": 557318144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 4.273508548736572, "created_at": "2025-01-15T08:32:33.128857+00:00"} {"global_step": 1064, "acc_step": 0, "speed/wps": 12876.811606701458, "speed/FLOPS": 202248020876654.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.7762742638587952, "optim/lr": 0.0006378, "optim/total_tokens": 557842432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 4.298902988433838, "created_at": "2025-01-15T08:32:43.314521+00:00"} {"global_step": 1065, "acc_step": 0, "speed/wps": 12879.11757593588, "speed/FLOPS": 202284239292215.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6288031935691833, "optim/lr": 0.0006384, "optim/total_tokens": 558366720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.366055011749268, "created_at": "2025-01-15T08:32:53.495446+00:00"} {"global_step": 1066, "acc_step": 0, "speed/wps": 12885.067463948104, "speed/FLOPS": 202377690459450.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6044327616691589, "optim/lr": 0.000639, "optim/total_tokens": 558891008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.278755187988281, "created_at": "2025-01-15T08:33:03.671879+00:00"} {"global_step": 1067, "acc_step": 0, "speed/wps": 12883.87891084541, "speed/FLOPS": 202359022599729.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42802536487579346, "optim/lr": 0.0006396, "optim/total_tokens": 559415296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.313035011291504, "created_at": "2025-01-15T08:33:13.849592+00:00"} {"global_step": 1068, "acc_step": 0, "speed/wps": 12884.024699734251, "speed/FLOPS": 202361312414563.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.402462899684906, "optim/lr": 0.0006402000000000001, "optim/total_tokens": 559939584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 4.266408920288086, "created_at": "2025-01-15T08:33:24.030018+00:00"} {"global_step": 1069, "acc_step": 0, "speed/wps": 12881.984088726254, "speed/FLOPS": 202329261814589.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41390034556388855, "optim/lr": 0.0006408000000000001, "optim/total_tokens": 560463872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 4.336125373840332, "created_at": "2025-01-15T08:33:34.208915+00:00"} {"global_step": 1070, "acc_step": 0, "speed/wps": 12879.281806170822, "speed/FLOPS": 202286818753731.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33214789628982544, "optim/lr": 0.0006414, "optim/total_tokens": 560988160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 4.2214131355285645, "created_at": "2025-01-15T08:33:44.394327+00:00"} {"global_step": 1071, "acc_step": 0, "speed/wps": 12883.778497062309, "speed/FLOPS": 202357445463282.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27345937490463257, "optim/lr": 0.000642, "optim/total_tokens": 561512448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 4.160396099090576, "created_at": "2025-01-15T08:33:54.571207+00:00"} {"global_step": 1072, "acc_step": 0, "speed/wps": 12879.579685479659, "speed/FLOPS": 202291497357605.38, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22674357891082764, "optim/lr": 0.0006426, "optim/total_tokens": 562036736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 4.290383338928223, "created_at": "2025-01-15T08:34:04.751735+00:00"} {"global_step": 1073, "acc_step": 0, "speed/wps": 12878.062438565936, "speed/FLOPS": 202267666909909.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24132589995861053, "optim/lr": 0.0006432, "optim/total_tokens": 562561024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.132595539093018, "created_at": "2025-01-15T08:34:14.936477+00:00"} {"global_step": 1074, "acc_step": 0, "speed/wps": 12883.730358304296, "speed/FLOPS": 202356689377938.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2050613909959793, "optim/lr": 0.0006438, "optim/total_tokens": 563085312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 4.080406665802002, "created_at": "2025-01-15T08:34:25.115430+00:00"} {"global_step": 1075, "acc_step": 0, "speed/wps": 12884.734966038317, "speed/FLOPS": 202372468122876.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2280307561159134, "optim/lr": 0.0006443999999999999, "optim/total_tokens": 563609600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.175434112548828, "created_at": "2025-01-15T08:34:35.292815+00:00"} {"global_step": 1076, "acc_step": 0, "speed/wps": 12881.031160542836, "speed/FLOPS": 202314294768008.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1938755363225937, "optim/lr": 0.000645, "optim/total_tokens": 564133888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 4.135444641113281, "created_at": "2025-01-15T08:34:45.474814+00:00"} {"global_step": 1077, "acc_step": 0, "speed/wps": 12879.11379620175, "speed/FLOPS": 202284179926296.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21441777050495148, "optim/lr": 0.0006456, "optim/total_tokens": 564658176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 4.037659168243408, "created_at": "2025-01-15T08:34:55.655594+00:00"} {"global_step": 1078, "acc_step": 0, "speed/wps": 12884.206871778038, "speed/FLOPS": 202364173676841.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18638074398040771, "optim/lr": 0.0006462, "optim/total_tokens": 565182464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399017, "loss/out": 4.082304954528809, "created_at": "2025-01-15T08:35:05.832840+00:00"} {"global_step": 1079, "acc_step": 0, "speed/wps": 12879.431599800237, "speed/FLOPS": 202289171468519.0, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20593859255313873, "optim/lr": 0.0006468000000000001, "optim/total_tokens": 565706752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285262, "loss/out": 4.061671257019043, "created_at": "2025-01-15T08:35:16.013671+00:00"} {"global_step": 1080, "acc_step": 0, "speed/wps": 12873.0290168884, "speed/FLOPS": 202188610105817.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20137667655944824, "optim/lr": 0.0006474, "optim/total_tokens": 566231040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 4.157942771911621, "created_at": "2025-01-15T08:35:26.202713+00:00"} {"global_step": 1081, "acc_step": 0, "speed/wps": 12881.81330461099, "speed/FLOPS": 202326579415377.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21262496709823608, "optim/lr": 0.000648, "optim/total_tokens": 566755328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.077033519744873, "created_at": "2025-01-15T08:35:36.382833+00:00"} {"global_step": 1082, "acc_step": 0, "speed/wps": 12879.243719205406, "speed/FLOPS": 202286220545602.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18253150582313538, "optim/lr": 0.0006486, "optim/total_tokens": 567279616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 4.009934425354004, "created_at": "2025-01-15T08:35:46.568490+00:00"} {"global_step": 1083, "acc_step": 0, "speed/wps": 12882.937512879585, "speed/FLOPS": 202344236651059.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1799621284008026, "optim/lr": 0.0006492000000000001, "optim/total_tokens": 567803904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.85746431350708, "created_at": "2025-01-15T08:35:56.746167+00:00"} {"global_step": 1084, "acc_step": 0, "speed/wps": 12883.490763204163, "speed/FLOPS": 202352926207652.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20063507556915283, "optim/lr": 0.0006498, "optim/total_tokens": 568328192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 4.080227851867676, "created_at": "2025-01-15T08:36:06.925699+00:00"} {"global_step": 1085, "acc_step": 0, "speed/wps": 12880.876828619925, "speed/FLOPS": 202311870773084.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22203350067138672, "optim/lr": 0.0006504, "optim/total_tokens": 568852480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.050049781799316, "created_at": "2025-01-15T08:36:17.107281+00:00"} {"global_step": 1086, "acc_step": 0, "speed/wps": 12878.24771336283, "speed/FLOPS": 202270576905189.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2157248854637146, "optim/lr": 0.000651, "optim/total_tokens": 569376768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.032010078430176, "created_at": "2025-01-15T08:36:27.292514+00:00"} {"global_step": 1087, "acc_step": 0, "speed/wps": 12870.813331464253, "speed/FLOPS": 202153809721560.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23575229942798615, "optim/lr": 0.0006516, "optim/total_tokens": 569901056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 4.177951812744141, "created_at": "2025-01-15T08:36:37.481425+00:00"} {"global_step": 1088, "acc_step": 0, "speed/wps": 12882.662338898996, "speed/FLOPS": 202339914665565.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2680341303348541, "optim/lr": 0.0006522, "optim/total_tokens": 570425344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500338, "loss/out": 4.0247979164123535, "created_at": "2025-01-15T08:36:47.662629+00:00"} {"global_step": 1089, "acc_step": 0, "speed/wps": 12881.553745433508, "speed/FLOPS": 202322502681823.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.24557240307331085, "optim/lr": 0.0006527999999999999, "optim/total_tokens": 570949632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378926, "loss/out": 4.047145843505859, "created_at": "2025-01-15T08:36:57.843767+00:00"} {"global_step": 1090, "acc_step": 0, "speed/wps": 12878.059663443973, "speed/FLOPS": 202267623322805.78, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22204767167568207, "optim/lr": 0.0006533999999999999, "optim/total_tokens": 571473920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 4.170636177062988, "created_at": "2025-01-15T08:37:08.033425+00:00"} {"global_step": 1091, "acc_step": 0, "speed/wps": 12882.180384686475, "speed/FLOPS": 202332344912383.62, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21763953566551208, "optim/lr": 0.000654, "optim/total_tokens": 571998208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 4.118115425109863, "created_at": "2025-01-15T08:37:18.214219+00:00"} {"global_step": 1092, "acc_step": 0, "speed/wps": 12884.31390355883, "speed/FLOPS": 202365854758036.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23620925843715668, "optim/lr": 0.0006546000000000001, "optim/total_tokens": 572522496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.1229119300842285, "created_at": "2025-01-15T08:37:28.396225+00:00"} {"global_step": 1093, "acc_step": 0, "speed/wps": 12882.909311439967, "speed/FLOPS": 202343793708697.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27790674567222595, "optim/lr": 0.0006552000000000001, "optim/total_tokens": 573046784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 4.0036091804504395, "created_at": "2025-01-15T08:37:38.574212+00:00"} {"global_step": 1094, "acc_step": 0, "speed/wps": 12880.435550196486, "speed/FLOPS": 202304939889064.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.25220930576324463, "optim/lr": 0.0006558, "optim/total_tokens": 573571072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.080329418182373, "created_at": "2025-01-15T08:37:48.755652+00:00"} {"global_step": 1095, "acc_step": 0, "speed/wps": 12874.887983772544, "speed/FLOPS": 202217807735221.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23787783086299896, "optim/lr": 0.0006564, "optim/total_tokens": 574095360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366413, "loss/out": 4.090814590454102, "created_at": "2025-01-15T08:37:58.941700+00:00"} {"global_step": 1096, "acc_step": 0, "speed/wps": 12881.697539791176, "speed/FLOPS": 202324761169804.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23024994134902954, "optim/lr": 0.000657, "optim/total_tokens": 574619648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331823, "loss/out": 4.098565101623535, "created_at": "2025-01-15T08:38:09.126587+00:00"} {"global_step": 1097, "acc_step": 0, "speed/wps": 12877.43147142349, "speed/FLOPS": 202257756703897.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.227910116314888, "optim/lr": 0.0006576, "optim/total_tokens": 575143936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486334, "loss/out": 4.014885902404785, "created_at": "2025-01-15T08:38:19.328444+00:00"} {"global_step": 1098, "acc_step": 0, "speed/wps": 12882.135024135217, "speed/FLOPS": 202331632462594.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.25077980756759644, "optim/lr": 0.0006582000000000001, "optim/total_tokens": 575668224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.150169372558594, "created_at": "2025-01-15T08:38:29.507463+00:00"} {"global_step": 1099, "acc_step": 0, "speed/wps": 12877.588598226626, "speed/FLOPS": 202260224596255.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25018924474716187, "optim/lr": 0.0006588, "optim/total_tokens": 576192512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 4.172697067260742, "created_at": "2025-01-15T08:38:39.689631+00:00"} {"global_step": 1100, "acc_step": 0, "speed/wps": 12879.040792492471, "speed/FLOPS": 202283033302725.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2884403169155121, "optim/lr": 0.0006594, "optim/total_tokens": 576716800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.9555890560150146, "created_at": "2025-01-15T08:38:49.871477+00:00"} {"global_step": 1101, "acc_step": 0, "speed/wps": 12880.448117718905, "speed/FLOPS": 202305137279273.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27711501717567444, "optim/lr": 0.00066, "optim/total_tokens": 577241088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.066589832305908, "created_at": "2025-01-15T08:39:00.058109+00:00"} {"global_step": 1102, "acc_step": 0, "speed/wps": 12880.763194381021, "speed/FLOPS": 202310085991212.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23721100389957428, "optim/lr": 0.0006606, "optim/total_tokens": 577765376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285796, "loss/out": 3.970487356185913, "created_at": "2025-01-15T08:39:10.237879+00:00"} {"global_step": 1103, "acc_step": 0, "speed/wps": 12873.596352884912, "speed/FLOPS": 202197520897244.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22706055641174316, "optim/lr": 0.0006612, "optim/total_tokens": 578289664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 4.099577903747559, "created_at": "2025-01-15T08:39:20.424629+00:00"} {"global_step": 1104, "acc_step": 0, "speed/wps": 12879.002237882427, "speed/FLOPS": 202282427749594.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2225649654865265, "optim/lr": 0.0006617999999999999, "optim/total_tokens": 578813952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 4.141251564025879, "created_at": "2025-01-15T08:39:30.606735+00:00"} {"global_step": 1105, "acc_step": 0, "speed/wps": 12878.65768567746, "speed/FLOPS": 202277016083751.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21859630942344666, "optim/lr": 0.0006624000000000001, "optim/total_tokens": 579338240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 3.9990429878234863, "created_at": "2025-01-15T08:39:40.788376+00:00"} {"global_step": 1106, "acc_step": 0, "speed/wps": 12877.8829124393, "speed/FLOPS": 202264847205396.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21052758395671844, "optim/lr": 0.0006630000000000001, "optim/total_tokens": 579862528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.049766540527344, "created_at": "2025-01-15T08:39:50.974385+00:00"} {"global_step": 1107, "acc_step": 0, "speed/wps": 12879.173996421781, "speed/FLOPS": 202285125453476.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24031029641628265, "optim/lr": 0.0006636000000000001, "optim/total_tokens": 580386816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.917092800140381, "created_at": "2025-01-15T08:40:01.157907+00:00"} {"global_step": 1108, "acc_step": 0, "speed/wps": 12873.967762533752, "speed/FLOPS": 202203354396149.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2243148684501648, "optim/lr": 0.0006642000000000001, "optim/total_tokens": 580911104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 3.9659266471862793, "created_at": "2025-01-15T08:40:11.343868+00:00"} {"global_step": 1109, "acc_step": 0, "speed/wps": 12874.831536052976, "speed/FLOPS": 202216921146218.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.22386564314365387, "optim/lr": 0.0006648, "optim/total_tokens": 581435392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.9852185249328613, "created_at": "2025-01-15T08:40:21.530977+00:00"} {"global_step": 1110, "acc_step": 0, "speed/wps": 12874.74748138552, "speed/FLOPS": 202215600952162.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.2159172147512436, "optim/lr": 0.0006654, "optim/total_tokens": 581959680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 4.060051918029785, "created_at": "2025-01-15T08:40:31.715502+00:00"} {"global_step": 1111, "acc_step": 0, "speed/wps": 12880.55404731148, "speed/FLOPS": 202306801049087.56, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23046065866947174, "optim/lr": 0.000666, "optim/total_tokens": 582483968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.951671600341797, "created_at": "2025-01-15T08:40:41.894948+00:00"} {"global_step": 1112, "acc_step": 0, "speed/wps": 12876.078437366776, "speed/FLOPS": 202236505444773.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2391768991947174, "optim/lr": 0.0006666, "optim/total_tokens": 583008256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.02958869934082, "created_at": "2025-01-15T08:40:52.080632+00:00"} {"global_step": 1113, "acc_step": 0, "speed/wps": 12872.703131846452, "speed/FLOPS": 202183491633420.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22150740027427673, "optim/lr": 0.0006672, "optim/total_tokens": 583532544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.9585418701171875, "created_at": "2025-01-15T08:41:02.270541+00:00"} {"global_step": 1114, "acc_step": 0, "speed/wps": 12879.793317733664, "speed/FLOPS": 202294852745716.0, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2549504041671753, "optim/lr": 0.0006678, "optim/total_tokens": 584056832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.997831344604492, "created_at": "2025-01-15T08:41:12.451043+00:00"} {"global_step": 1115, "acc_step": 0, "speed/wps": 12879.624065053862, "speed/FLOPS": 202292194399802.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2990822196006775, "optim/lr": 0.0006684, "optim/total_tokens": 584581120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.010623931884766, "created_at": "2025-01-15T08:41:22.631480+00:00"} {"global_step": 1116, "acc_step": 0, "speed/wps": 12874.165321696142, "speed/FLOPS": 202206457334269.72, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2917250096797943, "optim/lr": 0.000669, "optim/total_tokens": 585105408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 3.9793567657470703, "created_at": "2025-01-15T08:41:32.819167+00:00"} {"global_step": 1117, "acc_step": 0, "speed/wps": 12874.173671473283, "speed/FLOPS": 202206588478993.53, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2575361132621765, "optim/lr": 0.0006696, "optim/total_tokens": 585629696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.986196279525757, "created_at": "2025-01-15T08:41:43.008134+00:00"} {"global_step": 1118, "acc_step": 0, "speed/wps": 12876.919793981644, "speed/FLOPS": 202249720106555.1, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23165027797222137, "optim/lr": 0.0006702, "optim/total_tokens": 586153984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 4.085991859436035, "created_at": "2025-01-15T08:41:53.193465+00:00"} {"global_step": 1119, "acc_step": 0, "speed/wps": 12877.986708214468, "speed/FLOPS": 202266477460675.4, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22596575319766998, "optim/lr": 0.0006708, "optim/total_tokens": 586678272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.9275994300842285, "created_at": "2025-01-15T08:42:03.376686+00:00"} {"global_step": 1120, "acc_step": 0, "speed/wps": 12878.788504588963, "speed/FLOPS": 202279070774520.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22718626260757446, "optim/lr": 0.0006714000000000001, "optim/total_tokens": 587202560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.908599853515625, "created_at": "2025-01-15T08:42:13.563622+00:00"} {"global_step": 1121, "acc_step": 0, "speed/wps": 12878.106047614076, "speed/FLOPS": 202268351849935.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19586123526096344, "optim/lr": 0.0006720000000000001, "optim/total_tokens": 587726848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334895, "loss/out": 3.951784610748291, "created_at": "2025-01-15T08:42:23.745297+00:00"} {"global_step": 1122, "acc_step": 0, "speed/wps": 12878.598079531015, "speed/FLOPS": 202276079887314.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2672746777534485, "optim/lr": 0.0006726000000000001, "optim/total_tokens": 588251136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.027193546295166, "created_at": "2025-01-15T08:42:33.932417+00:00"} {"global_step": 1123, "acc_step": 0, "speed/wps": 12879.655802318204, "speed/FLOPS": 202292692877150.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2780255377292633, "optim/lr": 0.0006732, "optim/total_tokens": 588775424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.022977828979492, "created_at": "2025-01-15T08:42:44.112630+00:00"} {"global_step": 1124, "acc_step": 0, "speed/wps": 12875.273381124567, "speed/FLOPS": 202223860930226.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2605765461921692, "optim/lr": 0.0006738, "optim/total_tokens": 589299712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.999340057373047, "created_at": "2025-01-15T08:42:54.299268+00:00"} {"global_step": 1125, "acc_step": 0, "speed/wps": 12879.650627881681, "speed/FLOPS": 202292611605514.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2799552381038666, "optim/lr": 0.0006744, "optim/total_tokens": 589824000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.963930130004883, "created_at": "2025-01-15T08:43:04.481271+00:00"} {"global_step": 1126, "acc_step": 0, "speed/wps": 12875.325865566256, "speed/FLOPS": 202224685270504.25, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.297149658203125, "optim/lr": 0.000675, "optim/total_tokens": 590348288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.9644174575805664, "created_at": "2025-01-15T08:43:14.665315+00:00"} {"global_step": 1127, "acc_step": 0, "speed/wps": 12879.132929434369, "speed/FLOPS": 202284480440006.94, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2989320755004883, "optim/lr": 0.0006756, "optim/total_tokens": 590872576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.002774238586426, "created_at": "2025-01-15T08:43:24.847080+00:00"} {"global_step": 1128, "acc_step": 0, "speed/wps": 12874.352156732733, "speed/FLOPS": 202209391835257.62, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29180005192756653, "optim/lr": 0.0006762, "optim/total_tokens": 591396864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497719, "loss/out": 3.994600772857666, "created_at": "2025-01-15T08:43:35.032726+00:00"} {"global_step": 1129, "acc_step": 0, "speed/wps": 12876.166236759815, "speed/FLOPS": 202237884454890.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2509758770465851, "optim/lr": 0.0006768, "optim/total_tokens": 591921152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.9573707580566406, "created_at": "2025-01-15T08:43:45.218447+00:00"} {"global_step": 1130, "acc_step": 0, "speed/wps": 12873.113411785836, "speed/FLOPS": 202189935643652.7, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.230264812707901, "optim/lr": 0.0006774, "optim/total_tokens": 592445440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 4.036073207855225, "created_at": "2025-01-15T08:43:55.404494+00:00"} {"global_step": 1131, "acc_step": 0, "speed/wps": 12875.723861574996, "speed/FLOPS": 202230936344725.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28110358119010925, "optim/lr": 0.000678, "optim/total_tokens": 592969728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.009443283081055, "created_at": "2025-01-15T08:44:05.589362+00:00"} {"global_step": 1132, "acc_step": 0, "speed/wps": 12871.4964212735, "speed/FLOPS": 202164538585678.62, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2636690139770508, "optim/lr": 0.0006786, "optim/total_tokens": 593494016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 4.015100955963135, "created_at": "2025-01-15T08:44:15.780490+00:00"} {"global_step": 1133, "acc_step": 0, "speed/wps": 12878.304695414285, "speed/FLOPS": 202271471886608.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.232749804854393, "optim/lr": 0.0006792, "optim/total_tokens": 594018304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.971114158630371, "created_at": "2025-01-15T08:44:25.966469+00:00"} {"global_step": 1134, "acc_step": 0, "speed/wps": 12877.7361466265, "speed/FLOPS": 202262542046628.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2555249333381653, "optim/lr": 0.0006798, "optim/total_tokens": 594542592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.9431140422821045, "created_at": "2025-01-15T08:44:36.150176+00:00"} {"global_step": 1135, "acc_step": 0, "speed/wps": 12872.958969280979, "speed/FLOPS": 202187509911887.66, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.26532378792762756, "optim/lr": 0.0006804000000000001, "optim/total_tokens": 595066880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 3.9789228439331055, "created_at": "2025-01-15T08:44:46.337811+00:00"} {"global_step": 1136, "acc_step": 0, "speed/wps": 12875.159543018946, "speed/FLOPS": 202222072946347.62, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3104049265384674, "optim/lr": 0.0006810000000000001, "optim/total_tokens": 595591168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 3.96207332611084, "created_at": "2025-01-15T08:44:56.525998+00:00"} {"global_step": 1137, "acc_step": 0, "speed/wps": 12877.038275901594, "speed/FLOPS": 202251581027919.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3319779932498932, "optim/lr": 0.0006816000000000001, "optim/total_tokens": 596115456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 3.9256527423858643, "created_at": "2025-01-15T08:45:06.711711+00:00"} {"global_step": 1138, "acc_step": 0, "speed/wps": 12876.065397733466, "speed/FLOPS": 202236300639414.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2839930057525635, "optim/lr": 0.0006822, "optim/total_tokens": 596639744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 4.029468059539795, "created_at": "2025-01-15T08:45:16.895189+00:00"} {"global_step": 1139, "acc_step": 0, "speed/wps": 12871.467585954837, "speed/FLOPS": 202164085687375.16, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3092588782310486, "optim/lr": 0.0006828, "optim/total_tokens": 597164032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 3.937227487564087, "created_at": "2025-01-15T08:45:27.085292+00:00"} {"global_step": 1140, "acc_step": 0, "speed/wps": 12873.673235108938, "speed/FLOPS": 202198728438219.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2922815978527069, "optim/lr": 0.0006834, "optim/total_tokens": 597688320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.915292739868164, "created_at": "2025-01-15T08:45:37.270398+00:00"} {"global_step": 1141, "acc_step": 0, "speed/wps": 12873.299029347263, "speed/FLOPS": 202192851022520.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2304907590150833, "optim/lr": 0.000684, "optim/total_tokens": 598212608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.002154350280762, "created_at": "2025-01-15T08:45:47.455663+00:00"} {"global_step": 1142, "acc_step": 0, "speed/wps": 12875.527319427776, "speed/FLOPS": 202227849380226.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.29021400213241577, "optim/lr": 0.0006845999999999999, "optim/total_tokens": 598736896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366127, "loss/out": 3.931300401687622, "created_at": "2025-01-15T08:45:57.641276+00:00"} {"global_step": 1143, "acc_step": 0, "speed/wps": 12874.199693640885, "speed/FLOPS": 202206997192893.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2596348226070404, "optim/lr": 0.0006852, "optim/total_tokens": 599261184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 4.008906364440918, "created_at": "2025-01-15T08:46:07.828673+00:00"} {"global_step": 1144, "acc_step": 0, "speed/wps": 12873.944515828858, "speed/FLOPS": 202202989274704.97, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27723345160484314, "optim/lr": 0.0006858, "optim/total_tokens": 599785472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.950331211090088, "created_at": "2025-01-15T08:46:18.014682+00:00"} {"global_step": 1145, "acc_step": 0, "speed/wps": 12873.17538340671, "speed/FLOPS": 202190908993116.44, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3420192301273346, "optim/lr": 0.0006864, "optim/total_tokens": 600309760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.9937679767608643, "created_at": "2025-01-15T08:46:28.202842+00:00"} {"global_step": 1146, "acc_step": 0, "speed/wps": 12877.776204292886, "speed/FLOPS": 202263171207324.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.326661080121994, "optim/lr": 0.000687, "optim/total_tokens": 600834048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.104099750518799, "created_at": "2025-01-15T08:46:38.389466+00:00"} {"global_step": 1147, "acc_step": 0, "speed/wps": 12875.755871758414, "speed/FLOPS": 202231439108642.53, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2825077474117279, "optim/lr": 0.0006876, "optim/total_tokens": 601358336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.030341148376465, "created_at": "2025-01-15T08:46:48.578721+00:00"} {"global_step": 1148, "acc_step": 0, "speed/wps": 12873.182593254207, "speed/FLOPS": 202191022233678.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31705889105796814, "optim/lr": 0.0006882, "optim/total_tokens": 601882624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.9511311054229736, "created_at": "2025-01-15T08:46:58.766052+00:00"} {"global_step": 1149, "acc_step": 0, "speed/wps": 12876.900388094662, "speed/FLOPS": 202249415310433.56, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2514219582080841, "optim/lr": 0.0006888, "optim/total_tokens": 602406912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.9886035919189453, "created_at": "2025-01-15T08:47:08.950041+00:00"} {"global_step": 1150, "acc_step": 0, "speed/wps": 12872.143055001148, "speed/FLOPS": 202174694856939.22, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.29872408509254456, "optim/lr": 0.0006894000000000001, "optim/total_tokens": 602931200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463795, "loss/out": 4.072935104370117, "created_at": "2025-01-15T08:47:19.136277+00:00"} {"global_step": 1151, "acc_step": 0, "speed/wps": 12874.96455163269, "speed/FLOPS": 202219010338680.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2942362129688263, "optim/lr": 0.0006900000000000001, "optim/total_tokens": 603455488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344789, "loss/out": 4.070419788360596, "created_at": "2025-01-15T08:47:29.322380+00:00"} {"global_step": 1152, "acc_step": 0, "speed/wps": 12877.572665148302, "speed/FLOPS": 202259974345366.47, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.268237829208374, "optim/lr": 0.0006906, "optim/total_tokens": 603979776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 4.048871994018555, "created_at": "2025-01-15T08:47:39.505172+00:00"} {"global_step": 1153, "acc_step": 0, "speed/wps": 12872.447444687623, "speed/FLOPS": 202179475715240.56, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2510513365268707, "optim/lr": 0.0006912, "optim/total_tokens": 604504064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.066295623779297, "created_at": "2025-01-15T08:47:49.691766+00:00"} {"global_step": 1154, "acc_step": 0, "speed/wps": 12875.013066596995, "speed/FLOPS": 202219772332861.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2644539177417755, "optim/lr": 0.0006918, "optim/total_tokens": 605028352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.7937235832214355, "created_at": "2025-01-15T08:47:59.876634+00:00"} {"global_step": 1155, "acc_step": 0, "speed/wps": 12875.865801222119, "speed/FLOPS": 202233165701928.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3661915957927704, "optim/lr": 0.0006924, "optim/total_tokens": 605552640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.039838790893555, "created_at": "2025-01-15T08:48:10.060967+00:00"} {"global_step": 1156, "acc_step": 0, "speed/wps": 12870.87518677523, "speed/FLOPS": 202154781244217.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24762700498104095, "optim/lr": 0.000693, "optim/total_tokens": 606076928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452859, "loss/out": 4.047475337982178, "created_at": "2025-01-15T08:48:20.252867+00:00"} {"global_step": 1157, "acc_step": 0, "speed/wps": 12869.499211987433, "speed/FLOPS": 202133169669387.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2720883786678314, "optim/lr": 0.0006935999999999999, "optim/total_tokens": 606601216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416139, "loss/out": 3.996152400970459, "created_at": "2025-01-15T08:48:30.441679+00:00"} {"global_step": 1158, "acc_step": 0, "speed/wps": 12873.409173567887, "speed/FLOPS": 202194580988859.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32228848338127136, "optim/lr": 0.0006942, "optim/total_tokens": 607125504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 4.012554168701172, "created_at": "2025-01-15T08:48:40.630054+00:00"} {"global_step": 1159, "acc_step": 0, "speed/wps": 12874.353562927077, "speed/FLOPS": 202209413921472.12, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2963382601737976, "optim/lr": 0.0006948, "optim/total_tokens": 607649792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 4.036243438720703, "created_at": "2025-01-15T08:48:50.816194+00:00"} {"global_step": 1160, "acc_step": 0, "speed/wps": 12871.056684636023, "speed/FLOPS": 202157631917525.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2993369698524475, "optim/lr": 0.0006954, "optim/total_tokens": 608174080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.09380578994751, "created_at": "2025-01-15T08:49:01.003771+00:00"} {"global_step": 1161, "acc_step": 0, "speed/wps": 12877.141512542272, "speed/FLOPS": 202253202501223.0, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35110750794410706, "optim/lr": 0.000696, "optim/total_tokens": 608698368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.9240546226501465, "created_at": "2025-01-15T08:49:11.191919+00:00"} {"global_step": 1162, "acc_step": 0, "speed/wps": 12874.27322661812, "speed/FLOPS": 202208152129346.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3212457001209259, "optim/lr": 0.0006966, "optim/total_tokens": 609222656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 4.007586479187012, "created_at": "2025-01-15T08:49:21.382034+00:00"} {"global_step": 1163, "acc_step": 0, "speed/wps": 12872.62740093356, "speed/FLOPS": 202182302175368.78, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2789304852485657, "optim/lr": 0.0006972, "optim/total_tokens": 609746944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.9418272972106934, "created_at": "2025-01-15T08:49:31.568907+00:00"} {"global_step": 1164, "acc_step": 0, "speed/wps": 12874.797501118104, "speed/FLOPS": 202216386580797.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.35705482959747314, "optim/lr": 0.0006978, "optim/total_tokens": 610271232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.060065269470215, "created_at": "2025-01-15T08:49:41.755554+00:00"} {"global_step": 1165, "acc_step": 0, "speed/wps": 12872.174955551753, "speed/FLOPS": 202175195898922.4, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25898057222366333, "optim/lr": 0.0006984000000000001, "optim/total_tokens": 610795520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.077635765075684, "created_at": "2025-01-15T08:49:51.949458+00:00"} {"global_step": 1166, "acc_step": 0, "speed/wps": 12875.162859887072, "speed/FLOPS": 202222125042319.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.244674414396286, "optim/lr": 0.0006990000000000001, "optim/total_tokens": 611319808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 3.992328643798828, "created_at": "2025-01-15T08:50:02.135460+00:00"} {"global_step": 1167, "acc_step": 0, "speed/wps": 12876.894427667921, "speed/FLOPS": 202249321693741.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2539586126804352, "optim/lr": 0.0006996, "optim/total_tokens": 611844096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.993177890777588, "created_at": "2025-01-15T08:50:12.320054+00:00"} {"global_step": 1168, "acc_step": 0, "speed/wps": 12876.054456195061, "speed/FLOPS": 202236128787518.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22993852198123932, "optim/lr": 0.0007002, "optim/total_tokens": 612368384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.938966751098633, "created_at": "2025-01-15T08:50:22.504320+00:00"} {"global_step": 1169, "acc_step": 0, "speed/wps": 12874.004042399458, "speed/FLOPS": 202203924221294.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23144914209842682, "optim/lr": 0.0007008, "optim/total_tokens": 612892672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.907223701477051, "created_at": "2025-01-15T08:50:32.690158+00:00"} {"global_step": 1170, "acc_step": 0, "speed/wps": 12873.016233054868, "speed/FLOPS": 202188409318145.25, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2371072918176651, "optim/lr": 0.0007014, "optim/total_tokens": 613416960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.857438087463379, "created_at": "2025-01-15T08:50:42.876471+00:00"} {"global_step": 1171, "acc_step": 0, "speed/wps": 12870.925595467863, "speed/FLOPS": 202155572982004.25, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23023922741413116, "optim/lr": 0.000702, "optim/total_tokens": 613941248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.8587236404418945, "created_at": "2025-01-15T08:50:53.065661+00:00"} {"global_step": 1172, "acc_step": 0, "speed/wps": 12875.030016885934, "speed/FLOPS": 202220038560441.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23345470428466797, "optim/lr": 0.0007026, "optim/total_tokens": 614465536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.982717990875244, "created_at": "2025-01-15T08:51:03.249459+00:00"} {"global_step": 1173, "acc_step": 0, "speed/wps": 12874.85400860748, "speed/FLOPS": 202217274108567.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.24175229668617249, "optim/lr": 0.0007032, "optim/total_tokens": 614989824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.931159019470215, "created_at": "2025-01-15T08:51:13.434235+00:00"} {"global_step": 1174, "acc_step": 0, "speed/wps": 12872.299126206632, "speed/FLOPS": 202177146169685.62, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21834400296211243, "optim/lr": 0.0007038, "optim/total_tokens": 615514112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.8625776767730713, "created_at": "2025-01-15T08:51:23.623454+00:00"} {"global_step": 1175, "acc_step": 0, "speed/wps": 12873.521406318323, "speed/FLOPS": 202196343758428.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21861466765403748, "optim/lr": 0.0007044, "optim/total_tokens": 616038400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 4.028217792510986, "created_at": "2025-01-15T08:51:33.809571+00:00"} {"global_step": 1176, "acc_step": 0, "speed/wps": 12872.064998448948, "speed/FLOPS": 202173468871525.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21725696325302124, "optim/lr": 0.000705, "optim/total_tokens": 616562688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.9897007942199707, "created_at": "2025-01-15T08:51:43.996906+00:00"} {"global_step": 1177, "acc_step": 0, "speed/wps": 12873.142217976738, "speed/FLOPS": 202190388084464.66, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23025044798851013, "optim/lr": 0.0007056, "optim/total_tokens": 617086976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.8923654556274414, "created_at": "2025-01-15T08:51:54.182581+00:00"} {"global_step": 1178, "acc_step": 0, "speed/wps": 12876.62596323314, "speed/FLOPS": 202245105090884.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2460756003856659, "optim/lr": 0.0007062, "optim/total_tokens": 617611264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.9474520683288574, "created_at": "2025-01-15T08:52:04.367451+00:00"} {"global_step": 1179, "acc_step": 0, "speed/wps": 12874.198394649848, "speed/FLOPS": 202206976790454.7, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23490360379219055, "optim/lr": 0.0007068, "optim/total_tokens": 618135552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.906668186187744, "created_at": "2025-01-15T08:52:14.555296+00:00"} {"global_step": 1180, "acc_step": 0, "speed/wps": 12871.915756382003, "speed/FLOPS": 202171124819784.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23430946469306946, "optim/lr": 0.0007074000000000001, "optim/total_tokens": 618659840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405144, "loss/out": 3.8780875205993652, "created_at": "2025-01-15T08:52:24.741656+00:00"} {"global_step": 1181, "acc_step": 0, "speed/wps": 12880.189591944734, "speed/FLOPS": 202301076776738.7, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2525842487812042, "optim/lr": 0.000708, "optim/total_tokens": 619184128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.8260416984558105, "created_at": "2025-01-15T08:52:34.929442+00:00"} {"global_step": 1182, "acc_step": 0, "speed/wps": 12877.56353563709, "speed/FLOPS": 202259830953847.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.21896199882030487, "optim/lr": 0.0007086, "optim/total_tokens": 619708416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.055588245391846, "created_at": "2025-01-15T08:52:45.114456+00:00"} {"global_step": 1183, "acc_step": 0, "speed/wps": 12867.624878197332, "speed/FLOPS": 202103730681609.72, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22176320850849152, "optim/lr": 0.0007092, "optim/total_tokens": 620232704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.0737762451171875, "created_at": "2025-01-15T08:52:55.304348+00:00"} {"global_step": 1184, "acc_step": 0, "speed/wps": 12871.889629251598, "speed/FLOPS": 202170714457299.06, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3018299341201782, "optim/lr": 0.0007098, "optim/total_tokens": 620756992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.9225611686706543, "created_at": "2025-01-15T08:53:05.491862+00:00"} {"global_step": 1185, "acc_step": 0, "speed/wps": 12873.997315646817, "speed/FLOPS": 202203818568400.56, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23274849355220795, "optim/lr": 0.0007104, "optim/total_tokens": 621281280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.948045015335083, "created_at": "2025-01-15T08:53:15.679054+00:00"} {"global_step": 1186, "acc_step": 0, "speed/wps": 12871.661225927593, "speed/FLOPS": 202167127069236.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2296948879957199, "optim/lr": 0.0007109999999999999, "optim/total_tokens": 621805568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.945873260498047, "created_at": "2025-01-15T08:53:25.870427+00:00"} {"global_step": 1187, "acc_step": 0, "speed/wps": 12871.309627912446, "speed/FLOPS": 202161604739262.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23423506319522858, "optim/lr": 0.0007116, "optim/total_tokens": 622329856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404614, "loss/out": 3.98734188079834, "created_at": "2025-01-15T08:53:36.057251+00:00"} {"global_step": 1188, "acc_step": 0, "speed/wps": 12872.755973289026, "speed/FLOPS": 202184321580888.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3396643102169037, "optim/lr": 0.0007122, "optim/total_tokens": 622854144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313754, "loss/out": 3.953566551208496, "created_at": "2025-01-15T08:53:46.246048+00:00"} {"global_step": 1189, "acc_step": 0, "speed/wps": 12870.607273407091, "speed/FLOPS": 202150573296619.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3726457953453064, "optim/lr": 0.0007128, "optim/total_tokens": 623378432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.9812536239624023, "created_at": "2025-01-15T08:53:56.433539+00:00"} {"global_step": 1190, "acc_step": 0, "speed/wps": 12868.366954622548, "speed/FLOPS": 202115386011587.62, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41339945793151855, "optim/lr": 0.0007134000000000001, "optim/total_tokens": 623902720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.9230103492736816, "created_at": "2025-01-15T08:54:06.623369+00:00"} {"global_step": 1191, "acc_step": 0, "speed/wps": 12871.909120106731, "speed/FLOPS": 202171020587962.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3494146168231964, "optim/lr": 0.000714, "optim/total_tokens": 624427008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.069382667541504, "created_at": "2025-01-15T08:54:16.814646+00:00"} {"global_step": 1192, "acc_step": 0, "speed/wps": 12871.557951178256, "speed/FLOPS": 202165504997383.78, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37622326612472534, "optim/lr": 0.0007146, "optim/total_tokens": 624951296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.9556589126586914, "created_at": "2025-01-15T08:54:27.006034+00:00"} {"global_step": 1193, "acc_step": 0, "speed/wps": 12878.67444757489, "speed/FLOPS": 202277279352383.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3017038106918335, "optim/lr": 0.0007152, "optim/total_tokens": 625475584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.012144088745117, "created_at": "2025-01-15T08:54:37.187271+00:00"} {"global_step": 1194, "acc_step": 0, "speed/wps": 12871.883652271908, "speed/FLOPS": 202170620580619.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2411070615053177, "optim/lr": 0.0007158, "optim/total_tokens": 625999872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454196, "loss/out": 4.052092552185059, "created_at": "2025-01-15T08:54:47.374454+00:00"} {"global_step": 1195, "acc_step": 0, "speed/wps": 12872.546157207524, "speed/FLOPS": 202181026131010.8, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21889685094356537, "optim/lr": 0.0007164000000000001, "optim/total_tokens": 626524160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 3.870976448059082, "created_at": "2025-01-15T08:54:57.560289+00:00"} {"global_step": 1196, "acc_step": 0, "speed/wps": 12873.25543754123, "speed/FLOPS": 202192166353304.97, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26344555616378784, "optim/lr": 0.000717, "optim/total_tokens": 627048448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 4.02950382232666, "created_at": "2025-01-15T08:55:07.750761+00:00"} {"global_step": 1197, "acc_step": 0, "speed/wps": 12873.848244274544, "speed/FLOPS": 202201477197653.5, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.251914381980896, "optim/lr": 0.0007176, "optim/total_tokens": 627572736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.9309732913970947, "created_at": "2025-01-15T08:55:17.938160+00:00"} {"global_step": 1198, "acc_step": 0, "speed/wps": 12872.211901098428, "speed/FLOPS": 202175776179501.44, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25202706456184387, "optim/lr": 0.0007182, "optim/total_tokens": 628097024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.049332141876221, "created_at": "2025-01-15T08:55:28.130307+00:00"} {"global_step": 1199, "acc_step": 0, "speed/wps": 12874.108510523982, "speed/FLOPS": 202205565036744.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2699491083621979, "optim/lr": 0.0007188, "optim/total_tokens": 628621312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 4.030941963195801, "created_at": "2025-01-15T08:55:38.315131+00:00"} {"global_step": 1200, "acc_step": 0, "speed/wps": 12867.801128893396, "speed/FLOPS": 202106498940985.25, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2684243321418762, "optim/lr": 0.0007194, "optim/total_tokens": 629145600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.82181978225708, "created_at": "2025-01-15T08:55:48.507842+00:00"} {"global_step": 1201, "acc_step": 0, "speed/wps": 12873.834170494922, "speed/FLOPS": 202201256149605.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22234323620796204, "optim/lr": 0.0007199999999999999, "optim/total_tokens": 629669888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.009380340576172, "created_at": "2025-01-15T08:55:58.693897+00:00"} {"global_step": 1202, "acc_step": 0, "speed/wps": 12875.07341576535, "speed/FLOPS": 202220720199478.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.21609865128993988, "optim/lr": 0.0007206, "optim/total_tokens": 630194176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424102, "loss/out": 3.853790760040283, "created_at": "2025-01-15T08:56:08.881344+00:00"} {"global_step": 1203, "acc_step": 0, "speed/wps": 12873.23544923078, "speed/FLOPS": 202191852409422.56, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21043306589126587, "optim/lr": 0.0007212000000000001, "optim/total_tokens": 630718464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 3.92635178565979, "created_at": "2025-01-15T08:56:19.070007+00:00"} {"global_step": 1204, "acc_step": 0, "speed/wps": 12873.706950858645, "speed/FLOPS": 202199257990399.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21854329109191895, "optim/lr": 0.0007218000000000001, "optim/total_tokens": 631242752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.036693096160889, "created_at": "2025-01-15T08:56:29.254772+00:00"} {"global_step": 1205, "acc_step": 0, "speed/wps": 12876.654289150045, "speed/FLOPS": 202245549988332.97, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19713105261325836, "optim/lr": 0.0007224, "optim/total_tokens": 631767040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.9488396644592285, "created_at": "2025-01-15T08:56:39.437449+00:00"} {"global_step": 1206, "acc_step": 0, "speed/wps": 12873.05686776516, "speed/FLOPS": 202189047542108.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1790485382080078, "optim/lr": 0.000723, "optim/total_tokens": 632291328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.9024014472961426, "created_at": "2025-01-15T08:56:49.622941+00:00"} {"global_step": 1207, "acc_step": 0, "speed/wps": 12872.379097624442, "speed/FLOPS": 202178402230694.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18183378875255585, "optim/lr": 0.0007236, "optim/total_tokens": 632815616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.8739867210388184, "created_at": "2025-01-15T08:56:59.813117+00:00"} {"global_step": 1208, "acc_step": 0, "speed/wps": 12873.93133684515, "speed/FLOPS": 202202782280656.0, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1818459928035736, "optim/lr": 0.0007242, "optim/total_tokens": 633339904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.7743306159973145, "created_at": "2025-01-15T08:57:10.000282+00:00"} {"global_step": 1209, "acc_step": 0, "speed/wps": 12869.99439494316, "speed/FLOPS": 202140947198160.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16584695875644684, "optim/lr": 0.0007248, "optim/total_tokens": 633864192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 3.9062511920928955, "created_at": "2025-01-15T08:57:20.188332+00:00"} {"global_step": 1210, "acc_step": 0, "speed/wps": 12869.194631041326, "speed/FLOPS": 202128385807091.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19643843173980713, "optim/lr": 0.0007254, "optim/total_tokens": 634388480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349477, "loss/out": 3.953367233276367, "created_at": "2025-01-15T08:57:30.380011+00:00"} {"global_step": 1211, "acc_step": 0, "speed/wps": 12874.275413612286, "speed/FLOPS": 202208186479094.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23580743372440338, "optim/lr": 0.000726, "optim/total_tokens": 634912768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 4.006312370300293, "created_at": "2025-01-15T08:57:40.565408+00:00"} {"global_step": 1212, "acc_step": 0, "speed/wps": 12871.796190280853, "speed/FLOPS": 202169246869864.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24369250237941742, "optim/lr": 0.0007266, "optim/total_tokens": 635437056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 3.904481887817383, "created_at": "2025-01-15T08:57:50.752390+00:00"} {"global_step": 1213, "acc_step": 0, "speed/wps": 12870.535291238857, "speed/FLOPS": 202149442717753.5, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23939092457294464, "optim/lr": 0.0007272, "optim/total_tokens": 635961344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 3.8744025230407715, "created_at": "2025-01-15T08:58:00.943741+00:00"} {"global_step": 1214, "acc_step": 0, "speed/wps": 12873.683772218546, "speed/FLOPS": 202198893938005.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24689653515815735, "optim/lr": 0.0007278, "optim/total_tokens": 636485632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.01059103012085, "created_at": "2025-01-15T08:58:11.128643+00:00"} {"global_step": 1215, "acc_step": 0, "speed/wps": 12875.450055791865, "speed/FLOPS": 202226635848652.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3055775761604309, "optim/lr": 0.0007284, "optim/total_tokens": 637009920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.953991413116455, "created_at": "2025-01-15T08:58:21.312132+00:00"} {"global_step": 1216, "acc_step": 0, "speed/wps": 12867.757136651504, "speed/FLOPS": 202105807982374.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3723418414592743, "optim/lr": 0.000729, "optim/total_tokens": 637534208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.867173194885254, "created_at": "2025-01-15T08:58:31.505108+00:00"} {"global_step": 1217, "acc_step": 0, "speed/wps": 12870.734604814743, "speed/FLOPS": 202152573211348.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3328663408756256, "optim/lr": 0.0007296000000000001, "optim/total_tokens": 638058496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 3.7687647342681885, "created_at": "2025-01-15T08:58:41.692775+00:00"} {"global_step": 1218, "acc_step": 0, "speed/wps": 12871.933474340458, "speed/FLOPS": 202171403104669.06, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3360922336578369, "optim/lr": 0.0007302000000000001, "optim/total_tokens": 638582784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415290, "loss/out": 3.873692512512207, "created_at": "2025-01-15T08:58:51.880096+00:00"} {"global_step": 1219, "acc_step": 0, "speed/wps": 12872.62941075831, "speed/FLOPS": 202182333742428.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.27535492181777954, "optim/lr": 0.0007308000000000001, "optim/total_tokens": 639107072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.8666579723358154, "created_at": "2025-01-15T08:59:02.066004+00:00"} {"global_step": 1220, "acc_step": 0, "speed/wps": 12872.436471621388, "speed/FLOPS": 202179303368156.75, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.22876740992069244, "optim/lr": 0.0007314, "optim/total_tokens": 639631360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.9284071922302246, "created_at": "2025-01-15T08:59:12.253390+00:00"} {"global_step": 1221, "acc_step": 0, "speed/wps": 12873.01937337526, "speed/FLOPS": 202188458641192.28, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24321423470973969, "optim/lr": 0.000732, "optim/total_tokens": 640155648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.8910794258117676, "created_at": "2025-01-15T08:59:22.439483+00:00"} {"global_step": 1222, "acc_step": 0, "speed/wps": 12874.050906109002, "speed/FLOPS": 202204660280250.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23960916697978973, "optim/lr": 0.0007326, "optim/total_tokens": 640679936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.8644676208496094, "created_at": "2025-01-15T08:59:32.624857+00:00"} {"global_step": 1223, "acc_step": 0, "speed/wps": 12869.589866146152, "speed/FLOPS": 202134593517522.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2766664922237396, "optim/lr": 0.0007332, "optim/total_tokens": 641204224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.907843589782715, "created_at": "2025-01-15T08:59:42.813136+00:00"} {"global_step": 1224, "acc_step": 0, "speed/wps": 12871.266179065222, "speed/FLOPS": 202160922315411.7, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2515088617801666, "optim/lr": 0.0007338000000000001, "optim/total_tokens": 641728512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.8267555236816406, "created_at": "2025-01-15T08:59:53.001305+00:00"} {"global_step": 1225, "acc_step": 0, "speed/wps": 12850.006944740822, "speed/FLOPS": 201827016827097.7, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 0.24918177723884583, "optim/lr": 0.0007344, "optim/total_tokens": 642252800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.8895182609558105, "created_at": "2025-01-15T09:00:03.207626+00:00"} {"global_step": 1226, "acc_step": 0, "speed/wps": 12865.230954581833, "speed/FLOPS": 202066130821632.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23496326804161072, "optim/lr": 0.000735, "optim/total_tokens": 642777088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.8766074180603027, "created_at": "2025-01-15T09:00:13.400525+00:00"} {"global_step": 1227, "acc_step": 0, "speed/wps": 12864.056582922654, "speed/FLOPS": 202047685700969.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2262483537197113, "optim/lr": 0.0007356, "optim/total_tokens": 643301376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 3.8920986652374268, "created_at": "2025-01-15T09:00:23.594707+00:00"} {"global_step": 1228, "acc_step": 0, "speed/wps": 12870.29964342963, "speed/FLOPS": 202145741545095.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2911912202835083, "optim/lr": 0.0007362, "optim/total_tokens": 643825664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 3.8379135131835938, "created_at": "2025-01-15T09:00:33.783055+00:00"} {"global_step": 1229, "acc_step": 0, "speed/wps": 12864.208272026202, "speed/FLOPS": 202050068186783.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2981826066970825, "optim/lr": 0.0007368, "optim/total_tokens": 644349952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 3.857302665710449, "created_at": "2025-01-15T09:00:43.976075+00:00"} {"global_step": 1230, "acc_step": 0, "speed/wps": 12868.139586968897, "speed/FLOPS": 202111814890151.12, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.25214773416519165, "optim/lr": 0.0007374, "optim/total_tokens": 644874240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.8780055046081543, "created_at": "2025-01-15T09:00:54.166102+00:00"} {"global_step": 1231, "acc_step": 0, "speed/wps": 12875.951669313099, "speed/FLOPS": 202234514378292.9, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2484787404537201, "optim/lr": 0.000738, "optim/total_tokens": 645398528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.9384076595306396, "created_at": "2025-01-15T09:01:04.352299+00:00"} {"global_step": 1232, "acc_step": 0, "speed/wps": 12870.084919553648, "speed/FLOPS": 202142369011561.75, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1942637711763382, "optim/lr": 0.0007386000000000001, "optim/total_tokens": 645922816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.9088640213012695, "created_at": "2025-01-15T09:01:14.543123+00:00"} {"global_step": 1233, "acc_step": 0, "speed/wps": 12868.40006485943, "speed/FLOPS": 202115906053356.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22005371749401093, "optim/lr": 0.0007392000000000001, "optim/total_tokens": 646447104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.9917657375335693, "created_at": "2025-01-15T09:01:24.743429+00:00"} {"global_step": 1234, "acc_step": 0, "speed/wps": 12874.131176048846, "speed/FLOPS": 202205921029957.9, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26640719175338745, "optim/lr": 0.0007398000000000001, "optim/total_tokens": 646971392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471956, "loss/out": 3.736997127532959, "created_at": "2025-01-15T09:01:34.932470+00:00"} {"global_step": 1235, "acc_step": 0, "speed/wps": 12872.748092787891, "speed/FLOPS": 202184197806789.06, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2935281991958618, "optim/lr": 0.0007404, "optim/total_tokens": 647495680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500101, "loss/out": 3.9264490604400635, "created_at": "2025-01-15T09:01:45.123783+00:00"} {"global_step": 1236, "acc_step": 0, "speed/wps": 12871.947429768785, "speed/FLOPS": 202171622293847.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22311441600322723, "optim/lr": 0.000741, "optim/total_tokens": 648019968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.915599822998047, "created_at": "2025-01-15T09:01:55.311839+00:00"} {"global_step": 1237, "acc_step": 0, "speed/wps": 12872.303888554738, "speed/FLOPS": 202177220968906.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2180178016424179, "optim/lr": 0.0007416, "optim/total_tokens": 648544256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.8441643714904785, "created_at": "2025-01-15T09:02:05.498031+00:00"} {"global_step": 1238, "acc_step": 0, "speed/wps": 12870.274471498193, "speed/FLOPS": 202145346185322.47, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.227619469165802, "optim/lr": 0.0007422, "optim/total_tokens": 649068544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.9983649253845215, "created_at": "2025-01-15T09:02:15.688748+00:00"} {"global_step": 1239, "acc_step": 0, "speed/wps": 12873.615947826016, "speed/FLOPS": 202197828662720.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21988695859909058, "optim/lr": 0.0007428, "optim/total_tokens": 649592832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 3.859978437423706, "created_at": "2025-01-15T09:02:25.876497+00:00"} {"global_step": 1240, "acc_step": 0, "speed/wps": 12872.480651277998, "speed/FLOPS": 202179997270373.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20557813346385956, "optim/lr": 0.0007434, "optim/total_tokens": 650117120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.827589988708496, "created_at": "2025-01-15T09:02:36.064667+00:00"} {"global_step": 1241, "acc_step": 0, "speed/wps": 12869.623391044088, "speed/FLOPS": 202135120072112.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2001200169324875, "optim/lr": 0.000744, "optim/total_tokens": 650641408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.903045892715454, "created_at": "2025-01-15T09:02:46.255522+00:00"} {"global_step": 1242, "acc_step": 0, "speed/wps": 12872.226023133751, "speed/FLOPS": 202175997985471.9, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20690427720546722, "optim/lr": 0.0007446, "optim/total_tokens": 651165696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.8588192462921143, "created_at": "2025-01-15T09:02:56.443353+00:00"} {"global_step": 1243, "acc_step": 0, "speed/wps": 12871.473483676375, "speed/FLOPS": 202164178319196.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23610837757587433, "optim/lr": 0.0007452, "optim/total_tokens": 651689984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.729328155517578, "created_at": "2025-01-15T09:03:06.629933+00:00"} {"global_step": 1244, "acc_step": 0, "speed/wps": 12869.526553322308, "speed/FLOPS": 202133599102622.94, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24280880391597748, "optim/lr": 0.0007458, "optim/total_tokens": 652214272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.9064416885375977, "created_at": "2025-01-15T09:03:16.821845+00:00"} {"global_step": 1245, "acc_step": 0, "speed/wps": 12868.639707337208, "speed/FLOPS": 202119669967772.84, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21675433218479156, "optim/lr": 0.0007464, "optim/total_tokens": 652738560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.8252503871917725, "created_at": "2025-01-15T09:03:27.014703+00:00"} {"global_step": 1246, "acc_step": 0, "speed/wps": 12868.253023293377, "speed/FLOPS": 202113596563504.94, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19941958785057068, "optim/lr": 0.000747, "optim/total_tokens": 653262848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.8771486282348633, "created_at": "2025-01-15T09:03:37.211220+00:00"} {"global_step": 1247, "acc_step": 0, "speed/wps": 12869.644925119523, "speed/FLOPS": 202135458294358.1, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26264503598213196, "optim/lr": 0.0007476000000000001, "optim/total_tokens": 653787136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402750, "loss/out": 3.7904036045074463, "created_at": "2025-01-15T09:03:47.402258+00:00"} {"global_step": 1248, "acc_step": 0, "speed/wps": 12868.750277819117, "speed/FLOPS": 202121406629131.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31031307578086853, "optim/lr": 0.0007482000000000001, "optim/total_tokens": 654311424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410019, "loss/out": 3.9045770168304443, "created_at": "2025-01-15T09:03:57.600706+00:00"} {"global_step": 1249, "acc_step": 0, "speed/wps": 12875.050241627212, "speed/FLOPS": 202220356217794.6, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1993652582168579, "optim/lr": 0.0007488, "optim/total_tokens": 654835712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 3.946645736694336, "created_at": "2025-01-15T09:04:07.784482+00:00"} {"global_step": 1250, "acc_step": 0, "speed/wps": 12875.072477019734, "speed/FLOPS": 202220705455188.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19740702211856842, "optim/lr": 0.0007494, "optim/total_tokens": 655360000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.8917226791381836, "created_at": "2025-01-15T09:04:17.972026+00:00"} {"global_step": 1251, "acc_step": 0, "speed/wps": 12874.632927073862, "speed/FLOPS": 202213801719283.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2327556014060974, "optim/lr": 0.00075, "optim/total_tokens": 655884288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 3.81437349319458, "created_at": "2025-01-15T09:04:28.159489+00:00"} {"global_step": 1252, "acc_step": 0, "speed/wps": 12872.420652542947, "speed/FLOPS": 202179054907792.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19122125208377838, "optim/lr": 0.0007505999999999999, "optim/total_tokens": 656408576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.830626964569092, "created_at": "2025-01-15T09:04:38.347361+00:00"} {"global_step": 1253, "acc_step": 0, "speed/wps": 12871.887395977268, "speed/FLOPS": 202170679380656.84, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1953304260969162, "optim/lr": 0.0007512, "optim/total_tokens": 656932864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.873210906982422, "created_at": "2025-01-15T09:04:48.535079+00:00"} {"global_step": 1254, "acc_step": 0, "speed/wps": 12874.263059519286, "speed/FLOPS": 202207992441088.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1982596516609192, "optim/lr": 0.0007518, "optim/total_tokens": 657457152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.7691831588745117, "created_at": "2025-01-15T09:04:58.720744+00:00"} {"global_step": 1255, "acc_step": 0, "speed/wps": 12870.364456280424, "speed/FLOPS": 202146759519979.47, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.20148888230323792, "optim/lr": 0.0007524000000000001, "optim/total_tokens": 657981440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.7493174076080322, "created_at": "2025-01-15T09:05:08.913152+00:00"} {"global_step": 1256, "acc_step": 0, "speed/wps": 12875.20169658234, "speed/FLOPS": 202222735025985.8, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21075181663036346, "optim/lr": 0.000753, "optim/total_tokens": 658505728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.8167810440063477, "created_at": "2025-01-15T09:05:19.098702+00:00"} {"global_step": 1257, "acc_step": 0, "speed/wps": 12874.131985112756, "speed/FLOPS": 202205933737418.38, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2890034019947052, "optim/lr": 0.0007536, "optim/total_tokens": 659030016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365889, "loss/out": 3.924567699432373, "created_at": "2025-01-15T09:05:29.283222+00:00"} {"global_step": 1258, "acc_step": 0, "speed/wps": 12872.45077193222, "speed/FLOPS": 202179527974189.06, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27953726053237915, "optim/lr": 0.0007542, "optim/total_tokens": 659554304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 3.729079246520996, "created_at": "2025-01-15T09:05:39.472529+00:00"} {"global_step": 1259, "acc_step": 0, "speed/wps": 12876.720890830926, "speed/FLOPS": 202246596059251.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23669153451919556, "optim/lr": 0.0007548, "optim/total_tokens": 660078592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.768745183944702, "created_at": "2025-01-15T09:05:49.655174+00:00"} {"global_step": 1260, "acc_step": 0, "speed/wps": 12869.780100221835, "speed/FLOPS": 202137581405089.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2530469000339508, "optim/lr": 0.0007554, "optim/total_tokens": 660602880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.8249216079711914, "created_at": "2025-01-15T09:05:59.849348+00:00"} {"global_step": 1261, "acc_step": 0, "speed/wps": 12860.491416636562, "speed/FLOPS": 201991689865394.4, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.3048897683620453, "optim/lr": 0.000756, "optim/total_tokens": 661127168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 3.8679277896881104, "created_at": "2025-01-15T09:06:10.045733+00:00"} {"global_step": 1262, "acc_step": 0, "speed/wps": 12855.878251208214, "speed/FLOPS": 201919233763188.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.29783517122268677, "optim/lr": 0.0007566, "optim/total_tokens": 661651456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.8623828887939453, "created_at": "2025-01-15T09:06:20.245324+00:00"} {"global_step": 1263, "acc_step": 0, "speed/wps": 12858.608269668706, "speed/FLOPS": 201962112454549.16, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23091930150985718, "optim/lr": 0.0007572000000000001, "optim/total_tokens": 662175744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.805428981781006, "created_at": "2025-01-15T09:06:30.444878+00:00"} {"global_step": 1264, "acc_step": 0, "speed/wps": 12855.127634853023, "speed/FLOPS": 201907444301875.38, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.2824089825153351, "optim/lr": 0.0007578, "optim/total_tokens": 662700032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.9815425872802734, "created_at": "2025-01-15T09:06:40.644893+00:00"} {"global_step": 1265, "acc_step": 0, "speed/wps": 12862.054069263311, "speed/FLOPS": 202016233472207.25, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.21273373067378998, "optim/lr": 0.0007584000000000001, "optim/total_tokens": 663224320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.8328256607055664, "created_at": "2025-01-15T09:06:50.841248+00:00"} {"global_step": 1266, "acc_step": 0, "speed/wps": 12862.635911175146, "speed/FLOPS": 202025372098967.2, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2084011733531952, "optim/lr": 0.000759, "optim/total_tokens": 663748608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.7963452339172363, "created_at": "2025-01-15T09:07:01.039317+00:00"} {"global_step": 1267, "acc_step": 0, "speed/wps": 12865.897440727062, "speed/FLOPS": 202076598902391.28, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23698610067367554, "optim/lr": 0.0007595999999999999, "optim/total_tokens": 664272896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.753640651702881, "created_at": "2025-01-15T09:07:11.234589+00:00"} {"global_step": 1268, "acc_step": 0, "speed/wps": 12861.189428337022, "speed/FLOPS": 202002653098318.75, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2281150370836258, "optim/lr": 0.0007602, "optim/total_tokens": 664797184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.7408785820007324, "created_at": "2025-01-15T09:07:21.429645+00:00"} {"global_step": 1269, "acc_step": 0, "speed/wps": 12863.820042354866, "speed/FLOPS": 202043970506311.44, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25321072340011597, "optim/lr": 0.0007608, "optim/total_tokens": 665321472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.960850477218628, "created_at": "2025-01-15T09:07:31.624834+00:00"} {"global_step": 1270, "acc_step": 0, "speed/wps": 12873.878113621442, "speed/FLOPS": 202201946336791.3, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26906469464302063, "optim/lr": 0.0007614000000000001, "optim/total_tokens": 665845760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427332, "loss/out": 3.8862693309783936, "created_at": "2025-01-15T09:07:41.809560+00:00"} {"global_step": 1271, "acc_step": 0, "speed/wps": 12876.47044093656, "speed/FLOPS": 202242662399506.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22891274094581604, "optim/lr": 0.000762, "optim/total_tokens": 666370048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286083, "loss/out": 3.7272465229034424, "created_at": "2025-01-15T09:07:51.994426+00:00"} {"global_step": 1272, "acc_step": 0, "speed/wps": 12872.563420074219, "speed/FLOPS": 202181297268053.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24461570382118225, "optim/lr": 0.0007626, "optim/total_tokens": 666894336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 3.866969108581543, "created_at": "2025-01-15T09:08:02.181640+00:00"} {"global_step": 1273, "acc_step": 0, "speed/wps": 12868.817935634757, "speed/FLOPS": 202122469288097.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23990598320960999, "optim/lr": 0.0007632, "optim/total_tokens": 667418624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.872619152069092, "created_at": "2025-01-15T09:08:12.375686+00:00"} {"global_step": 1274, "acc_step": 0, "speed/wps": 12863.37291400244, "speed/FLOPS": 202036947741117.88, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22113248705863953, "optim/lr": 0.0007638, "optim/total_tokens": 667942912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.8676137924194336, "created_at": "2025-01-15T09:08:22.572280+00:00"} {"global_step": 1275, "acc_step": 0, "speed/wps": 12860.489631534527, "speed/FLOPS": 201991661827913.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.256952166557312, "optim/lr": 0.0007644000000000001, "optim/total_tokens": 668467200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500101, "loss/out": 3.8549647331237793, "created_at": "2025-01-15T09:08:32.772461+00:00"} {"global_step": 1276, "acc_step": 0, "speed/wps": 12875.624910663168, "speed/FLOPS": 202229382184682.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2674631178379059, "optim/lr": 0.0007650000000000001, "optim/total_tokens": 668991488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.8800694942474365, "created_at": "2025-01-15T09:08:42.959175+00:00"} {"global_step": 1277, "acc_step": 0, "speed/wps": 12870.073988176795, "speed/FLOPS": 202142197319266.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36302945017814636, "optim/lr": 0.0007656, "optim/total_tokens": 669515776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.896864891052246, "created_at": "2025-01-15T09:08:53.149314+00:00"} {"global_step": 1278, "acc_step": 0, "speed/wps": 12868.748731645483, "speed/FLOPS": 202121382344349.4, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2862870991230011, "optim/lr": 0.0007662000000000001, "optim/total_tokens": 670040064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 3.9428319931030273, "created_at": "2025-01-15T09:09:03.341546+00:00"} {"global_step": 1279, "acc_step": 0, "speed/wps": 12873.401149784258, "speed/FLOPS": 202194454964311.34, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.31655293703079224, "optim/lr": 0.0007668, "optim/total_tokens": 670564352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.847935914993286, "created_at": "2025-01-15T09:09:13.527308+00:00"} {"global_step": 1280, "acc_step": 0, "speed/wps": 12874.596510576506, "speed/FLOPS": 202213229748151.47, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2749224007129669, "optim/lr": 0.0007674000000000001, "optim/total_tokens": 671088640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283388, "loss/out": 3.76198148727417, "created_at": "2025-01-15T09:09:23.711455+00:00"} {"global_step": 1281, "acc_step": 0, "speed/wps": 12874.234565112063, "speed/FLOPS": 202207544897267.2, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2596968114376068, "optim/lr": 0.000768, "optim/total_tokens": 671612928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.8840696811676025, "created_at": "2025-01-15T09:09:33.897256+00:00"} {"global_step": 1282, "acc_step": 0, "speed/wps": 12866.456430751316, "speed/FLOPS": 202085378608855.0, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2481490522623062, "optim/lr": 0.0007685999999999999, "optim/total_tokens": 672137216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.8420066833496094, "created_at": "2025-01-15T09:09:44.091046+00:00"} {"global_step": 1283, "acc_step": 0, "speed/wps": 12874.091317706096, "speed/FLOPS": 202205294999913.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22036026418209076, "optim/lr": 0.0007692, "optim/total_tokens": 672661504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.841947555541992, "created_at": "2025-01-15T09:09:54.275966+00:00"} {"global_step": 1284, "acc_step": 0, "speed/wps": 12871.126120624785, "speed/FLOPS": 202158722505144.16, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20981945097446442, "optim/lr": 0.0007698, "optim/total_tokens": 673185792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.762622833251953, "created_at": "2025-01-15T09:10:04.464044+00:00"} {"global_step": 1285, "acc_step": 0, "speed/wps": 12855.674971540655, "speed/FLOPS": 201916040976674.1, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21942391991615295, "optim/lr": 0.0007704, "optim/total_tokens": 673710080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.795586585998535, "created_at": "2025-01-15T09:10:14.664973+00:00"} {"global_step": 1286, "acc_step": 0, "speed/wps": 12869.190121646208, "speed/FLOPS": 202128314980844.22, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22255246341228485, "optim/lr": 0.000771, "optim/total_tokens": 674234368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.789099931716919, "created_at": "2025-01-15T09:10:24.862960+00:00"} {"global_step": 1287, "acc_step": 0, "speed/wps": 12868.717224614344, "speed/FLOPS": 202120887483130.16, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23690181970596313, "optim/lr": 0.0007716, "optim/total_tokens": 674758656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.840980052947998, "created_at": "2025-01-15T09:10:35.054225+00:00"} {"global_step": 1288, "acc_step": 0, "speed/wps": 12870.259339197775, "speed/FLOPS": 202145108511750.47, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21683597564697266, "optim/lr": 0.0007722000000000001, "optim/total_tokens": 675282944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.8003134727478027, "created_at": "2025-01-15T09:10:45.243428+00:00"} {"global_step": 1289, "acc_step": 0, "speed/wps": 12864.866267453608, "speed/FLOPS": 202060402909151.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22625751793384552, "optim/lr": 0.0007728, "optim/total_tokens": 675807232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.9050850868225098, "created_at": "2025-01-15T09:10:55.440200+00:00"} {"global_step": 1290, "acc_step": 0, "speed/wps": 12869.200500791061, "speed/FLOPS": 202128477999576.56, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.26476728916168213, "optim/lr": 0.0007733999999999999, "optim/total_tokens": 676331520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 3.8674466609954834, "created_at": "2025-01-15T09:11:05.630297+00:00"} {"global_step": 1291, "acc_step": 0, "speed/wps": 12873.579320407642, "speed/FLOPS": 202197253378783.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2601609230041504, "optim/lr": 0.0007740000000000001, "optim/total_tokens": 676855808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.8934946060180664, "created_at": "2025-01-15T09:11:15.818052+00:00"} {"global_step": 1292, "acc_step": 0, "speed/wps": 12867.78820484697, "speed/FLOPS": 202106295951076.62, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23460564017295837, "optim/lr": 0.0007746, "optim/total_tokens": 677380096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449664, "loss/out": 3.9096059799194336, "created_at": "2025-01-15T09:11:26.010742+00:00"} {"global_step": 1293, "acc_step": 0, "speed/wps": 12866.192942771835, "speed/FLOPS": 202081240168066.8, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28759193420410156, "optim/lr": 0.0007752000000000001, "optim/total_tokens": 677904384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 3.8144168853759766, "created_at": "2025-01-15T09:11:36.205014+00:00"} {"global_step": 1294, "acc_step": 0, "speed/wps": 12859.434271720416, "speed/FLOPS": 201975085951817.03, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.001, "optim/grad_norm": 0.28733304142951965, "optim/lr": 0.0007758, "optim/total_tokens": 678428672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 3.9030356407165527, "created_at": "2025-01-15T09:11:46.402377+00:00"} {"global_step": 1295, "acc_step": 0, "speed/wps": 12868.51006172534, "speed/FLOPS": 202117633705286.72, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22502538561820984, "optim/lr": 0.0007763999999999999, "optim/total_tokens": 678952960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 3.9225754737854004, "created_at": "2025-01-15T09:11:56.592939+00:00"} {"global_step": 1296, "acc_step": 0, "speed/wps": 12872.347366777793, "speed/FLOPS": 202177903854145.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24276290833950043, "optim/lr": 0.000777, "optim/total_tokens": 679477248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.8122477531433105, "created_at": "2025-01-15T09:12:06.782873+00:00"} {"global_step": 1297, "acc_step": 0, "speed/wps": 12864.626412711956, "speed/FLOPS": 202056635660838.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22684796154499054, "optim/lr": 0.0007775999999999999, "optim/total_tokens": 680001536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.8869946002960205, "created_at": "2025-01-15T09:12:16.980786+00:00"} {"global_step": 1298, "acc_step": 0, "speed/wps": 12864.27617370753, "speed/FLOPS": 202051134675995.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22661249339580536, "optim/lr": 0.0007782, "optim/total_tokens": 680525824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.7706494331359863, "created_at": "2025-01-15T09:12:27.174331+00:00"} {"global_step": 1299, "acc_step": 0, "speed/wps": 12866.30393935669, "speed/FLOPS": 202082983521957.28, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.20514030754566193, "optim/lr": 0.0007788, "optim/total_tokens": 681050112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.8880386352539062, "created_at": "2025-01-15T09:12:37.367509+00:00"} {"global_step": 1300, "acc_step": 0, "speed/wps": 12871.190400316073, "speed/FLOPS": 202159732106025.38, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.21182945370674133, "optim/lr": 0.0007794, "optim/total_tokens": 681574400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.8279221057891846, "created_at": "2025-01-15T09:12:47.555046+00:00"} {"global_step": 1301, "acc_step": 0, "speed/wps": 12867.506780026273, "speed/FLOPS": 202101875787549.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20630930364131927, "optim/lr": 0.0007800000000000001, "optim/total_tokens": 682098688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.967339515686035, "created_at": "2025-01-15T09:12:57.745268+00:00"} {"global_step": 1302, "acc_step": 0, "speed/wps": 12871.88113410151, "speed/FLOPS": 202170581029293.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24841901659965515, "optim/lr": 0.0007806, "optim/total_tokens": 682622976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.78149676322937, "created_at": "2025-01-15T09:13:07.934663+00:00"} {"global_step": 1303, "acc_step": 0, "speed/wps": 12868.39125309567, "speed/FLOPS": 202115767652497.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20527686178684235, "optim/lr": 0.0007812000000000001, "optim/total_tokens": 683147264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.8060507774353027, "created_at": "2025-01-15T09:13:18.124650+00:00"} {"global_step": 1304, "acc_step": 0, "speed/wps": 12872.694234124436, "speed/FLOPS": 202183351882469.9, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23692059516906738, "optim/lr": 0.0007818, "optim/total_tokens": 683671552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356561, "loss/out": 3.870594024658203, "created_at": "2025-01-15T09:13:28.313511+00:00"} {"global_step": 1305, "acc_step": 0, "speed/wps": 12865.904314272584, "speed/FLOPS": 202076706860868.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29983067512512207, "optim/lr": 0.0007823999999999999, "optim/total_tokens": 684195840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332882, "loss/out": 3.896411895751953, "created_at": "2025-01-15T09:13:38.507067+00:00"} {"global_step": 1306, "acc_step": 0, "speed/wps": 12871.355719796704, "speed/FLOPS": 202162328675641.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2732115387916565, "optim/lr": 0.0007830000000000001, "optim/total_tokens": 684720128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 3.9540414810180664, "created_at": "2025-01-15T09:13:48.693878+00:00"} {"global_step": 1307, "acc_step": 0, "speed/wps": 12867.671764888883, "speed/FLOPS": 202104467101529.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26250335574150085, "optim/lr": 0.0007836, "optim/total_tokens": 685244416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.898130178451538, "created_at": "2025-01-15T09:13:58.891503+00:00"} {"global_step": 1308, "acc_step": 0, "speed/wps": 12867.209534184583, "speed/FLOPS": 202097207133147.03, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26932331919670105, "optim/lr": 0.0007842000000000001, "optim/total_tokens": 685768704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.887606620788574, "created_at": "2025-01-15T09:14:09.081680+00:00"} {"global_step": 1309, "acc_step": 0, "speed/wps": 12870.497047655563, "speed/FLOPS": 202148842049725.66, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.27000242471694946, "optim/lr": 0.0007848, "optim/total_tokens": 686292992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.868901252746582, "created_at": "2025-01-15T09:14:19.271347+00:00"} {"global_step": 1310, "acc_step": 0, "speed/wps": 12874.02722454816, "speed/FLOPS": 202204288328795.28, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2730284631252289, "optim/lr": 0.0007853999999999999, "optim/total_tokens": 686817280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.7706727981567383, "created_at": "2025-01-15T09:14:29.456442+00:00"} {"global_step": 1311, "acc_step": 0, "speed/wps": 12870.784010724357, "speed/FLOPS": 202153349199049.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2650928497314453, "optim/lr": 0.000786, "optim/total_tokens": 687341568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.8210678100585938, "created_at": "2025-01-15T09:14:39.643636+00:00"} {"global_step": 1312, "acc_step": 0, "speed/wps": 12869.110406370784, "speed/FLOPS": 202127062942903.4, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22216370701789856, "optim/lr": 0.0007865999999999999, "optim/total_tokens": 687865856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.790468215942383, "created_at": "2025-01-15T09:14:49.833170+00:00"} {"global_step": 1313, "acc_step": 0, "speed/wps": 12871.9513406847, "speed/FLOPS": 202171683720156.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20044319331645966, "optim/lr": 0.0007872, "optim/total_tokens": 688390144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.9067037105560303, "created_at": "2025-01-15T09:15:00.025007+00:00"} {"global_step": 1314, "acc_step": 0, "speed/wps": 12870.79821908388, "speed/FLOPS": 202153572360860.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18640638887882233, "optim/lr": 0.0007878000000000001, "optim/total_tokens": 688914432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.819225311279297, "created_at": "2025-01-15T09:15:10.212683+00:00"} {"global_step": 1315, "acc_step": 0, "speed/wps": 12874.89608478998, "speed/FLOPS": 202217934972832.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1972288340330124, "optim/lr": 0.0007884, "optim/total_tokens": 689438720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.8379523754119873, "created_at": "2025-01-15T09:15:20.400747+00:00"} {"global_step": 1316, "acc_step": 0, "speed/wps": 12872.641745232564, "speed/FLOPS": 202182527472295.78, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18631364405155182, "optim/lr": 0.0007890000000000001, "optim/total_tokens": 689963008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.701387405395508, "created_at": "2025-01-15T09:15:30.589052+00:00"} {"global_step": 1317, "acc_step": 0, "speed/wps": 12871.63536352775, "speed/FLOPS": 202166720864708.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17396984994411469, "optim/lr": 0.0007896, "optim/total_tokens": 690487296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.870187997817993, "created_at": "2025-01-15T09:15:40.779265+00:00"} {"global_step": 1318, "acc_step": 0, "speed/wps": 12869.328539863956, "speed/FLOPS": 202130489029160.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2042226642370224, "optim/lr": 0.0007902000000000001, "optim/total_tokens": 691011584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.7503466606140137, "created_at": "2025-01-15T09:15:50.971529+00:00"} {"global_step": 1319, "acc_step": 0, "speed/wps": 12870.388436270841, "speed/FLOPS": 202147136158680.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.2192932367324829, "optim/lr": 0.0007908, "optim/total_tokens": 691535872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.9054715633392334, "created_at": "2025-01-15T09:16:01.168667+00:00"} {"global_step": 1320, "acc_step": 0, "speed/wps": 12871.44170667536, "speed/FLOPS": 202163679217729.1, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20414668321609497, "optim/lr": 0.0007913999999999999, "optim/total_tokens": 692060160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400011, "loss/out": 3.7550883293151855, "created_at": "2025-01-15T09:16:11.360333+00:00"} {"global_step": 1321, "acc_step": 0, "speed/wps": 12869.492533484216, "speed/FLOPS": 202133064774317.53, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1891404688358307, "optim/lr": 0.0007920000000000001, "optim/total_tokens": 692584448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400880, "loss/out": 3.815030574798584, "created_at": "2025-01-15T09:16:21.548787+00:00"} {"global_step": 1322, "acc_step": 0, "speed/wps": 12873.060994834266, "speed/FLOPS": 202189112363399.84, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18482695519924164, "optim/lr": 0.0007926, "optim/total_tokens": 693108736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328312, "loss/out": 3.8081021308898926, "created_at": "2025-01-15T09:16:31.734268+00:00"} {"global_step": 1323, "acc_step": 0, "speed/wps": 12872.232630302562, "speed/FLOPS": 202176101760137.28, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20345456898212433, "optim/lr": 0.0007932000000000001, "optim/total_tokens": 693633024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 3.727086305618286, "created_at": "2025-01-15T09:16:41.923172+00:00"} {"global_step": 1324, "acc_step": 0, "speed/wps": 12874.727606219989, "speed/FLOPS": 202215288785376.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19925928115844727, "optim/lr": 0.0007938, "optim/total_tokens": 694157312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.914459228515625, "created_at": "2025-01-15T09:16:52.113480+00:00"} {"global_step": 1325, "acc_step": 0, "speed/wps": 12875.217673264897, "speed/FLOPS": 202222985961739.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19110780954360962, "optim/lr": 0.0007943999999999999, "optim/total_tokens": 694681600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.781754732131958, "created_at": "2025-01-15T09:17:02.298132+00:00"} {"global_step": 1326, "acc_step": 0, "speed/wps": 12870.474735080237, "speed/FLOPS": 202148491600069.5, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21662463247776031, "optim/lr": 0.000795, "optim/total_tokens": 695205888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.844212293624878, "created_at": "2025-01-15T09:17:12.487424+00:00"} {"global_step": 1327, "acc_step": 0, "speed/wps": 12874.135567606565, "speed/FLOPS": 202205990005406.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2344834804534912, "optim/lr": 0.0007956, "optim/total_tokens": 695730176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.9195852279663086, "created_at": "2025-01-15T09:17:22.673859+00:00"} {"global_step": 1328, "acc_step": 0, "speed/wps": 12875.368395969796, "speed/FLOPS": 202225353268934.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24102695286273956, "optim/lr": 0.0007962000000000001, "optim/total_tokens": 696254464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.8041253089904785, "created_at": "2025-01-15T09:17:32.858342+00:00"} {"global_step": 1329, "acc_step": 0, "speed/wps": 12872.897922865268, "speed/FLOPS": 202186551094042.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23803411424160004, "optim/lr": 0.0007968000000000001, "optim/total_tokens": 696778752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416344, "loss/out": 3.7538251876831055, "created_at": "2025-01-15T09:17:43.051892+00:00"} {"global_step": 1330, "acc_step": 0, "speed/wps": 12870.878347238578, "speed/FLOPS": 202154830883637.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.298067569732666, "optim/lr": 0.0007974, "optim/total_tokens": 697303040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371206, "loss/out": 3.7459464073181152, "created_at": "2025-01-15T09:17:53.239680+00:00"} {"global_step": 1331, "acc_step": 0, "speed/wps": 12861.476302077763, "speed/FLOPS": 202007158844622.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2253417819738388, "optim/lr": 0.0007980000000000001, "optim/total_tokens": 697827328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363445, "loss/out": 3.801762104034424, "created_at": "2025-01-15T09:18:03.435545+00:00"} {"global_step": 1332, "acc_step": 0, "speed/wps": 12874.049318620871, "speed/FLOPS": 202204635346567.7, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21865276992321014, "optim/lr": 0.0007986, "optim/total_tokens": 698351616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 3.7762420177459717, "created_at": "2025-01-15T09:18:13.622615+00:00"} {"global_step": 1333, "acc_step": 0, "speed/wps": 12873.805840497334, "speed/FLOPS": 202200811188063.53, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2288828343153, "optim/lr": 0.0007992000000000001, "optim/total_tokens": 698875904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.795994281768799, "created_at": "2025-01-15T09:18:23.810782+00:00"} {"global_step": 1334, "acc_step": 0, "speed/wps": 12867.02504740481, "speed/FLOPS": 202094309514759.28, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23818568885326385, "optim/lr": 0.0007998, "optim/total_tokens": 699400192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.7520785331726074, "created_at": "2025-01-15T09:18:34.003538+00:00"} {"global_step": 1335, "acc_step": 0, "speed/wps": 12868.58852818407, "speed/FLOPS": 202118866128845.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34990885853767395, "optim/lr": 0.0008003999999999999, "optim/total_tokens": 699924480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483664, "loss/out": 3.841357707977295, "created_at": "2025-01-15T09:18:44.193028+00:00"} {"global_step": 1336, "acc_step": 0, "speed/wps": 12871.401777872927, "speed/FLOPS": 202163052081018.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2533741891384125, "optim/lr": 0.0008010000000000001, "optim/total_tokens": 700448768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.921146869659424, "created_at": "2025-01-15T09:18:54.386453+00:00"} {"global_step": 1337, "acc_step": 0, "speed/wps": 12859.850902366705, "speed/FLOPS": 201981629708627.47, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19491077959537506, "optim/lr": 0.0008016, "optim/total_tokens": 700973056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.8564200401306152, "created_at": "2025-01-15T09:19:04.584921+00:00"} {"global_step": 1338, "acc_step": 0, "speed/wps": 12862.653949215579, "speed/FLOPS": 202025655411179.06, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.21977341175079346, "optim/lr": 0.0008022000000000001, "optim/total_tokens": 701497344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.8941636085510254, "created_at": "2025-01-15T09:19:14.785952+00:00"} {"global_step": 1339, "acc_step": 0, "speed/wps": 12873.768519563511, "speed/FLOPS": 202200225011513.5, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.33256834745407104, "optim/lr": 0.0008028, "optim/total_tokens": 702021632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.701542377471924, "created_at": "2025-01-15T09:19:24.975537+00:00"} {"global_step": 1340, "acc_step": 0, "speed/wps": 12869.75105369378, "speed/FLOPS": 202137125189452.75, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.2988961935043335, "optim/lr": 0.0008034, "optim/total_tokens": 702545920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364786, "loss/out": 3.8071560859680176, "created_at": "2025-01-15T09:19:35.170833+00:00"} {"global_step": 1341, "acc_step": 0, "speed/wps": 12865.898732162672, "speed/FLOPS": 202076619186162.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.44004425406455994, "optim/lr": 0.000804, "optim/total_tokens": 703070208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418245, "loss/out": 3.917369842529297, "created_at": "2025-01-15T09:19:45.367245+00:00"} {"global_step": 1342, "acc_step": 0, "speed/wps": 12873.774587792248, "speed/FLOPS": 202200320321384.4, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2923851013183594, "optim/lr": 0.0008046, "optim/total_tokens": 703594496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 3.8558192253112793, "created_at": "2025-01-15T09:19:55.555286+00:00"} {"global_step": 1343, "acc_step": 0, "speed/wps": 12873.288914420533, "speed/FLOPS": 202192692153696.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5146127343177795, "optim/lr": 0.0008052000000000001, "optim/total_tokens": 704118784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 3.9191689491271973, "created_at": "2025-01-15T09:20:05.745744+00:00"} {"global_step": 1344, "acc_step": 0, "speed/wps": 12872.769597288869, "speed/FLOPS": 202184535564527.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6291230916976929, "optim/lr": 0.0008058000000000001, "optim/total_tokens": 704643072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.8334572315216064, "created_at": "2025-01-15T09:20:15.937474+00:00"} {"global_step": 1345, "acc_step": 0, "speed/wps": 12873.490784988513, "speed/FLOPS": 202195862808365.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9481505155563354, "optim/lr": 0.0008064, "optim/total_tokens": 705167360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 4.079676628112793, "created_at": "2025-01-15T09:20:26.132558+00:00"} {"global_step": 1346, "acc_step": 0, "speed/wps": 12875.193075739207, "speed/FLOPS": 202222599623798.12, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6812074184417725, "optim/lr": 0.0008070000000000001, "optim/total_tokens": 705691648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 4.098267555236816, "created_at": "2025-01-15T09:20:36.320945+00:00"} {"global_step": 1347, "acc_step": 0, "speed/wps": 12876.859938831414, "speed/FLOPS": 202248779999171.16, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6278400421142578, "optim/lr": 0.0008076, "optim/total_tokens": 706215936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.027122497558594, "created_at": "2025-01-15T09:20:46.503301+00:00"} {"global_step": 1348, "acc_step": 0, "speed/wps": 12877.36125571346, "speed/FLOPS": 202256653869685.25, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4300215244293213, "optim/lr": 0.0008081999999999999, "optim/total_tokens": 706740224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506128, "loss/out": 3.9938712120056152, "created_at": "2025-01-15T09:20:56.685419+00:00"} {"global_step": 1349, "acc_step": 0, "speed/wps": 12875.454842852205, "speed/FLOPS": 202226711036013.44, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6334638595581055, "optim/lr": 0.0008088, "optim/total_tokens": 707264512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.9845197200775146, "created_at": "2025-01-15T09:21:06.872011+00:00"} {"global_step": 1350, "acc_step": 0, "speed/wps": 12875.025160648915, "speed/FLOPS": 202219962286565.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5541133284568787, "optim/lr": 0.0008093999999999999, "optim/total_tokens": 707788800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391306, "loss/out": 4.031055927276611, "created_at": "2025-01-15T09:21:17.059427+00:00"} {"global_step": 1351, "acc_step": 0, "speed/wps": 12874.742606517018, "speed/FLOPS": 202215524385654.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.6852818131446838, "optim/lr": 0.0008100000000000001, "optim/total_tokens": 708313088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 4.057586193084717, "created_at": "2025-01-15T09:21:27.246024+00:00"} {"global_step": 1352, "acc_step": 0, "speed/wps": 12877.628484119181, "speed/FLOPS": 202260851059007.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.1293843984603882, "optim/lr": 0.0008106, "optim/total_tokens": 708837376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.151702880859375, "created_at": "2025-01-15T09:21:37.431647+00:00"} {"global_step": 1353, "acc_step": 0, "speed/wps": 12874.430713296557, "speed/FLOPS": 202210625674040.53, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1614882946014404, "optim/lr": 0.0008112, "optim/total_tokens": 709361664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 4.444836139678955, "created_at": "2025-01-15T09:21:47.619036+00:00"} {"global_step": 1354, "acc_step": 0, "speed/wps": 12876.745464740749, "speed/FLOPS": 202246982026273.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.136250615119934, "optim/lr": 0.0008118, "optim/total_tokens": 709885952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 4.306551933288574, "created_at": "2025-01-15T09:21:57.804898+00:00"} {"global_step": 1355, "acc_step": 0, "speed/wps": 12873.706508081423, "speed/FLOPS": 202199251035974.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8925189971923828, "optim/lr": 0.0008124, "optim/total_tokens": 710410240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.247636795043945, "created_at": "2025-01-15T09:22:07.992835+00:00"} {"global_step": 1356, "acc_step": 0, "speed/wps": 12875.512054623923, "speed/FLOPS": 202227609625505.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8558861613273621, "optim/lr": 0.000813, "optim/total_tokens": 710934528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.2162394523620605, "created_at": "2025-01-15T09:22:18.176948+00:00"} {"global_step": 1357, "acc_step": 0, "speed/wps": 12873.683084503222, "speed/FLOPS": 202198883136491.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7392107844352722, "optim/lr": 0.0008136, "optim/total_tokens": 711458816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489248, "loss/out": 4.172072410583496, "created_at": "2025-01-15T09:22:28.365118+00:00"} {"global_step": 1358, "acc_step": 0, "speed/wps": 12870.749544018194, "speed/FLOPS": 202152807852067.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1326619386672974, "optim/lr": 0.0008142, "optim/total_tokens": 711983104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374328, "loss/out": 4.341674327850342, "created_at": "2025-01-15T09:22:38.553554+00:00"} {"global_step": 1359, "acc_step": 0, "speed/wps": 12867.074681291426, "speed/FLOPS": 202095089083152.7, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.6405337452888489, "optim/lr": 0.0008148000000000001, "optim/total_tokens": 712507392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 4.231547832489014, "created_at": "2025-01-15T09:22:48.744485+00:00"} {"global_step": 1360, "acc_step": 0, "speed/wps": 12864.026107724103, "speed/FLOPS": 202047207046098.8, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.5332977771759033, "optim/lr": 0.0008154, "optim/total_tokens": 713031680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 4.233148574829102, "created_at": "2025-01-15T09:22:58.940119+00:00"} {"global_step": 1361, "acc_step": 0, "speed/wps": 12879.099095685331, "speed/FLOPS": 202283949034485.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5217122435569763, "optim/lr": 0.0008160000000000001, "optim/total_tokens": 713555968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 4.201972484588623, "created_at": "2025-01-15T09:23:09.123047+00:00"} {"global_step": 1362, "acc_step": 0, "speed/wps": 12876.898203743127, "speed/FLOPS": 202249381002191.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4542238116264343, "optim/lr": 0.0008166, "optim/total_tokens": 714080256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.149981498718262, "created_at": "2025-01-15T09:23:19.311800+00:00"} {"global_step": 1363, "acc_step": 0, "speed/wps": 12876.258984047332, "speed/FLOPS": 202239341178488.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34332403540611267, "optim/lr": 0.0008171999999999999, "optim/total_tokens": 714604544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 4.044626235961914, "created_at": "2025-01-15T09:23:29.495041+00:00"} {"global_step": 1364, "acc_step": 0, "speed/wps": 12874.332550128689, "speed/FLOPS": 202209083886598.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3011914789676666, "optim/lr": 0.0008178, "optim/total_tokens": 715128832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 4.027914524078369, "created_at": "2025-01-15T09:23:39.683591+00:00"} {"global_step": 1365, "acc_step": 0, "speed/wps": 12871.031869938792, "speed/FLOPS": 202157242168606.6, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31036967039108276, "optim/lr": 0.0008183999999999999, "optim/total_tokens": 715653120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470430, "loss/out": 4.009799003601074, "created_at": "2025-01-15T09:23:49.871035+00:00"} {"global_step": 1366, "acc_step": 0, "speed/wps": 12865.593871162831, "speed/FLOPS": 202071830925236.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.27087005972862244, "optim/lr": 0.0008190000000000001, "optim/total_tokens": 716177408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.009030342102051, "created_at": "2025-01-15T09:24:00.064017+00:00"} {"global_step": 1367, "acc_step": 0, "speed/wps": 12870.985144739241, "speed/FLOPS": 202156508285140.5, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2516094148159027, "optim/lr": 0.0008196, "optim/total_tokens": 716701696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.996833324432373, "created_at": "2025-01-15T09:24:10.253132+00:00"} {"global_step": 1368, "acc_step": 0, "speed/wps": 12875.262841412608, "speed/FLOPS": 202223695389566.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3051396608352661, "optim/lr": 0.0008202, "optim/total_tokens": 717225984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.9844648838043213, "created_at": "2025-01-15T09:24:20.437030+00:00"} {"global_step": 1369, "acc_step": 0, "speed/wps": 12874.03063504556, "speed/FLOPS": 202204341895343.47, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2719227075576782, "optim/lr": 0.0008208, "optim/total_tokens": 717750272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.9158236980438232, "created_at": "2025-01-15T09:24:30.626834+00:00"} {"global_step": 1370, "acc_step": 0, "speed/wps": 12876.321186230914, "speed/FLOPS": 202240318149256.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23268592357635498, "optim/lr": 0.0008214, "optim/total_tokens": 718274560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.9736413955688477, "created_at": "2025-01-15T09:24:40.812699+00:00"} {"global_step": 1371, "acc_step": 0, "speed/wps": 12877.516190350401, "speed/FLOPS": 202259087331060.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2107606828212738, "optim/lr": 0.000822, "optim/total_tokens": 718798848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.9225637912750244, "created_at": "2025-01-15T09:24:51.004804+00:00"} {"global_step": 1372, "acc_step": 0, "speed/wps": 12875.877218350315, "speed/FLOPS": 202233345023616.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2218877375125885, "optim/lr": 0.0008226, "optim/total_tokens": 719323136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 3.9188899993896484, "created_at": "2025-01-15T09:25:01.191672+00:00"} {"global_step": 1373, "acc_step": 0, "speed/wps": 12875.45533993397, "speed/FLOPS": 202226718843365.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2527163624763489, "optim/lr": 0.0008232, "optim/total_tokens": 719847424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 3.870251178741455, "created_at": "2025-01-15T09:25:11.376407+00:00"} {"global_step": 1374, "acc_step": 0, "speed/wps": 12877.29580209865, "speed/FLOPS": 202255625830721.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1937631517648697, "optim/lr": 0.0008238000000000001, "optim/total_tokens": 720371712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.8752377033233643, "created_at": "2025-01-15T09:25:21.561670+00:00"} {"global_step": 1375, "acc_step": 0, "speed/wps": 12873.997911538836, "speed/FLOPS": 202203827927703.56, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23982085287570953, "optim/lr": 0.0008244, "optim/total_tokens": 720896000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.998359203338623, "created_at": "2025-01-15T09:25:31.749512+00:00"} {"global_step": 1376, "acc_step": 0, "speed/wps": 12875.901712470075, "speed/FLOPS": 202233729737425.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.19854119420051575, "optim/lr": 0.0008250000000000001, "optim/total_tokens": 721420288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.889629364013672, "created_at": "2025-01-15T09:25:41.932816+00:00"} {"global_step": 1377, "acc_step": 0, "speed/wps": 12870.481147421862, "speed/FLOPS": 202148592314706.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.18297818303108215, "optim/lr": 0.0008256, "optim/total_tokens": 721944576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.8084828853607178, "created_at": "2025-01-15T09:25:52.122839+00:00"} {"global_step": 1378, "acc_step": 0, "speed/wps": 12872.449847697437, "speed/FLOPS": 202179513457811.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1775505095720291, "optim/lr": 0.0008261999999999999, "optim/total_tokens": 722468864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.8921923637390137, "created_at": "2025-01-15T09:26:02.312563+00:00"} {"global_step": 1379, "acc_step": 0, "speed/wps": 12880.000878254192, "speed/FLOPS": 202298112768908.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16412748396396637, "optim/lr": 0.0008268, "optim/total_tokens": 722993152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 3.8906779289245605, "created_at": "2025-01-15T09:26:12.493000+00:00"} {"global_step": 1380, "acc_step": 0, "speed/wps": 12875.099681852513, "speed/FLOPS": 202221132744470.94, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17595219612121582, "optim/lr": 0.0008273999999999999, "optim/total_tokens": 723517440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.786318302154541, "created_at": "2025-01-15T09:26:22.676992+00:00"} {"global_step": 1381, "acc_step": 0, "speed/wps": 12876.949599822517, "speed/FLOPS": 202250188248243.53, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19590991735458374, "optim/lr": 0.0008280000000000001, "optim/total_tokens": 724041728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.8354995250701904, "created_at": "2025-01-15T09:26:32.863692+00:00"} {"global_step": 1382, "acc_step": 0, "speed/wps": 12878.581977787344, "speed/FLOPS": 202275826987303.53, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2233635038137436, "optim/lr": 0.0008286, "optim/total_tokens": 724566016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.884831666946411, "created_at": "2025-01-15T09:26:43.045264+00:00"} {"global_step": 1383, "acc_step": 0, "speed/wps": 12876.368750101528, "speed/FLOPS": 202241065205204.25, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.235566645860672, "optim/lr": 0.0008292, "optim/total_tokens": 725090304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.7676808834075928, "created_at": "2025-01-15T09:26:53.228854+00:00"} {"global_step": 1384, "acc_step": 0, "speed/wps": 12877.608713175709, "speed/FLOPS": 202260540529172.06, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2002788633108139, "optim/lr": 0.0008298, "optim/total_tokens": 725614592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497150, "loss/out": 3.883784532546997, "created_at": "2025-01-15T09:27:03.413568+00:00"} {"global_step": 1385, "acc_step": 0, "speed/wps": 12874.701624318359, "speed/FLOPS": 202214880703908.7, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19182419776916504, "optim/lr": 0.0008304, "optim/total_tokens": 726138880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347485, "loss/out": 3.901607036590576, "created_at": "2025-01-15T09:27:13.597678+00:00"} {"global_step": 1386, "acc_step": 0, "speed/wps": 12874.218782833166, "speed/FLOPS": 202207297014890.0, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20418615639209747, "optim/lr": 0.0008310000000000001, "optim/total_tokens": 726663168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.7890686988830566, "created_at": "2025-01-15T09:27:23.784927+00:00"} {"global_step": 1387, "acc_step": 0, "speed/wps": 12881.497691994673, "speed/FLOPS": 202321622285540.8, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.219549760222435, "optim/lr": 0.0008316, "optim/total_tokens": 727187456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 3.7125144004821777, "created_at": "2025-01-15T09:27:33.964697+00:00"} {"global_step": 1388, "acc_step": 0, "speed/wps": 12874.098603950524, "speed/FLOPS": 202205409440395.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2200983166694641, "optim/lr": 0.0008322, "optim/total_tokens": 727711744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.8444130420684814, "created_at": "2025-01-15T09:27:44.152497+00:00"} {"global_step": 1389, "acc_step": 0, "speed/wps": 12877.020894992245, "speed/FLOPS": 202251308036853.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1779063493013382, "optim/lr": 0.0008328000000000001, "optim/total_tokens": 728236032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 3.7988057136535645, "created_at": "2025-01-15T09:27:54.337951+00:00"} {"global_step": 1390, "acc_step": 0, "speed/wps": 12872.471176950761, "speed/FLOPS": 202179848463044.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20639647543430328, "optim/lr": 0.0008334, "optim/total_tokens": 728760320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.8311028480529785, "created_at": "2025-01-15T09:28:04.527111+00:00"} {"global_step": 1391, "acc_step": 0, "speed/wps": 12874.270447240497, "speed/FLOPS": 202208108475401.44, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1869211345911026, "optim/lr": 0.0008340000000000001, "optim/total_tokens": 729284608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.944714307785034, "created_at": "2025-01-15T09:28:14.714882+00:00"} {"global_step": 1392, "acc_step": 0, "speed/wps": 12878.886719219996, "speed/FLOPS": 202280613370261.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20423761010169983, "optim/lr": 0.0008346, "optim/total_tokens": 729808896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457361, "loss/out": 3.693011522293091, "created_at": "2025-01-15T09:28:24.897913+00:00"} {"global_step": 1393, "acc_step": 0, "speed/wps": 12875.59073724798, "speed/FLOPS": 202228845444237.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1868078112602234, "optim/lr": 0.0008351999999999999, "optim/total_tokens": 730333184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.8915982246398926, "created_at": "2025-01-15T09:28:35.087521+00:00"} {"global_step": 1394, "acc_step": 0, "speed/wps": 12877.275572658398, "speed/FLOPS": 202255308099564.66, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19905047118663788, "optim/lr": 0.0008358, "optim/total_tokens": 730857472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486618, "loss/out": 3.8322761058807373, "created_at": "2025-01-15T09:28:45.271402+00:00"} {"global_step": 1395, "acc_step": 0, "speed/wps": 12873.271209215682, "speed/FLOPS": 202192414069124.8, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19432280957698822, "optim/lr": 0.0008364, "optim/total_tokens": 731381760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.8645858764648438, "created_at": "2025-01-15T09:28:55.459188+00:00"} {"global_step": 1396, "acc_step": 0, "speed/wps": 12878.075885256638, "speed/FLOPS": 202267878108664.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21709471940994263, "optim/lr": 0.0008370000000000001, "optim/total_tokens": 731906048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.627704381942749, "created_at": "2025-01-15T09:29:05.643206+00:00"} {"global_step": 1397, "acc_step": 0, "speed/wps": 12872.880077442163, "speed/FLOPS": 202186270807150.3, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1818871796131134, "optim/lr": 0.0008376, "optim/total_tokens": 732430336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464555, "loss/out": 3.79697322845459, "created_at": "2025-01-15T09:29:15.831173+00:00"} {"global_step": 1398, "acc_step": 0, "speed/wps": 12878.40480705519, "speed/FLOPS": 202273044277496.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1896507441997528, "optim/lr": 0.0008382, "optim/total_tokens": 732954624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 3.676626682281494, "created_at": "2025-01-15T09:29:26.012429+00:00"} {"global_step": 1399, "acc_step": 0, "speed/wps": 12875.074250860373, "speed/FLOPS": 202220733315793.53, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1678706854581833, "optim/lr": 0.0008388000000000001, "optim/total_tokens": 733478912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.7918057441711426, "created_at": "2025-01-15T09:29:36.199683+00:00"} {"global_step": 1400, "acc_step": 0, "speed/wps": 12872.639617663583, "speed/FLOPS": 202182494055901.34, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.15644627809524536, "optim/lr": 0.0008394, "optim/total_tokens": 734003200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.7140278816223145, "created_at": "2025-01-15T09:29:46.389388+00:00"} {"global_step": 1401, "acc_step": 0, "speed/wps": 12869.888626017795, "speed/FLOPS": 202139285951845.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16558489203453064, "optim/lr": 0.0008400000000000001, "optim/total_tokens": 734527488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.799557685852051, "created_at": "2025-01-15T09:29:56.579751+00:00"} {"global_step": 1402, "acc_step": 0, "speed/wps": 12876.19943662224, "speed/FLOPS": 202238405904350.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16405050456523895, "optim/lr": 0.0008406, "optim/total_tokens": 735051776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.8822150230407715, "created_at": "2025-01-15T09:30:06.763390+00:00"} {"global_step": 1403, "acc_step": 0, "speed/wps": 12871.007270747488, "speed/FLOPS": 202156855804504.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18679623305797577, "optim/lr": 0.0008412, "optim/total_tokens": 735576064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 3.841763496398926, "created_at": "2025-01-15T09:30:16.950975+00:00"} {"global_step": 1404, "acc_step": 0, "speed/wps": 12873.035095017292, "speed/FLOPS": 202188705571184.2, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20100820064544678, "optim/lr": 0.0008418000000000001, "optim/total_tokens": 736100352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323067, "loss/out": 3.7351646423339844, "created_at": "2025-01-15T09:30:27.140691+00:00"} {"global_step": 1405, "acc_step": 0, "speed/wps": 12876.749623629195, "speed/FLOPS": 202247047347331.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22510939836502075, "optim/lr": 0.0008424, "optim/total_tokens": 736624640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.8433361053466797, "created_at": "2025-01-15T09:30:37.323516+00:00"} {"global_step": 1406, "acc_step": 0, "speed/wps": 12869.434580842932, "speed/FLOPS": 202132154548450.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21043233573436737, "optim/lr": 0.0008430000000000001, "optim/total_tokens": 737148928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.700173854827881, "created_at": "2025-01-15T09:30:47.513763+00:00"} {"global_step": 1407, "acc_step": 0, "speed/wps": 12872.448961138592, "speed/FLOPS": 202179499533186.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22464686632156372, "optim/lr": 0.0008436, "optim/total_tokens": 737673216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.686281681060791, "created_at": "2025-01-15T09:30:57.703209+00:00"} {"global_step": 1408, "acc_step": 0, "speed/wps": 12877.5058861599, "speed/FLOPS": 202258925489589.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24828138947486877, "optim/lr": 0.0008441999999999999, "optim/total_tokens": 738197504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.7763478755950928, "created_at": "2025-01-15T09:31:07.893439+00:00"} {"global_step": 1409, "acc_step": 0, "speed/wps": 12871.634240459894, "speed/FLOPS": 202166703225384.38, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2450607270002365, "optim/lr": 0.0008448, "optim/total_tokens": 738721792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.9349312782287598, "created_at": "2025-01-15T09:31:18.080997+00:00"} {"global_step": 1410, "acc_step": 0, "speed/wps": 12872.97022818438, "speed/FLOPS": 202187686748436.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20016241073608398, "optim/lr": 0.0008454, "optim/total_tokens": 739246080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 3.6312050819396973, "created_at": "2025-01-15T09:31:28.266635+00:00"} {"global_step": 1411, "acc_step": 0, "speed/wps": 12872.779988095448, "speed/FLOPS": 202184698766422.97, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1863977611064911, "optim/lr": 0.000846, "optim/total_tokens": 739770368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457391, "loss/out": 3.8365020751953125, "created_at": "2025-01-15T09:31:38.452299+00:00"} {"global_step": 1412, "acc_step": 0, "speed/wps": 12870.165673203861, "speed/FLOPS": 202143637358605.47, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19422975182533264, "optim/lr": 0.0008466000000000001, "optim/total_tokens": 740294656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.854325771331787, "created_at": "2025-01-15T09:31:48.640507+00:00"} {"global_step": 1413, "acc_step": 0, "speed/wps": 12872.17972016262, "speed/FLOPS": 202175270733683.3, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18900448083877563, "optim/lr": 0.0008472, "optim/total_tokens": 740818944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.737981081008911, "created_at": "2025-01-15T09:31:58.827867+00:00"} {"global_step": 1414, "acc_step": 0, "speed/wps": 12876.13965649517, "speed/FLOPS": 202237466975308.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18675446510314941, "optim/lr": 0.0008478000000000001, "optim/total_tokens": 741343232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.852020740509033, "created_at": "2025-01-15T09:32:09.013927+00:00"} {"global_step": 1415, "acc_step": 0, "speed/wps": 12873.572139555108, "speed/FLOPS": 202197140593627.12, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2379002869129181, "optim/lr": 0.0008484, "optim/total_tokens": 741867520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7222471237182617, "created_at": "2025-01-15T09:32:19.203359+00:00"} {"global_step": 1416, "acc_step": 0, "speed/wps": 12874.228677814634, "speed/FLOPS": 202207452429171.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24281714856624603, "optim/lr": 0.0008489999999999999, "optim/total_tokens": 742391808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.747832775115967, "created_at": "2025-01-15T09:32:29.389382+00:00"} {"global_step": 1417, "acc_step": 0, "speed/wps": 12876.955089022978, "speed/FLOPS": 202250274463679.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2116052508354187, "optim/lr": 0.0008496, "optim/total_tokens": 742916096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.707524299621582, "created_at": "2025-01-15T09:32:39.573958+00:00"} {"global_step": 1418, "acc_step": 0, "speed/wps": 12875.928857194993, "speed/FLOPS": 202234156082630.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2071988731622696, "optim/lr": 0.0008502, "optim/total_tokens": 743440384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431584, "loss/out": 3.7737183570861816, "created_at": "2025-01-15T09:32:49.757536+00:00"} {"global_step": 1419, "acc_step": 0, "speed/wps": 12876.133005261432, "speed/FLOPS": 202237362508543.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20915935933589935, "optim/lr": 0.0008508000000000001, "optim/total_tokens": 743964672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.848724842071533, "created_at": "2025-01-15T09:32:59.945176+00:00"} {"global_step": 1420, "acc_step": 0, "speed/wps": 12873.675398347401, "speed/FLOPS": 202198762414852.53, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20442330837249756, "optim/lr": 0.0008514, "optim/total_tokens": 744488960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.716087579727173, "created_at": "2025-01-15T09:33:10.131859+00:00"} {"global_step": 1421, "acc_step": 0, "speed/wps": 12876.745428217942, "speed/FLOPS": 202246981452632.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21390148997306824, "optim/lr": 0.0008519999999999999, "optim/total_tokens": 745013248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.8271851539611816, "created_at": "2025-01-15T09:33:20.314707+00:00"} {"global_step": 1422, "acc_step": 0, "speed/wps": 12866.583985878653, "speed/FLOPS": 202087382037408.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22877606749534607, "optim/lr": 0.0008526, "optim/total_tokens": 745537536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.7611947059631348, "created_at": "2025-01-15T09:33:30.506957+00:00"} {"global_step": 1423, "acc_step": 0, "speed/wps": 12875.318854737669, "speed/FLOPS": 202224575155807.4, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.223724365234375, "optim/lr": 0.0008531999999999999, "optim/total_tokens": 746061824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.794412136077881, "created_at": "2025-01-15T09:33:40.691911+00:00"} {"global_step": 1424, "acc_step": 0, "speed/wps": 12874.057797790496, "speed/FLOPS": 202204768523578.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23299206793308258, "optim/lr": 0.0008538, "optim/total_tokens": 746586112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.920063018798828, "created_at": "2025-01-15T09:33:50.880661+00:00"} {"global_step": 1425, "acc_step": 0, "speed/wps": 12876.202856517022, "speed/FLOPS": 202238459618497.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20782260596752167, "optim/lr": 0.0008544000000000001, "optim/total_tokens": 747110400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 3.7169179916381836, "created_at": "2025-01-15T09:34:01.064097+00:00"} {"global_step": 1426, "acc_step": 0, "speed/wps": 12872.206402990512, "speed/FLOPS": 202175689824161.44, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19231288135051727, "optim/lr": 0.000855, "optim/total_tokens": 747634688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.5777010917663574, "created_at": "2025-01-15T09:34:11.254001+00:00"} {"global_step": 1427, "acc_step": 0, "speed/wps": 12873.191608224322, "speed/FLOPS": 202191163826172.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1860007345676422, "optim/lr": 0.0008556000000000001, "optim/total_tokens": 748158976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.846038579940796, "created_at": "2025-01-15T09:34:21.441118+00:00"} {"global_step": 1428, "acc_step": 0, "speed/wps": 12872.306789526978, "speed/FLOPS": 202177266532662.2, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25492748618125916, "optim/lr": 0.0008562, "optim/total_tokens": 748683264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.716062545776367, "created_at": "2025-01-15T09:34:31.631213+00:00"} {"global_step": 1429, "acc_step": 0, "speed/wps": 12872.839887742985, "speed/FLOPS": 202185639572698.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24577929079532623, "optim/lr": 0.0008568000000000001, "optim/total_tokens": 749207552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.798248767852783, "created_at": "2025-01-15T09:34:41.824743+00:00"} {"global_step": 1430, "acc_step": 0, "speed/wps": 12870.215370162776, "speed/FLOPS": 202144417917636.0, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25309160351753235, "optim/lr": 0.0008574, "optim/total_tokens": 749731840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.836223602294922, "created_at": "2025-01-15T09:34:52.013592+00:00"} {"global_step": 1431, "acc_step": 0, "speed/wps": 12877.12842951784, "speed/FLOPS": 202252997014346.3, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23234984278678894, "optim/lr": 0.0008579999999999999, "optim/total_tokens": 750256128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.7470226287841797, "created_at": "2025-01-15T09:35:02.196727+00:00"} {"global_step": 1432, "acc_step": 0, "speed/wps": 12874.898672453557, "speed/FLOPS": 202217975615644.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19817999005317688, "optim/lr": 0.0008586, "optim/total_tokens": 750780416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310495, "loss/out": 3.813887596130371, "created_at": "2025-01-15T09:35:12.380903+00:00"} {"global_step": 1433, "acc_step": 0, "speed/wps": 12876.578359762296, "speed/FLOPS": 202244357412960.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19728362560272217, "optim/lr": 0.0008592, "optim/total_tokens": 751304704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.751941204071045, "created_at": "2025-01-15T09:35:22.563645+00:00"} {"global_step": 1434, "acc_step": 0, "speed/wps": 12876.767836764666, "speed/FLOPS": 202247333409651.38, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20995816588401794, "optim/lr": 0.0008598000000000001, "optim/total_tokens": 751828992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.754016637802124, "created_at": "2025-01-15T09:35:32.746279+00:00"} {"global_step": 1435, "acc_step": 0, "speed/wps": 12876.433855000263, "speed/FLOPS": 202242087767102.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25564220547676086, "optim/lr": 0.0008604, "optim/total_tokens": 752353280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.7399888038635254, "created_at": "2025-01-15T09:35:42.938621+00:00"} {"global_step": 1436, "acc_step": 0, "speed/wps": 12874.25321748142, "speed/FLOPS": 202207837858359.0, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21601660549640656, "optim/lr": 0.000861, "optim/total_tokens": 752877568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.6926746368408203, "created_at": "2025-01-15T09:35:53.123984+00:00"} {"global_step": 1437, "acc_step": 0, "speed/wps": 12873.299014039447, "speed/FLOPS": 202192850782089.88, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2195006012916565, "optim/lr": 0.0008616, "optim/total_tokens": 753401856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.8454179763793945, "created_at": "2025-01-15T09:36:03.310379+00:00"} {"global_step": 1438, "acc_step": 0, "speed/wps": 12872.396601401722, "speed/FLOPS": 202178677151569.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19847027957439423, "optim/lr": 0.0008622, "optim/total_tokens": 753926144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.8074262142181396, "created_at": "2025-01-15T09:36:13.502644+00:00"} {"global_step": 1439, "acc_step": 0, "speed/wps": 12873.43460740308, "speed/FLOPS": 202194980462190.53, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20387299358844757, "optim/lr": 0.0008628, "optim/total_tokens": 754450432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277225, "loss/out": 3.7806529998779297, "created_at": "2025-01-15T09:36:23.689605+00:00"} {"global_step": 1440, "acc_step": 0, "speed/wps": 12874.347807425036, "speed/FLOPS": 202209323523403.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21843791007995605, "optim/lr": 0.0008634000000000001, "optim/total_tokens": 754974720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.834723949432373, "created_at": "2025-01-15T09:36:33.881870+00:00"} {"global_step": 1441, "acc_step": 0, "speed/wps": 12872.742394060077, "speed/FLOPS": 202184108300437.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1872173249721527, "optim/lr": 0.000864, "optim/total_tokens": 755499008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 3.6823763847351074, "created_at": "2025-01-15T09:36:44.067951+00:00"} {"global_step": 1442, "acc_step": 0, "speed/wps": 12870.866231548967, "speed/FLOPS": 202154640590083.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17196431756019592, "optim/lr": 0.0008646000000000001, "optim/total_tokens": 756023296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.7244627475738525, "created_at": "2025-01-15T09:36:54.255776+00:00"} {"global_step": 1443, "acc_step": 0, "speed/wps": 12875.350571966104, "speed/FLOPS": 202225073318464.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20812435448169708, "optim/lr": 0.0008652, "optim/total_tokens": 756547584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 3.712038993835449, "created_at": "2025-01-15T09:37:04.443129+00:00"} {"global_step": 1444, "acc_step": 0, "speed/wps": 12876.4415091073, "speed/FLOPS": 202242207985371.25, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2138432115316391, "optim/lr": 0.0008658000000000001, "optim/total_tokens": 757071872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 3.7603015899658203, "created_at": "2025-01-15T09:37:14.628120+00:00"} {"global_step": 1445, "acc_step": 0, "speed/wps": 12874.713852052473, "speed/FLOPS": 202215072757275.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23262356221675873, "optim/lr": 0.0008664, "optim/total_tokens": 757596160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464342, "loss/out": 3.692673683166504, "created_at": "2025-01-15T09:37:24.813663+00:00"} {"global_step": 1446, "acc_step": 0, "speed/wps": 12872.25254485899, "speed/FLOPS": 202176414545611.44, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2306285798549652, "optim/lr": 0.0008669999999999999, "optim/total_tokens": 758120448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 3.7178730964660645, "created_at": "2025-01-15T09:37:35.001676+00:00"} {"global_step": 1447, "acc_step": 0, "speed/wps": 12874.221367869108, "speed/FLOPS": 202207337616431.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21496964991092682, "optim/lr": 0.0008676000000000001, "optim/total_tokens": 758644736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 3.7705001831054688, "created_at": "2025-01-15T09:37:45.188473+00:00"} {"global_step": 1448, "acc_step": 0, "speed/wps": 12872.478687417299, "speed/FLOPS": 202179966425242.12, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.23297983407974243, "optim/lr": 0.0008682, "optim/total_tokens": 759169024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.7767720222473145, "created_at": "2025-01-15T09:37:55.377260+00:00"} {"global_step": 1449, "acc_step": 0, "speed/wps": 12877.274468636375, "speed/FLOPS": 202255290759381.7, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2678036689758301, "optim/lr": 0.0008688000000000001, "optim/total_tokens": 759693312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.805767059326172, "created_at": "2025-01-15T09:38:05.560429+00:00"} {"global_step": 1450, "acc_step": 0, "speed/wps": 12874.458792831023, "speed/FLOPS": 202211066701714.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2549842298030853, "optim/lr": 0.0008694, "optim/total_tokens": 760217600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.7561347484588623, "created_at": "2025-01-15T09:38:15.745515+00:00"} {"global_step": 1451, "acc_step": 0, "speed/wps": 12874.53431445494, "speed/FLOPS": 202212252872595.84, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22516565024852753, "optim/lr": 0.00087, "optim/total_tokens": 760741888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.6679089069366455, "created_at": "2025-01-15T09:38:25.930843+00:00"} {"global_step": 1452, "acc_step": 0, "speed/wps": 12876.790549355486, "speed/FLOPS": 202247690142100.56, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17852947115898132, "optim/lr": 0.0008706, "optim/total_tokens": 761266176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430484, "loss/out": 3.7050905227661133, "created_at": "2025-01-15T09:38:36.116911+00:00"} {"global_step": 1453, "acc_step": 0, "speed/wps": 12874.544934239097, "speed/FLOPS": 202212419670899.12, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22902074456214905, "optim/lr": 0.0008712, "optim/total_tokens": 761790464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 3.6229403018951416, "created_at": "2025-01-15T09:38:46.301989+00:00"} {"global_step": 1454, "acc_step": 0, "speed/wps": 12871.488528188127, "speed/FLOPS": 202164414613926.62, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23313085734844208, "optim/lr": 0.0008718, "optim/total_tokens": 762314752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 3.6950111389160156, "created_at": "2025-01-15T09:38:56.491100+00:00"} {"global_step": 1455, "acc_step": 0, "speed/wps": 12868.114729395886, "speed/FLOPS": 202111424467809.1, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19849824905395508, "optim/lr": 0.0008724000000000001, "optim/total_tokens": 762839040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 3.7275009155273438, "created_at": "2025-01-15T09:39:06.683830+00:00"} {"global_step": 1456, "acc_step": 0, "speed/wps": 12864.041643966515, "speed/FLOPS": 202047451064134.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20210805535316467, "optim/lr": 0.000873, "optim/total_tokens": 763363328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 3.7376914024353027, "created_at": "2025-01-15T09:39:16.877543+00:00"} {"global_step": 1457, "acc_step": 0, "speed/wps": 12866.573913261525, "speed/FLOPS": 202087223833115.72, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25253960490226746, "optim/lr": 0.0008736000000000001, "optim/total_tokens": 763887616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 3.749675989151001, "created_at": "2025-01-15T09:39:27.068770+00:00"} {"global_step": 1458, "acc_step": 0, "speed/wps": 12871.854257196917, "speed/FLOPS": 202170158890574.0, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.212968111038208, "optim/lr": 0.0008742, "optim/total_tokens": 764411904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.7557730674743652, "created_at": "2025-01-15T09:39:37.262074+00:00"} {"global_step": 1459, "acc_step": 0, "speed/wps": 12868.369929124427, "speed/FLOPS": 202115432730227.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22462771832942963, "optim/lr": 0.0008748000000000001, "optim/total_tokens": 764936192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.8123250007629395, "created_at": "2025-01-15T09:39:47.452129+00:00"} {"global_step": 1460, "acc_step": 0, "speed/wps": 12874.74595732621, "speed/FLOPS": 202215577014716.9, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16920001804828644, "optim/lr": 0.0008754, "optim/total_tokens": 765460480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.7524471282958984, "created_at": "2025-01-15T09:39:57.637368+00:00"} {"global_step": 1461, "acc_step": 0, "speed/wps": 12871.239231745505, "speed/FLOPS": 202160499070726.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1945428103208542, "optim/lr": 0.0008759999999999999, "optim/total_tokens": 765984768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.786620616912842, "created_at": "2025-01-15T09:40:07.824638+00:00"} {"global_step": 1462, "acc_step": 0, "speed/wps": 12869.351127367447, "speed/FLOPS": 202130843796941.3, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18967095017433167, "optim/lr": 0.0008766000000000001, "optim/total_tokens": 766509056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.7158985137939453, "created_at": "2025-01-15T09:40:18.015744+00:00"} {"global_step": 1463, "acc_step": 0, "speed/wps": 12871.105165445293, "speed/FLOPS": 202158393375254.78, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19498972594738007, "optim/lr": 0.0008772, "optim/total_tokens": 767033344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.736891746520996, "created_at": "2025-01-15T09:40:28.207033+00:00"} {"global_step": 1464, "acc_step": 0, "speed/wps": 12869.189959251558, "speed/FLOPS": 202128312430213.1, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22094377875328064, "optim/lr": 0.0008778000000000001, "optim/total_tokens": 767557632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 3.745652198791504, "created_at": "2025-01-15T09:40:38.403720+00:00"} {"global_step": 1465, "acc_step": 0, "speed/wps": 12875.657177148956, "speed/FLOPS": 202229888974180.4, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23953935503959656, "optim/lr": 0.0008784, "optim/total_tokens": 768081920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.6681504249572754, "created_at": "2025-01-15T09:40:48.587298+00:00"} {"global_step": 1466, "acc_step": 0, "speed/wps": 12873.717493911452, "speed/FLOPS": 202199423583531.3, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21968437731266022, "optim/lr": 0.000879, "optim/total_tokens": 768606208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.8073854446411133, "created_at": "2025-01-15T09:40:58.774615+00:00"} {"global_step": 1467, "acc_step": 0, "speed/wps": 12872.029561856652, "speed/FLOPS": 202172912291148.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26098790764808655, "optim/lr": 0.0008796, "optim/total_tokens": 769130496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.8283169269561768, "created_at": "2025-01-15T09:41:08.964394+00:00"} {"global_step": 1468, "acc_step": 0, "speed/wps": 12870.360520435032, "speed/FLOPS": 202146697702119.16, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22463713586330414, "optim/lr": 0.0008802, "optim/total_tokens": 769654784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.7794790267944336, "created_at": "2025-01-15T09:41:19.154878+00:00"} {"global_step": 1469, "acc_step": 0, "speed/wps": 12867.72740864552, "speed/FLOPS": 202105341063190.12, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21937565505504608, "optim/lr": 0.0008808, "optim/total_tokens": 770179072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7664475440979004, "created_at": "2025-01-15T09:41:29.347282+00:00"} {"global_step": 1470, "acc_step": 0, "speed/wps": 12876.498922857845, "speed/FLOPS": 202243109747218.3, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1906002312898636, "optim/lr": 0.0008814000000000001, "optim/total_tokens": 770703360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.6342709064483643, "created_at": "2025-01-15T09:41:39.532361+00:00"} {"global_step": 1471, "acc_step": 0, "speed/wps": 12870.375006806084, "speed/FLOPS": 202146925230482.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18383851647377014, "optim/lr": 0.000882, "optim/total_tokens": 771227648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.7186496257781982, "created_at": "2025-01-15T09:41:49.721759+00:00"} {"global_step": 1472, "acc_step": 0, "speed/wps": 12872.8293896928, "speed/FLOPS": 202185474686394.3, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23594927787780762, "optim/lr": 0.0008826000000000001, "optim/total_tokens": 771751936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.7593584060668945, "created_at": "2025-01-15T09:41:59.907546+00:00"} {"global_step": 1473, "acc_step": 0, "speed/wps": 12851.619142199093, "speed/FLOPS": 201852338603571.3, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.2131044566631317, "optim/lr": 0.0008832, "optim/total_tokens": 772276224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302801, "loss/out": 3.765256881713867, "created_at": "2025-01-15T09:42:10.116365+00:00"} {"global_step": 1474, "acc_step": 0, "speed/wps": 12859.85141117008, "speed/FLOPS": 201981637700083.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.22346459329128265, "optim/lr": 0.0008837999999999999, "optim/total_tokens": 772800512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 3.6997625827789307, "created_at": "2025-01-15T09:42:20.312364+00:00"} {"global_step": 1475, "acc_step": 0, "speed/wps": 12859.088718988707, "speed/FLOPS": 201969658571326.3, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.2548677325248718, "optim/lr": 0.0008844, "optim/total_tokens": 773324800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290558, "loss/out": 3.608945369720459, "created_at": "2025-01-15T09:42:30.511133+00:00"} {"global_step": 1476, "acc_step": 0, "speed/wps": 12866.394937847754, "speed/FLOPS": 202084412778304.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2134331315755844, "optim/lr": 0.0008849999999999999, "optim/total_tokens": 773849088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 3.747030019760132, "created_at": "2025-01-15T09:42:40.702960+00:00"} {"global_step": 1477, "acc_step": 0, "speed/wps": 12867.734913577555, "speed/FLOPS": 202105458938460.6, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23096969723701477, "optim/lr": 0.0008856000000000001, "optim/total_tokens": 774373376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.774320125579834, "created_at": "2025-01-15T09:42:50.895283+00:00"} {"global_step": 1478, "acc_step": 0, "speed/wps": 12867.030898706289, "speed/FLOPS": 202094401417489.53, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2185780107975006, "optim/lr": 0.0008862, "optim/total_tokens": 774897664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.7657899856567383, "created_at": "2025-01-15T09:43:01.085712+00:00"} {"global_step": 1479, "acc_step": 0, "speed/wps": 12867.413221150104, "speed/FLOPS": 202100406316835.75, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20519709587097168, "optim/lr": 0.0008868, "optim/total_tokens": 775421952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.7589569091796875, "created_at": "2025-01-15T09:43:11.277880+00:00"} {"global_step": 1480, "acc_step": 0, "speed/wps": 12865.143371288805, "speed/FLOPS": 202064755205663.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21227119863033295, "optim/lr": 0.0008874, "optim/total_tokens": 775946240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.7976365089416504, "created_at": "2025-01-15T09:43:21.472085+00:00"} {"global_step": 1481, "acc_step": 0, "speed/wps": 12868.586218395358, "speed/FLOPS": 202118829850439.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18944047391414642, "optim/lr": 0.000888, "optim/total_tokens": 776470528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.587484359741211, "created_at": "2025-01-15T09:43:31.663404+00:00"} {"global_step": 1482, "acc_step": 0, "speed/wps": 12869.873106248413, "speed/FLOPS": 202139042192541.03, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18939603865146637, "optim/lr": 0.0008886, "optim/total_tokens": 776994816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.821009635925293, "created_at": "2025-01-15T09:43:41.854737+00:00"} {"global_step": 1483, "acc_step": 0, "speed/wps": 12871.846825158089, "speed/FLOPS": 202170042160191.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2124432623386383, "optim/lr": 0.0008892, "optim/total_tokens": 777519104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505601, "loss/out": 3.7179486751556396, "created_at": "2025-01-15T09:43:52.041177+00:00"} {"global_step": 1484, "acc_step": 0, "speed/wps": 12871.460175551903, "speed/FLOPS": 202163969296814.12, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24905216693878174, "optim/lr": 0.0008897999999999999, "optim/total_tokens": 778043392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.847041606903076, "created_at": "2025-01-15T09:44:02.235159+00:00"} {"global_step": 1485, "acc_step": 0, "speed/wps": 12868.397120935944, "speed/FLOPS": 202115859814992.34, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2534213364124298, "optim/lr": 0.0008904000000000001, "optim/total_tokens": 778567680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 272116, "loss/out": 3.768455982208252, "created_at": "2025-01-15T09:44:12.424499+00:00"} {"global_step": 1486, "acc_step": 0, "speed/wps": 12869.610999980445, "speed/FLOPS": 202134925453430.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.23755580186843872, "optim/lr": 0.000891, "optim/total_tokens": 779091968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.9610419273376465, "created_at": "2025-01-15T09:44:22.612687+00:00"} {"global_step": 1487, "acc_step": 0, "speed/wps": 12864.948783127336, "speed/FLOPS": 202061698931195.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.200016587972641, "optim/lr": 0.0008916000000000001, "optim/total_tokens": 779616256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.8673667907714844, "created_at": "2025-01-15T09:44:32.807067+00:00"} {"global_step": 1488, "acc_step": 0, "speed/wps": 12869.739322610183, "speed/FLOPS": 202136940936664.66, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2238396406173706, "optim/lr": 0.0008922, "optim/total_tokens": 780140544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.962216854095459, "created_at": "2025-01-15T09:44:42.999492+00:00"} {"global_step": 1489, "acc_step": 0, "speed/wps": 12872.024992787306, "speed/FLOPS": 202172840527636.34, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19049377739429474, "optim/lr": 0.0008927999999999999, "optim/total_tokens": 780664832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.779055595397949, "created_at": "2025-01-15T09:44:53.187198+00:00"} {"global_step": 1490, "acc_step": 0, "speed/wps": 12868.964718286315, "speed/FLOPS": 202124774711340.44, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.21017040312290192, "optim/lr": 0.0008934, "optim/total_tokens": 781189120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 3.7420222759246826, "created_at": "2025-01-15T09:45:03.378781+00:00"} {"global_step": 1491, "acc_step": 0, "speed/wps": 12876.534689194821, "speed/FLOPS": 202243671506688.62, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.2140142172574997, "optim/lr": 0.0008939999999999999, "optim/total_tokens": 781713408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.7613580226898193, "created_at": "2025-01-15T09:45:13.561414+00:00"} {"global_step": 1492, "acc_step": 0, "speed/wps": 12875.536630987679, "speed/FLOPS": 202227995631069.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1922537386417389, "optim/lr": 0.0008946000000000001, "optim/total_tokens": 782237696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.787092447280884, "created_at": "2025-01-15T09:45:23.747165+00:00"} {"global_step": 1493, "acc_step": 0, "speed/wps": 12870.422551474514, "speed/FLOPS": 202147671984833.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20356041193008423, "optim/lr": 0.0008952, "optim/total_tokens": 782761984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 3.792767286300659, "created_at": "2025-01-15T09:45:33.937912+00:00"} {"global_step": 1494, "acc_step": 0, "speed/wps": 12869.670211097662, "speed/FLOPS": 202135855445391.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.42542386054992676, "optim/lr": 0.0008958, "optim/total_tokens": 783286272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 3.6709682941436768, "created_at": "2025-01-15T09:45:44.135866+00:00"} {"global_step": 1495, "acc_step": 0, "speed/wps": 12858.85153043703, "speed/FLOPS": 201965933199191.0, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25819069147109985, "optim/lr": 0.0008964, "optim/total_tokens": 783810560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.7422399520874023, "created_at": "2025-01-15T09:45:54.332565+00:00"} {"global_step": 1496, "acc_step": 0, "speed/wps": 12867.34141117088, "speed/FLOPS": 202099278442434.78, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18192420899868011, "optim/lr": 0.000897, "optim/total_tokens": 784334848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415004, "loss/out": 3.8747594356536865, "created_at": "2025-01-15T09:46:04.524306+00:00"} {"global_step": 1497, "acc_step": 0, "speed/wps": 12868.7213818486, "speed/FLOPS": 202120952778206.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18267332017421722, "optim/lr": 0.0008976000000000001, "optim/total_tokens": 784859136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 3.6843719482421875, "created_at": "2025-01-15T09:46:14.716455+00:00"} {"global_step": 1498, "acc_step": 0, "speed/wps": 12873.433772520742, "speed/FLOPS": 202194967349216.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1968553215265274, "optim/lr": 0.0008982, "optim/total_tokens": 785383424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.750845432281494, "created_at": "2025-01-15T09:46:24.901549+00:00"} {"global_step": 1499, "acc_step": 0, "speed/wps": 12875.56649756829, "speed/FLOPS": 202228464726759.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6153015494346619, "optim/lr": 0.0008988, "optim/total_tokens": 785907712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.8599376678466797, "created_at": "2025-01-15T09:46:35.090084+00:00"} {"global_step": 1500, "acc_step": 0, "speed/wps": 12872.453473995178, "speed/FLOPS": 202179570413800.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4153260886669159, "optim/lr": 0.0008994000000000001, "optim/total_tokens": 786432000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.895615339279175, "created_at": "2025-01-15T09:46:45.278489+00:00"} {"global_step": 1501, "acc_step": 0, "speed/wps": 12870.506588520848, "speed/FLOPS": 202148991902125.5, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.48808491230010986, "optim/lr": 0.0009, "optim/total_tokens": 786956288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.849905014038086, "created_at": "2025-01-15T09:46:55.467308+00:00"} {"global_step": 1502, "acc_step": 0, "speed/wps": 12886.266881392414, "speed/FLOPS": 202396528958585.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3407962322235107, "optim/lr": 0.0009006000000000001, "optim/total_tokens": 787480576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 5.320475101470947, "created_at": "2025-01-15T09:47:05.646362+00:00"} {"global_step": 1503, "acc_step": 0, "speed/wps": 12870.287402893324, "speed/FLOPS": 202145549290652.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.7922794818878174, "optim/lr": 0.0009012, "optim/total_tokens": 788004864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 4.1022491455078125, "created_at": "2025-01-15T09:47:15.835000+00:00"} {"global_step": 1504, "acc_step": 0, "speed/wps": 12877.127048638209, "speed/FLOPS": 202252975325734.2, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5882577300071716, "optim/lr": 0.0009017999999999999, "optim/total_tokens": 788529152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 3.9838218688964844, "created_at": "2025-01-15T09:47:26.018175+00:00"} {"global_step": 1505, "acc_step": 0, "speed/wps": 12871.444750875595, "speed/FLOPS": 202163727031076.97, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2154645919799805, "optim/lr": 0.0009024, "optim/total_tokens": 789053440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.23267936706543, "created_at": "2025-01-15T09:47:36.207028+00:00"} {"global_step": 1506, "acc_step": 0, "speed/wps": 12872.197451798316, "speed/FLOPS": 202175549233387.75, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1659116744995117, "optim/lr": 0.0009029999999999999, "optim/total_tokens": 789577728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499580, "loss/out": 4.489143371582031, "created_at": "2025-01-15T09:47:46.393135+00:00"} {"global_step": 1507, "acc_step": 0, "speed/wps": 12875.989157226048, "speed/FLOPS": 202235103177479.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.06905996799469, "optim/lr": 0.0009036000000000001, "optim/total_tokens": 790102016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.131089210510254, "created_at": "2025-01-15T09:47:56.579867+00:00"} {"global_step": 1508, "acc_step": 0, "speed/wps": 12873.848956737638, "speed/FLOPS": 202201488387865.4, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.17064368724823, "optim/lr": 0.0009042, "optim/total_tokens": 790626304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 4.344634532928467, "created_at": "2025-01-15T09:48:06.767948+00:00"} {"global_step": 1509, "acc_step": 0, "speed/wps": 12875.527723458496, "speed/FLOPS": 202227855726083.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.1000902652740479, "optim/lr": 0.0009048, "optim/total_tokens": 791150592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.16154670715332, "created_at": "2025-01-15T09:48:16.952681+00:00"} {"global_step": 1510, "acc_step": 0, "speed/wps": 12879.085294461383, "speed/FLOPS": 202283732267298.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1483639478683472, "optim/lr": 0.0009054000000000001, "optim/total_tokens": 791674880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 4.612285614013672, "created_at": "2025-01-15T09:48:27.137560+00:00"} {"global_step": 1511, "acc_step": 0, "speed/wps": 12869.177643159073, "speed/FLOPS": 202128118989056.53, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0308899879455566, "optim/lr": 0.000906, "optim/total_tokens": 792199168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 4.394337177276611, "created_at": "2025-01-15T09:48:37.355643+00:00"} {"global_step": 1512, "acc_step": 0, "speed/wps": 12873.233322643055, "speed/FLOPS": 202191819008440.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8944618701934814, "optim/lr": 0.0009066000000000001, "optim/total_tokens": 792723456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 4.283363342285156, "created_at": "2025-01-15T09:48:47.544654+00:00"} {"global_step": 1513, "acc_step": 0, "speed/wps": 12880.55008872761, "speed/FLOPS": 202306738874088.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7920730710029602, "optim/lr": 0.0009072, "optim/total_tokens": 793247744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 4.159777641296387, "created_at": "2025-01-15T09:48:57.724772+00:00"} {"global_step": 1514, "acc_step": 0, "speed/wps": 12874.175647628048, "speed/FLOPS": 202206619517219.62, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9008435606956482, "optim/lr": 0.0009078, "optim/total_tokens": 793772032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.279607772827148, "created_at": "2025-01-15T09:49:07.919479+00:00"} {"global_step": 1515, "acc_step": 0, "speed/wps": 12876.573583647714, "speed/FLOPS": 202244282397517.88, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5888593792915344, "optim/lr": 0.0009084000000000001, "optim/total_tokens": 794296320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.1896891593933105, "created_at": "2025-01-15T09:49:18.103707+00:00"} {"global_step": 1516, "acc_step": 0, "speed/wps": 12873.356499860169, "speed/FLOPS": 202193753675898.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8476935029029846, "optim/lr": 0.000909, "optim/total_tokens": 794820608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 4.174004077911377, "created_at": "2025-01-15T09:49:28.291562+00:00"} {"global_step": 1517, "acc_step": 0, "speed/wps": 12880.230389833561, "speed/FLOPS": 202301717563644.88, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.511565625667572, "optim/lr": 0.0009096000000000001, "optim/total_tokens": 795344896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 4.148304462432861, "created_at": "2025-01-15T09:49:38.471585+00:00"} {"global_step": 1518, "acc_step": 0, "speed/wps": 12875.463236707248, "speed/FLOPS": 202226842873041.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46927610039711, "optim/lr": 0.0009102, "optim/total_tokens": 795869184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 4.088199615478516, "created_at": "2025-01-15T09:49:48.655246+00:00"} {"global_step": 1519, "acc_step": 0, "speed/wps": 12874.054228294255, "speed/FLOPS": 202204712459734.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.48225948214530945, "optim/lr": 0.0009107999999999999, "optim/total_tokens": 796393472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387505, "loss/out": 4.001717567443848, "created_at": "2025-01-15T09:49:58.840169+00:00"} {"global_step": 1520, "acc_step": 0, "speed/wps": 12880.514731537736, "speed/FLOPS": 202306183540835.4, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37074801325798035, "optim/lr": 0.0009114, "optim/total_tokens": 796917760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311884, "loss/out": 4.097212791442871, "created_at": "2025-01-15T09:50:09.020035+00:00"} {"global_step": 1521, "acc_step": 0, "speed/wps": 12877.066221891291, "speed/FLOPS": 202252019958089.16, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25695621967315674, "optim/lr": 0.000912, "optim/total_tokens": 797442048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.031909942626953, "created_at": "2025-01-15T09:50:19.203858+00:00"} {"global_step": 1522, "acc_step": 0, "speed/wps": 12881.878744023907, "speed/FLOPS": 202327607231280.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27290061116218567, "optim/lr": 0.0009126000000000001, "optim/total_tokens": 797966336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.9662113189697266, "created_at": "2025-01-15T09:50:29.387879+00:00"} {"global_step": 1523, "acc_step": 0, "speed/wps": 12883.75173321301, "speed/FLOPS": 202357025100252.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.224742591381073, "optim/lr": 0.0009132000000000001, "optim/total_tokens": 798490624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 4.011579990386963, "created_at": "2025-01-15T09:50:39.568231+00:00"} {"global_step": 1524, "acc_step": 0, "speed/wps": 12878.971146155365, "speed/FLOPS": 202281939411296.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2771546542644501, "optim/lr": 0.0009138, "optim/total_tokens": 799014912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.9525036811828613, "created_at": "2025-01-15T09:50:49.753638+00:00"} {"global_step": 1525, "acc_step": 0, "speed/wps": 12878.512403585692, "speed/FLOPS": 202274734228861.6, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20627494156360626, "optim/lr": 0.0009144000000000001, "optim/total_tokens": 799539200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 3.880711317062378, "created_at": "2025-01-15T09:50:59.937952+00:00"} {"global_step": 1526, "acc_step": 0, "speed/wps": 12883.184635132811, "speed/FLOPS": 202348118045628.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2530534267425537, "optim/lr": 0.000915, "optim/total_tokens": 800063488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.785618543624878, "created_at": "2025-01-15T09:51:10.118034+00:00"} {"global_step": 1527, "acc_step": 0, "speed/wps": 12877.866444835514, "speed/FLOPS": 202264588559049.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27254053950309753, "optim/lr": 0.0009156000000000001, "optim/total_tokens": 800587776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.8055059909820557, "created_at": "2025-01-15T09:51:20.302864+00:00"} {"global_step": 1528, "acc_step": 0, "speed/wps": 12878.193492853037, "speed/FLOPS": 202269725297576.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3013436198234558, "optim/lr": 0.0009162, "optim/total_tokens": 801112064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.9547533988952637, "created_at": "2025-01-15T09:51:30.488011+00:00"} {"global_step": 1529, "acc_step": 0, "speed/wps": 12876.570200077811, "speed/FLOPS": 202244229253903.03, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23356443643569946, "optim/lr": 0.0009168, "optim/total_tokens": 801636352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.8593032360076904, "created_at": "2025-01-15T09:51:40.674769+00:00"} {"global_step": 1530, "acc_step": 0, "speed/wps": 12875.31041866675, "speed/FLOPS": 202224442655721.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20880332589149475, "optim/lr": 0.0009174000000000001, "optim/total_tokens": 802160640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.858283281326294, "created_at": "2025-01-15T09:51:50.861720+00:00"} {"global_step": 1531, "acc_step": 0, "speed/wps": 12878.357053730891, "speed/FLOPS": 202272294245918.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22839534282684326, "optim/lr": 0.000918, "optim/total_tokens": 802684928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 3.98024845123291, "created_at": "2025-01-15T09:52:01.048257+00:00"} {"global_step": 1532, "acc_step": 0, "speed/wps": 12879.46308152715, "speed/FLOPS": 202289665932300.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19772803783416748, "optim/lr": 0.0009186000000000001, "optim/total_tokens": 803209216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.8643267154693604, "created_at": "2025-01-15T09:52:11.228576+00:00"} {"global_step": 1533, "acc_step": 0, "speed/wps": 12877.773655536164, "speed/FLOPS": 202263131175597.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1731799989938736, "optim/lr": 0.0009192, "optim/total_tokens": 803733504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.8162379264831543, "created_at": "2025-01-15T09:52:21.412806+00:00"} {"global_step": 1534, "acc_step": 0, "speed/wps": 12873.982053292943, "speed/FLOPS": 202203578852160.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1887086182832718, "optim/lr": 0.0009198, "optim/total_tokens": 804257792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.7275047302246094, "created_at": "2025-01-15T09:52:31.597420+00:00"} {"global_step": 1535, "acc_step": 0, "speed/wps": 12877.458583715394, "speed/FLOPS": 202258182539698.3, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1743256151676178, "optim/lr": 0.0009204, "optim/total_tokens": 804782080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.677643299102783, "created_at": "2025-01-15T09:52:41.784073+00:00"} {"global_step": 1536, "acc_step": 0, "speed/wps": 12880.174734551818, "speed/FLOPS": 202300843420966.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1718636155128479, "optim/lr": 0.000921, "optim/total_tokens": 805306368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.8803412914276123, "created_at": "2025-01-15T09:52:51.964121+00:00"} {"global_step": 1537, "acc_step": 0, "speed/wps": 12876.414795759574, "speed/FLOPS": 202241788415537.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18617698550224304, "optim/lr": 0.0009216, "optim/total_tokens": 805830656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407300, "loss/out": 3.69929838180542, "created_at": "2025-01-15T09:53:02.149233+00:00"} {"global_step": 1538, "acc_step": 0, "speed/wps": 12874.00638946309, "speed/FLOPS": 202203961085153.88, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19611035287380219, "optim/lr": 0.0009222000000000001, "optim/total_tokens": 806354944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.8321290016174316, "created_at": "2025-01-15T09:53:12.333937+00:00"} {"global_step": 1539, "acc_step": 0, "speed/wps": 12879.316230192542, "speed/FLOPS": 202287359430295.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18228718638420105, "optim/lr": 0.0009228, "optim/total_tokens": 806879232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.776425838470459, "created_at": "2025-01-15T09:53:22.515141+00:00"} {"global_step": 1540, "acc_step": 0, "speed/wps": 12877.693492172839, "speed/FLOPS": 202261872099820.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8554378151893616, "optim/lr": 0.0009234000000000001, "optim/total_tokens": 807403520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.68904709815979, "created_at": "2025-01-15T09:53:32.700385+00:00"} {"global_step": 1541, "acc_step": 0, "speed/wps": 12874.809272047643, "speed/FLOPS": 202216571459420.56, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18565607070922852, "optim/lr": 0.000924, "optim/total_tokens": 807927808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.7781357765197754, "created_at": "2025-01-15T09:53:42.892397+00:00"} {"global_step": 1542, "acc_step": 0, "speed/wps": 12878.254799265711, "speed/FLOPS": 202270688199030.88, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0019, "optim/grad_norm": 5.127058982849121, "optim/lr": 0.0009245999999999999, "optim/total_tokens": 808452096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.852895736694336, "created_at": "2025-01-15T09:53:53.074535+00:00"} {"global_step": 1543, "acc_step": 0, "speed/wps": 12876.21437909011, "speed/FLOPS": 202238640596341.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2684730589389801, "optim/lr": 0.0009252, "optim/total_tokens": 808976384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.7881627082824707, "created_at": "2025-01-15T09:54:03.259706+00:00"} {"global_step": 1544, "acc_step": 0, "speed/wps": 12879.758952454304, "speed/FLOPS": 202294312991780.8, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.550841748714447, "optim/lr": 0.0009258, "optim/total_tokens": 809500672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.7623403072357178, "created_at": "2025-01-15T09:54:13.440467+00:00"} {"global_step": 1545, "acc_step": 0, "speed/wps": 12876.440916523346, "speed/FLOPS": 202242198678025.94, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31511595845222473, "optim/lr": 0.0009264000000000001, "optim/total_tokens": 810024960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.760554790496826, "created_at": "2025-01-15T09:54:23.631130+00:00"} {"global_step": 1546, "acc_step": 0, "speed/wps": 12879.736520530194, "speed/FLOPS": 202293960667587.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21225151419639587, "optim/lr": 0.000927, "optim/total_tokens": 810549248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.8405556678771973, "created_at": "2025-01-15T09:54:33.811523+00:00"} {"global_step": 1547, "acc_step": 0, "speed/wps": 12876.559580486608, "speed/FLOPS": 202244062458630.34, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1781265139579773, "optim/lr": 0.0009276, "optim/total_tokens": 811073536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.8231704235076904, "created_at": "2025-01-15T09:54:43.996512+00:00"} {"global_step": 1548, "acc_step": 0, "speed/wps": 12875.421378265264, "speed/FLOPS": 202226185428690.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19993147253990173, "optim/lr": 0.0009282, "optim/total_tokens": 811597824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.8947858810424805, "created_at": "2025-01-15T09:54:54.183731+00:00"} {"global_step": 1549, "acc_step": 0, "speed/wps": 12880.246773895225, "speed/FLOPS": 202301974897847.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18191741406917572, "optim/lr": 0.0009288, "optim/total_tokens": 812122112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 3.8973450660705566, "created_at": "2025-01-15T09:55:04.363381+00:00"} {"global_step": 1550, "acc_step": 0, "speed/wps": 12872.820939222269, "speed/FLOPS": 202185341960142.44, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18968288600444794, "optim/lr": 0.0009294, "optim/total_tokens": 812646400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 3.742480754852295, "created_at": "2025-01-15T09:55:14.551263+00:00"} {"global_step": 1551, "acc_step": 0, "speed/wps": 12877.52629542505, "speed/FLOPS": 202259246045144.12, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17437060177326202, "optim/lr": 0.00093, "optim/total_tokens": 813170688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.844242572784424, "created_at": "2025-01-15T09:55:24.733374+00:00"} {"global_step": 1552, "acc_step": 0, "speed/wps": 12877.336176334884, "speed/FLOPS": 202256259963582.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17321178317070007, "optim/lr": 0.0009306, "optim/total_tokens": 813694976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.746338367462158, "created_at": "2025-01-15T09:55:34.917480+00:00"} {"global_step": 1553, "acc_step": 0, "speed/wps": 12875.358911501735, "speed/FLOPS": 202225204302330.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2640013098716736, "optim/lr": 0.0009312000000000001, "optim/total_tokens": 814219264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.772700309753418, "created_at": "2025-01-15T09:55:45.107446+00:00"} {"global_step": 1554, "acc_step": 0, "speed/wps": 12876.07084377098, "speed/FLOPS": 202236386176916.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1753588616847992, "optim/lr": 0.0009318, "optim/total_tokens": 814743552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.7614731788635254, "created_at": "2025-01-15T09:55:55.294620+00:00"} {"global_step": 1555, "acc_step": 0, "speed/wps": 12877.781551618406, "speed/FLOPS": 202263255194419.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17818868160247803, "optim/lr": 0.0009324000000000001, "optim/total_tokens": 815267840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.6641650199890137, "created_at": "2025-01-15T09:56:05.478841+00:00"} {"global_step": 1556, "acc_step": 0, "speed/wps": 12876.199450758902, "speed/FLOPS": 202238406126386.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1926799863576889, "optim/lr": 0.000933, "optim/total_tokens": 815792128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364786, "loss/out": 3.7071080207824707, "created_at": "2025-01-15T09:56:15.661719+00:00"} {"global_step": 1557, "acc_step": 0, "speed/wps": 12877.30602229666, "speed/FLOPS": 202255786352975.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2216195911169052, "optim/lr": 0.0009335999999999999, "optim/total_tokens": 816316416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.842942476272583, "created_at": "2025-01-15T09:56:25.847092+00:00"} {"global_step": 1558, "acc_step": 0, "speed/wps": 12874.112098893094, "speed/FLOPS": 202205621397012.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1930520087480545, "optim/lr": 0.0009342, "optim/total_tokens": 816840704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 3.830613136291504, "created_at": "2025-01-15T09:56:36.033000+00:00"} {"global_step": 1559, "acc_step": 0, "speed/wps": 12875.787447426472, "speed/FLOPS": 202231935047898.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20279137790203094, "optim/lr": 0.0009348, "optim/total_tokens": 817364992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.7569713592529297, "created_at": "2025-01-15T09:56:46.216380+00:00"} {"global_step": 1560, "acc_step": 0, "speed/wps": 12875.121339087593, "speed/FLOPS": 202221472901108.03, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19101659953594208, "optim/lr": 0.0009354000000000001, "optim/total_tokens": 817889280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.683769702911377, "created_at": "2025-01-15T09:56:56.401381+00:00"} {"global_step": 1561, "acc_step": 0, "speed/wps": 12877.959528887943, "speed/FLOPS": 202266050572004.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7037358283996582, "optim/lr": 0.000936, "optim/total_tokens": 818413568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.851227045059204, "created_at": "2025-01-15T09:57:06.582931+00:00"} {"global_step": 1562, "acc_step": 0, "speed/wps": 12876.496947167892, "speed/FLOPS": 202243078716292.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0641142129898071, "optim/lr": 0.0009366, "optim/total_tokens": 818937856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347191, "loss/out": 3.688206672668457, "created_at": "2025-01-15T09:57:16.768916+00:00"} {"global_step": 1563, "acc_step": 0, "speed/wps": 12876.555005844499, "speed/FLOPS": 202243990607589.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0398452281951904, "optim/lr": 0.0009372, "optim/total_tokens": 819462144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 3.8896095752716064, "created_at": "2025-01-15T09:57:26.953016+00:00"} {"global_step": 1564, "acc_step": 0, "speed/wps": 12876.999933574301, "speed/FLOPS": 202250978808981.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9313351511955261, "optim/lr": 0.0009378, "optim/total_tokens": 819986432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.017971515655518, "created_at": "2025-01-15T09:57:37.142038+00:00"} {"global_step": 1565, "acc_step": 0, "speed/wps": 12878.367932011193, "speed/FLOPS": 202272465104259.34, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2076237201690674, "optim/lr": 0.0009384, "optim/total_tokens": 820510720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 4.306048393249512, "created_at": "2025-01-15T09:57:47.325540+00:00"} {"global_step": 1566, "acc_step": 0, "speed/wps": 12877.883799746802, "speed/FLOPS": 202264861141779.66, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9858514070510864, "optim/lr": 0.0009390000000000001, "optim/total_tokens": 821035008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.201653003692627, "created_at": "2025-01-15T09:57:57.507112+00:00"} {"global_step": 1567, "acc_step": 0, "speed/wps": 12880.607878635137, "speed/FLOPS": 202307646543997.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.334072470664978, "optim/lr": 0.0009396, "optim/total_tokens": 821559296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440101, "loss/out": 5.305238723754883, "created_at": "2025-01-15T09:58:07.686654+00:00"} {"global_step": 1568, "acc_step": 0, "speed/wps": 12872.744080129434, "speed/FLOPS": 202184134782474.0, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2272342443466187, "optim/lr": 0.0009402000000000001, "optim/total_tokens": 822083584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346784, "loss/out": 4.48122501373291, "created_at": "2025-01-15T09:58:17.872545+00:00"} {"global_step": 1569, "acc_step": 0, "speed/wps": 12876.143030414452, "speed/FLOPS": 202237519967347.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.2978885173797607, "optim/lr": 0.0009408, "optim/total_tokens": 822607872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 5.686699867248535, "created_at": "2025-01-15T09:58:28.059320+00:00"} {"global_step": 1570, "acc_step": 0, "speed/wps": 12883.019630787967, "speed/FLOPS": 202345526425652.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2708314657211304, "optim/lr": 0.0009414000000000001, "optim/total_tokens": 823132160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 4.805840015411377, "created_at": "2025-01-15T09:58:38.237743+00:00"} {"global_step": 1571, "acc_step": 0, "speed/wps": 12874.77368254195, "speed/FLOPS": 202216012477328.56, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8798648118972778, "optim/lr": 0.000942, "optim/total_tokens": 823656448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 4.576964378356934, "created_at": "2025-01-15T09:58:48.424332+00:00"} {"global_step": 1572, "acc_step": 0, "speed/wps": 12877.13125490217, "speed/FLOPS": 202253041390889.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1694011688232422, "optim/lr": 0.0009425999999999999, "optim/total_tokens": 824180736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.734555244445801, "created_at": "2025-01-15T09:58:58.607558+00:00"} {"global_step": 1573, "acc_step": 0, "speed/wps": 12877.849227838766, "speed/FLOPS": 202264318142457.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0567090511322021, "optim/lr": 0.0009432, "optim/total_tokens": 824705024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393639, "loss/out": 4.614162921905518, "created_at": "2025-01-15T09:59:08.791326+00:00"} {"global_step": 1574, "acc_step": 0, "speed/wps": 12871.006991773564, "speed/FLOPS": 202156851422835.53, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1677148342132568, "optim/lr": 0.0009438, "optim/total_tokens": 825229312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 4.562348365783691, "created_at": "2025-01-15T09:59:18.983019+00:00"} {"global_step": 1575, "acc_step": 0, "speed/wps": 12878.988969577675, "speed/FLOPS": 202282219352636.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.6801322102546692, "optim/lr": 0.0009444000000000001, "optim/total_tokens": 825753600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.37291145324707, "created_at": "2025-01-15T09:59:29.165113+00:00"} {"global_step": 1576, "acc_step": 0, "speed/wps": 12877.467860363691, "speed/FLOPS": 202258328242207.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5123823285102844, "optim/lr": 0.000945, "optim/total_tokens": 826277888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 4.297385215759277, "created_at": "2025-01-15T09:59:39.351983+00:00"} {"global_step": 1577, "acc_step": 0, "speed/wps": 12876.576713922706, "speed/FLOPS": 202244331562788.1, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4968167543411255, "optim/lr": 0.0009456, "optim/total_tokens": 826802176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463722, "loss/out": 4.099585056304932, "created_at": "2025-01-15T09:59:49.541829+00:00"} {"global_step": 1578, "acc_step": 0, "speed/wps": 12879.94791366369, "speed/FLOPS": 202297280887234.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48978307843208313, "optim/lr": 0.0009462, "optim/total_tokens": 827326464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 4.335935592651367, "created_at": "2025-01-15T09:59:59.725609+00:00"} {"global_step": 1579, "acc_step": 0, "speed/wps": 12882.22961083838, "speed/FLOPS": 202333118076743.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3761884570121765, "optim/lr": 0.0009468, "optim/total_tokens": 827850752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372261, "loss/out": 4.17808723449707, "created_at": "2025-01-15T10:00:09.903812+00:00"} {"global_step": 1580, "acc_step": 0, "speed/wps": 12879.779315819791, "speed/FLOPS": 202294632826417.97, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4356081187725067, "optim/lr": 0.0009474, "optim/total_tokens": 828375040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295655, "loss/out": 3.9976823329925537, "created_at": "2025-01-15T10:00:20.085961+00:00"} {"global_step": 1581, "acc_step": 0, "speed/wps": 12876.332670139162, "speed/FLOPS": 202240498519816.22, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.37743449211120605, "optim/lr": 0.0009480000000000001, "optim/total_tokens": 828899328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.026479721069336, "created_at": "2025-01-15T10:00:30.271105+00:00"} {"global_step": 1582, "acc_step": 0, "speed/wps": 12878.009142292589, "speed/FLOPS": 202266829818699.94, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3027488589286804, "optim/lr": 0.0009486, "optim/total_tokens": 829423616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.017154693603516, "created_at": "2025-01-15T10:00:40.458219+00:00"} {"global_step": 1583, "acc_step": 0, "speed/wps": 12880.28963505231, "speed/FLOPS": 202302648091216.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3845765292644501, "optim/lr": 0.0009492000000000001, "optim/total_tokens": 829947904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 3.9916841983795166, "created_at": "2025-01-15T10:00:50.643189+00:00"} {"global_step": 1584, "acc_step": 0, "speed/wps": 12877.624633381762, "speed/FLOPS": 202260790577884.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40081024169921875, "optim/lr": 0.0009498, "optim/total_tokens": 830472192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.989163875579834, "created_at": "2025-01-15T10:01:00.829184+00:00"} {"global_step": 1585, "acc_step": 0, "speed/wps": 12878.641752310528, "speed/FLOPS": 202276765828328.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.32386675477027893, "optim/lr": 0.0009504000000000001, "optim/total_tokens": 830996480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402750, "loss/out": 3.9448065757751465, "created_at": "2025-01-15T10:01:11.012976+00:00"} {"global_step": 1586, "acc_step": 0, "speed/wps": 12875.17852080454, "speed/FLOPS": 202222371018548.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2912043333053589, "optim/lr": 0.000951, "optim/total_tokens": 831520768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 3.934042453765869, "created_at": "2025-01-15T10:01:21.200064+00:00"} {"global_step": 1587, "acc_step": 0, "speed/wps": 12877.409424654381, "speed/FLOPS": 202257410429093.5, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29188546538352966, "optim/lr": 0.0009515999999999999, "optim/total_tokens": 832045056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.9473695755004883, "created_at": "2025-01-15T10:01:31.382362+00:00"} {"global_step": 1588, "acc_step": 0, "speed/wps": 12876.109204126353, "speed/FLOPS": 202236988679010.56, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.27815544605255127, "optim/lr": 0.0009522, "optim/total_tokens": 832569344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.9623184204101562, "created_at": "2025-01-15T10:01:41.568437+00:00"} {"global_step": 1589, "acc_step": 0, "speed/wps": 12880.079106524101, "speed/FLOPS": 202299341451384.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.34720778465270996, "optim/lr": 0.0009528, "optim/total_tokens": 833093632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.908524513244629, "created_at": "2025-01-15T10:01:51.751644+00:00"} {"global_step": 1590, "acc_step": 0, "speed/wps": 12877.163944853974, "speed/FLOPS": 202253554831503.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2416757196187973, "optim/lr": 0.0009534000000000001, "optim/total_tokens": 833617920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.9789462089538574, "created_at": "2025-01-15T10:02:01.943001+00:00"} {"global_step": 1591, "acc_step": 0, "speed/wps": 12877.744320750451, "speed/FLOPS": 202262670432477.72, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20353710651397705, "optim/lr": 0.000954, "optim/total_tokens": 834142208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 3.7945504188537598, "created_at": "2025-01-15T10:02:12.127427+00:00"} {"global_step": 1592, "acc_step": 0, "speed/wps": 12877.061683413784, "speed/FLOPS": 202251948675063.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19662374258041382, "optim/lr": 0.0009546, "optim/total_tokens": 834666496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 3.8163959980010986, "created_at": "2025-01-15T10:02:22.309642+00:00"} {"global_step": 1593, "acc_step": 0, "speed/wps": 12873.822769943426, "speed/FLOPS": 202201077088277.9, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21759086847305298, "optim/lr": 0.0009552, "optim/total_tokens": 835190784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 3.910269260406494, "created_at": "2025-01-15T10:02:32.495114+00:00"} {"global_step": 1594, "acc_step": 0, "speed/wps": 12873.369091253422, "speed/FLOPS": 202193951441031.53, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.23618797957897186, "optim/lr": 0.0009558, "optim/total_tokens": 835715072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.8460516929626465, "created_at": "2025-01-15T10:02:42.682790+00:00"} {"global_step": 1595, "acc_step": 0, "speed/wps": 12879.261212191228, "speed/FLOPS": 202286495296982.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19712239503860474, "optim/lr": 0.0009563999999999999, "optim/total_tokens": 836239360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.788144111633301, "created_at": "2025-01-15T10:02:52.865906+00:00"} {"global_step": 1596, "acc_step": 0, "speed/wps": 12879.319130790445, "speed/FLOPS": 202287404988171.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19611996412277222, "optim/lr": 0.0009570000000000001, "optim/total_tokens": 836763648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 3.825793981552124, "created_at": "2025-01-15T10:03:03.050931+00:00"} {"global_step": 1597, "acc_step": 0, "speed/wps": 12876.595885564853, "speed/FLOPS": 202244632679772.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1676894873380661, "optim/lr": 0.0009576, "optim/total_tokens": 837287936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.886324644088745, "created_at": "2025-01-15T10:03:13.233768+00:00"} {"global_step": 1598, "acc_step": 0, "speed/wps": 12876.83241198772, "speed/FLOPS": 202248347652265.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16432808339595795, "optim/lr": 0.0009582000000000001, "optim/total_tokens": 837812224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 3.8323564529418945, "created_at": "2025-01-15T10:03:23.422604+00:00"} {"global_step": 1599, "acc_step": 0, "speed/wps": 12877.712605867362, "speed/FLOPS": 202262172306658.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16055788099765778, "optim/lr": 0.0009588, "optim/total_tokens": 838336512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.8157856464385986, "created_at": "2025-01-15T10:03:33.604361+00:00"} {"global_step": 1600, "acc_step": 0, "speed/wps": 12879.051578820321, "speed/FLOPS": 202283202716825.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16610753536224365, "optim/lr": 0.0009593999999999999, "optim/total_tokens": 838860800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.911186933517456, "created_at": "2025-01-15T10:03:43.789318+00:00"} {"global_step": 1601, "acc_step": 0, "speed/wps": 12875.477047838334, "speed/FLOPS": 202227059795833.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16341359913349152, "optim/lr": 0.00096, "optim/total_tokens": 839385088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.8874592781066895, "created_at": "2025-01-15T10:03:53.974981+00:00"} {"global_step": 1602, "acc_step": 0, "speed/wps": 12878.6758558921, "speed/FLOPS": 202277301471940.75, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1837143450975418, "optim/lr": 0.0009605999999999999, "optim/total_tokens": 839909376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.8424570560455322, "created_at": "2025-01-15T10:04:04.158356+00:00"} {"global_step": 1603, "acc_step": 0, "speed/wps": 12876.961490150612, "speed/FLOPS": 202250375002185.28, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.2440420389175415, "optim/lr": 0.0009612, "optim/total_tokens": 840433664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.854992151260376, "created_at": "2025-01-15T10:04:14.343468+00:00"} {"global_step": 1604, "acc_step": 0, "speed/wps": 12876.378433976168, "speed/FLOPS": 202241217303762.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3248006999492645, "optim/lr": 0.0009618, "optim/total_tokens": 840957952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 3.754582405090332, "created_at": "2025-01-15T10:04:24.528525+00:00"} {"global_step": 1605, "acc_step": 0, "speed/wps": 12880.785627523692, "speed/FLOPS": 202310438334544.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22611990571022034, "optim/lr": 0.0009624, "optim/total_tokens": 841482240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.850451946258545, "created_at": "2025-01-15T10:04:34.710974+00:00"} {"global_step": 1606, "acc_step": 0, "speed/wps": 12881.66070366215, "speed/FLOPS": 202324182607780.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25463271141052246, "optim/lr": 0.000963, "optim/total_tokens": 842006528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.744032144546509, "created_at": "2025-01-15T10:04:44.894101+00:00"} {"global_step": 1607, "acc_step": 0, "speed/wps": 12872.86857732009, "speed/FLOPS": 202186090181930.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.23050902783870697, "optim/lr": 0.0009636, "optim/total_tokens": 842530816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.8985369205474854, "created_at": "2025-01-15T10:04:55.079611+00:00"} {"global_step": 1608, "acc_step": 0, "speed/wps": 12876.761499450087, "speed/FLOPS": 202247233873417.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1982613503932953, "optim/lr": 0.0009642000000000001, "optim/total_tokens": 843055104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.8214993476867676, "created_at": "2025-01-15T10:05:05.271087+00:00"} {"global_step": 1609, "acc_step": 0, "speed/wps": 12876.900233752647, "speed/FLOPS": 202249412886280.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21805739402770996, "optim/lr": 0.0009648, "optim/total_tokens": 843579392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.9473462104797363, "created_at": "2025-01-15T10:05:15.458480+00:00"} {"global_step": 1610, "acc_step": 0, "speed/wps": 12876.471596657337, "speed/FLOPS": 202242680551689.2, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19642284512519836, "optim/lr": 0.0009653999999999999, "optim/total_tokens": 844103680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411072, "loss/out": 3.7974958419799805, "created_at": "2025-01-15T10:05:25.654151+00:00"} {"global_step": 1611, "acc_step": 0, "speed/wps": 12875.912839873305, "speed/FLOPS": 202233904508583.7, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20498669147491455, "optim/lr": 0.0009660000000000001, "optim/total_tokens": 844627968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 3.742406129837036, "created_at": "2025-01-15T10:05:35.838630+00:00"} {"global_step": 1612, "acc_step": 0, "speed/wps": 12875.0666619825, "speed/FLOPS": 202220614122038.34, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21158303320407867, "optim/lr": 0.0009666, "optim/total_tokens": 845152256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.8410120010375977, "created_at": "2025-01-15T10:05:46.022532+00:00"} {"global_step": 1613, "acc_step": 0, "speed/wps": 12875.164227388952, "speed/FLOPS": 202222146520815.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18225069344043732, "optim/lr": 0.0009672000000000001, "optim/total_tokens": 845676544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.6893210411071777, "created_at": "2025-01-15T10:05:56.209224+00:00"} {"global_step": 1614, "acc_step": 0, "speed/wps": 12875.482267216783, "speed/FLOPS": 202227141773344.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1831800937652588, "optim/lr": 0.0009678, "optim/total_tokens": 846200832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.7886502742767334, "created_at": "2025-01-15T10:06:06.394907+00:00"} {"global_step": 1615, "acc_step": 0, "speed/wps": 12881.12788852752, "speed/FLOPS": 202315814013925.97, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20351383090019226, "optim/lr": 0.0009683999999999999, "optim/total_tokens": 846725120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 3.6687917709350586, "created_at": "2025-01-15T10:06:16.575260+00:00"} {"global_step": 1616, "acc_step": 0, "speed/wps": 12874.24861974626, "speed/FLOPS": 202207765644610.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1709391176700592, "optim/lr": 0.000969, "optim/total_tokens": 847249408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.8023102283477783, "created_at": "2025-01-15T10:06:26.760169+00:00"} {"global_step": 1617, "acc_step": 0, "speed/wps": 12875.957448631454, "speed/FLOPS": 202234605150429.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16216284036636353, "optim/lr": 0.0009695999999999999, "optim/total_tokens": 847773696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 3.5892646312713623, "created_at": "2025-01-15T10:06:36.944987+00:00"} {"global_step": 1618, "acc_step": 0, "speed/wps": 12878.558966611756, "speed/FLOPS": 202275465565170.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16885383427143097, "optim/lr": 0.0009702000000000001, "optim/total_tokens": 848297984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.8179047107696533, "created_at": "2025-01-15T10:06:47.128660+00:00"} {"global_step": 1619, "acc_step": 0, "speed/wps": 12880.494034592411, "speed/FLOPS": 202305858466868.5, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2602653205394745, "optim/lr": 0.0009708000000000001, "optim/total_tokens": 848822272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.6663761138916016, "created_at": "2025-01-15T10:06:57.308140+00:00"} {"global_step": 1620, "acc_step": 0, "speed/wps": 12875.006767522465, "speed/FLOPS": 202219673397239.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1725597381591797, "optim/lr": 0.0009714, "optim/total_tokens": 849346560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.6878914833068848, "created_at": "2025-01-15T10:07:07.497555+00:00"} {"global_step": 1621, "acc_step": 0, "speed/wps": 12880.690621123898, "speed/FLOPS": 202308946128483.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2132956087589264, "optim/lr": 0.0009720000000000001, "optim/total_tokens": 849870848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.6749143600463867, "created_at": "2025-01-15T10:07:17.676938+00:00"} {"global_step": 1622, "acc_step": 0, "speed/wps": 12880.456408454136, "speed/FLOPS": 202305267496663.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26830899715423584, "optim/lr": 0.0009726, "optim/total_tokens": 850395136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.6520466804504395, "created_at": "2025-01-15T10:07:27.861054+00:00"} {"global_step": 1623, "acc_step": 0, "speed/wps": 12869.795509139849, "speed/FLOPS": 202137823423321.44, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23820552229881287, "optim/lr": 0.0009732000000000001, "optim/total_tokens": 850919424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374420, "loss/out": 3.6892199516296387, "created_at": "2025-01-15T10:07:38.050702+00:00"} {"global_step": 1624, "acc_step": 0, "speed/wps": 12882.491422416399, "speed/FLOPS": 202337230187342.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23147235810756683, "optim/lr": 0.0009738, "optim/total_tokens": 851443712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.772831916809082, "created_at": "2025-01-15T10:07:48.233634+00:00"} {"global_step": 1625, "acc_step": 0, "speed/wps": 12873.016422627972, "speed/FLOPS": 202188412295651.34, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21565526723861694, "optim/lr": 0.0009743999999999999, "optim/total_tokens": 851968000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.704575538635254, "created_at": "2025-01-15T10:07:58.419458+00:00"} {"global_step": 1626, "acc_step": 0, "speed/wps": 12875.77469816495, "speed/FLOPS": 202231734803227.38, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0122103691101074, "optim/lr": 0.0009750000000000001, "optim/total_tokens": 852492288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.7683703899383545, "created_at": "2025-01-15T10:08:08.605650+00:00"} {"global_step": 1627, "acc_step": 0, "speed/wps": 12867.626775870507, "speed/FLOPS": 202103760487174.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2878783941268921, "optim/lr": 0.0009756, "optim/total_tokens": 853016576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.67919921875, "created_at": "2025-01-15T10:08:18.795580+00:00"} {"global_step": 1628, "acc_step": 0, "speed/wps": 12863.710179737733, "speed/FLOPS": 202042244962942.3, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2803136110305786, "optim/lr": 0.0009762000000000001, "optim/total_tokens": 853540864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.7896342277526855, "created_at": "2025-01-15T10:08:28.994877+00:00"} {"global_step": 1629, "acc_step": 0, "speed/wps": 12861.16552374904, "speed/FLOPS": 202002277643916.12, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.2670479118824005, "optim/lr": 0.0009768, "optim/total_tokens": 854065152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.792613983154297, "created_at": "2025-01-15T10:08:39.197038+00:00"} {"global_step": 1630, "acc_step": 0, "speed/wps": 12871.678210949538, "speed/FLOPS": 202167393842346.0, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2667272388935089, "optim/lr": 0.0009774, "optim/total_tokens": 854589440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.7579221725463867, "created_at": "2025-01-15T10:08:49.383708+00:00"} {"global_step": 1631, "acc_step": 0, "speed/wps": 12869.163731392533, "speed/FLOPS": 202127900485646.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.2193068116903305, "optim/lr": 0.0009780000000000001, "optim/total_tokens": 855113728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.8747246265411377, "created_at": "2025-01-15T10:08:59.572437+00:00"} {"global_step": 1632, "acc_step": 0, "speed/wps": 12873.410613705044, "speed/FLOPS": 202194603608192.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24842016398906708, "optim/lr": 0.0009786, "optim/total_tokens": 855638016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 3.8535006046295166, "created_at": "2025-01-15T10:09:09.757658+00:00"} {"global_step": 1633, "acc_step": 0, "speed/wps": 12866.732940824217, "speed/FLOPS": 202089721579514.56, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.22392964363098145, "optim/lr": 0.0009792000000000002, "optim/total_tokens": 856162304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.769449234008789, "created_at": "2025-01-15T10:09:19.948256+00:00"} {"global_step": 1634, "acc_step": 0, "speed/wps": 12869.795445588137, "speed/FLOPS": 202137822425154.47, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.236276775598526, "optim/lr": 0.0009798, "optim/total_tokens": 856686592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.802574396133423, "created_at": "2025-01-15T10:09:30.139544+00:00"} {"global_step": 1635, "acc_step": 0, "speed/wps": 12872.018125650537, "speed/FLOPS": 202172732669816.94, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.2729749083518982, "optim/lr": 0.0009804, "optim/total_tokens": 857210880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.8449411392211914, "created_at": "2025-01-15T10:09:40.326523+00:00"} {"global_step": 1636, "acc_step": 0, "speed/wps": 12863.338843215184, "speed/FLOPS": 202036412612585.84, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28760823607444763, "optim/lr": 0.000981, "optim/total_tokens": 857735168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.714729070663452, "created_at": "2025-01-15T10:09:50.519735+00:00"} {"global_step": 1637, "acc_step": 0, "speed/wps": 12870.322117821093, "speed/FLOPS": 202146094536296.53, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2754208445549011, "optim/lr": 0.0009816, "optim/total_tokens": 858259456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.7699527740478516, "created_at": "2025-01-15T10:10:00.707321+00:00"} {"global_step": 1638, "acc_step": 0, "speed/wps": 12868.11633659915, "speed/FLOPS": 202111449711144.9, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2479606568813324, "optim/lr": 0.0009822000000000001, "optim/total_tokens": 858783744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.8318746089935303, "created_at": "2025-01-15T10:10:10.897014+00:00"} {"global_step": 1639, "acc_step": 0, "speed/wps": 12866.965572415224, "speed/FLOPS": 202093375378320.62, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.25494125485420227, "optim/lr": 0.0009828, "optim/total_tokens": 859308032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.7049877643585205, "created_at": "2025-01-15T10:10:21.089489+00:00"} {"global_step": 1640, "acc_step": 0, "speed/wps": 12873.801917855877, "speed/FLOPS": 202200749577589.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2556031346321106, "optim/lr": 0.0009834, "optim/total_tokens": 859832320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354275, "loss/out": 3.756207227706909, "created_at": "2025-01-15T10:10:31.274283+00:00"} {"global_step": 1641, "acc_step": 0, "speed/wps": 12872.500384108218, "speed/FLOPS": 202180307201587.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.229261577129364, "optim/lr": 0.000984, "optim/total_tokens": 860356608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.838742733001709, "created_at": "2025-01-15T10:10:41.462554+00:00"} {"global_step": 1642, "acc_step": 0, "speed/wps": 12872.71934252207, "speed/FLOPS": 202183746244357.3, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18627867102622986, "optim/lr": 0.0009846, "optim/total_tokens": 860880896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.7152419090270996, "created_at": "2025-01-15T10:10:51.649198+00:00"} {"global_step": 1643, "acc_step": 0, "speed/wps": 12869.10358706025, "speed/FLOPS": 202126955836260.75, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21971236169338226, "optim/lr": 0.0009852, "optim/total_tokens": 861405184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.7443771362304688, "created_at": "2025-01-15T10:11:01.840789+00:00"} {"global_step": 1644, "acc_step": 0, "speed/wps": 12867.202021630512, "speed/FLOPS": 202097089138162.0, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.178153395652771, "optim/lr": 0.0009858, "optim/total_tokens": 861929472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.7936105728149414, "created_at": "2025-01-15T10:11:12.032920+00:00"} {"global_step": 1645, "acc_step": 0, "speed/wps": 12856.323496348885, "speed/FLOPS": 201926226949953.8, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18536324799060822, "optim/lr": 0.0009864, "optim/total_tokens": 862453760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 3.8391246795654297, "created_at": "2025-01-15T10:11:22.238637+00:00"} {"global_step": 1646, "acc_step": 0, "speed/wps": 12870.381510853615, "speed/FLOPS": 202147027385486.88, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16898958384990692, "optim/lr": 0.000987, "optim/total_tokens": 862978048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.7355830669403076, "created_at": "2025-01-15T10:11:32.429659+00:00"} {"global_step": 1647, "acc_step": 0, "speed/wps": 12870.912602758774, "speed/FLOPS": 202155368913654.16, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1840173751115799, "optim/lr": 0.0009876, "optim/total_tokens": 863502336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.6118221282958984, "created_at": "2025-01-15T10:11:42.617001+00:00"} {"global_step": 1648, "acc_step": 0, "speed/wps": 12874.365291824697, "speed/FLOPS": 202209598139926.44, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17982056736946106, "optim/lr": 0.0009882, "optim/total_tokens": 864026624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.806845188140869, "created_at": "2025-01-15T10:11:52.806870+00:00"} {"global_step": 1649, "acc_step": 0, "speed/wps": 12871.110561379834, "speed/FLOPS": 202158478125821.53, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20350223779678345, "optim/lr": 0.0009888, "optim/total_tokens": 864550912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.880093574523926, "created_at": "2025-01-15T10:12:02.999431+00:00"} {"global_step": 1650, "acc_step": 0, "speed/wps": 12872.302509886975, "speed/FLOPS": 202177199315035.1, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18436993658542633, "optim/lr": 0.0009893999999999999, "optim/total_tokens": 865075200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280980, "loss/out": 3.7230241298675537, "created_at": "2025-01-15T10:12:13.191344+00:00"} {"global_step": 1651, "acc_step": 0, "speed/wps": 12872.809171928155, "speed/FLOPS": 202185157138618.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23300080001354218, "optim/lr": 0.00099, "optim/total_tokens": 865599488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 3.5939548015594482, "created_at": "2025-01-15T10:12:23.376954+00:00"} {"global_step": 1652, "acc_step": 0, "speed/wps": 12871.701779095547, "speed/FLOPS": 202167764012465.1, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2449444830417633, "optim/lr": 0.0009906, "optim/total_tokens": 866123776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 3.7465860843658447, "created_at": "2025-01-15T10:12:33.569320+00:00"} {"global_step": 1653, "acc_step": 0, "speed/wps": 12869.781949101747, "speed/FLOPS": 202137610444289.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.17904707789421082, "optim/lr": 0.0009912, "optim/total_tokens": 866648064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.668161153793335, "created_at": "2025-01-15T10:12:43.757507+00:00"} {"global_step": 1654, "acc_step": 0, "speed/wps": 12872.59307388306, "speed/FLOPS": 202181763021870.38, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18556109070777893, "optim/lr": 0.0009918, "optim/total_tokens": 867172352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.5512895584106445, "created_at": "2025-01-15T10:12:53.944608+00:00"} {"global_step": 1655, "acc_step": 0, "speed/wps": 12873.16168907546, "speed/FLOPS": 202190693904826.16, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22121737897396088, "optim/lr": 0.0009924, "optim/total_tokens": 867696640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 3.6779041290283203, "created_at": "2025-01-15T10:13:04.130101+00:00"} {"global_step": 1656, "acc_step": 0, "speed/wps": 12873.076311523568, "speed/FLOPS": 202189352933052.66, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19573059678077698, "optim/lr": 0.0009930000000000002, "optim/total_tokens": 868220928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.743925094604492, "created_at": "2025-01-15T10:13:14.315420+00:00"} {"global_step": 1657, "acc_step": 0, "speed/wps": 12877.759544892493, "speed/FLOPS": 202262909548548.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1986474245786667, "optim/lr": 0.0009936, "optim/total_tokens": 868745216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460830, "loss/out": 3.606903076171875, "created_at": "2025-01-15T10:13:24.497125+00:00"} {"global_step": 1658, "acc_step": 0, "speed/wps": 12868.396449081492, "speed/FLOPS": 202115849262594.97, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2163630574941635, "optim/lr": 0.0009942, "optim/total_tokens": 869269504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 3.6300854682922363, "created_at": "2025-01-15T10:13:34.686625+00:00"} {"global_step": 1659, "acc_step": 0, "speed/wps": 12871.753702309557, "speed/FLOPS": 202168579537891.1, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20540958642959595, "optim/lr": 0.0009948000000000001, "optim/total_tokens": 869793792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 3.6782877445220947, "created_at": "2025-01-15T10:13:44.882450+00:00"} {"global_step": 1660, "acc_step": 0, "speed/wps": 12868.33356452139, "speed/FLOPS": 202114861574166.12, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.22757351398468018, "optim/lr": 0.0009954, "optim/total_tokens": 870318080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.685842275619507, "created_at": "2025-01-15T10:13:55.076170+00:00"} {"global_step": 1661, "acc_step": 0, "speed/wps": 12874.55580490244, "speed/FLOPS": 202212590409604.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24698001146316528, "optim/lr": 0.0009960000000000001, "optim/total_tokens": 870842368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.8059005737304688, "created_at": "2025-01-15T10:14:05.261676+00:00"} {"global_step": 1662, "acc_step": 0, "speed/wps": 12873.099873014848, "speed/FLOPS": 202189722998650.1, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31622862815856934, "optim/lr": 0.0009966, "optim/total_tokens": 871366656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.8187947273254395, "created_at": "2025-01-15T10:14:15.450323+00:00"} {"global_step": 1663, "acc_step": 0, "speed/wps": 12871.271904737763, "speed/FLOPS": 202161012244966.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.29568326473236084, "optim/lr": 0.0009972, "optim/total_tokens": 871890944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.7665610313415527, "created_at": "2025-01-15T10:14:25.639667+00:00"} {"global_step": 1664, "acc_step": 0, "speed/wps": 12872.093507976591, "speed/FLOPS": 202173916652833.78, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.23739206790924072, "optim/lr": 0.0009978, "optim/total_tokens": 872415232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.7511534690856934, "created_at": "2025-01-15T10:14:35.841929+00:00"} {"global_step": 1665, "acc_step": 0, "speed/wps": 12870.813936478116, "speed/FLOPS": 202153819224134.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2312377244234085, "optim/lr": 0.0009984, "optim/total_tokens": 872939520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.774122714996338, "created_at": "2025-01-15T10:14:46.030533+00:00"} {"global_step": 1666, "acc_step": 0, "speed/wps": 12869.835452691877, "speed/FLOPS": 202138450791694.25, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2655467689037323, "optim/lr": 0.000999, "optim/total_tokens": 873463808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.8127565383911133, "created_at": "2025-01-15T10:14:56.226433+00:00"} {"global_step": 1667, "acc_step": 0, "speed/wps": 12871.546745364885, "speed/FLOPS": 202165328994686.38, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.32437410950660706, "optim/lr": 0.0009996, "optim/total_tokens": 873988096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.631598711013794, "created_at": "2025-01-15T10:15:06.414259+00:00"} {"global_step": 1668, "acc_step": 0, "speed/wps": 12871.5685260277, "speed/FLOPS": 202165671089925.66, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26291975378990173, "optim/lr": 0.0010002, "optim/total_tokens": 874512384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.880549192428589, "created_at": "2025-01-15T10:15:16.601383+00:00"} {"global_step": 1669, "acc_step": 0, "speed/wps": 12872.268400021667, "speed/FLOPS": 202176663572728.56, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27011656761169434, "optim/lr": 0.0010008, "optim/total_tokens": 875036672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467757, "loss/out": 3.701103687286377, "created_at": "2025-01-15T10:15:26.787433+00:00"} {"global_step": 1670, "acc_step": 0, "speed/wps": 12867.728976927348, "speed/FLOPS": 202105365695211.28, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30387255549430847, "optim/lr": 0.0010014, "optim/total_tokens": 875560960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.672657012939453, "created_at": "2025-01-15T10:15:36.977877+00:00"} {"global_step": 1671, "acc_step": 0, "speed/wps": 12873.4641321622, "speed/FLOPS": 202195444189103.78, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25544995069503784, "optim/lr": 0.001002, "optim/total_tokens": 876085248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.801877498626709, "created_at": "2025-01-15T10:15:47.164233+00:00"} {"global_step": 1672, "acc_step": 0, "speed/wps": 12869.272830480311, "speed/FLOPS": 202129614036737.66, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25112712383270264, "optim/lr": 0.0010026, "optim/total_tokens": 876609536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.811497211456299, "created_at": "2025-01-15T10:15:57.355054+00:00"} {"global_step": 1673, "acc_step": 0, "speed/wps": 12870.499373431436, "speed/FLOPS": 202148878579231.7, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21944765746593475, "optim/lr": 0.0010031999999999999, "optim/total_tokens": 877133824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.6682589054107666, "created_at": "2025-01-15T10:16:07.545624+00:00"} {"global_step": 1674, "acc_step": 0, "speed/wps": 12873.702894031245, "speed/FLOPS": 202199194272350.3, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23852317035198212, "optim/lr": 0.0010038, "optim/total_tokens": 877658112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.7795681953430176, "created_at": "2025-01-15T10:16:17.732043+00:00"} {"global_step": 1675, "acc_step": 0, "speed/wps": 12867.576409484709, "speed/FLOPS": 202102969413875.0, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23883388936519623, "optim/lr": 0.0010044, "optim/total_tokens": 878182400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.7165002822875977, "created_at": "2025-01-15T10:16:27.926760+00:00"} {"global_step": 1676, "acc_step": 0, "speed/wps": 12871.495460682727, "speed/FLOPS": 202164523498280.56, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2707459330558777, "optim/lr": 0.001005, "optim/total_tokens": 878706688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 3.7705564498901367, "created_at": "2025-01-15T10:16:38.116402+00:00"} {"global_step": 1677, "acc_step": 0, "speed/wps": 12869.944174788556, "speed/FLOPS": 202140158421622.38, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2603784203529358, "optim/lr": 0.0010056, "optim/total_tokens": 879230976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342633, "loss/out": 3.7443180084228516, "created_at": "2025-01-15T10:16:48.312137+00:00"} {"global_step": 1678, "acc_step": 0, "speed/wps": 12866.648288435947, "speed/FLOPS": 202088391997432.22, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24003998935222626, "optim/lr": 0.0010061999999999998, "optim/total_tokens": 879755264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348945, "loss/out": 3.7399072647094727, "created_at": "2025-01-15T10:16:58.503512+00:00"} {"global_step": 1679, "acc_step": 0, "speed/wps": 12861.407453234446, "speed/FLOPS": 202006077478930.38, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.20668502151966095, "optim/lr": 0.0010068, "optim/total_tokens": 880279552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 3.625606060028076, "created_at": "2025-01-15T10:17:08.699266+00:00"} {"global_step": 1680, "acc_step": 0, "speed/wps": 12864.528014469548, "speed/FLOPS": 202055090181228.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.177489772439003, "optim/lr": 0.0010073999999999999, "optim/total_tokens": 880803840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 3.735050678253174, "created_at": "2025-01-15T10:17:18.894052+00:00"} {"global_step": 1681, "acc_step": 0, "speed/wps": 12872.817664770944, "speed/FLOPS": 202185290530384.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19688746333122253, "optim/lr": 0.001008, "optim/total_tokens": 881328128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 3.729731321334839, "created_at": "2025-01-15T10:17:29.079591+00:00"} {"global_step": 1682, "acc_step": 0, "speed/wps": 12871.149859787369, "speed/FLOPS": 202159095361313.56, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18799225986003876, "optim/lr": 0.0010086000000000001, "optim/total_tokens": 881852416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.7110419273376465, "created_at": "2025-01-15T10:17:39.267564+00:00"} {"global_step": 1683, "acc_step": 0, "speed/wps": 12866.596227605645, "speed/FLOPS": 202087574310553.25, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20774245262145996, "optim/lr": 0.0010092, "optim/total_tokens": 882376704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.6983649730682373, "created_at": "2025-01-15T10:17:49.463196+00:00"} {"global_step": 1684, "acc_step": 0, "speed/wps": 12871.851970958176, "speed/FLOPS": 202170122982052.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17488642036914825, "optim/lr": 0.0010098000000000002, "optim/total_tokens": 882900992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.7529544830322266, "created_at": "2025-01-15T10:17:59.652028+00:00"} {"global_step": 1685, "acc_step": 0, "speed/wps": 12873.164187729426, "speed/FLOPS": 202190733149620.22, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17366501688957214, "optim/lr": 0.0010104, "optim/total_tokens": 883425280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.741016387939453, "created_at": "2025-01-15T10:18:09.838494+00:00"} {"global_step": 1686, "acc_step": 0, "speed/wps": 12877.09817983854, "speed/FLOPS": 202252521901565.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19501502811908722, "optim/lr": 0.0010110000000000002, "optim/total_tokens": 883949568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.698531150817871, "created_at": "2025-01-15T10:18:20.023888+00:00"} {"global_step": 1687, "acc_step": 0, "speed/wps": 12872.986534907564, "speed/FLOPS": 202187942867932.3, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1542826145887375, "optim/lr": 0.0010116, "optim/total_tokens": 884473856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.7518553733825684, "created_at": "2025-01-15T10:18:30.210093+00:00"} {"global_step": 1688, "acc_step": 0, "speed/wps": 12873.634219220276, "speed/FLOPS": 202198115640074.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15662908554077148, "optim/lr": 0.0010122, "optim/total_tokens": 884998144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 3.634047508239746, "created_at": "2025-01-15T10:18:40.395608+00:00"} {"global_step": 1689, "acc_step": 0, "speed/wps": 12869.62277672733, "speed/FLOPS": 202135110423423.62, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17152263224124908, "optim/lr": 0.0010128000000000001, "optim/total_tokens": 885522432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 3.7188615798950195, "created_at": "2025-01-15T10:18:50.588739+00:00"} {"global_step": 1690, "acc_step": 0, "speed/wps": 12877.69622000383, "speed/FLOPS": 202261914944154.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15739643573760986, "optim/lr": 0.0010134, "optim/total_tokens": 886046720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.658520460128784, "created_at": "2025-01-15T10:19:00.772343+00:00"} {"global_step": 1691, "acc_step": 0, "speed/wps": 12872.627606978833, "speed/FLOPS": 202182305411592.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1546056717634201, "optim/lr": 0.0010140000000000001, "optim/total_tokens": 886571008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.667963743209839, "created_at": "2025-01-15T10:19:10.959954+00:00"} {"global_step": 1692, "acc_step": 0, "speed/wps": 12877.1780565489, "speed/FLOPS": 202253776475063.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18910828232765198, "optim/lr": 0.0010146, "optim/total_tokens": 887095296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.608942985534668, "created_at": "2025-01-15T10:19:21.142873+00:00"} {"global_step": 1693, "acc_step": 0, "speed/wps": 12878.232286555505, "speed/FLOPS": 202270334605981.38, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2007654905319214, "optim/lr": 0.0010152, "optim/total_tokens": 887619584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.7554869651794434, "created_at": "2025-01-15T10:19:31.326566+00:00"} {"global_step": 1694, "acc_step": 0, "speed/wps": 12871.215835979097, "speed/FLOPS": 202160131608065.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25410670042037964, "optim/lr": 0.0010158, "optim/total_tokens": 888143872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465818, "loss/out": 3.5823748111724854, "created_at": "2025-01-15T10:19:41.513462+00:00"} {"global_step": 1695, "acc_step": 0, "speed/wps": 12878.002645852534, "speed/FLOPS": 202266727783181.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35728949308395386, "optim/lr": 0.0010164, "optim/total_tokens": 888668160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380792, "loss/out": 3.756197929382324, "created_at": "2025-01-15T10:19:51.699332+00:00"} {"global_step": 1696, "acc_step": 0, "speed/wps": 12873.091201999232, "speed/FLOPS": 202189586808434.62, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5147317051887512, "optim/lr": 0.0010170000000000001, "optim/total_tokens": 889192448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 3.813894748687744, "created_at": "2025-01-15T10:20:01.888377+00:00"} {"global_step": 1697, "acc_step": 0, "speed/wps": 12872.545683898094, "speed/FLOPS": 202181018697035.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.546157717704773, "optim/lr": 0.0010176, "optim/total_tokens": 889716736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.8574018478393555, "created_at": "2025-01-15T10:20:12.075095+00:00"} {"global_step": 1698, "acc_step": 0, "speed/wps": 12877.75011935325, "speed/FLOPS": 202262761507502.8, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.7278385758399963, "optim/lr": 0.0010182, "optim/total_tokens": 890241024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.9186136722564697, "created_at": "2025-01-15T10:20:22.262083+00:00"} {"global_step": 1699, "acc_step": 0, "speed/wps": 12877.389568369625, "speed/FLOPS": 202257098558856.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9725061058998108, "optim/lr": 0.0010188, "optim/total_tokens": 890765312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.9475889205932617, "created_at": "2025-01-15T10:20:32.446941+00:00"} {"global_step": 1700, "acc_step": 0, "speed/wps": 12878.637968141675, "speed/FLOPS": 202276706392757.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6969351172447205, "optim/lr": 0.0010194, "optim/total_tokens": 891289600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.021972179412842, "created_at": "2025-01-15T10:20:42.630775+00:00"} {"global_step": 1701, "acc_step": 0, "speed/wps": 12877.877059512257, "speed/FLOPS": 202264755277134.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6032267212867737, "optim/lr": 0.00102, "optim/total_tokens": 891813888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.9687581062316895, "created_at": "2025-01-15T10:20:52.815921+00:00"} {"global_step": 1702, "acc_step": 0, "speed/wps": 12881.64380075764, "speed/FLOPS": 202323917124437.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5107849836349487, "optim/lr": 0.0010206, "optim/total_tokens": 892338176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 4.026865005493164, "created_at": "2025-01-15T10:21:02.996068+00:00"} {"global_step": 1703, "acc_step": 0, "speed/wps": 12876.955917292955, "speed/FLOPS": 202250287472797.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.632682204246521, "optim/lr": 0.0010211999999999999, "optim/total_tokens": 892862464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.000759124755859, "created_at": "2025-01-15T10:21:13.180912+00:00"} {"global_step": 1704, "acc_step": 0, "speed/wps": 12879.969538883752, "speed/FLOPS": 202297620541031.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4829217493534088, "optim/lr": 0.0010218, "optim/total_tokens": 893386752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.003664493560791, "created_at": "2025-01-15T10:21:23.364145+00:00"} {"global_step": 1705, "acc_step": 0, "speed/wps": 12878.630271354148, "speed/FLOPS": 202276585504132.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4767678678035736, "optim/lr": 0.0010224, "optim/total_tokens": 893911040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.9987735748291016, "created_at": "2025-01-15T10:21:33.545203+00:00"} {"global_step": 1706, "acc_step": 0, "speed/wps": 12874.344380271568, "speed/FLOPS": 202209269695249.06, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5052871108055115, "optim/lr": 0.001023, "optim/total_tokens": 894435328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 3.944194793701172, "created_at": "2025-01-15T10:21:43.731386+00:00"} {"global_step": 1707, "acc_step": 0, "speed/wps": 12878.650480302354, "speed/FLOPS": 202276902913434.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45707273483276367, "optim/lr": 0.0010236, "optim/total_tokens": 894959616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369341, "loss/out": 3.9263172149658203, "created_at": "2025-01-15T10:21:53.915941+00:00"} {"global_step": 1708, "acc_step": 0, "speed/wps": 12874.065251225351, "speed/FLOPS": 202204885590014.72, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.431283175945282, "optim/lr": 0.0010242, "optim/total_tokens": 895483904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 3.9439029693603516, "created_at": "2025-01-15T10:22:04.103653+00:00"} {"global_step": 1709, "acc_step": 0, "speed/wps": 12877.648190315751, "speed/FLOPS": 202261160571903.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4145227372646332, "optim/lr": 0.0010248, "optim/total_tokens": 896008192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.9027013778686523, "created_at": "2025-01-15T10:22:14.288344+00:00"} {"global_step": 1710, "acc_step": 0, "speed/wps": 12878.371990600817, "speed/FLOPS": 202272528849986.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2866702079772949, "optim/lr": 0.0010254, "optim/total_tokens": 896532480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.9544425010681152, "created_at": "2025-01-15T10:22:24.470666+00:00"} {"global_step": 1711, "acc_step": 0, "speed/wps": 12876.408602519878, "speed/FLOPS": 202241691142197.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24366872012615204, "optim/lr": 0.001026, "optim/total_tokens": 897056768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.7309961318969727, "created_at": "2025-01-15T10:22:34.654782+00:00"} {"global_step": 1712, "acc_step": 0, "speed/wps": 12878.52805144011, "speed/FLOPS": 202274979999917.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2688400149345398, "optim/lr": 0.0010266000000000001, "optim/total_tokens": 897581056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433975, "loss/out": 3.8328967094421387, "created_at": "2025-01-15T10:22:44.836616+00:00"} {"global_step": 1713, "acc_step": 0, "speed/wps": 12878.427581987315, "speed/FLOPS": 202273401989102.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20691817998886108, "optim/lr": 0.0010272, "optim/total_tokens": 898105344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290025, "loss/out": 3.725604772567749, "created_at": "2025-01-15T10:22:55.020584+00:00"} {"global_step": 1714, "acc_step": 0, "speed/wps": 12879.639611885565, "speed/FLOPS": 202292438584157.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23054341971874237, "optim/lr": 0.0010278000000000002, "optim/total_tokens": 898629632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.8950719833374023, "created_at": "2025-01-15T10:23:05.200736+00:00"} {"global_step": 1715, "acc_step": 0, "speed/wps": 12880.354011018684, "speed/FLOPS": 202303659204230.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21980056166648865, "optim/lr": 0.0010284, "optim/total_tokens": 899153920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.8268792629241943, "created_at": "2025-01-15T10:23:15.381277+00:00"} {"global_step": 1716, "acc_step": 0, "speed/wps": 12883.048789129873, "speed/FLOPS": 202345984397479.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18195198476314545, "optim/lr": 0.0010290000000000002, "optim/total_tokens": 899678208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 4.004312515258789, "created_at": "2025-01-15T10:23:25.558809+00:00"} {"global_step": 1717, "acc_step": 0, "speed/wps": 12879.086184291753, "speed/FLOPS": 202283746243307.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19201557338237762, "optim/lr": 0.0010296, "optim/total_tokens": 900202496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280980, "loss/out": 3.779003143310547, "created_at": "2025-01-15T10:23:35.739662+00:00"} {"global_step": 1718, "acc_step": 0, "speed/wps": 12876.156265760826, "speed/FLOPS": 202237727846649.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18707874417304993, "optim/lr": 0.0010302, "optim/total_tokens": 900726784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.78670597076416, "created_at": "2025-01-15T10:23:45.922596+00:00"} {"global_step": 1719, "acc_step": 0, "speed/wps": 12874.406903143872, "speed/FLOPS": 202210251702874.03, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19749626517295837, "optim/lr": 0.0010308000000000001, "optim/total_tokens": 901251072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 3.723853826522827, "created_at": "2025-01-15T10:23:56.108977+00:00"} {"global_step": 1720, "acc_step": 0, "speed/wps": 12877.793531858973, "speed/FLOPS": 202263443360560.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18864098191261292, "optim/lr": 0.0010314, "optim/total_tokens": 901775360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.755201816558838, "created_at": "2025-01-15T10:24:06.292537+00:00"} {"global_step": 1721, "acc_step": 0, "speed/wps": 12876.727957381445, "speed/FLOPS": 202246707049137.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1930576115846634, "optim/lr": 0.001032, "optim/total_tokens": 902299648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.739546298980713, "created_at": "2025-01-15T10:24:16.476138+00:00"} {"global_step": 1722, "acc_step": 0, "speed/wps": 12877.54404650701, "speed/FLOPS": 202259524850278.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2105047106742859, "optim/lr": 0.0010326, "optim/total_tokens": 902823936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.8046207427978516, "created_at": "2025-01-15T10:24:26.660809+00:00"} {"global_step": 1723, "acc_step": 0, "speed/wps": 12881.84797444705, "speed/FLOPS": 202327123952793.9, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20965875685214996, "optim/lr": 0.0010332, "optim/total_tokens": 903348224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314047, "loss/out": 3.773611307144165, "created_at": "2025-01-15T10:24:36.840608+00:00"} {"global_step": 1724, "acc_step": 0, "speed/wps": 12874.566225681241, "speed/FLOPS": 202212754082255.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22652903199195862, "optim/lr": 0.0010338, "optim/total_tokens": 903872512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.804222583770752, "created_at": "2025-01-15T10:24:47.026845+00:00"} {"global_step": 1725, "acc_step": 0, "speed/wps": 12881.150190896864, "speed/FLOPS": 202316164303283.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16979944705963135, "optim/lr": 0.0010344, "optim/total_tokens": 904396800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.727980136871338, "created_at": "2025-01-15T10:24:57.209045+00:00"} {"global_step": 1726, "acc_step": 0, "speed/wps": 12879.640257804784, "speed/FLOPS": 202292448729206.56, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.001, "optim/grad_norm": 0.15587888658046722, "optim/lr": 0.001035, "optim/total_tokens": 904921088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.847783088684082, "created_at": "2025-01-15T10:25:07.390681+00:00"} {"global_step": 1727, "acc_step": 0, "speed/wps": 12874.679089873636, "speed/FLOPS": 202214526769488.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15168780088424683, "optim/lr": 0.0010356, "optim/total_tokens": 905445376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.647946357727051, "created_at": "2025-01-15T10:25:17.579302+00:00"} {"global_step": 1728, "acc_step": 0, "speed/wps": 12881.220417279215, "speed/FLOPS": 202317267305119.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15262097120285034, "optim/lr": 0.0010362, "optim/total_tokens": 905969664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.6263699531555176, "created_at": "2025-01-15T10:25:27.758200+00:00"} {"global_step": 1729, "acc_step": 0, "speed/wps": 12876.501435766695, "speed/FLOPS": 202243149215904.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1413809359073639, "optim/lr": 0.0010368, "optim/total_tokens": 906493952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.8152666091918945, "created_at": "2025-01-15T10:25:37.940723+00:00"} {"global_step": 1730, "acc_step": 0, "speed/wps": 12878.222222796143, "speed/FLOPS": 202270176540811.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15230625867843628, "optim/lr": 0.0010374, "optim/total_tokens": 907018240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.7400565147399902, "created_at": "2025-01-15T10:25:48.122152+00:00"} {"global_step": 1731, "acc_step": 0, "speed/wps": 12878.772569719808, "speed/FLOPS": 202278820495503.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1695607453584671, "optim/lr": 0.0010379999999999999, "optim/total_tokens": 907542528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.6165590286254883, "created_at": "2025-01-15T10:25:58.309970+00:00"} {"global_step": 1732, "acc_step": 0, "speed/wps": 12880.70922383014, "speed/FLOPS": 202309238309548.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17576266825199127, "optim/lr": 0.0010386, "optim/total_tokens": 908066816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.66450834274292, "created_at": "2025-01-15T10:26:08.489329+00:00"} {"global_step": 1733, "acc_step": 0, "speed/wps": 12874.54195099309, "speed/FLOPS": 202212372814921.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1562671810388565, "optim/lr": 0.0010391999999999999, "optim/total_tokens": 908591104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.7376346588134766, "created_at": "2025-01-15T10:26:18.675678+00:00"} {"global_step": 1734, "acc_step": 0, "speed/wps": 12880.990173667635, "speed/FLOPS": 202313651012811.4, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1745089739561081, "optim/lr": 0.0010398, "optim/total_tokens": 909115392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 3.6222243309020996, "created_at": "2025-01-15T10:26:28.854829+00:00"} {"global_step": 1735, "acc_step": 0, "speed/wps": 12880.318098507476, "speed/FLOPS": 202303095148893.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21306383609771729, "optim/lr": 0.0010404, "optim/total_tokens": 909639680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.5205061435699463, "created_at": "2025-01-15T10:26:39.044136+00:00"} {"global_step": 1736, "acc_step": 0, "speed/wps": 12866.534906280041, "speed/FLOPS": 202086611174870.3, "speed/curr_iter_time": 1.2885, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.18378807604312897, "optim/lr": 0.001041, "optim/total_tokens": 910163968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.6483688354492188, "created_at": "2025-01-15T10:26:49.237505+00:00"} {"global_step": 1737, "acc_step": 0, "speed/wps": 12864.876550250363, "speed/FLOPS": 202060564414604.22, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17787282168865204, "optim/lr": 0.0010416, "optim/total_tokens": 910688256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 3.7197530269622803, "created_at": "2025-01-15T10:26:59.434951+00:00"} {"global_step": 1738, "acc_step": 0, "speed/wps": 12878.7333319962, "speed/FLOPS": 202278204213136.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1656579226255417, "optim/lr": 0.0010422, "optim/total_tokens": 911212544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.6770429611206055, "created_at": "2025-01-15T10:27:09.624221+00:00"} {"global_step": 1739, "acc_step": 0, "speed/wps": 12874.286062411953, "speed/FLOPS": 202208353733126.66, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17494304478168488, "optim/lr": 0.0010428, "optim/total_tokens": 911736832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.6601762771606445, "created_at": "2025-01-15T10:27:19.810846+00:00"} {"global_step": 1740, "acc_step": 0, "speed/wps": 12878.493618668343, "speed/FLOPS": 202274439185921.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20374475419521332, "optim/lr": 0.0010434, "optim/total_tokens": 912261120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 3.717604398727417, "created_at": "2025-01-15T10:27:29.992867+00:00"} {"global_step": 1741, "acc_step": 0, "speed/wps": 12876.721505825511, "speed/FLOPS": 202246605718586.56, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21345742046833038, "optim/lr": 0.001044, "optim/total_tokens": 912785408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.6958706378936768, "created_at": "2025-01-15T10:27:40.182686+00:00"} {"global_step": 1742, "acc_step": 0, "speed/wps": 12879.628757387814, "speed/FLOPS": 202292268099354.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2186489701271057, "optim/lr": 0.0010446000000000001, "optim/total_tokens": 913309696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 3.689657211303711, "created_at": "2025-01-15T10:27:50.363656+00:00"} {"global_step": 1743, "acc_step": 0, "speed/wps": 12875.918295205796, "speed/FLOPS": 202233990192076.88, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22417119145393372, "optim/lr": 0.0010452, "optim/total_tokens": 913833984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.5439696311950684, "created_at": "2025-01-15T10:28:00.547464+00:00"} {"global_step": 1744, "acc_step": 0, "speed/wps": 12877.524168597736, "speed/FLOPS": 202259212640398.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21203815937042236, "optim/lr": 0.0010458000000000002, "optim/total_tokens": 914358272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.7468671798706055, "created_at": "2025-01-15T10:28:10.731611+00:00"} {"global_step": 1745, "acc_step": 0, "speed/wps": 12866.556571176052, "speed/FLOPS": 202086951451832.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21269473433494568, "optim/lr": 0.0010464, "optim/total_tokens": 914882560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.574336290359497, "created_at": "2025-01-15T10:28:20.922803+00:00"} {"global_step": 1746, "acc_step": 0, "speed/wps": 12875.526625625811, "speed/FLOPS": 202227838483112.8, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1882334202528, "optim/lr": 0.001047, "optim/total_tokens": 915406848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.6288084983825684, "created_at": "2025-01-15T10:28:31.109447+00:00"} {"global_step": 1747, "acc_step": 0, "speed/wps": 12875.880391872624, "speed/FLOPS": 202233394868144.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17958921194076538, "optim/lr": 0.0010476, "optim/total_tokens": 915931136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.6039509773254395, "created_at": "2025-01-15T10:28:41.293789+00:00"} {"global_step": 1748, "acc_step": 0, "speed/wps": 12879.51642873145, "speed/FLOPS": 202290503823450.84, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1860019564628601, "optim/lr": 0.0010482, "optim/total_tokens": 916455424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.6669883728027344, "created_at": "2025-01-15T10:28:51.478293+00:00"} {"global_step": 1749, "acc_step": 0, "speed/wps": 12878.838331803905, "speed/FLOPS": 202279853379400.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1789012849330902, "optim/lr": 0.0010488000000000001, "optim/total_tokens": 916979712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 3.67838454246521, "created_at": "2025-01-15T10:29:01.662264+00:00"} {"global_step": 1750, "acc_step": 0, "speed/wps": 12876.201924675033, "speed/FLOPS": 202238444982638.97, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18767982721328735, "optim/lr": 0.0010494, "optim/total_tokens": 917504000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.6373486518859863, "created_at": "2025-01-15T10:29:11.847836+00:00"} {"global_step": 1751, "acc_step": 0, "speed/wps": 12876.7135592046, "speed/FLOPS": 202246480905985.34, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16854532063007355, "optim/lr": 0.00105, "optim/total_tokens": 918028288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.6802568435668945, "created_at": "2025-01-15T10:29:22.031576+00:00"} {"global_step": 1752, "acc_step": 0, "speed/wps": 12861.312864115815, "speed/FLOPS": 202004591826843.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16289299726486206, "optim/lr": 0.0010506, "optim/total_tokens": 918552576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.707463264465332, "created_at": "2025-01-15T10:29:32.230140+00:00"} {"global_step": 1753, "acc_step": 0, "speed/wps": 12870.62872955973, "speed/FLOPS": 202150910294980.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15518683195114136, "optim/lr": 0.0010512, "optim/total_tokens": 919076864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.55910587310791, "created_at": "2025-01-15T10:29:42.418494+00:00"} {"global_step": 1754, "acc_step": 0, "speed/wps": 12870.876393283403, "speed/FLOPS": 202154800194086.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18418407440185547, "optim/lr": 0.0010518, "optim/total_tokens": 919601152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.6952896118164062, "created_at": "2025-01-15T10:29:52.606742+00:00"} {"global_step": 1755, "acc_step": 0, "speed/wps": 12874.776309023551, "speed/FLOPS": 202216053729831.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19673548638820648, "optim/lr": 0.0010524, "optim/total_tokens": 920125440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.6751437187194824, "created_at": "2025-01-15T10:30:02.791484+00:00"} {"global_step": 1756, "acc_step": 0, "speed/wps": 12866.74549926629, "speed/FLOPS": 202089918827104.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20334653556346893, "optim/lr": 0.001053, "optim/total_tokens": 920649728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350240, "loss/out": 3.780146598815918, "created_at": "2025-01-15T10:30:12.985431+00:00"} {"global_step": 1757, "acc_step": 0, "speed/wps": 12865.837628017907, "speed/FLOPS": 202075659461602.75, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23221537470817566, "optim/lr": 0.0010536, "optim/total_tokens": 921174016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 3.691530704498291, "created_at": "2025-01-15T10:30:23.177440+00:00"} {"global_step": 1758, "acc_step": 0, "speed/wps": 12879.909914699774, "speed/FLOPS": 202296684061289.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20189251005649567, "optim/lr": 0.0010542, "optim/total_tokens": 921698304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.5881497859954834, "created_at": "2025-01-15T10:30:33.358009+00:00"} {"global_step": 1759, "acc_step": 0, "speed/wps": 12875.61849437301, "speed/FLOPS": 202229281408028.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16446194052696228, "optim/lr": 0.0010548, "optim/total_tokens": 922222592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.6846654415130615, "created_at": "2025-01-15T10:30:43.544485+00:00"} {"global_step": 1760, "acc_step": 0, "speed/wps": 12876.148647334272, "speed/FLOPS": 202237608188791.6, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20465080440044403, "optim/lr": 0.0010554, "optim/total_tokens": 922746880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.606191635131836, "created_at": "2025-01-15T10:30:53.728161+00:00"} {"global_step": 1761, "acc_step": 0, "speed/wps": 12881.502949268202, "speed/FLOPS": 202321704858245.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1784995198249817, "optim/lr": 0.0010559999999999999, "optim/total_tokens": 923271168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 3.6043202877044678, "created_at": "2025-01-15T10:31:03.910330+00:00"} {"global_step": 1762, "acc_step": 0, "speed/wps": 12882.010754439938, "speed/FLOPS": 202329680636265.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1683260202407837, "optim/lr": 0.0010566, "optim/total_tokens": 923795456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.632608413696289, "created_at": "2025-01-15T10:31:14.092806+00:00"} {"global_step": 1763, "acc_step": 0, "speed/wps": 12879.56322885881, "speed/FLOPS": 202291238883761.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1561361849308014, "optim/lr": 0.0010572, "optim/total_tokens": 924319744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 3.625349998474121, "created_at": "2025-01-15T10:31:24.277428+00:00"} {"global_step": 1764, "acc_step": 0, "speed/wps": 12880.530104871441, "speed/FLOPS": 202306425000166.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1721445620059967, "optim/lr": 0.0010578, "optim/total_tokens": 924844032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.6462841033935547, "created_at": "2025-01-15T10:31:34.458919+00:00"} {"global_step": 1765, "acc_step": 0, "speed/wps": 12882.205882670716, "speed/FLOPS": 202332745393264.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18838639557361603, "optim/lr": 0.0010584, "optim/total_tokens": 925368320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.6694281101226807, "created_at": "2025-01-15T10:31:44.637324+00:00"} {"global_step": 1766, "acc_step": 0, "speed/wps": 12877.37946704437, "speed/FLOPS": 202256939903662.2, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.20114946365356445, "optim/lr": 0.001059, "optim/total_tokens": 925892608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.5804500579833984, "created_at": "2025-01-15T10:31:54.820593+00:00"} {"global_step": 1767, "acc_step": 0, "speed/wps": 12875.650226035237, "speed/FLOPS": 202229779797387.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.25567859411239624, "optim/lr": 0.0010596000000000002, "optim/total_tokens": 926416896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.7156519889831543, "created_at": "2025-01-15T10:32:05.006225+00:00"} {"global_step": 1768, "acc_step": 0, "speed/wps": 12876.753204045508, "speed/FLOPS": 202247103582689.53, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.26115795969963074, "optim/lr": 0.0010602, "optim/total_tokens": 926941184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.6978907585144043, "created_at": "2025-01-15T10:32:15.188708+00:00"} {"global_step": 1769, "acc_step": 0, "speed/wps": 12879.247117145234, "speed/FLOPS": 202286273914916.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22202272713184357, "optim/lr": 0.0010608000000000002, "optim/total_tokens": 927465472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 3.7115540504455566, "created_at": "2025-01-15T10:32:25.371754+00:00"} {"global_step": 1770, "acc_step": 0, "speed/wps": 12879.105634482205, "speed/FLOPS": 202284051735276.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2133978009223938, "optim/lr": 0.0010614, "optim/total_tokens": 927989760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.7485201358795166, "created_at": "2025-01-15T10:32:35.552753+00:00"} {"global_step": 1771, "acc_step": 0, "speed/wps": 12881.03909009382, "speed/FLOPS": 202314419312502.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23979225754737854, "optim/lr": 0.001062, "optim/total_tokens": 928514048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.657639980316162, "created_at": "2025-01-15T10:32:45.731874+00:00"} {"global_step": 1772, "acc_step": 0, "speed/wps": 12877.748851460787, "speed/FLOPS": 202262741593509.4, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20429906249046326, "optim/lr": 0.0010626000000000001, "optim/total_tokens": 929038336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.658910036087036, "created_at": "2025-01-15T10:32:55.917542+00:00"} {"global_step": 1773, "acc_step": 0, "speed/wps": 12882.292547564562, "speed/FLOPS": 202334106584511.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1758631318807602, "optim/lr": 0.0010632, "optim/total_tokens": 929562624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.612308979034424, "created_at": "2025-01-15T10:33:06.096622+00:00"} {"global_step": 1774, "acc_step": 0, "speed/wps": 12878.931030376798, "speed/FLOPS": 202281309337869.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20212087035179138, "optim/lr": 0.0010638000000000002, "optim/total_tokens": 930086912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 3.564927101135254, "created_at": "2025-01-15T10:33:16.277471+00:00"} {"global_step": 1775, "acc_step": 0, "speed/wps": 12880.717549089084, "speed/FLOPS": 202309369069179.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18713928759098053, "optim/lr": 0.0010644, "optim/total_tokens": 930611200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.712394952774048, "created_at": "2025-01-15T10:33:26.461288+00:00"} {"global_step": 1776, "acc_step": 0, "speed/wps": 12874.159680597548, "speed/FLOPS": 202206368733064.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17796826362609863, "optim/lr": 0.001065, "optim/total_tokens": 931135488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.553377866744995, "created_at": "2025-01-15T10:33:36.645805+00:00"} {"global_step": 1777, "acc_step": 0, "speed/wps": 12880.879384499707, "speed/FLOPS": 202311910916688.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2573736608028412, "optim/lr": 0.0010656, "optim/total_tokens": 931659776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.680530071258545, "created_at": "2025-01-15T10:33:46.825840+00:00"} {"global_step": 1778, "acc_step": 0, "speed/wps": 12876.120039734154, "speed/FLOPS": 202237158867120.78, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3396252393722534, "optim/lr": 0.0010662, "optim/total_tokens": 932184064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.7207417488098145, "created_at": "2025-01-15T10:33:57.009019+00:00"} {"global_step": 1779, "acc_step": 0, "speed/wps": 12877.956191724696, "speed/FLOPS": 202265998157269.47, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2846192419528961, "optim/lr": 0.0010668000000000001, "optim/total_tokens": 932708352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.6758017539978027, "created_at": "2025-01-15T10:34:07.195134+00:00"} {"global_step": 1780, "acc_step": 0, "speed/wps": 12884.464431178285, "speed/FLOPS": 202368219001146.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3741290271282196, "optim/lr": 0.0010674, "optim/total_tokens": 933232640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357380, "loss/out": 3.741307497024536, "created_at": "2025-01-15T10:34:17.373822+00:00"} {"global_step": 1781, "acc_step": 0, "speed/wps": 12877.258308933338, "speed/FLOPS": 202255036949039.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4340894818305969, "optim/lr": 0.001068, "optim/total_tokens": 933756928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 3.799473524093628, "created_at": "2025-01-15T10:34:27.558464+00:00"} {"global_step": 1782, "acc_step": 0, "speed/wps": 12878.390631391707, "speed/FLOPS": 202272821629222.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.34553080797195435, "optim/lr": 0.0010686, "optim/total_tokens": 934281216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.907357931137085, "created_at": "2025-01-15T10:34:37.741624+00:00"} {"global_step": 1783, "acc_step": 0, "speed/wps": 12877.122128373267, "speed/FLOPS": 202252898046212.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4056456983089447, "optim/lr": 0.0010692, "optim/total_tokens": 934805504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.6670451164245605, "created_at": "2025-01-15T10:34:47.924391+00:00"} {"global_step": 1784, "acc_step": 0, "speed/wps": 12878.271389241454, "speed/FLOPS": 202270948767396.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36474278569221497, "optim/lr": 0.0010697999999999999, "optim/total_tokens": 935329792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.8460981845855713, "created_at": "2025-01-15T10:34:58.106309+00:00"} {"global_step": 1785, "acc_step": 0, "speed/wps": 12876.99401075782, "speed/FLOPS": 202250885783009.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4427381455898285, "optim/lr": 0.0010704, "optim/total_tokens": 935854080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.7234082221984863, "created_at": "2025-01-15T10:35:08.292215+00:00"} {"global_step": 1786, "acc_step": 0, "speed/wps": 12878.7977925817, "speed/FLOPS": 202279216655209.47, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4065672755241394, "optim/lr": 0.001071, "optim/total_tokens": 936378368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329371, "loss/out": 3.872889995574951, "created_at": "2025-01-15T10:35:18.474659+00:00"} {"global_step": 1787, "acc_step": 0, "speed/wps": 12877.543938103307, "speed/FLOPS": 202259523147649.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.299867182970047, "optim/lr": 0.0010716, "optim/total_tokens": 936902656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.788236141204834, "created_at": "2025-01-15T10:35:28.660896+00:00"} {"global_step": 1788, "acc_step": 0, "speed/wps": 12877.178151985867, "speed/FLOPS": 202253777974032.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.26633575558662415, "optim/lr": 0.0010722, "optim/total_tokens": 937426944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.8457627296447754, "created_at": "2025-01-15T10:35:38.847885+00:00"} {"global_step": 1789, "acc_step": 0, "speed/wps": 12876.70563144386, "speed/FLOPS": 202246356389609.03, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2644821107387543, "optim/lr": 0.0010727999999999998, "optim/total_tokens": 937951232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.700545310974121, "created_at": "2025-01-15T10:35:49.031543+00:00"} {"global_step": 1790, "acc_step": 0, "speed/wps": 12879.313473388187, "speed/FLOPS": 202287316130895.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21958787739276886, "optim/lr": 0.0010734, "optim/total_tokens": 938475520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 3.7577173709869385, "created_at": "2025-01-15T10:35:59.218852+00:00"} {"global_step": 1791, "acc_step": 0, "speed/wps": 12880.781884535025, "speed/FLOPS": 202310379545764.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.23328305780887604, "optim/lr": 0.001074, "optim/total_tokens": 938999808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417514, "loss/out": 3.889756202697754, "created_at": "2025-01-15T10:36:09.401083+00:00"} {"global_step": 1792, "acc_step": 0, "speed/wps": 12876.44683176244, "speed/FLOPS": 202242291584984.3, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2223329395055771, "optim/lr": 0.0010746, "optim/total_tokens": 939524096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 3.7859203815460205, "created_at": "2025-01-15T10:36:19.586567+00:00"} {"global_step": 1793, "acc_step": 0, "speed/wps": 12874.988938630831, "speed/FLOPS": 202219393369997.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1643466204404831, "optim/lr": 0.0010752000000000001, "optim/total_tokens": 940048384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 3.67291259765625, "created_at": "2025-01-15T10:36:29.770316+00:00"} {"global_step": 1794, "acc_step": 0, "speed/wps": 12879.752635769331, "speed/FLOPS": 202294213779563.28, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15376374125480652, "optim/lr": 0.0010758, "optim/total_tokens": 940572672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.7784738540649414, "created_at": "2025-01-15T10:36:39.951358+00:00"} {"global_step": 1795, "acc_step": 0, "speed/wps": 12879.464537162949, "speed/FLOPS": 202289688795061.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19091330468654633, "optim/lr": 0.0010764000000000001, "optim/total_tokens": 941096960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.6682653427124023, "created_at": "2025-01-15T10:36:50.140112+00:00"} {"global_step": 1796, "acc_step": 0, "speed/wps": 12877.453498237253, "speed/FLOPS": 202258102665275.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1759835034608841, "optim/lr": 0.001077, "optim/total_tokens": 941621248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.7046594619750977, "created_at": "2025-01-15T10:37:00.323090+00:00"} {"global_step": 1797, "acc_step": 0, "speed/wps": 12877.99685644939, "speed/FLOPS": 202266636852649.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16443899273872375, "optim/lr": 0.0010776000000000002, "optim/total_tokens": 942145536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 3.739745616912842, "created_at": "2025-01-15T10:37:10.506080+00:00"} {"global_step": 1798, "acc_step": 0, "speed/wps": 12881.443240145543, "speed/FLOPS": 202320767044425.53, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22314396500587463, "optim/lr": 0.0010782, "optim/total_tokens": 942669824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470104, "loss/out": 3.6189732551574707, "created_at": "2025-01-15T10:37:20.685537+00:00"} {"global_step": 1799, "acc_step": 0, "speed/wps": 12878.636869778536, "speed/FLOPS": 202276689141454.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31136614084243774, "optim/lr": 0.0010788, "optim/total_tokens": 943194112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.8598647117614746, "created_at": "2025-01-15T10:37:30.873549+00:00"} {"global_step": 1800, "acc_step": 0, "speed/wps": 12878.11741450883, "speed/FLOPS": 202268530382637.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.227690652012825, "optim/lr": 0.0010794, "optim/total_tokens": 943718400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.734175682067871, "created_at": "2025-01-15T10:37:41.056484+00:00"} {"global_step": 1801, "acc_step": 0, "speed/wps": 12882.18045543546, "speed/FLOPS": 202332346023593.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.241115540266037, "optim/lr": 0.00108, "optim/total_tokens": 944242688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.6750779151916504, "created_at": "2025-01-15T10:37:51.238900+00:00"} {"global_step": 1802, "acc_step": 0, "speed/wps": 12881.249161908425, "speed/FLOPS": 202317718779020.16, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24019622802734375, "optim/lr": 0.0010806000000000001, "optim/total_tokens": 944766976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402407, "loss/out": 3.713223934173584, "created_at": "2025-01-15T10:38:01.420221+00:00"} {"global_step": 1803, "acc_step": 0, "speed/wps": 12876.05247946347, "speed/FLOPS": 202236097740232.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19488759338855743, "optim/lr": 0.0010812, "optim/total_tokens": 945291264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.6702396869659424, "created_at": "2025-01-15T10:38:11.604586+00:00"} {"global_step": 1804, "acc_step": 0, "speed/wps": 12876.480554973019, "speed/FLOPS": 202242821254347.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1738516390323639, "optim/lr": 0.0010818, "optim/total_tokens": 945815552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.70039439201355, "created_at": "2025-01-15T10:38:21.787817+00:00"} {"global_step": 1805, "acc_step": 0, "speed/wps": 12879.730875726958, "speed/FLOPS": 202293872008196.38, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19390971958637238, "optim/lr": 0.0010824, "optim/total_tokens": 946339840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.7026400566101074, "created_at": "2025-01-15T10:38:31.968217+00:00"} {"global_step": 1806, "acc_step": 0, "speed/wps": 12878.493840221628, "speed/FLOPS": 202274442665720.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.16822756826877594, "optim/lr": 0.001083, "optim/total_tokens": 946864128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.6494851112365723, "created_at": "2025-01-15T10:38:42.150206+00:00"} {"global_step": 1807, "acc_step": 0, "speed/wps": 12879.890074277637, "speed/FLOPS": 202296372440196.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2197442501783371, "optim/lr": 0.0010836, "optim/total_tokens": 947388416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.708963632583618, "created_at": "2025-01-15T10:38:52.332213+00:00"} {"global_step": 1808, "acc_step": 0, "speed/wps": 12880.168441034639, "speed/FLOPS": 202300744572631.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1823442429304123, "optim/lr": 0.0010842, "optim/total_tokens": 947912704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432352, "loss/out": 3.684110164642334, "created_at": "2025-01-15T10:39:02.512324+00:00"} {"global_step": 1809, "acc_step": 0, "speed/wps": 12884.782661686657, "speed/FLOPS": 202373217248574.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18048551678657532, "optim/lr": 0.0010848, "optim/total_tokens": 948436992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.7517504692077637, "created_at": "2025-01-15T10:39:12.691884+00:00"} {"global_step": 1810, "acc_step": 0, "speed/wps": 12879.000876636543, "speed/FLOPS": 202282406369357.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16245964169502258, "optim/lr": 0.0010854, "optim/total_tokens": 948961280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.6179373264312744, "created_at": "2025-01-15T10:39:22.872667+00:00"} {"global_step": 1811, "acc_step": 0, "speed/wps": 12875.443022447613, "speed/FLOPS": 202226525380316.4, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1479317545890808, "optim/lr": 0.001086, "optim/total_tokens": 949485568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.65948486328125, "created_at": "2025-01-15T10:39:33.056397+00:00"} {"global_step": 1812, "acc_step": 0, "speed/wps": 12884.161644393464, "speed/FLOPS": 202363463318618.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1473681926727295, "optim/lr": 0.0010866, "optim/total_tokens": 950009856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.635744094848633, "created_at": "2025-01-15T10:39:43.234097+00:00"} {"global_step": 1813, "acc_step": 0, "speed/wps": 12880.10017934205, "speed/FLOPS": 202299672428947.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1555107980966568, "optim/lr": 0.0010872, "optim/total_tokens": 950534144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 3.683432102203369, "created_at": "2025-01-15T10:39:53.413881+00:00"} {"global_step": 1814, "acc_step": 0, "speed/wps": 12884.65653611165, "speed/FLOPS": 202371236273104.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14902296662330627, "optim/lr": 0.0010877999999999999, "optim/total_tokens": 951058432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.6074700355529785, "created_at": "2025-01-15T10:40:03.593329+00:00"} {"global_step": 1815, "acc_step": 0, "speed/wps": 12881.33631971898, "speed/FLOPS": 202319087712202.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.15549172461032867, "optim/lr": 0.0010884, "optim/total_tokens": 951582720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.723620653152466, "created_at": "2025-01-15T10:40:13.775059+00:00"} {"global_step": 1816, "acc_step": 0, "speed/wps": 12883.621427927015, "speed/FLOPS": 202354978476677.22, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1687215119600296, "optim/lr": 0.001089, "optim/total_tokens": 952107008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 3.6589088439941406, "created_at": "2025-01-15T10:40:23.954444+00:00"} {"global_step": 1817, "acc_step": 0, "speed/wps": 12881.079603343118, "speed/FLOPS": 202315055628754.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17202015221118927, "optim/lr": 0.0010896, "optim/total_tokens": 952631296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361573, "loss/out": 3.6742148399353027, "created_at": "2025-01-15T10:40:34.133855+00:00"} {"global_step": 1818, "acc_step": 0, "speed/wps": 12881.440843208926, "speed/FLOPS": 202320729397242.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17822004854679108, "optim/lr": 0.0010902, "optim/total_tokens": 953155584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 3.5676772594451904, "created_at": "2025-01-15T10:40:44.314093+00:00"} {"global_step": 1819, "acc_step": 0, "speed/wps": 12880.388610522601, "speed/FLOPS": 202304202636984.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19557325541973114, "optim/lr": 0.0010908, "optim/total_tokens": 953679872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.6255710124969482, "created_at": "2025-01-15T10:40:54.493861+00:00"} {"global_step": 1820, "acc_step": 0, "speed/wps": 12883.2702930591, "speed/FLOPS": 202349463421066.75, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18293915688991547, "optim/lr": 0.0010914, "optim/total_tokens": 954204160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501150, "loss/out": 3.7426626682281494, "created_at": "2025-01-15T10:41:04.673737+00:00"} {"global_step": 1821, "acc_step": 0, "speed/wps": 12882.070695216791, "speed/FLOPS": 202330622088533.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16932693123817444, "optim/lr": 0.001092, "optim/total_tokens": 954728448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 3.6155710220336914, "created_at": "2025-01-15T10:41:14.854321+00:00"} {"global_step": 1822, "acc_step": 0, "speed/wps": 12878.878794675875, "speed/FLOPS": 202280488904407.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15895254909992218, "optim/lr": 0.0010926, "optim/total_tokens": 955252736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 3.6978697776794434, "created_at": "2025-01-15T10:41:25.042191+00:00"} {"global_step": 1823, "acc_step": 0, "speed/wps": 12880.997763635374, "speed/FLOPS": 202313770223684.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.16050508618354797, "optim/lr": 0.0010932000000000001, "optim/total_tokens": 955777024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 3.697502851486206, "created_at": "2025-01-15T10:41:35.225393+00:00"} {"global_step": 1824, "acc_step": 0, "speed/wps": 12884.619964981313, "speed/FLOPS": 202370661873248.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1928686797618866, "optim/lr": 0.0010938, "optim/total_tokens": 956301312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.5111122131347656, "created_at": "2025-01-15T10:41:45.405093+00:00"} {"global_step": 1825, "acc_step": 0, "speed/wps": 12882.241951925598, "speed/FLOPS": 202333311910476.3, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21363091468811035, "optim/lr": 0.0010944000000000001, "optim/total_tokens": 956825600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.732635736465454, "created_at": "2025-01-15T10:41:55.590733+00:00"} {"global_step": 1826, "acc_step": 0, "speed/wps": 12884.403509143414, "speed/FLOPS": 202367262136873.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1906590461730957, "optim/lr": 0.001095, "optim/total_tokens": 957349888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.6170151233673096, "created_at": "2025-01-15T10:42:05.767607+00:00"} {"global_step": 1827, "acc_step": 0, "speed/wps": 12877.211537337504, "speed/FLOPS": 202254302336856.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.183735653758049, "optim/lr": 0.0010956000000000002, "optim/total_tokens": 957874176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.769725799560547, "created_at": "2025-01-15T10:42:15.949841+00:00"} {"global_step": 1828, "acc_step": 0, "speed/wps": 12882.022137661705, "speed/FLOPS": 202329859425405.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.158920556306839, "optim/lr": 0.0010962, "optim/total_tokens": 958398464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361049, "loss/out": 3.688300848007202, "created_at": "2025-01-15T10:42:26.130878+00:00"} {"global_step": 1829, "acc_step": 0, "speed/wps": 12884.387786850733, "speed/FLOPS": 202367015196661.75, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16816532611846924, "optim/lr": 0.0010968, "optim/total_tokens": 958922752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.652888059616089, "created_at": "2025-01-15T10:42:36.311827+00:00"} {"global_step": 1830, "acc_step": 0, "speed/wps": 12882.543189797316, "speed/FLOPS": 202338043265195.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17167828977108002, "optim/lr": 0.0010974000000000001, "optim/total_tokens": 959447040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 3.707479476928711, "created_at": "2025-01-15T10:42:46.494778+00:00"} {"global_step": 1831, "acc_step": 0, "speed/wps": 12884.831467724649, "speed/FLOPS": 202373983814468.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19019225239753723, "optim/lr": 0.001098, "optim/total_tokens": 959971328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.5394186973571777, "created_at": "2025-01-15T10:42:56.670800+00:00"} {"global_step": 1832, "acc_step": 0, "speed/wps": 12885.3365164407, "speed/FLOPS": 202381916298563.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19560766220092773, "optim/lr": 0.0010986000000000001, "optim/total_tokens": 960495616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482614, "loss/out": 3.660590887069702, "created_at": "2025-01-15T10:43:06.849416+00:00"} {"global_step": 1833, "acc_step": 0, "speed/wps": 12885.701611181888, "speed/FLOPS": 202387650613166.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21071757376194, "optim/lr": 0.0010992, "optim/total_tokens": 961019904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 3.756805896759033, "created_at": "2025-01-15T10:43:17.027879+00:00"} {"global_step": 1834, "acc_step": 0, "speed/wps": 12879.210113523839, "speed/FLOPS": 202285692722194.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19381392002105713, "optim/lr": 0.0010998, "optim/total_tokens": 961544192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.731325149536133, "created_at": "2025-01-15T10:43:27.208512+00:00"} {"global_step": 1835, "acc_step": 0, "speed/wps": 12884.761443785694, "speed/FLOPS": 202372883992283.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18949083983898163, "optim/lr": 0.0011004, "optim/total_tokens": 962068480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.6972899436950684, "created_at": "2025-01-15T10:43:37.394008+00:00"} {"global_step": 1836, "acc_step": 0, "speed/wps": 12882.678153678717, "speed/FLOPS": 202340163058413.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19000525772571564, "optim/lr": 0.001101, "optim/total_tokens": 962592768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.5190470218658447, "created_at": "2025-01-15T10:43:47.574264+00:00"} {"global_step": 1837, "acc_step": 0, "speed/wps": 12877.617635367607, "speed/FLOPS": 202260680664456.22, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18629547953605652, "optim/lr": 0.0011016, "optim/total_tokens": 963117056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473062, "loss/out": 3.6427676677703857, "created_at": "2025-01-15T10:43:57.758441+00:00"} {"global_step": 1838, "acc_step": 0, "speed/wps": 12882.596597897336, "speed/FLOPS": 202338882112797.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2202809602022171, "optim/lr": 0.0011022, "optim/total_tokens": 963641344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 3.619997262954712, "created_at": "2025-01-15T10:44:07.939061+00:00"} {"global_step": 1839, "acc_step": 0, "speed/wps": 12878.078240875575, "speed/FLOPS": 202267915106897.12, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2176683098077774, "optim/lr": 0.0011028, "optim/total_tokens": 964165632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.6458520889282227, "created_at": "2025-01-15T10:44:18.123256+00:00"} {"global_step": 1840, "acc_step": 0, "speed/wps": 12884.872052018656, "speed/FLOPS": 202374621246574.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.18344710767269135, "optim/lr": 0.0011034, "optim/total_tokens": 964689920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.662992477416992, "created_at": "2025-01-15T10:44:28.301118+00:00"} {"global_step": 1841, "acc_step": 0, "speed/wps": 12880.56356536973, "speed/FLOPS": 202306950543271.6, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24806459248065948, "optim/lr": 0.001104, "optim/total_tokens": 965214208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.62412691116333, "created_at": "2025-01-15T10:44:38.481845+00:00"} {"global_step": 1842, "acc_step": 0, "speed/wps": 12878.195986390188, "speed/FLOPS": 202269764462003.44, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2808964252471924, "optim/lr": 0.0011046, "optim/total_tokens": 965738496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420635, "loss/out": 3.6832733154296875, "created_at": "2025-01-15T10:44:48.663754+00:00"} {"global_step": 1843, "acc_step": 0, "speed/wps": 12877.816611000371, "speed/FLOPS": 202263805850189.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19178174436092377, "optim/lr": 0.0011052, "optim/total_tokens": 966262784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 3.707691192626953, "created_at": "2025-01-15T10:44:58.846505+00:00"} {"global_step": 1844, "acc_step": 0, "speed/wps": 12882.287263704779, "speed/FLOPS": 202334023594233.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.213431715965271, "optim/lr": 0.0011057999999999999, "optim/total_tokens": 966787072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.6344008445739746, "created_at": "2025-01-15T10:45:09.025907+00:00"} {"global_step": 1845, "acc_step": 0, "speed/wps": 12880.170203313131, "speed/FLOPS": 202300772251636.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23501840233802795, "optim/lr": 0.0011064, "optim/total_tokens": 967311360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.596163034439087, "created_at": "2025-01-15T10:45:19.207480+00:00"} {"global_step": 1846, "acc_step": 0, "speed/wps": 12883.352911722328, "speed/FLOPS": 202350761060702.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27527090907096863, "optim/lr": 0.001107, "optim/total_tokens": 967835648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.664978504180908, "created_at": "2025-01-15T10:45:29.384653+00:00"} {"global_step": 1847, "acc_step": 0, "speed/wps": 12880.67668563372, "speed/FLOPS": 202308727252461.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3006916046142578, "optim/lr": 0.0011076, "optim/total_tokens": 968359936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.684147834777832, "created_at": "2025-01-15T10:45:39.564437+00:00"} {"global_step": 1848, "acc_step": 0, "speed/wps": 12876.762829590518, "speed/FLOPS": 202247254765100.78, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21685443818569183, "optim/lr": 0.0011082, "optim/total_tokens": 968884224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402039, "loss/out": 3.7635457515716553, "created_at": "2025-01-15T10:45:49.749394+00:00"} {"global_step": 1849, "acc_step": 0, "speed/wps": 12876.789952026753, "speed/FLOPS": 202247680760232.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2536993622779846, "optim/lr": 0.0011088, "optim/total_tokens": 969408512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.656306743621826, "created_at": "2025-01-15T10:45:59.933113+00:00"} {"global_step": 1850, "acc_step": 0, "speed/wps": 12873.632312709262, "speed/FLOPS": 202198085695699.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.199655219912529, "optim/lr": 0.0011094, "optim/total_tokens": 969932800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 3.692023277282715, "created_at": "2025-01-15T10:46:10.122339+00:00"} {"global_step": 1851, "acc_step": 0, "speed/wps": 12879.107045250894, "speed/FLOPS": 202284073893336.88, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2647704780101776, "optim/lr": 0.00111, "optim/total_tokens": 970457088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.6274678707122803, "created_at": "2025-01-15T10:46:20.304738+00:00"} {"global_step": 1852, "acc_step": 0, "speed/wps": 12879.077402673176, "speed/FLOPS": 202283608315920.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17403478920459747, "optim/lr": 0.0011106, "optim/total_tokens": 970981376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.733285903930664, "created_at": "2025-01-15T10:46:30.487332+00:00"} {"global_step": 1853, "acc_step": 0, "speed/wps": 12885.478696418524, "speed/FLOPS": 202384149430490.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20940090715885162, "optim/lr": 0.0011112000000000001, "optim/total_tokens": 971505664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.5981392860412598, "created_at": "2025-01-15T10:46:40.667490+00:00"} {"global_step": 1854, "acc_step": 0, "speed/wps": 12883.719567676275, "speed/FLOPS": 202356519896297.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1860455870628357, "optim/lr": 0.0011118, "optim/total_tokens": 972029952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382419, "loss/out": 3.587677001953125, "created_at": "2025-01-15T10:46:50.846848+00:00"} {"global_step": 1855, "acc_step": 0, "speed/wps": 12880.898673879743, "speed/FLOPS": 202312213882908.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16622720658779144, "optim/lr": 0.0011124000000000002, "optim/total_tokens": 972554240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.634242057800293, "created_at": "2025-01-15T10:47:01.026106+00:00"} {"global_step": 1856, "acc_step": 0, "speed/wps": 12885.503228261985, "speed/FLOPS": 202384534736802.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21277287602424622, "optim/lr": 0.001113, "optim/total_tokens": 973078528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.4565088748931885, "created_at": "2025-01-15T10:47:11.204246+00:00"} {"global_step": 1857, "acc_step": 0, "speed/wps": 12881.11469719047, "speed/FLOPS": 202315606825850.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19349539279937744, "optim/lr": 0.0011136, "optim/total_tokens": 973602816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.7400662899017334, "created_at": "2025-01-15T10:47:21.383823+00:00"} {"global_step": 1858, "acc_step": 0, "speed/wps": 12885.550809262068, "speed/FLOPS": 202385282061792.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20459207892417908, "optim/lr": 0.0011142, "optim/total_tokens": 974127104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371444, "loss/out": 3.6527087688446045, "created_at": "2025-01-15T10:47:31.568321+00:00"} {"global_step": 1859, "acc_step": 0, "speed/wps": 12879.495601805838, "speed/FLOPS": 202290176707964.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.20197969675064087, "optim/lr": 0.0011148, "optim/total_tokens": 974651392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 3.601778984069824, "created_at": "2025-01-15T10:47:41.752105+00:00"} {"global_step": 1860, "acc_step": 0, "speed/wps": 12882.459935316807, "speed/FLOPS": 202336735639175.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19749398529529572, "optim/lr": 0.0011154000000000001, "optim/total_tokens": 975175680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.5706186294555664, "created_at": "2025-01-15T10:47:51.935991+00:00"} {"global_step": 1861, "acc_step": 0, "speed/wps": 12878.12993744213, "speed/FLOPS": 202268727072513.4, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22133496403694153, "optim/lr": 0.001116, "optim/total_tokens": 975699968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.695589065551758, "created_at": "2025-01-15T10:48:02.117458+00:00"} {"global_step": 1862, "acc_step": 0, "speed/wps": 12879.085391105205, "speed/FLOPS": 202283733785222.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2816696763038635, "optim/lr": 0.0011166, "optim/total_tokens": 976224256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.579087495803833, "created_at": "2025-01-15T10:48:12.301554+00:00"} {"global_step": 1863, "acc_step": 0, "speed/wps": 12881.795151481921, "speed/FLOPS": 202326294295540.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23931792378425598, "optim/lr": 0.0011172, "optim/total_tokens": 976748544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430246, "loss/out": 3.6058905124664307, "created_at": "2025-01-15T10:48:22.480277+00:00"} {"global_step": 1864, "acc_step": 0, "speed/wps": 12885.002357830703, "speed/FLOPS": 202376667878411.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1938374638557434, "optim/lr": 0.0011178, "optim/total_tokens": 977272832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.693208932876587, "created_at": "2025-01-15T10:48:32.658689+00:00"} {"global_step": 1865, "acc_step": 0, "speed/wps": 12881.547371573364, "speed/FLOPS": 202322402571591.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15152566134929657, "optim/lr": 0.0011184, "optim/total_tokens": 977797120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.5931506156921387, "created_at": "2025-01-15T10:48:42.839945+00:00"} {"global_step": 1866, "acc_step": 0, "speed/wps": 12882.844467701603, "speed/FLOPS": 202342775248681.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16503456234931946, "optim/lr": 0.001119, "optim/total_tokens": 978321408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 3.758298397064209, "created_at": "2025-01-15T10:48:53.017540+00:00"} {"global_step": 1867, "acc_step": 0, "speed/wps": 12881.419725917542, "speed/FLOPS": 202320397721162.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1535956710577011, "optim/lr": 0.0011196, "optim/total_tokens": 978845696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.618096351623535, "created_at": "2025-01-15T10:49:03.196774+00:00"} {"global_step": 1868, "acc_step": 0, "speed/wps": 12885.42639225413, "speed/FLOPS": 202383327921715.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15444619953632355, "optim/lr": 0.0011202, "optim/total_tokens": 979369984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 3.6018784046173096, "created_at": "2025-01-15T10:49:13.376705+00:00"} {"global_step": 1869, "acc_step": 0, "speed/wps": 12882.06780281241, "speed/FLOPS": 202330576659347.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15695329010486603, "optim/lr": 0.0011208, "optim/total_tokens": 979894272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.5149619579315186, "created_at": "2025-01-15T10:49:23.555132+00:00"} {"global_step": 1870, "acc_step": 0, "speed/wps": 12885.787890126852, "speed/FLOPS": 202389005742556.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16546492278575897, "optim/lr": 0.0011214, "optim/total_tokens": 980418560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.6070940494537354, "created_at": "2025-01-15T10:49:33.735389+00:00"} {"global_step": 1871, "acc_step": 0, "speed/wps": 12883.246543367384, "speed/FLOPS": 202349090399522.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18799830973148346, "optim/lr": 0.001122, "optim/total_tokens": 980942848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.7236123085021973, "created_at": "2025-01-15T10:49:43.915216+00:00"} {"global_step": 1872, "acc_step": 0, "speed/wps": 12883.26227466844, "speed/FLOPS": 202349337481223.0, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1414787322282791, "optim/lr": 0.0011225999999999999, "optim/total_tokens": 981467136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.6017277240753174, "created_at": "2025-01-15T10:49:54.094875+00:00"} {"global_step": 1873, "acc_step": 0, "speed/wps": 12874.779887164783, "speed/FLOPS": 202216109929456.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.15157467126846313, "optim/lr": 0.0011232, "optim/total_tokens": 981991424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.6054763793945312, "created_at": "2025-01-15T10:50:04.280284+00:00"} {"global_step": 1874, "acc_step": 0, "speed/wps": 12874.25478264355, "speed/FLOPS": 202207862441380.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.20620085299015045, "optim/lr": 0.0011237999999999999, "optim/total_tokens": 982515712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 3.752077579498291, "created_at": "2025-01-15T10:50:14.474030+00:00"} {"global_step": 1875, "acc_step": 0, "speed/wps": 12883.596727516342, "speed/FLOPS": 202354590522785.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20607535541057587, "optim/lr": 0.0011244, "optim/total_tokens": 983040000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.594825506210327, "created_at": "2025-01-15T10:50:24.651389+00:00"} {"global_step": 1876, "acc_step": 0, "speed/wps": 12884.645589404412, "speed/FLOPS": 202371064340025.1, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1606827974319458, "optim/lr": 0.0011250000000000001, "optim/total_tokens": 983564288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 3.5645530223846436, "created_at": "2025-01-15T10:50:34.830665+00:00"} {"global_step": 1877, "acc_step": 0, "speed/wps": 12885.452656949263, "speed/FLOPS": 202383740444844.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14660273492336273, "optim/lr": 0.0011256, "optim/total_tokens": 984088576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.6676688194274902, "created_at": "2025-01-15T10:50:45.006330+00:00"} {"global_step": 1878, "acc_step": 0, "speed/wps": 12878.293186776335, "speed/FLOPS": 202271291127635.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16903124749660492, "optim/lr": 0.0011262000000000002, "optim/total_tokens": 984612864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 3.5775325298309326, "created_at": "2025-01-15T10:50:55.187974+00:00"} {"global_step": 1879, "acc_step": 0, "speed/wps": 12878.420960208152, "speed/FLOPS": 202273297984960.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.28193795680999756, "optim/lr": 0.0011268, "optim/total_tokens": 985137152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294596, "loss/out": 3.7789719104766846, "created_at": "2025-01-15T10:51:05.369890+00:00"} {"global_step": 1880, "acc_step": 0, "speed/wps": 12874.070975855453, "speed/FLOPS": 202204975503196.94, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3763306438922882, "optim/lr": 0.0011274000000000002, "optim/total_tokens": 985661440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.6572747230529785, "created_at": "2025-01-15T10:51:15.554822+00:00"} {"global_step": 1881, "acc_step": 0, "speed/wps": 12876.827787668608, "speed/FLOPS": 202248275020978.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3037668764591217, "optim/lr": 0.001128, "optim/total_tokens": 986185728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 3.5490918159484863, "created_at": "2025-01-15T10:51:25.738304+00:00"} {"global_step": 1882, "acc_step": 0, "speed/wps": 12876.571036545136, "speed/FLOPS": 202244242391771.8, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30159229040145874, "optim/lr": 0.0011286, "optim/total_tokens": 986710016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.6528360843658447, "created_at": "2025-01-15T10:51:35.920936+00:00"} {"global_step": 1883, "acc_step": 0, "speed/wps": 12881.168324523744, "speed/FLOPS": 202316449116811.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2852349281311035, "optim/lr": 0.0011292000000000001, "optim/total_tokens": 987234304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 3.635763645172119, "created_at": "2025-01-15T10:51:46.105729+00:00"} {"global_step": 1884, "acc_step": 0, "speed/wps": 12876.919526533226, "speed/FLOPS": 202249715905910.2, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.30997762084007263, "optim/lr": 0.0011298, "optim/total_tokens": 987758592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283675, "loss/out": 3.761139392852783, "created_at": "2025-01-15T10:51:56.289298+00:00"} {"global_step": 1885, "acc_step": 0, "speed/wps": 12873.55556168443, "speed/FLOPS": 202196880215387.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2881687879562378, "optim/lr": 0.0011304000000000002, "optim/total_tokens": 988282880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.6660618782043457, "created_at": "2025-01-15T10:52:06.474171+00:00"} {"global_step": 1886, "acc_step": 0, "speed/wps": 12879.454625880851, "speed/FLOPS": 202289533124756.2, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27291667461395264, "optim/lr": 0.001131, "optim/total_tokens": 988807168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.6646180152893066, "created_at": "2025-01-15T10:52:16.654631+00:00"} {"global_step": 1887, "acc_step": 0, "speed/wps": 12878.209756247663, "speed/FLOPS": 202269980736536.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.31226298213005066, "optim/lr": 0.0011316, "optim/total_tokens": 989331456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 3.728926181793213, "created_at": "2025-01-15T10:52:26.836545+00:00"} {"global_step": 1888, "acc_step": 0, "speed/wps": 12877.134526835445, "speed/FLOPS": 202253092781097.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3138037323951721, "optim/lr": 0.0011322, "optim/total_tokens": 989855744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.642058849334717, "created_at": "2025-01-15T10:52:37.021222+00:00"} {"global_step": 1889, "acc_step": 0, "speed/wps": 12881.52261662623, "speed/FLOPS": 202322013761129.44, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25959768891334534, "optim/lr": 0.0011328, "optim/total_tokens": 990380032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.634307622909546, "created_at": "2025-01-15T10:52:47.199935+00:00"} {"global_step": 1890, "acc_step": 0, "speed/wps": 12873.53930886531, "speed/FLOPS": 202196624942529.75, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.17806954681873322, "optim/lr": 0.0011334000000000001, "optim/total_tokens": 990904320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 3.584408760070801, "created_at": "2025-01-15T10:52:57.386650+00:00"} {"global_step": 1891, "acc_step": 0, "speed/wps": 12884.183166993296, "speed/FLOPS": 202363801360623.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18411381542682648, "optim/lr": 0.001134, "optim/total_tokens": 991428608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.5486884117126465, "created_at": "2025-01-15T10:53:07.565549+00:00"} {"global_step": 1892, "acc_step": 0, "speed/wps": 12880.412683286324, "speed/FLOPS": 202304580732817.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15530318021774292, "optim/lr": 0.0011346, "optim/total_tokens": 991952896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.6746106147766113, "created_at": "2025-01-15T10:53:17.751287+00:00"} {"global_step": 1893, "acc_step": 0, "speed/wps": 12882.380096834964, "speed/FLOPS": 202335481666108.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1617424339056015, "optim/lr": 0.0011352, "optim/total_tokens": 992477184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.666055202484131, "created_at": "2025-01-15T10:53:27.930034+00:00"} {"global_step": 1894, "acc_step": 0, "speed/wps": 12870.133263578588, "speed/FLOPS": 202143128320905.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18967418372631073, "optim/lr": 0.0011358, "optim/total_tokens": 993001472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.6601786613464355, "created_at": "2025-01-15T10:53:38.122038+00:00"} {"global_step": 1895, "acc_step": 0, "speed/wps": 12872.1976307514, "speed/FLOPS": 202175552044091.84, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18369793891906738, "optim/lr": 0.0011364, "optim/total_tokens": 993525760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 3.571998119354248, "created_at": "2025-01-15T10:53:48.314956+00:00"} {"global_step": 1896, "acc_step": 0, "speed/wps": 12873.130656156622, "speed/FLOPS": 202190206490192.12, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1637224555015564, "optim/lr": 0.001137, "optim/total_tokens": 994050048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 3.7508440017700195, "created_at": "2025-01-15T10:53:58.505673+00:00"} {"global_step": 1897, "acc_step": 0, "speed/wps": 12883.62079812021, "speed/FLOPS": 202354968584695.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18130095303058624, "optim/lr": 0.0011376, "optim/total_tokens": 994574336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.691265821456909, "created_at": "2025-01-15T10:54:08.683537+00:00"} {"global_step": 1898, "acc_step": 0, "speed/wps": 12872.65720339046, "speed/FLOPS": 202182770263907.25, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19887588918209076, "optim/lr": 0.0011382, "optim/total_tokens": 995098624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.5887579917907715, "created_at": "2025-01-15T10:54:18.875128+00:00"} {"global_step": 1899, "acc_step": 0, "speed/wps": 12876.164208159762, "speed/FLOPS": 202237852592938.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21883109211921692, "optim/lr": 0.0011388, "optim/total_tokens": 995622912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.734408378601074, "created_at": "2025-01-15T10:54:29.059648+00:00"} {"global_step": 1900, "acc_step": 0, "speed/wps": 12874.267727926857, "speed/FLOPS": 202208065764844.03, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.232122540473938, "optim/lr": 0.0011394, "optim/total_tokens": 996147200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.5518860816955566, "created_at": "2025-01-15T10:54:39.246081+00:00"} {"global_step": 1901, "acc_step": 0, "speed/wps": 12869.633281316588, "speed/FLOPS": 202135275412432.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21859513223171234, "optim/lr": 0.00114, "optim/total_tokens": 996671488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.7887816429138184, "created_at": "2025-01-15T10:54:49.437346+00:00"} {"global_step": 1902, "acc_step": 0, "speed/wps": 12880.046265033734, "speed/FLOPS": 202298825630649.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18852630257606506, "optim/lr": 0.0011406, "optim/total_tokens": 997195776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 3.7127480506896973, "created_at": "2025-01-15T10:54:59.618712+00:00"} {"global_step": 1903, "acc_step": 0, "speed/wps": 12878.443577282997, "speed/FLOPS": 202273653217200.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16741931438446045, "optim/lr": 0.0011412, "optim/total_tokens": 997720064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.771071434020996, "created_at": "2025-01-15T10:55:09.800966+00:00"} {"global_step": 1904, "acc_step": 0, "speed/wps": 12883.98462913386, "speed/FLOPS": 202360683050721.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21826054155826569, "optim/lr": 0.0011418, "optim/total_tokens": 998244352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.6510605812072754, "created_at": "2025-01-15T10:55:19.978224+00:00"} {"global_step": 1905, "acc_step": 0, "speed/wps": 12882.04064870513, "speed/FLOPS": 202330150166779.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23004451394081116, "optim/lr": 0.0011424, "optim/total_tokens": 998768640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.5676779747009277, "created_at": "2025-01-15T10:55:30.158874+00:00"} {"global_step": 1906, "acc_step": 0, "speed/wps": 12878.522102469738, "speed/FLOPS": 202274886563163.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17558299005031586, "optim/lr": 0.0011430000000000001, "optim/total_tokens": 999292928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.6118743419647217, "created_at": "2025-01-15T10:55:40.341263+00:00"} {"global_step": 1907, "acc_step": 0, "speed/wps": 12878.1268488346, "speed/FLOPS": 202268678561687.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21921215951442719, "optim/lr": 0.0011436, "optim/total_tokens": 999817216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.629779100418091, "created_at": "2025-01-15T10:55:50.523147+00:00"} {"global_step": 1908, "acc_step": 0, "speed/wps": 12879.862183217228, "speed/FLOPS": 202295934372766.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15067310631275177, "optim/lr": 0.0011442000000000002, "optim/total_tokens": 1000341504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.56185245513916, "created_at": "2025-01-15T10:56:00.705100+00:00"} {"global_step": 1909, "acc_step": 0, "speed/wps": 12880.356263735319, "speed/FLOPS": 202303694586240.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19414794445037842, "optim/lr": 0.0011448, "optim/total_tokens": 1000865792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.5549044609069824, "created_at": "2025-01-15T10:56:10.884819+00:00"} {"global_step": 1910, "acc_step": 0, "speed/wps": 12880.770650866512, "speed/FLOPS": 202310203105563.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20123739540576935, "optim/lr": 0.0011454, "optim/total_tokens": 1001390080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499005, "loss/out": 3.658949375152588, "created_at": "2025-01-15T10:56:21.067736+00:00"} {"global_step": 1911, "acc_step": 0, "speed/wps": 12879.201915145477, "speed/FLOPS": 202285563955396.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16915465891361237, "optim/lr": 0.001146, "optim/total_tokens": 1001914368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.5695042610168457, "created_at": "2025-01-15T10:56:31.249287+00:00"} {"global_step": 1912, "acc_step": 0, "speed/wps": 12882.264358381919, "speed/FLOPS": 202333663834662.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1816377490758896, "optim/lr": 0.0011466, "optim/total_tokens": 1002438656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490582, "loss/out": 3.63873028755188, "created_at": "2025-01-15T10:56:41.428055+00:00"} {"global_step": 1913, "acc_step": 0, "speed/wps": 12872.423797282223, "speed/FLOPS": 202179104300243.8, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1879555881023407, "optim/lr": 0.0011472000000000001, "optim/total_tokens": 1002962944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 3.567188262939453, "created_at": "2025-01-15T10:56:51.614305+00:00"} {"global_step": 1914, "acc_step": 0, "speed/wps": 12879.267819512479, "speed/FLOPS": 202286599074042.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17411530017852783, "optim/lr": 0.0011478, "optim/total_tokens": 1003487232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.6049773693084717, "created_at": "2025-01-15T10:57:01.795025+00:00"} {"global_step": 1915, "acc_step": 0, "speed/wps": 12883.145812767596, "speed/FLOPS": 202347508287034.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15335460007190704, "optim/lr": 0.0011484, "optim/total_tokens": 1004011520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 3.541261672973633, "created_at": "2025-01-15T10:57:11.975521+00:00"} {"global_step": 1916, "acc_step": 0, "speed/wps": 12875.419280403077, "speed/FLOPS": 202226152478881.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1472369283437729, "optim/lr": 0.001149, "optim/total_tokens": 1004535808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.6602187156677246, "created_at": "2025-01-15T10:57:22.159439+00:00"} {"global_step": 1917, "acc_step": 0, "speed/wps": 12882.549473843239, "speed/FLOPS": 202338141964772.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16645178198814392, "optim/lr": 0.0011496, "optim/total_tokens": 1005060096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 3.624356269836426, "created_at": "2025-01-15T10:57:32.337735+00:00"} {"global_step": 1918, "acc_step": 0, "speed/wps": 12878.964820814615, "speed/FLOPS": 202281840063128.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1633213460445404, "optim/lr": 0.0011502, "optim/total_tokens": 1005584384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.5740818977355957, "created_at": "2025-01-15T10:57:42.521533+00:00"} {"global_step": 1919, "acc_step": 0, "speed/wps": 12877.859765893798, "speed/FLOPS": 202264483657092.25, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14062443375587463, "optim/lr": 0.0011508, "optim/total_tokens": 1006108672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 3.5699355602264404, "created_at": "2025-01-15T10:57:52.708451+00:00"} {"global_step": 1920, "acc_step": 0, "speed/wps": 12876.375407469257, "speed/FLOPS": 202241169768312.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.14456163346767426, "optim/lr": 0.0011514, "optim/total_tokens": 1006632960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471672, "loss/out": 3.47654128074646, "created_at": "2025-01-15T10:58:02.894747+00:00"} {"global_step": 1921, "acc_step": 0, "speed/wps": 12874.815229366395, "speed/FLOPS": 202216665027297.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16418440639972687, "optim/lr": 0.001152, "optim/total_tokens": 1007157248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.5733134746551514, "created_at": "2025-01-15T10:58:13.094552+00:00"} {"global_step": 1922, "acc_step": 0, "speed/wps": 12877.021670251437, "speed/FLOPS": 202251320213364.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16752035915851593, "optim/lr": 0.0011526, "optim/total_tokens": 1007681536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.601350784301758, "created_at": "2025-01-15T10:58:23.276817+00:00"} {"global_step": 1923, "acc_step": 0, "speed/wps": 12880.630972580664, "speed/FLOPS": 202308009266146.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1617867648601532, "optim/lr": 0.0011532, "optim/total_tokens": 1008205824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.4978270530700684, "created_at": "2025-01-15T10:58:33.457868+00:00"} {"global_step": 1924, "acc_step": 0, "speed/wps": 12880.968863330727, "speed/FLOPS": 202313316304686.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1581268310546875, "optim/lr": 0.0011538, "optim/total_tokens": 1008730112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 3.635139226913452, "created_at": "2025-01-15T10:58:43.638937+00:00"} {"global_step": 1925, "acc_step": 0, "speed/wps": 12879.778291520695, "speed/FLOPS": 202294616738393.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19204498827457428, "optim/lr": 0.0011543999999999999, "optim/total_tokens": 1009254400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.622807264328003, "created_at": "2025-01-15T10:58:53.819109+00:00"} {"global_step": 1926, "acc_step": 0, "speed/wps": 12877.010049620778, "speed/FLOPS": 202251137695391.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18542195856571198, "optim/lr": 0.001155, "optim/total_tokens": 1009778688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506365, "loss/out": 3.583263397216797, "created_at": "2025-01-15T10:59:04.001519+00:00"} {"global_step": 1927, "acc_step": 0, "speed/wps": 12874.961840270622, "speed/FLOPS": 202218967753013.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15601466596126556, "optim/lr": 0.0011556, "optim/total_tokens": 1010302976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.563711166381836, "created_at": "2025-01-15T10:59:14.186172+00:00"} {"global_step": 1928, "acc_step": 0, "speed/wps": 12876.035908162368, "speed/FLOPS": 202235837465176.97, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18606886267662048, "optim/lr": 0.0011562, "optim/total_tokens": 1010827264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431584, "loss/out": 3.607231616973877, "created_at": "2025-01-15T10:59:24.373367+00:00"} {"global_step": 1929, "acc_step": 0, "speed/wps": 12883.915057878734, "speed/FLOPS": 202359590338558.8, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18685093522071838, "optim/lr": 0.0011568, "optim/total_tokens": 1011351552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328312, "loss/out": 3.561880588531494, "created_at": "2025-01-15T10:59:34.553053+00:00"} {"global_step": 1930, "acc_step": 0, "speed/wps": 12873.72269184421, "speed/FLOPS": 202199505224207.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19006794691085815, "optim/lr": 0.0011574, "optim/total_tokens": 1011875840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.6372995376586914, "created_at": "2025-01-15T10:59:44.740318+00:00"} {"global_step": 1931, "acc_step": 0, "speed/wps": 12878.918696810988, "speed/FLOPS": 202281115622270.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20676381886005402, "optim/lr": 0.001158, "optim/total_tokens": 1012400128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.610121250152588, "created_at": "2025-01-15T10:59:54.921223+00:00"} {"global_step": 1932, "acc_step": 0, "speed/wps": 12876.511902107397, "speed/FLOPS": 202243313604167.88, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31115296483039856, "optim/lr": 0.0011586, "optim/total_tokens": 1012924416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.7092275619506836, "created_at": "2025-01-15T11:00:05.106580+00:00"} {"global_step": 1933, "acc_step": 0, "speed/wps": 12873.14960676664, "speed/FLOPS": 202190504135563.3, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27760186791419983, "optim/lr": 0.0011592, "optim/total_tokens": 1013448704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465818, "loss/out": 3.6415486335754395, "created_at": "2025-01-15T11:00:15.296562+00:00"} {"global_step": 1934, "acc_step": 0, "speed/wps": 12872.081161602631, "speed/FLOPS": 202173722736065.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.2130511850118637, "optim/lr": 0.0011598000000000001, "optim/total_tokens": 1013972992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.6664366722106934, "created_at": "2025-01-15T11:00:25.484085+00:00"} {"global_step": 1935, "acc_step": 0, "speed/wps": 12875.489873075552, "speed/FLOPS": 202227261233807.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19405533373355865, "optim/lr": 0.0011604, "optim/total_tokens": 1014497280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 3.6836071014404297, "created_at": "2025-01-15T11:00:35.667588+00:00"} {"global_step": 1936, "acc_step": 0, "speed/wps": 12881.929810809606, "speed/FLOPS": 202328409305322.8, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20829543471336365, "optim/lr": 0.0011610000000000001, "optim/total_tokens": 1015021568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.6584675312042236, "created_at": "2025-01-15T11:00:45.846603+00:00"} {"global_step": 1937, "acc_step": 0, "speed/wps": 12879.892830150107, "speed/FLOPS": 202296415724960.62, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2183500975370407, "optim/lr": 0.0011616, "optim/total_tokens": 1015545856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.4890213012695312, "created_at": "2025-01-15T11:00:56.028904+00:00"} {"global_step": 1938, "acc_step": 0, "speed/wps": 12880.398988877554, "speed/FLOPS": 202304365643310.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20521798729896545, "optim/lr": 0.0011622000000000002, "optim/total_tokens": 1016070144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.6279666423797607, "created_at": "2025-01-15T11:01:06.208564+00:00"} {"global_step": 1939, "acc_step": 0, "speed/wps": 12878.050154977833, "speed/FLOPS": 202267473979279.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.206497922539711, "optim/lr": 0.0011628, "optim/total_tokens": 1016594432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.6074230670928955, "created_at": "2025-01-15T11:01:16.391355+00:00"} {"global_step": 1940, "acc_step": 0, "speed/wps": 12885.40876926174, "speed/FLOPS": 202383051128403.44, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21321359276771545, "optim/lr": 0.0011634, "optim/total_tokens": 1017118720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 3.661433219909668, "created_at": "2025-01-15T11:01:26.568696+00:00"} {"global_step": 1941, "acc_step": 0, "speed/wps": 12883.821444238303, "speed/FLOPS": 202358120007697.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19616913795471191, "optim/lr": 0.001164, "optim/total_tokens": 1017643008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.636647939682007, "created_at": "2025-01-15T11:01:36.745828+00:00"} {"global_step": 1942, "acc_step": 0, "speed/wps": 12877.5723858897, "speed/FLOPS": 202259969959226.38, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16892504692077637, "optim/lr": 0.0011646, "optim/total_tokens": 1018167296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.7325074672698975, "created_at": "2025-01-15T11:01:46.935960+00:00"} {"global_step": 1943, "acc_step": 0, "speed/wps": 12871.801728064862, "speed/FLOPS": 202169333848372.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18055002391338348, "optim/lr": 0.0011652000000000001, "optim/total_tokens": 1018691584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.642637252807617, "created_at": "2025-01-15T11:01:57.126280+00:00"} {"global_step": 1944, "acc_step": 0, "speed/wps": 12878.07046344764, "speed/FLOPS": 202267792951704.12, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.18337365984916687, "optim/lr": 0.0011658, "optim/total_tokens": 1019215872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.5768375396728516, "created_at": "2025-01-15T11:02:07.308293+00:00"} {"global_step": 1945, "acc_step": 0, "speed/wps": 12875.30234657517, "speed/FLOPS": 202224315872430.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18028143048286438, "optim/lr": 0.0011664, "optim/total_tokens": 1019740160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.455972671508789, "created_at": "2025-01-15T11:02:17.494682+00:00"} {"global_step": 1946, "acc_step": 0, "speed/wps": 12881.52200824713, "speed/FLOPS": 202322004205699.66, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1626998484134674, "optim/lr": 0.001167, "optim/total_tokens": 1020264448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.7647740840911865, "created_at": "2025-01-15T11:02:27.673443+00:00"} {"global_step": 1947, "acc_step": 0, "speed/wps": 12881.911913216994, "speed/FLOPS": 202328128199036.97, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.173804372549057, "optim/lr": 0.0011676, "optim/total_tokens": 1020788736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 3.540785789489746, "created_at": "2025-01-15T11:02:37.854922+00:00"} {"global_step": 1948, "acc_step": 0, "speed/wps": 12880.806022429288, "speed/FLOPS": 202310758664562.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1770104169845581, "optim/lr": 0.0011682, "optim/total_tokens": 1021313024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.646111249923706, "created_at": "2025-01-15T11:02:48.036192+00:00"} {"global_step": 1949, "acc_step": 0, "speed/wps": 12881.72971881363, "speed/FLOPS": 202325266585570.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18230648338794708, "optim/lr": 0.0011688, "optim/total_tokens": 1021837312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.5130233764648438, "created_at": "2025-01-15T11:02:58.216627+00:00"} {"global_step": 1950, "acc_step": 0, "speed/wps": 12880.165896046014, "speed/FLOPS": 202300704600087.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1718803495168686, "optim/lr": 0.0011694, "optim/total_tokens": 1022361600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.558393955230713, "created_at": "2025-01-15T11:03:08.399500+00:00"} {"global_step": 1951, "acc_step": 0, "speed/wps": 12878.21125402341, "speed/FLOPS": 202270004261163.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16193704307079315, "optim/lr": 0.00117, "optim/total_tokens": 1022885888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 3.6114501953125, "created_at": "2025-01-15T11:03:18.589069+00:00"} {"global_step": 1952, "acc_step": 0, "speed/wps": 12881.068548355124, "speed/FLOPS": 202314881994976.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15896275639533997, "optim/lr": 0.0011706, "optim/total_tokens": 1023410176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366026, "loss/out": 3.5386557579040527, "created_at": "2025-01-15T11:03:28.768847+00:00"} {"global_step": 1953, "acc_step": 0, "speed/wps": 12877.487419960613, "speed/FLOPS": 202258635452554.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15010415017604828, "optim/lr": 0.0011712, "optim/total_tokens": 1023934464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302403, "loss/out": 3.3936767578125, "created_at": "2025-01-15T11:03:38.953233+00:00"} {"global_step": 1954, "acc_step": 0, "speed/wps": 12878.151647295663, "speed/FLOPS": 202269068055595.6, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17814959585666656, "optim/lr": 0.0011718, "optim/total_tokens": 1024458752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.676905632019043, "created_at": "2025-01-15T11:03:49.136490+00:00"} {"global_step": 1955, "acc_step": 0, "speed/wps": 12881.012819809403, "speed/FLOPS": 202314006701586.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18875567615032196, "optim/lr": 0.0011723999999999999, "optim/total_tokens": 1024983040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.655292510986328, "created_at": "2025-01-15T11:03:59.317844+00:00"} {"global_step": 1956, "acc_step": 0, "speed/wps": 12882.04014050264, "speed/FLOPS": 202330142184760.8, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1700502336025238, "optim/lr": 0.001173, "optim/total_tokens": 1025507328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.639334201812744, "created_at": "2025-01-15T11:04:09.498853+00:00"} {"global_step": 1957, "acc_step": 0, "speed/wps": 12877.521188684084, "speed/FLOPS": 202259165836759.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17485427856445312, "optim/lr": 0.0011736, "optim/total_tokens": 1026031616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 3.632265329360962, "created_at": "2025-01-15T11:04:19.684080+00:00"} {"global_step": 1958, "acc_step": 0, "speed/wps": 12878.706487751519, "speed/FLOPS": 202277782587385.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23415935039520264, "optim/lr": 0.0011742, "optim/total_tokens": 1026555904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 3.5844850540161133, "created_at": "2025-01-15T11:04:29.868686+00:00"} {"global_step": 1959, "acc_step": 0, "speed/wps": 12875.789996575018, "speed/FLOPS": 202231975085779.72, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19414204359054565, "optim/lr": 0.0011748, "optim/total_tokens": 1027080192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 3.5259785652160645, "created_at": "2025-01-15T11:04:40.055474+00:00"} {"global_step": 1960, "acc_step": 0, "speed/wps": 12875.214140817441, "speed/FLOPS": 202222930479798.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2020183801651001, "optim/lr": 0.0011754, "optim/total_tokens": 1027604480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.5254273414611816, "created_at": "2025-01-15T11:04:50.242151+00:00"} {"global_step": 1961, "acc_step": 0, "speed/wps": 12885.425752832893, "speed/FLOPS": 202383317878726.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15350091457366943, "optim/lr": 0.001176, "optim/total_tokens": 1028128768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.5530948638916016, "created_at": "2025-01-15T11:05:00.419393+00:00"} {"global_step": 1962, "acc_step": 0, "speed/wps": 12874.741247349217, "speed/FLOPS": 202215503038056.06, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.1620948165655136, "optim/lr": 0.0011766, "optim/total_tokens": 1028653056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.629680871963501, "created_at": "2025-01-15T11:05:10.603848+00:00"} {"global_step": 1963, "acc_step": 0, "speed/wps": 12879.290123691584, "speed/FLOPS": 202286949391824.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19402632117271423, "optim/lr": 0.0011772000000000002, "optim/total_tokens": 1029177344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 3.5619008541107178, "created_at": "2025-01-15T11:05:20.784990+00:00"} {"global_step": 1964, "acc_step": 0, "speed/wps": 12879.16168360536, "speed/FLOPS": 202284932063774.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1749548614025116, "optim/lr": 0.0011778000000000001, "optim/total_tokens": 1029701632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 3.5752601623535156, "created_at": "2025-01-15T11:05:30.971673+00:00"} {"global_step": 1965, "acc_step": 0, "speed/wps": 12879.863729705117, "speed/FLOPS": 202295958662483.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19576029479503632, "optim/lr": 0.0011784, "optim/total_tokens": 1030225920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.5176777839660645, "created_at": "2025-01-15T11:05:41.153774+00:00"} {"global_step": 1966, "acc_step": 0, "speed/wps": 12877.163426432106, "speed/FLOPS": 202253546688975.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.223979651927948, "optim/lr": 0.0011790000000000001, "optim/total_tokens": 1030750208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.704174518585205, "created_at": "2025-01-15T11:05:51.339916+00:00"} {"global_step": 1967, "acc_step": 0, "speed/wps": 12871.91774714819, "speed/FLOPS": 202171156087503.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2062726765871048, "optim/lr": 0.0011796, "optim/total_tokens": 1031274496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.59452486038208, "created_at": "2025-01-15T11:06:01.530007+00:00"} {"global_step": 1968, "acc_step": 0, "speed/wps": 12876.456416793948, "speed/FLOPS": 202242442131075.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24269849061965942, "optim/lr": 0.0011802000000000002, "optim/total_tokens": 1031798784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.5498483180999756, "created_at": "2025-01-15T11:06:11.713234+00:00"} {"global_step": 1969, "acc_step": 0, "speed/wps": 12871.12245505257, "speed/FLOPS": 202158664932295.47, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22098547220230103, "optim/lr": 0.0011808, "optim/total_tokens": 1032323072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487147, "loss/out": 3.733992099761963, "created_at": "2025-01-15T11:06:21.900256+00:00"} {"global_step": 1970, "acc_step": 0, "speed/wps": 12875.81279406528, "speed/FLOPS": 202232333151691.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20918919146060944, "optim/lr": 0.0011814, "optim/total_tokens": 1032847360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345442, "loss/out": 3.532041311264038, "created_at": "2025-01-15T11:06:32.085236+00:00"} {"global_step": 1971, "acc_step": 0, "speed/wps": 12880.238798167937, "speed/FLOPS": 202301849628090.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.25171926617622375, "optim/lr": 0.001182, "optim/total_tokens": 1033371648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 3.603902816772461, "created_at": "2025-01-15T11:06:42.272654+00:00"} {"global_step": 1972, "acc_step": 0, "speed/wps": 12876.813999559263, "speed/FLOPS": 202248058459774.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22060735523700714, "optim/lr": 0.0011826, "optim/total_tokens": 1033895936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.606635570526123, "created_at": "2025-01-15T11:06:52.455421+00:00"} {"global_step": 1973, "acc_step": 0, "speed/wps": 12872.047067037942, "speed/FLOPS": 202173187234075.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13911642134189606, "optim/lr": 0.0011832, "optim/total_tokens": 1034420224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.550748586654663, "created_at": "2025-01-15T11:07:02.642938+00:00"} {"global_step": 1974, "acc_step": 0, "speed/wps": 12870.703192816105, "speed/FLOPS": 202152079842745.12, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1800829917192459, "optim/lr": 0.0011838, "optim/total_tokens": 1034944512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.774271249771118, "created_at": "2025-01-15T11:07:12.837560+00:00"} {"global_step": 1975, "acc_step": 0, "speed/wps": 12881.234633363383, "speed/FLOPS": 202317490588255.97, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1766590178012848, "optim/lr": 0.0011844, "optim/total_tokens": 1035468800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506128, "loss/out": 3.6234230995178223, "created_at": "2025-01-15T11:07:23.020380+00:00"} {"global_step": 1976, "acc_step": 0, "speed/wps": 12873.233523997675, "speed/FLOPS": 202191822170991.1, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17086507380008698, "optim/lr": 0.001185, "optim/total_tokens": 1035993088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.529437303543091, "created_at": "2025-01-15T11:07:33.209192+00:00"} {"global_step": 1977, "acc_step": 0, "speed/wps": 12877.523568844344, "speed/FLOPS": 202259203220447.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20260083675384521, "optim/lr": 0.0011856, "optim/total_tokens": 1036517376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428074, "loss/out": 3.7769501209259033, "created_at": "2025-01-15T11:07:43.391386+00:00"} {"global_step": 1978, "acc_step": 0, "speed/wps": 12879.67597323084, "speed/FLOPS": 202293009689050.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1659594625234604, "optim/lr": 0.0011861999999999999, "optim/total_tokens": 1037041664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 3.6299784183502197, "created_at": "2025-01-15T11:07:53.574401+00:00"} {"global_step": 1979, "acc_step": 0, "speed/wps": 12876.756341476776, "speed/FLOPS": 202247152860358.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17898105084896088, "optim/lr": 0.0011868, "optim/total_tokens": 1037565952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.5966944694519043, "created_at": "2025-01-15T11:08:03.760235+00:00"} {"global_step": 1980, "acc_step": 0, "speed/wps": 12876.940537173403, "speed/FLOPS": 202250045906885.8, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18170739710330963, "optim/lr": 0.0011874, "optim/total_tokens": 1038090240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.536822557449341, "created_at": "2025-01-15T11:08:13.948057+00:00"} {"global_step": 1981, "acc_step": 0, "speed/wps": 12870.125385111345, "speed/FLOPS": 202143004578750.88, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17064502835273743, "optim/lr": 0.001188, "optim/total_tokens": 1038614528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.601330041885376, "created_at": "2025-01-15T11:08:24.137119+00:00"} {"global_step": 1982, "acc_step": 0, "speed/wps": 12871.665380355284, "speed/FLOPS": 202167192320232.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2094472348690033, "optim/lr": 0.0011886, "optim/total_tokens": 1039138816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.5147931575775146, "created_at": "2025-01-15T11:08:34.326456+00:00"} {"global_step": 1983, "acc_step": 0, "speed/wps": 12881.301327236048, "speed/FLOPS": 202318538107173.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23915506899356842, "optim/lr": 0.0011891999999999998, "optim/total_tokens": 1039663104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.5816121101379395, "created_at": "2025-01-15T11:08:44.505480+00:00"} {"global_step": 1984, "acc_step": 0, "speed/wps": 12884.125455873285, "speed/FLOPS": 202362894928180.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15662947297096252, "optim/lr": 0.0011898, "optim/total_tokens": 1040187392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329132, "loss/out": 3.5089797973632812, "created_at": "2025-01-15T11:08:54.682774+00:00"} {"global_step": 1985, "acc_step": 0, "speed/wps": 12882.13865942926, "speed/FLOPS": 202331689559882.56, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14403539896011353, "optim/lr": 0.0011903999999999999, "optim/total_tokens": 1040711680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.555492401123047, "created_at": "2025-01-15T11:09:04.865386+00:00"} {"global_step": 1986, "acc_step": 0, "speed/wps": 12878.078725197534, "speed/FLOPS": 202267922713839.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16023962199687958, "optim/lr": 0.001191, "optim/total_tokens": 1041235968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.607394218444824, "created_at": "2025-01-15T11:09:15.048063+00:00"} {"global_step": 1987, "acc_step": 0, "speed/wps": 12880.372290935125, "speed/FLOPS": 202303946315437.5, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1778828352689743, "optim/lr": 0.0011916000000000001, "optim/total_tokens": 1041760256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.539055824279785, "created_at": "2025-01-15T11:09:25.236867+00:00"} {"global_step": 1988, "acc_step": 0, "speed/wps": 12881.677722158673, "speed/FLOPS": 202324449906653.78, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16418302059173584, "optim/lr": 0.0011922, "optim/total_tokens": 1042284544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.5244059562683105, "created_at": "2025-01-15T11:09:35.418936+00:00"} {"global_step": 1989, "acc_step": 0, "speed/wps": 12877.449567481182, "speed/FLOPS": 202258040927350.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22795352339744568, "optim/lr": 0.0011928000000000002, "optim/total_tokens": 1042808832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369048, "loss/out": 3.559507369995117, "created_at": "2025-01-15T11:09:45.602120+00:00"} {"global_step": 1990, "acc_step": 0, "speed/wps": 12878.780399851015, "speed/FLOPS": 202278943478473.97, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30187317728996277, "optim/lr": 0.0011934, "optim/total_tokens": 1043333120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 3.5460493564605713, "created_at": "2025-01-15T11:09:55.789474+00:00"} {"global_step": 1991, "acc_step": 0, "speed/wps": 12877.17216773732, "speed/FLOPS": 202253683983185.88, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31943491101264954, "optim/lr": 0.0011940000000000002, "optim/total_tokens": 1043857408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.6519179344177246, "created_at": "2025-01-15T11:10:05.971957+00:00"} {"global_step": 1992, "acc_step": 0, "speed/wps": 12877.260742017641, "speed/FLOPS": 202255075163972.12, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2377350926399231, "optim/lr": 0.0011946, "optim/total_tokens": 1044381696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 3.6160733699798584, "created_at": "2025-01-15T11:10:16.158126+00:00"} {"global_step": 1993, "acc_step": 0, "speed/wps": 12876.907914334215, "speed/FLOPS": 202249533520367.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.28177863359451294, "optim/lr": 0.0011952, "optim/total_tokens": 1044905984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.5855770111083984, "created_at": "2025-01-15T11:10:26.343914+00:00"} {"global_step": 1994, "acc_step": 0, "speed/wps": 12882.163001684605, "speed/FLOPS": 202332071888452.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3011600077152252, "optim/lr": 0.0011958000000000001, "optim/total_tokens": 1045430272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415820, "loss/out": 3.6966028213500977, "created_at": "2025-01-15T11:10:36.523645+00:00"} {"global_step": 1995, "acc_step": 0, "speed/wps": 12875.17421216385, "speed/FLOPS": 202222303345425.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.28987374901771545, "optim/lr": 0.0011964, "optim/total_tokens": 1045954560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 3.5824804306030273, "created_at": "2025-01-15T11:10:46.707671+00:00"} {"global_step": 1996, "acc_step": 0, "speed/wps": 12874.94122005791, "speed/FLOPS": 202218643884237.38, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23489849269390106, "optim/lr": 0.0011970000000000001, "optim/total_tokens": 1046478848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.6576976776123047, "created_at": "2025-01-15T11:10:56.892085+00:00"} {"global_step": 1997, "acc_step": 0, "speed/wps": 12880.169734158007, "speed/FLOPS": 202300764882910.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22492113709449768, "optim/lr": 0.0011976, "optim/total_tokens": 1047003136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.5150749683380127, "created_at": "2025-01-15T11:11:07.073028+00:00"} {"global_step": 1998, "acc_step": 0, "speed/wps": 12878.166260803398, "speed/FLOPS": 202269297580815.66, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21783578395843506, "optim/lr": 0.0011982, "optim/total_tokens": 1047527424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.6605453491210938, "created_at": "2025-01-15T11:11:17.256679+00:00"} {"global_step": 1999, "acc_step": 0, "speed/wps": 12876.588257193062, "speed/FLOPS": 202244512865711.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1892382800579071, "optim/lr": 0.0011988, "optim/total_tokens": 1048051712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.7222089767456055, "created_at": "2025-01-15T11:11:27.439857+00:00"} {"global_step": 2000, "acc_step": 0, "speed/wps": 12875.495013543343, "speed/FLOPS": 202227341971918.1, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1661573350429535, "optim/lr": 0.0011994, "optim/total_tokens": 1048576000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 3.68118953704834, "created_at": "2025-01-15T11:11:37.623347+00:00"} {"global_step": 2001, "acc_step": 0, "speed/wps": 12511.522799442624, "speed/FLOPS": 196510658199232.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.3038, "optim/grad_norm": 0.1674470603466034, "optim/lr": 0.0012000000000000001, "optim/total_tokens": 1049100288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.54093599319458, "created_at": "2025-01-15T11:11:48.103931+00:00"} {"global_step": 2002, "acc_step": 0, "speed/wps": 12896.09300243129, "speed/FLOPS": 202550861692006.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15823808312416077, "optim/lr": 0.0012006, "optim/total_tokens": 1049624576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.530731201171875, "created_at": "2025-01-15T11:11:58.274804+00:00"} {"global_step": 2003, "acc_step": 0, "speed/wps": 12898.737549095233, "speed/FLOPS": 202592397931351.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17652572691440582, "optim/lr": 0.0012012, "optim/total_tokens": 1050148864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.620293140411377, "created_at": "2025-01-15T11:12:08.440570+00:00"} {"global_step": 2004, "acc_step": 0, "speed/wps": 12893.757734208519, "speed/FLOPS": 202514183095572.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1562589854001999, "optim/lr": 0.0012018, "optim/total_tokens": 1050673152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 3.5740301609039307, "created_at": "2025-01-15T11:12:18.612060+00:00"} {"global_step": 2005, "acc_step": 0, "speed/wps": 12892.152558452928, "speed/FLOPS": 202488971604588.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.185689777135849, "optim/lr": 0.0012024, "optim/total_tokens": 1051197440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.5583972930908203, "created_at": "2025-01-15T11:12:28.783400+00:00"} {"global_step": 2006, "acc_step": 0, "speed/wps": 12886.567352951319, "speed/FLOPS": 202401248277309.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16097667813301086, "optim/lr": 0.001203, "optim/total_tokens": 1051721728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.781851291656494, "created_at": "2025-01-15T11:12:38.964735+00:00"} {"global_step": 2007, "acc_step": 0, "speed/wps": 12878.361144131826, "speed/FLOPS": 202272358491286.38, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15850764513015747, "optim/lr": 0.0012036, "optim/total_tokens": 1052246016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467757, "loss/out": 3.6337385177612305, "created_at": "2025-01-15T11:12:49.148616+00:00"} {"global_step": 2008, "acc_step": 0, "speed/wps": 12878.549241745512, "speed/FLOPS": 202275312822782.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1657436192035675, "optim/lr": 0.0012041999999999999, "optim/total_tokens": 1052770304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 3.819753885269165, "created_at": "2025-01-15T11:12:59.332329+00:00"} {"global_step": 2009, "acc_step": 0, "speed/wps": 12879.544786225326, "speed/FLOPS": 202290949216859.47, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20359882712364197, "optim/lr": 0.0012048, "optim/total_tokens": 1053294592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.5971357822418213, "created_at": "2025-01-15T11:13:09.514472+00:00"} {"global_step": 2010, "acc_step": 0, "speed/wps": 12880.945777604953, "speed/FLOPS": 202312953711640.4, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19462190568447113, "optim/lr": 0.0012054, "optim/total_tokens": 1053818880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.6128921508789062, "created_at": "2025-01-15T11:13:19.694642+00:00"} {"global_step": 2011, "acc_step": 0, "speed/wps": 12881.022188820592, "speed/FLOPS": 202314153854782.03, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19431988894939423, "optim/lr": 0.001206, "optim/total_tokens": 1054343168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.653980016708374, "created_at": "2025-01-15T11:13:29.877479+00:00"} {"global_step": 2012, "acc_step": 0, "speed/wps": 12881.704144896185, "speed/FLOPS": 202324864912055.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14860238134860992, "optim/lr": 0.0012066, "optim/total_tokens": 1054867456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497719, "loss/out": 3.653928756713867, "created_at": "2025-01-15T11:13:40.056286+00:00"} {"global_step": 2013, "acc_step": 0, "speed/wps": 12878.892272535133, "speed/FLOPS": 202280700592707.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17319296300411224, "optim/lr": 0.0012072, "optim/total_tokens": 1055391744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408354, "loss/out": 3.525108575820923, "created_at": "2025-01-15T11:13:50.238877+00:00"} {"global_step": 2014, "acc_step": 0, "speed/wps": 12877.346198648473, "speed/FLOPS": 202256417377789.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16002480685710907, "optim/lr": 0.0012078, "optim/total_tokens": 1055916032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.4826948642730713, "created_at": "2025-01-15T11:14:00.420825+00:00"} {"global_step": 2015, "acc_step": 0, "speed/wps": 12875.960729392347, "speed/FLOPS": 202234656679287.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1531478613615036, "optim/lr": 0.0012084, "optim/total_tokens": 1056440320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.5231964588165283, "created_at": "2025-01-15T11:14:10.605058+00:00"} {"global_step": 2016, "acc_step": 0, "speed/wps": 12882.613660139865, "speed/FLOPS": 202339150098761.97, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15416045486927032, "optim/lr": 0.001209, "optim/total_tokens": 1056964608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.5908215045928955, "created_at": "2025-01-15T11:14:20.782997+00:00"} {"global_step": 2017, "acc_step": 0, "speed/wps": 12877.303471369536, "speed/FLOPS": 202255746287159.94, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1677500605583191, "optim/lr": 0.0012096000000000001, "optim/total_tokens": 1057488896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.5055370330810547, "created_at": "2025-01-15T11:14:30.965324+00:00"} {"global_step": 2018, "acc_step": 0, "speed/wps": 12868.607619566737, "speed/FLOPS": 202119165985244.4, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16912995278835297, "optim/lr": 0.0012102, "optim/total_tokens": 1058013184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.5091326236724854, "created_at": "2025-01-15T11:14:41.154995+00:00"} {"global_step": 2019, "acc_step": 0, "speed/wps": 12875.164645531273, "speed/FLOPS": 202222153088315.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12977363169193268, "optim/lr": 0.0012108000000000002, "optim/total_tokens": 1058537472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414278, "loss/out": 3.5483131408691406, "created_at": "2025-01-15T11:14:51.342548+00:00"} {"global_step": 2020, "acc_step": 0, "speed/wps": 12880.00143580333, "speed/FLOPS": 202298121525984.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17559833824634552, "optim/lr": 0.0012114, "optim/total_tokens": 1059061760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 3.5788440704345703, "created_at": "2025-01-15T11:15:01.525058+00:00"} {"global_step": 2021, "acc_step": 0, "speed/wps": 12877.444942718796, "speed/FLOPS": 202257968289102.06, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.18581131100654602, "optim/lr": 0.0012120000000000002, "optim/total_tokens": 1059586048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.5603737831115723, "created_at": "2025-01-15T11:15:11.708515+00:00"} {"global_step": 2022, "acc_step": 0, "speed/wps": 12879.422974470392, "speed/FLOPS": 202289035995861.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1867728978395462, "optim/lr": 0.0012126, "optim/total_tokens": 1060110336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.4970157146453857, "created_at": "2025-01-15T11:15:21.889195+00:00"} {"global_step": 2023, "acc_step": 0, "speed/wps": 12878.994968479174, "speed/FLOPS": 202282313573627.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18335361778736115, "optim/lr": 0.0012132, "optim/total_tokens": 1060634624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.5356602668762207, "created_at": "2025-01-15T11:15:32.070148+00:00"} {"global_step": 2024, "acc_step": 0, "speed/wps": 12886.447804764759, "speed/FLOPS": 202399370608762.78, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18382196128368378, "optim/lr": 0.0012138000000000001, "optim/total_tokens": 1061158912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 3.6003494262695312, "created_at": "2025-01-15T11:15:42.245347+00:00"} {"global_step": 2025, "acc_step": 0, "speed/wps": 12884.867255593887, "speed/FLOPS": 202374545912132.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1861480325460434, "optim/lr": 0.0012144, "optim/total_tokens": 1061683200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.4391062259674072, "created_at": "2025-01-15T11:15:52.426025+00:00"} {"global_step": 2026, "acc_step": 0, "speed/wps": 12886.071523930766, "speed/FLOPS": 202393460601200.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19690223038196564, "optim/lr": 0.0012150000000000002, "optim/total_tokens": 1062207488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.488560438156128, "created_at": "2025-01-15T11:16:02.601187+00:00"} {"global_step": 2027, "acc_step": 0, "speed/wps": 12881.032653080232, "speed/FLOPS": 202314318210359.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19403940439224243, "optim/lr": 0.0012156, "optim/total_tokens": 1062731776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.5720746517181396, "created_at": "2025-01-15T11:16:12.785398+00:00"} {"global_step": 2028, "acc_step": 0, "speed/wps": 12883.604719196888, "speed/FLOPS": 202354716043110.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.21737605333328247, "optim/lr": 0.0012162, "optim/total_tokens": 1063256064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335952, "loss/out": 3.476290225982666, "created_at": "2025-01-15T11:16:22.964972+00:00"} {"global_step": 2029, "acc_step": 0, "speed/wps": 12889.201148541477, "speed/FLOPS": 202442615656266.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1999969482421875, "optim/lr": 0.0012168, "optim/total_tokens": 1063780352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.5125889778137207, "created_at": "2025-01-15T11:16:33.140376+00:00"} {"global_step": 2030, "acc_step": 0, "speed/wps": 12885.18106561234, "speed/FLOPS": 202379474729692.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1707456260919571, "optim/lr": 0.0012174, "optim/total_tokens": 1064304640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.5226426124572754, "created_at": "2025-01-15T11:16:43.322271+00:00"} {"global_step": 2031, "acc_step": 0, "speed/wps": 12884.547356514977, "speed/FLOPS": 202369521457510.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1583757996559143, "optim/lr": 0.0012180000000000001, "optim/total_tokens": 1064828928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.4948577880859375, "created_at": "2025-01-15T11:16:53.502578+00:00"} {"global_step": 2032, "acc_step": 0, "speed/wps": 12879.728923797818, "speed/FLOPS": 202293841350467.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17051008343696594, "optim/lr": 0.0012186, "optim/total_tokens": 1065353216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.5499846935272217, "created_at": "2025-01-15T11:17:03.682786+00:00"} {"global_step": 2033, "acc_step": 0, "speed/wps": 12881.231063423224, "speed/FLOPS": 202317434517440.16, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18449266254901886, "optim/lr": 0.0012192, "optim/total_tokens": 1065877504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379658, "loss/out": 3.6038835048675537, "created_at": "2025-01-15T11:17:13.861976+00:00"} {"global_step": 2034, "acc_step": 0, "speed/wps": 12881.37865639731, "speed/FLOPS": 202319752667912.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21956941485404968, "optim/lr": 0.0012198, "optim/total_tokens": 1066401792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.6238996982574463, "created_at": "2025-01-15T11:17:24.049998+00:00"} {"global_step": 2035, "acc_step": 0, "speed/wps": 12881.308797313479, "speed/FLOPS": 202318655435004.16, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2104920893907547, "optim/lr": 0.0012204, "optim/total_tokens": 1066926080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.53527569770813, "created_at": "2025-01-15T11:17:34.230627+00:00"} {"global_step": 2036, "acc_step": 0, "speed/wps": 12879.796964672032, "speed/FLOPS": 202294910025894.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20889025926589966, "optim/lr": 0.0012209999999999999, "optim/total_tokens": 1067450368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.6555070877075195, "created_at": "2025-01-15T11:17:44.415847+00:00"} {"global_step": 2037, "acc_step": 0, "speed/wps": 12887.540510055635, "speed/FLOPS": 202416533046890.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22284440696239471, "optim/lr": 0.0012216, "optim/total_tokens": 1067974656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.5842905044555664, "created_at": "2025-01-15T11:17:54.593742+00:00"} {"global_step": 2038, "acc_step": 0, "speed/wps": 12879.628816321949, "speed/FLOPS": 202292269024996.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1665431261062622, "optim/lr": 0.0012222, "optim/total_tokens": 1068498944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416139, "loss/out": 3.5505478382110596, "created_at": "2025-01-15T11:18:04.775542+00:00"} {"global_step": 2039, "acc_step": 0, "speed/wps": 12883.839850764001, "speed/FLOPS": 202358409107476.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19103683531284332, "optim/lr": 0.0012228, "optim/total_tokens": 1069023232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 3.5237908363342285, "created_at": "2025-01-15T11:18:14.955605+00:00"} {"global_step": 2040, "acc_step": 0, "speed/wps": 12884.598648530415, "speed/FLOPS": 202370327069095.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.1602875292301178, "optim/lr": 0.0012234, "optim/total_tokens": 1069547520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.4791202545166016, "created_at": "2025-01-15T11:18:25.134495+00:00"} {"global_step": 2041, "acc_step": 0, "speed/wps": 12879.601769147535, "speed/FLOPS": 202291844211955.5, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2376634031534195, "optim/lr": 0.001224, "optim/total_tokens": 1070071808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.531648635864258, "created_at": "2025-01-15T11:18:35.318043+00:00"} {"global_step": 2042, "acc_step": 0, "speed/wps": 12882.008654429916, "speed/FLOPS": 202329647652722.25, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.19884732365608215, "optim/lr": 0.0012246, "optim/total_tokens": 1070596096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432875, "loss/out": 3.486290693283081, "created_at": "2025-01-15T11:18:45.501666+00:00"} {"global_step": 2043, "acc_step": 0, "speed/wps": 12882.507632990915, "speed/FLOPS": 202337484796691.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16446924209594727, "optim/lr": 0.0012252, "optim/total_tokens": 1071120384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 3.5147793292999268, "created_at": "2025-01-15T11:18:55.681509+00:00"} {"global_step": 2044, "acc_step": 0, "speed/wps": 12882.870969520161, "speed/FLOPS": 202343191496159.7, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17302373051643372, "optim/lr": 0.0012258, "optim/total_tokens": 1071644672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.4591522216796875, "created_at": "2025-01-15T11:19:05.861938+00:00"} {"global_step": 2045, "acc_step": 0, "speed/wps": 12884.870123303397, "speed/FLOPS": 202374590953450.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12869477272033691, "optim/lr": 0.0012264, "optim/total_tokens": 1072168960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.7268803119659424, "created_at": "2025-01-15T11:19:16.045746+00:00"} {"global_step": 2046, "acc_step": 0, "speed/wps": 12881.923111144963, "speed/FLOPS": 202328304077883.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15755735337734222, "optim/lr": 0.001227, "optim/total_tokens": 1072693248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.4958062171936035, "created_at": "2025-01-15T11:19:26.228541+00:00"} {"global_step": 2047, "acc_step": 0, "speed/wps": 12884.562292413302, "speed/FLOPS": 202369756046317.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1424751877784729, "optim/lr": 0.0012276000000000001, "optim/total_tokens": 1073217536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 3.5071892738342285, "created_at": "2025-01-15T11:19:36.404811+00:00"} {"global_step": 2048, "acc_step": 0, "speed/wps": 12877.317794278757, "speed/FLOPS": 202255971248130.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15264426171779633, "optim/lr": 0.0012282, "optim/total_tokens": 1073741824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.4152019023895264, "created_at": "2025-01-15T11:19:46.589146+00:00"} {"global_step": 2049, "acc_step": 0, "speed/wps": 12877.565638911305, "speed/FLOPS": 202259863988659.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1601969301700592, "optim/lr": 0.0012288000000000002, "optim/total_tokens": 1074266112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.4142489433288574, "created_at": "2025-01-15T11:19:56.774331+00:00"} {"global_step": 2050, "acc_step": 0, "speed/wps": 12882.094237762409, "speed/FLOPS": 202330991856562.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16688060760498047, "optim/lr": 0.0012294, "optim/total_tokens": 1074790400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.522653579711914, "created_at": "2025-01-15T11:20:06.953156+00:00"} {"global_step": 2051, "acc_step": 0, "speed/wps": 12877.052076273352, "speed/FLOPS": 202251797781721.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16542337834835052, "optim/lr": 0.00123, "optim/total_tokens": 1075314688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.5818471908569336, "created_at": "2025-01-15T11:20:17.137579+00:00"} {"global_step": 2052, "acc_step": 0, "speed/wps": 12879.35742210539, "speed/FLOPS": 202288006405890.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17014531791210175, "optim/lr": 0.0012306, "optim/total_tokens": 1075838976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.588627338409424, "created_at": "2025-01-15T11:20:27.319348+00:00"} {"global_step": 2053, "acc_step": 0, "speed/wps": 12878.817495285935, "speed/FLOPS": 202279526113253.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23235377669334412, "optim/lr": 0.0012312, "optim/total_tokens": 1076363264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372460, "loss/out": 3.5075063705444336, "created_at": "2025-01-15T11:20:37.503398+00:00"} {"global_step": 2054, "acc_step": 0, "speed/wps": 12884.38514700972, "speed/FLOPS": 202366973734331.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22854657471179962, "optim/lr": 0.0012318000000000001, "optim/total_tokens": 1076887552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507696, "loss/out": 3.5740036964416504, "created_at": "2025-01-15T11:20:47.687357+00:00"} {"global_step": 2055, "acc_step": 0, "speed/wps": 12880.471038981375, "speed/FLOPS": 202305497289197.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1860293745994568, "optim/lr": 0.0012324, "optim/total_tokens": 1077411840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.477078914642334, "created_at": "2025-01-15T11:20:57.866858+00:00"} {"global_step": 2056, "acc_step": 0, "speed/wps": 12877.63686549044, "speed/FLOPS": 202260982699960.25, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2026756852865219, "optim/lr": 0.001233, "optim/total_tokens": 1077936128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.5576679706573486, "created_at": "2025-01-15T11:21:08.049321+00:00"} {"global_step": 2057, "acc_step": 0, "speed/wps": 12882.994902022992, "speed/FLOPS": 202345138026417.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20372527837753296, "optim/lr": 0.0012336, "optim/total_tokens": 1078460416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.5229368209838867, "created_at": "2025-01-15T11:21:18.232614+00:00"} {"global_step": 2058, "acc_step": 0, "speed/wps": 12877.224212858859, "speed/FLOPS": 202254501423337.12, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20555800199508667, "optim/lr": 0.0012342, "optim/total_tokens": 1078984704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 3.62906813621521, "created_at": "2025-01-15T11:21:28.416034+00:00"} {"global_step": 2059, "acc_step": 0, "speed/wps": 12862.057091392633, "speed/FLOPS": 202016280938901.1, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1785079389810562, "optim/lr": 0.0012348, "optim/total_tokens": 1079508992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412610, "loss/out": 3.546677589416504, "created_at": "2025-01-15T11:21:38.610410+00:00"} {"global_step": 2060, "acc_step": 0, "speed/wps": 12879.459946312703, "speed/FLOPS": 202289616689449.47, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1726122796535492, "optim/lr": 0.0012354, "optim/total_tokens": 1080033280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.5460214614868164, "created_at": "2025-01-15T11:21:48.791398+00:00"} {"global_step": 2061, "acc_step": 0, "speed/wps": 12884.844995794243, "speed/FLOPS": 202374196291390.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2011597901582718, "optim/lr": 0.001236, "optim/total_tokens": 1080557568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.6164355278015137, "created_at": "2025-01-15T11:21:58.968885+00:00"} {"global_step": 2062, "acc_step": 0, "speed/wps": 12877.886484058958, "speed/FLOPS": 202264903302590.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.15867304801940918, "optim/lr": 0.0012366, "optim/total_tokens": 1081081856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.414726734161377, "created_at": "2025-01-15T11:22:09.151322+00:00"} {"global_step": 2063, "acc_step": 0, "speed/wps": 12887.094116110711, "speed/FLOPS": 202409521816575.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21426159143447876, "optim/lr": 0.0012372, "optim/total_tokens": 1081606144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 3.7022016048431396, "created_at": "2025-01-15T11:22:19.329277+00:00"} {"global_step": 2064, "acc_step": 0, "speed/wps": 12883.828445458335, "speed/FLOPS": 202358229971478.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26566416025161743, "optim/lr": 0.0012378, "optim/total_tokens": 1082130432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.585557460784912, "created_at": "2025-01-15T11:22:29.508840+00:00"} {"global_step": 2065, "acc_step": 0, "speed/wps": 12881.431978206138, "speed/FLOPS": 202320590160191.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2325708419084549, "optim/lr": 0.0012384, "optim/total_tokens": 1082654720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.510402202606201, "created_at": "2025-01-15T11:22:39.687912+00:00"} {"global_step": 2066, "acc_step": 0, "speed/wps": 12883.23989540352, "speed/FLOPS": 202348985984115.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1578541398048401, "optim/lr": 0.0012389999999999999, "optim/total_tokens": 1083179008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.534935474395752, "created_at": "2025-01-15T11:22:49.868386+00:00"} {"global_step": 2067, "acc_step": 0, "speed/wps": 12884.782358522738, "speed/FLOPS": 202373212486968.88, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2349705547094345, "optim/lr": 0.0012396, "optim/total_tokens": 1083703296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 3.5689258575439453, "created_at": "2025-01-15T11:23:00.044747+00:00"} {"global_step": 2068, "acc_step": 0, "speed/wps": 12880.019708825257, "speed/FLOPS": 202298408528903.44, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2612764835357666, "optim/lr": 0.0012402, "optim/total_tokens": 1084227584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.630002021789551, "created_at": "2025-01-15T11:23:10.224783+00:00"} {"global_step": 2069, "acc_step": 0, "speed/wps": 12871.529925498004, "speed/FLOPS": 202165064815564.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21107150614261627, "optim/lr": 0.0012408, "optim/total_tokens": 1084751872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.4563333988189697, "created_at": "2025-01-15T11:23:20.411808+00:00"} {"global_step": 2070, "acc_step": 0, "speed/wps": 12886.083175080294, "speed/FLOPS": 202393643598514.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1445920467376709, "optim/lr": 0.0012414, "optim/total_tokens": 1085276160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.7232155799865723, "created_at": "2025-01-15T11:23:30.589689+00:00"} {"global_step": 2071, "acc_step": 0, "speed/wps": 12888.349019091436, "speed/FLOPS": 202429231792305.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18080738186836243, "optim/lr": 0.001242, "optim/total_tokens": 1085800448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.645404815673828, "created_at": "2025-01-15T11:23:40.764177+00:00"} {"global_step": 2072, "acc_step": 0, "speed/wps": 12881.972008716748, "speed/FLOPS": 202329072081440.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18067066371440887, "optim/lr": 0.0012426000000000002, "optim/total_tokens": 1086324736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.5966193675994873, "created_at": "2025-01-15T11:23:50.943500+00:00"} {"global_step": 2073, "acc_step": 0, "speed/wps": 12883.06646699172, "speed/FLOPS": 202346262052592.2, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19587720930576324, "optim/lr": 0.0012432, "optim/total_tokens": 1086849024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315844, "loss/out": 3.533662796020508, "created_at": "2025-01-15T11:24:01.124027+00:00"} {"global_step": 2074, "acc_step": 0, "speed/wps": 12886.459789335913, "speed/FLOPS": 202399558842921.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1645590364933014, "optim/lr": 0.0012438000000000002, "optim/total_tokens": 1087373312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.587510585784912, "created_at": "2025-01-15T11:24:11.300947+00:00"} {"global_step": 2075, "acc_step": 0, "speed/wps": 12886.270786855815, "speed/FLOPS": 202396590299254.72, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1736660599708557, "optim/lr": 0.0012444, "optim/total_tokens": 1087897600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.495537042617798, "created_at": "2025-01-15T11:24:21.477930+00:00"} {"global_step": 2076, "acc_step": 0, "speed/wps": 12884.728984201076, "speed/FLOPS": 202372374169903.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17862991988658905, "optim/lr": 0.001245, "optim/total_tokens": 1088421888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.6005263328552246, "created_at": "2025-01-15T11:24:31.656427+00:00"} {"global_step": 2077, "acc_step": 0, "speed/wps": 12881.788839874012, "speed/FLOPS": 202326195163064.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18303067982196808, "optim/lr": 0.0012456000000000001, "optim/total_tokens": 1088946176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 3.5338594913482666, "created_at": "2025-01-15T11:24:41.841677+00:00"} {"global_step": 2078, "acc_step": 0, "speed/wps": 12879.920464378783, "speed/FLOPS": 202296849758495.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12800903618335724, "optim/lr": 0.0012462, "optim/total_tokens": 1089470464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.4622044563293457, "created_at": "2025-01-15T11:24:52.023222+00:00"} {"global_step": 2079, "acc_step": 0, "speed/wps": 12883.190931597886, "speed/FLOPS": 202348216940264.3, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1595613658428192, "optim/lr": 0.0012468000000000002, "optim/total_tokens": 1089994752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.500988721847534, "created_at": "2025-01-15T11:25:02.206482+00:00"} {"global_step": 2080, "acc_step": 0, "speed/wps": 12885.4205643212, "speed/FLOPS": 202383236386020.4, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1583152711391449, "optim/lr": 0.0012474, "optim/total_tokens": 1090519040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.572826385498047, "created_at": "2025-01-15T11:25:12.384179+00:00"} {"global_step": 2081, "acc_step": 0, "speed/wps": 12881.234368093654, "speed/FLOPS": 202317486421830.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17330960929393768, "optim/lr": 0.001248, "optim/total_tokens": 1091043328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446221, "loss/out": 3.543487548828125, "created_at": "2025-01-15T11:25:22.564696+00:00"} {"global_step": 2082, "acc_step": 0, "speed/wps": 12880.755499768602, "speed/FLOPS": 202309965136751.38, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1805369108915329, "optim/lr": 0.0012486, "optim/total_tokens": 1091567616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.434530735015869, "created_at": "2025-01-15T11:25:32.744443+00:00"} {"global_step": 2083, "acc_step": 0, "speed/wps": 12877.953474388949, "speed/FLOPS": 202265955477777.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1972435861825943, "optim/lr": 0.0012492, "optim/total_tokens": 1092091904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.6125004291534424, "created_at": "2025-01-15T11:25:42.926099+00:00"} {"global_step": 2084, "acc_step": 0, "speed/wps": 12882.879779898356, "speed/FLOPS": 202343329875255.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18757018446922302, "optim/lr": 0.0012498000000000001, "optim/total_tokens": 1092616192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.57293701171875, "created_at": "2025-01-15T11:25:53.105449+00:00"} {"global_step": 2085, "acc_step": 0, "speed/wps": 12880.688210320843, "speed/FLOPS": 202308908263508.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15367119014263153, "optim/lr": 0.0012504, "optim/total_tokens": 1093140480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.416926145553589, "created_at": "2025-01-15T11:26:03.286113+00:00"} {"global_step": 2086, "acc_step": 0, "speed/wps": 12877.57467062604, "speed/FLOPS": 202260005844150.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18386608362197876, "optim/lr": 0.001251, "optim/total_tokens": 1093664768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.6183645725250244, "created_at": "2025-01-15T11:26:13.470503+00:00"} {"global_step": 2087, "acc_step": 0, "speed/wps": 12884.376691976613, "speed/FLOPS": 202366840936417.72, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18447841703891754, "optim/lr": 0.0012516, "optim/total_tokens": 1094189056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.4825544357299805, "created_at": "2025-01-15T11:26:23.654667+00:00"} {"global_step": 2088, "acc_step": 0, "speed/wps": 12886.662230128743, "speed/FLOPS": 202402738453755.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17878639698028564, "optim/lr": 0.0012522, "optim/total_tokens": 1094713344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.5291712284088135, "created_at": "2025-01-15T11:26:33.830812+00:00"} {"global_step": 2089, "acc_step": 0, "speed/wps": 12884.199641319947, "speed/FLOPS": 202364060112561.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1637677401304245, "optim/lr": 0.0012528, "optim/total_tokens": 1095237632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.668186664581299, "created_at": "2025-01-15T11:26:44.007799+00:00"} {"global_step": 2090, "acc_step": 0, "speed/wps": 12881.725778372154, "speed/FLOPS": 202325204695522.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1613173633813858, "optim/lr": 0.0012534, "optim/total_tokens": 1095761920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.555144786834717, "created_at": "2025-01-15T11:26:54.192345+00:00"} {"global_step": 2091, "acc_step": 0, "speed/wps": 12886.772605500675, "speed/FLOPS": 202404472050643.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18429048359394073, "optim/lr": 0.001254, "optim/total_tokens": 1096286208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.703939437866211, "created_at": "2025-01-15T11:27:04.369022+00:00"} {"global_step": 2092, "acc_step": 0, "speed/wps": 12880.054147400502, "speed/FLOPS": 202298949434051.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20077314972877502, "optim/lr": 0.0012546, "optim/total_tokens": 1096810496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 3.6525843143463135, "created_at": "2025-01-15T11:27:14.551194+00:00"} {"global_step": 2093, "acc_step": 0, "speed/wps": 12884.156220997449, "speed/FLOPS": 202363378136731.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16290418803691864, "optim/lr": 0.0012552, "optim/total_tokens": 1097334784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.4608216285705566, "created_at": "2025-01-15T11:27:24.728758+00:00"} {"global_step": 2094, "acc_step": 0, "speed/wps": 12885.30233510435, "speed/FLOPS": 202381379433705.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16485735774040222, "optim/lr": 0.0012558, "optim/total_tokens": 1097859072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.5450212955474854, "created_at": "2025-01-15T11:27:34.909069+00:00"} {"global_step": 2095, "acc_step": 0, "speed/wps": 12885.136107435761, "speed/FLOPS": 202378768599749.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1411285698413849, "optim/lr": 0.0012564, "optim/total_tokens": 1098383360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.539294719696045, "created_at": "2025-01-15T11:27:45.094140+00:00"} {"global_step": 2096, "acc_step": 0, "speed/wps": 12884.962325975905, "speed/FLOPS": 202376039123122.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1879328191280365, "optim/lr": 0.0012569999999999999, "optim/total_tokens": 1098907648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 3.542585849761963, "created_at": "2025-01-15T11:27:55.270317+00:00"} {"global_step": 2097, "acc_step": 0, "speed/wps": 12878.377963001582, "speed/FLOPS": 202272622654747.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1773076057434082, "optim/lr": 0.0012576, "optim/total_tokens": 1099431936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 3.679983615875244, "created_at": "2025-01-15T11:28:05.456189+00:00"} {"global_step": 2098, "acc_step": 0, "speed/wps": 12869.35326091035, "speed/FLOPS": 202130877307164.4, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19966569542884827, "optim/lr": 0.0012582000000000001, "optim/total_tokens": 1099956224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.498750686645508, "created_at": "2025-01-15T11:28:15.646374+00:00"} {"global_step": 2099, "acc_step": 0, "speed/wps": 12861.491385523008, "speed/FLOPS": 202007395750857.3, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18362943828105927, "optim/lr": 0.0012588, "optim/total_tokens": 1100480512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.437887191772461, "created_at": "2025-01-15T11:28:25.843789+00:00"} {"global_step": 2100, "acc_step": 0, "speed/wps": 12876.504377508867, "speed/FLOPS": 202243195420008.1, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18769854307174683, "optim/lr": 0.0012594000000000001, "optim/total_tokens": 1101004800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.5081799030303955, "created_at": "2025-01-15T11:28:36.029563+00:00"} {"global_step": 2101, "acc_step": 0, "speed/wps": 12885.184765149841, "speed/FLOPS": 202379532836012.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21369606256484985, "optim/lr": 0.00126, "optim/total_tokens": 1101529088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.435424327850342, "created_at": "2025-01-15T11:28:46.208159+00:00"} {"global_step": 2102, "acc_step": 0, "speed/wps": 12873.833518092195, "speed/FLOPS": 202201245902723.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23078706860542297, "optim/lr": 0.0012606000000000002, "optim/total_tokens": 1102053376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.6602580547332764, "created_at": "2025-01-15T11:28:56.400001+00:00"} {"global_step": 2103, "acc_step": 0, "speed/wps": 12881.105561477394, "speed/FLOPS": 202315463336923.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2294706106185913, "optim/lr": 0.0012612, "optim/total_tokens": 1102577664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425156, "loss/out": 3.58771014213562, "created_at": "2025-01-15T11:29:06.579308+00:00"} {"global_step": 2104, "acc_step": 0, "speed/wps": 12884.602843162626, "speed/FLOPS": 202370392951558.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15687158703804016, "optim/lr": 0.0012618, "optim/total_tokens": 1103101952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 3.562272071838379, "created_at": "2025-01-15T11:29:16.757836+00:00"} {"global_step": 2105, "acc_step": 0, "speed/wps": 12876.58748080793, "speed/FLOPS": 202244500671515.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1813228279352188, "optim/lr": 0.0012624000000000001, "optim/total_tokens": 1103626240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.497079849243164, "created_at": "2025-01-15T11:29:26.942548+00:00"} {"global_step": 2106, "acc_step": 0, "speed/wps": 12884.433806129673, "speed/FLOPS": 202367737992674.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1644197553396225, "optim/lr": 0.001263, "optim/total_tokens": 1104150528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.6330113410949707, "created_at": "2025-01-15T11:29:37.119567+00:00"} {"global_step": 2107, "acc_step": 0, "speed/wps": 12878.660395060864, "speed/FLOPS": 202277058638340.78, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2282828688621521, "optim/lr": 0.0012636000000000001, "optim/total_tokens": 1104674816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448897, "loss/out": 3.489193916320801, "created_at": "2025-01-15T11:29:47.300549+00:00"} {"global_step": 2108, "acc_step": 0, "speed/wps": 12883.981332487961, "speed/FLOPS": 202360631272367.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.180454283952713, "optim/lr": 0.0012642, "optim/total_tokens": 1105199104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 3.537019968032837, "created_at": "2025-01-15T11:29:57.483011+00:00"} {"global_step": 2109, "acc_step": 0, "speed/wps": 12883.441400040356, "speed/FLOPS": 202352150891333.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18656398355960846, "optim/lr": 0.0012648, "optim/total_tokens": 1105723392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.5200982093811035, "created_at": "2025-01-15T11:30:07.665807+00:00"} {"global_step": 2110, "acc_step": 0, "speed/wps": 12887.098389071161, "speed/FLOPS": 202409588929291.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16343018412590027, "optim/lr": 0.0012654, "optim/total_tokens": 1106247680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.630450963973999, "created_at": "2025-01-15T11:30:17.842552+00:00"} {"global_step": 2111, "acc_step": 0, "speed/wps": 12885.319656975855, "speed/FLOPS": 202381651497500.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18146878480911255, "optim/lr": 0.001266, "optim/total_tokens": 1106771968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.4684829711914062, "created_at": "2025-01-15T11:30:28.027514+00:00"} {"global_step": 2112, "acc_step": 0, "speed/wps": 12885.253562070218, "speed/FLOPS": 202380613386183.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1677967607975006, "optim/lr": 0.0012666, "optim/total_tokens": 1107296256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.5446481704711914, "created_at": "2025-01-15T11:30:38.206438+00:00"} {"global_step": 2113, "acc_step": 0, "speed/wps": 12886.467652416073, "speed/FLOPS": 202399682343400.2, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1808372139930725, "optim/lr": 0.0012672, "optim/total_tokens": 1107820544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 3.5937485694885254, "created_at": "2025-01-15T11:30:48.382604+00:00"} {"global_step": 2114, "acc_step": 0, "speed/wps": 12884.774040984381, "speed/FLOPS": 202373081848599.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1690741330385208, "optim/lr": 0.0012678, "optim/total_tokens": 1108344832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.5931918621063232, "created_at": "2025-01-15T11:30:58.561345+00:00"} {"global_step": 2115, "acc_step": 0, "speed/wps": 12885.24879013535, "speed/FLOPS": 202380538436389.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17988012731075287, "optim/lr": 0.0012684, "optim/total_tokens": 1108869120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.4809350967407227, "created_at": "2025-01-15T11:31:08.739139+00:00"} {"global_step": 2116, "acc_step": 0, "speed/wps": 12879.256082859105, "speed/FLOPS": 202286414733772.97, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.22148892283439636, "optim/lr": 0.001269, "optim/total_tokens": 1109393408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.6187005043029785, "created_at": "2025-01-15T11:31:18.920146+00:00"} {"global_step": 2117, "acc_step": 0, "speed/wps": 12885.478244572903, "speed/FLOPS": 202384142333634.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19780656695365906, "optim/lr": 0.0012696, "optim/total_tokens": 1109917696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.574850559234619, "created_at": "2025-01-15T11:31:29.098052+00:00"} {"global_step": 2118, "acc_step": 0, "speed/wps": 12884.752519949296, "speed/FLOPS": 202372743831170.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19165556132793427, "optim/lr": 0.0012702, "optim/total_tokens": 1110441984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356847, "loss/out": 3.595864772796631, "created_at": "2025-01-15T11:31:39.277296+00:00"} {"global_step": 2119, "acc_step": 0, "speed/wps": 12875.164488875133, "speed/FLOPS": 202222150627815.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1500757336616516, "optim/lr": 0.0012707999999999999, "optim/total_tokens": 1110966272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 3.6556849479675293, "created_at": "2025-01-15T11:31:49.461149+00:00"} {"global_step": 2120, "acc_step": 0, "speed/wps": 12884.434946766512, "speed/FLOPS": 202367755907943.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.175379678606987, "optim/lr": 0.0012714, "optim/total_tokens": 1111490560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.5817384719848633, "created_at": "2025-01-15T11:31:59.643366+00:00"} {"global_step": 2121, "acc_step": 0, "speed/wps": 12885.615149346842, "speed/FLOPS": 202386292611235.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16839516162872314, "optim/lr": 0.001272, "optim/total_tokens": 1112014848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.483468532562256, "created_at": "2025-01-15T11:32:09.821301+00:00"} {"global_step": 2122, "acc_step": 0, "speed/wps": 12886.345481953744, "speed/FLOPS": 202397763488409.5, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1824573576450348, "optim/lr": 0.0012726, "optim/total_tokens": 1112539136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426204, "loss/out": 3.4465954303741455, "created_at": "2025-01-15T11:32:19.996196+00:00"} {"global_step": 2123, "acc_step": 0, "speed/wps": 12888.407179975631, "speed/FLOPS": 202430145288913.22, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17154459655284882, "optim/lr": 0.0012732, "optim/total_tokens": 1113063424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 3.5306947231292725, "created_at": "2025-01-15T11:32:30.171896+00:00"} {"global_step": 2124, "acc_step": 0, "speed/wps": 12881.367362737808, "speed/FLOPS": 202319575285470.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2408054769039154, "optim/lr": 0.0012738, "optim/total_tokens": 1113587712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.587780714035034, "created_at": "2025-01-15T11:32:40.352192+00:00"} {"global_step": 2125, "acc_step": 0, "speed/wps": 12882.7099767511, "speed/FLOPS": 202340662883494.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1853773146867752, "optim/lr": 0.0012744, "optim/total_tokens": 1114112000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.448000431060791, "created_at": "2025-01-15T11:32:50.530051+00:00"} {"global_step": 2126, "acc_step": 0, "speed/wps": 12887.217326251033, "speed/FLOPS": 202411457001138.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21617832779884338, "optim/lr": 0.001275, "optim/total_tokens": 1114636288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.5340542793273926, "created_at": "2025-01-15T11:33:00.709144+00:00"} {"global_step": 2127, "acc_step": 0, "speed/wps": 12887.170977909558, "speed/FLOPS": 202410729036745.88, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19601835310459137, "optim/lr": 0.0012756, "optim/total_tokens": 1115160576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 3.517578125, "created_at": "2025-01-15T11:33:10.887340+00:00"} {"global_step": 2128, "acc_step": 0, "speed/wps": 12883.301903235008, "speed/FLOPS": 202349959902316.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22400392591953278, "optim/lr": 0.0012762000000000001, "optim/total_tokens": 1115684864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.4683876037597656, "created_at": "2025-01-15T11:33:21.065099+00:00"} {"global_step": 2129, "acc_step": 0, "speed/wps": 12881.925285410585, "speed/FLOPS": 202328338227712.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14869308471679688, "optim/lr": 0.0012768, "optim/total_tokens": 1116209152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.5419039726257324, "created_at": "2025-01-15T11:33:31.243585+00:00"} {"global_step": 2130, "acc_step": 0, "speed/wps": 12888.379372404985, "speed/FLOPS": 202429708532804.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1498400717973709, "optim/lr": 0.0012774000000000001, "optim/total_tokens": 1116733440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.577404499053955, "created_at": "2025-01-15T11:33:41.416855+00:00"} {"global_step": 2131, "acc_step": 0, "speed/wps": 12881.836965303604, "speed/FLOPS": 202326951039067.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1431502401828766, "optim/lr": 0.001278, "optim/total_tokens": 1117257728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453393, "loss/out": 3.523615598678589, "created_at": "2025-01-15T11:33:51.596858+00:00"} {"global_step": 2132, "acc_step": 0, "speed/wps": 12879.43372018577, "speed/FLOPS": 202289204772087.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12819480895996094, "optim/lr": 0.0012786000000000002, "optim/total_tokens": 1117782016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 3.574000597000122, "created_at": "2025-01-15T11:34:01.779195+00:00"} {"global_step": 2133, "acc_step": 0, "speed/wps": 12882.793299322904, "speed/FLOPS": 202341971578981.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12790116667747498, "optim/lr": 0.0012792, "optim/total_tokens": 1118306304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.6092000007629395, "created_at": "2025-01-15T11:34:11.959964+00:00"} {"global_step": 2134, "acc_step": 0, "speed/wps": 12879.565054619929, "speed/FLOPS": 202291267559848.4, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17949563264846802, "optim/lr": 0.0012798, "optim/total_tokens": 1118830592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.4722414016723633, "created_at": "2025-01-15T11:34:22.142732+00:00"} {"global_step": 2135, "acc_step": 0, "speed/wps": 12883.50459627998, "speed/FLOPS": 202353143475117.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19247372448444366, "optim/lr": 0.0012804000000000001, "optim/total_tokens": 1119354880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.5781753063201904, "created_at": "2025-01-15T11:34:32.322841+00:00"} {"global_step": 2136, "acc_step": 0, "speed/wps": 12880.03008776449, "speed/FLOPS": 202298571544406.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15355807542800903, "optim/lr": 0.001281, "optim/total_tokens": 1119879168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.4337148666381836, "created_at": "2025-01-15T11:34:42.502647+00:00"} {"global_step": 2137, "acc_step": 0, "speed/wps": 12879.598075172624, "speed/FLOPS": 202291786193003.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15730169415473938, "optim/lr": 0.0012816000000000001, "optim/total_tokens": 1120403456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.535153865814209, "created_at": "2025-01-15T11:34:52.682945+00:00"} {"global_step": 2138, "acc_step": 0, "speed/wps": 12881.465101474838, "speed/FLOPS": 202321110406642.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15029430389404297, "optim/lr": 0.0012822, "optim/total_tokens": 1120927744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.6648147106170654, "created_at": "2025-01-15T11:35:02.864275+00:00"} {"global_step": 2139, "acc_step": 0, "speed/wps": 12884.444884588256, "speed/FLOPS": 202367911995090.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13518983125686646, "optim/lr": 0.0012828, "optim/total_tokens": 1121452032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.504147529602051, "created_at": "2025-01-15T11:35:13.045523+00:00"} {"global_step": 2140, "acc_step": 0, "speed/wps": 12884.46409618101, "speed/FLOPS": 202368213739553.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16416001319885254, "optim/lr": 0.0012834, "optim/total_tokens": 1121976320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410502, "loss/out": 3.4919533729553223, "created_at": "2025-01-15T11:35:23.223810+00:00"} {"global_step": 2141, "acc_step": 0, "speed/wps": 12892.34085011322, "speed/FLOPS": 202491928983853.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17007708549499512, "optim/lr": 0.001284, "optim/total_tokens": 1122500608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 3.5280356407165527, "created_at": "2025-01-15T11:35:33.394215+00:00"} {"global_step": 2142, "acc_step": 0, "speed/wps": 12883.25849014247, "speed/FLOPS": 202349278040042.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16109751164913177, "optim/lr": 0.0012846, "optim/total_tokens": 1123024896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.523545503616333, "created_at": "2025-01-15T11:35:43.571782+00:00"} {"global_step": 2143, "acc_step": 0, "speed/wps": 12876.648136914766, "speed/FLOPS": 202245453359023.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15143702924251556, "optim/lr": 0.0012852, "optim/total_tokens": 1123549184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.4814095497131348, "created_at": "2025-01-15T11:35:53.757687+00:00"} {"global_step": 2144, "acc_step": 0, "speed/wps": 12887.823250910284, "speed/FLOPS": 202420973880539.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.160052090883255, "optim/lr": 0.0012858, "optim/total_tokens": 1124073472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 3.471067428588867, "created_at": "2025-01-15T11:36:03.935314+00:00"} {"global_step": 2145, "acc_step": 0, "speed/wps": 12884.389505460947, "speed/FLOPS": 202367042189796.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15835422277450562, "optim/lr": 0.0012864, "optim/total_tokens": 1124597760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.5625808238983154, "created_at": "2025-01-15T11:36:14.116201+00:00"} {"global_step": 2146, "acc_step": 0, "speed/wps": 12875.718491220889, "speed/FLOPS": 202230851995934.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1736774444580078, "optim/lr": 0.001287, "optim/total_tokens": 1125122048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343445, "loss/out": 3.5466432571411133, "created_at": "2025-01-15T11:36:24.301712+00:00"} {"global_step": 2147, "acc_step": 0, "speed/wps": 12880.601008231799, "speed/FLOPS": 202307538634872.25, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22818826138973236, "optim/lr": 0.0012876, "optim/total_tokens": 1125646336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.5823755264282227, "created_at": "2025-01-15T11:36:34.481681+00:00"} {"global_step": 2148, "acc_step": 0, "speed/wps": 12878.407842765386, "speed/FLOPS": 202273091957497.06, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20670589804649353, "optim/lr": 0.0012882, "optim/total_tokens": 1126170624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.515044689178467, "created_at": "2025-01-15T11:36:44.663298+00:00"} {"global_step": 2149, "acc_step": 0, "speed/wps": 12884.225589619888, "speed/FLOPS": 202364467666268.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.213467076420784, "optim/lr": 0.0012887999999999999, "optim/total_tokens": 1126694912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.5013864040374756, "created_at": "2025-01-15T11:36:54.840483+00:00"} {"global_step": 2150, "acc_step": 0, "speed/wps": 12885.807480905629, "speed/FLOPS": 202389313442657.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23238754272460938, "optim/lr": 0.0012894, "optim/total_tokens": 1127219200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.5220277309417725, "created_at": "2025-01-15T11:37:05.016864+00:00"} {"global_step": 2151, "acc_step": 0, "speed/wps": 12883.937444207231, "speed/FLOPS": 202359941946609.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.20859293639659882, "optim/lr": 0.00129, "optim/total_tokens": 1127743488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 3.6504065990448, "created_at": "2025-01-15T11:37:15.198026+00:00"} {"global_step": 2152, "acc_step": 0, "speed/wps": 12889.775721549031, "speed/FLOPS": 202451640114896.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.216761976480484, "optim/lr": 0.0012906, "optim/total_tokens": 1128267776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.52376651763916, "created_at": "2025-01-15T11:37:25.372291+00:00"} {"global_step": 2153, "acc_step": 0, "speed/wps": 12891.172055657353, "speed/FLOPS": 202473571460831.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2316872477531433, "optim/lr": 0.0012912, "optim/total_tokens": 1128792064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.672299385070801, "created_at": "2025-01-15T11:37:35.543655+00:00"} {"global_step": 2154, "acc_step": 0, "speed/wps": 12884.194489181187, "speed/FLOPS": 202363979191142.7, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1508013755083084, "optim/lr": 0.0012918, "optim/total_tokens": 1129316352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389966, "loss/out": 3.5336897373199463, "created_at": "2025-01-15T11:37:45.720391+00:00"} {"global_step": 2155, "acc_step": 0, "speed/wps": 12884.85884829327, "speed/FLOPS": 202374413863923.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16996876895427704, "optim/lr": 0.0012924, "optim/total_tokens": 1129840640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 3.4910836219787598, "created_at": "2025-01-15T11:37:55.896652+00:00"} {"global_step": 2156, "acc_step": 0, "speed/wps": 12892.08380948204, "speed/FLOPS": 202487891807529.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13246998190879822, "optim/lr": 0.001293, "optim/total_tokens": 1130364928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.494473695755005, "created_at": "2025-01-15T11:38:06.068140+00:00"} {"global_step": 2157, "acc_step": 0, "speed/wps": 12884.50357053969, "speed/FLOPS": 202368833738600.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20773926377296448, "optim/lr": 0.0012936000000000002, "optim/total_tokens": 1130889216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.497138738632202, "created_at": "2025-01-15T11:38:16.246008+00:00"} {"global_step": 2158, "acc_step": 0, "speed/wps": 12886.775625097234, "speed/FLOPS": 202404519477556.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2259557545185089, "optim/lr": 0.0012942000000000001, "optim/total_tokens": 1131413504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.5742757320404053, "created_at": "2025-01-15T11:38:26.420623+00:00"} {"global_step": 2159, "acc_step": 0, "speed/wps": 12882.015587649208, "speed/FLOPS": 202329756548458.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2543260157108307, "optim/lr": 0.0012948, "optim/total_tokens": 1131937792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.640789031982422, "created_at": "2025-01-15T11:38:36.599273+00:00"} {"global_step": 2160, "acc_step": 0, "speed/wps": 12882.019075485816, "speed/FLOPS": 202329811329725.2, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2105264812707901, "optim/lr": 0.0012954000000000002, "optim/total_tokens": 1132462080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.502120018005371, "created_at": "2025-01-15T11:38:46.780900+00:00"} {"global_step": 2161, "acc_step": 0, "speed/wps": 12884.450618452976, "speed/FLOPS": 202368002053314.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15639473497867584, "optim/lr": 0.001296, "optim/total_tokens": 1132986368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.566361427307129, "created_at": "2025-01-15T11:38:56.959303+00:00"} {"global_step": 2162, "acc_step": 0, "speed/wps": 12890.905064232613, "speed/FLOPS": 202469377993621.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21447426080703735, "optim/lr": 0.0012966, "optim/total_tokens": 1133510656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.451291561126709, "created_at": "2025-01-15T11:39:07.132549+00:00"} {"global_step": 2163, "acc_step": 0, "speed/wps": 12883.576521912528, "speed/FLOPS": 202354273166012.38, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17184795439243317, "optim/lr": 0.0012972, "optim/total_tokens": 1134034944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348781, "loss/out": 3.5121190547943115, "created_at": "2025-01-15T11:39:17.309676+00:00"} {"global_step": 2164, "acc_step": 0, "speed/wps": 12886.954256988523, "speed/FLOPS": 202407325136875.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14579306542873383, "optim/lr": 0.0012978, "optim/total_tokens": 1134559232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.604471206665039, "created_at": "2025-01-15T11:39:27.491692+00:00"} {"global_step": 2165, "acc_step": 0, "speed/wps": 12886.864635981929, "speed/FLOPS": 202405917515814.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14887864887714386, "optim/lr": 0.0012984000000000001, "optim/total_tokens": 1135083520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.5760087966918945, "created_at": "2025-01-15T11:39:37.667835+00:00"} {"global_step": 2166, "acc_step": 0, "speed/wps": 12886.897880416656, "speed/FLOPS": 202406439665344.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13989901542663574, "optim/lr": 0.001299, "optim/total_tokens": 1135607808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.467632532119751, "created_at": "2025-01-15T11:39:47.842374+00:00"} {"global_step": 2167, "acc_step": 0, "speed/wps": 12879.35733960108, "speed/FLOPS": 202288005110047.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16419163346290588, "optim/lr": 0.0012996, "optim/total_tokens": 1136132096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 3.5281331539154053, "created_at": "2025-01-15T11:39:58.024777+00:00"} {"global_step": 2168, "acc_step": 0, "speed/wps": 12877.314366722145, "speed/FLOPS": 202255917413644.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1560555100440979, "optim/lr": 0.0013002, "optim/total_tokens": 1136656384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 3.468350410461426, "created_at": "2025-01-15T11:40:08.210670+00:00"} {"global_step": 2169, "acc_step": 0, "speed/wps": 12889.88669332028, "speed/FLOPS": 202453383079055.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15734367072582245, "optim/lr": 0.0013008, "optim/total_tokens": 1137180672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.505906581878662, "created_at": "2025-01-15T11:40:18.382717+00:00"} {"global_step": 2170, "acc_step": 0, "speed/wps": 12889.864340618553, "speed/FLOPS": 202453031999158.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14177444577217102, "optim/lr": 0.0013014, "optim/total_tokens": 1137704960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.521479368209839, "created_at": "2025-01-15T11:40:28.556116+00:00"} {"global_step": 2171, "acc_step": 0, "speed/wps": 12886.576675766522, "speed/FLOPS": 202401394704932.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12939536571502686, "optim/lr": 0.001302, "optim/total_tokens": 1138229248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.491598129272461, "created_at": "2025-01-15T11:40:38.733801+00:00"} {"global_step": 2172, "acc_step": 0, "speed/wps": 12889.296815273558, "speed/FLOPS": 202444118233754.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15200917422771454, "optim/lr": 0.0013026, "optim/total_tokens": 1138753536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 3.4989099502563477, "created_at": "2025-01-15T11:40:48.909137+00:00"} {"global_step": 2173, "acc_step": 0, "speed/wps": 12885.193191751001, "speed/FLOPS": 202379665187362.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19052717089653015, "optim/lr": 0.0013032, "optim/total_tokens": 1139277824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.547628879547119, "created_at": "2025-01-15T11:40:59.087958+00:00"} {"global_step": 2174, "acc_step": 0, "speed/wps": 12887.812370812593, "speed/FLOPS": 202420802993654.25, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1858646720647812, "optim/lr": 0.0013038, "optim/total_tokens": 1139802112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.546206474304199, "created_at": "2025-01-15T11:41:09.263971+00:00"} {"global_step": 2175, "acc_step": 0, "speed/wps": 12886.936839860333, "speed/FLOPS": 202407051576943.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1979880928993225, "optim/lr": 0.0013044, "optim/total_tokens": 1140326400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.5667343139648438, "created_at": "2025-01-15T11:41:19.441546+00:00"} {"global_step": 2176, "acc_step": 0, "speed/wps": 12883.47574369474, "speed/FLOPS": 202352690305618.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1875641942024231, "optim/lr": 0.001305, "optim/total_tokens": 1140850688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 3.5027525424957275, "created_at": "2025-01-15T11:41:29.622128+00:00"} {"global_step": 2177, "acc_step": 0, "speed/wps": 12886.121248234203, "speed/FLOPS": 202394241589714.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17032462358474731, "optim/lr": 0.0013055999999999999, "optim/total_tokens": 1141374976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.6491873264312744, "created_at": "2025-01-15T11:41:39.800344+00:00"} {"global_step": 2178, "acc_step": 0, "speed/wps": 12882.979177271369, "speed/FLOPS": 202344891047585.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17243802547454834, "optim/lr": 0.0013062, "optim/total_tokens": 1141899264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.429264783859253, "created_at": "2025-01-15T11:41:49.983112+00:00"} {"global_step": 2179, "acc_step": 0, "speed/wps": 12890.2809584214, "speed/FLOPS": 202459575554245.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2206783890724182, "optim/lr": 0.0013067999999999999, "optim/total_tokens": 1142423552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.5316267013549805, "created_at": "2025-01-15T11:42:00.156686+00:00"} {"global_step": 2180, "acc_step": 0, "speed/wps": 12888.193304367733, "speed/FLOPS": 202426786078595.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23585502803325653, "optim/lr": 0.0013074, "optim/total_tokens": 1142947840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344503, "loss/out": 3.49576997756958, "created_at": "2025-01-15T11:42:10.330523+00:00"} {"global_step": 2181, "acc_step": 0, "speed/wps": 12889.434984041573, "speed/FLOPS": 202446288364119.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18137894570827484, "optim/lr": 0.001308, "optim/total_tokens": 1143472128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.4346766471862793, "created_at": "2025-01-15T11:42:20.507136+00:00"} {"global_step": 2182, "acc_step": 0, "speed/wps": 12888.171122773252, "speed/FLOPS": 202426437686173.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.235053151845932, "optim/lr": 0.0013086, "optim/total_tokens": 1143996416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.5286037921905518, "created_at": "2025-01-15T11:42:30.680627+00:00"} {"global_step": 2183, "acc_step": 0, "speed/wps": 12885.716218233387, "speed/FLOPS": 202387880036982.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22349980473518372, "optim/lr": 0.0013092000000000002, "optim/total_tokens": 1144520704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.4648752212524414, "created_at": "2025-01-15T11:42:40.861229+00:00"} {"global_step": 2184, "acc_step": 0, "speed/wps": 12888.263275913525, "speed/FLOPS": 202427885077872.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17517845332622528, "optim/lr": 0.0013098, "optim/total_tokens": 1145044992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.557973861694336, "created_at": "2025-01-15T11:42:51.036778+00:00"} {"global_step": 2185, "acc_step": 0, "speed/wps": 12891.09090239454, "speed/FLOPS": 202472296837323.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17490220069885254, "optim/lr": 0.0013104000000000002, "optim/total_tokens": 1145569280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 3.569730520248413, "created_at": "2025-01-15T11:43:01.208292+00:00"} {"global_step": 2186, "acc_step": 0, "speed/wps": 12891.886314090734, "speed/FLOPS": 202484789871022.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16864831745624542, "optim/lr": 0.001311, "optim/total_tokens": 1146093568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 3.4648690223693848, "created_at": "2025-01-15T11:43:11.385908+00:00"} {"global_step": 2187, "acc_step": 0, "speed/wps": 12890.361835841028, "speed/FLOPS": 202460845845257.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17111878097057343, "optim/lr": 0.0013116, "optim/total_tokens": 1146617856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 3.498319625854492, "created_at": "2025-01-15T11:43:21.558112+00:00"} {"global_step": 2188, "acc_step": 0, "speed/wps": 12889.910828752954, "speed/FLOPS": 202453762159191.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17135310173034668, "optim/lr": 0.0013122000000000001, "optim/total_tokens": 1147142144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 3.559453010559082, "created_at": "2025-01-15T11:43:31.736502+00:00"} {"global_step": 2189, "acc_step": 0, "speed/wps": 12882.482311872094, "speed/FLOPS": 202337087093725.22, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21967682242393494, "optim/lr": 0.0013128, "optim/total_tokens": 1147666432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.6128711700439453, "created_at": "2025-01-15T11:43:41.917228+00:00"} {"global_step": 2190, "acc_step": 0, "speed/wps": 12886.569637340684, "speed/FLOPS": 202401284156783.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.21043501794338226, "optim/lr": 0.0013134000000000002, "optim/total_tokens": 1148190720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.5756783485412598, "created_at": "2025-01-15T11:43:52.092283+00:00"} {"global_step": 2191, "acc_step": 0, "speed/wps": 12882.569183311336, "speed/FLOPS": 202338451529052.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16347195208072662, "optim/lr": 0.001314, "optim/total_tokens": 1148715008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469813, "loss/out": 3.5028839111328125, "created_at": "2025-01-15T11:44:02.271548+00:00"} {"global_step": 2192, "acc_step": 0, "speed/wps": 12891.21151563161, "speed/FLOPS": 202474191233951.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18073974549770355, "optim/lr": 0.0013146, "optim/total_tokens": 1149239296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384743, "loss/out": 3.504328727722168, "created_at": "2025-01-15T11:44:12.445105+00:00"} {"global_step": 2193, "acc_step": 0, "speed/wps": 12890.263799122145, "speed/FLOPS": 202459306043871.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15641091763973236, "optim/lr": 0.0013152, "optim/total_tokens": 1149763584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.4285836219787598, "created_at": "2025-01-15T11:44:22.616914+00:00"} {"global_step": 2194, "acc_step": 0, "speed/wps": 12886.195530998586, "speed/FLOPS": 202395408302604.53, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2280445694923401, "optim/lr": 0.0013158, "optim/total_tokens": 1150287872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.546794891357422, "created_at": "2025-01-15T11:44:32.797919+00:00"} {"global_step": 2195, "acc_step": 0, "speed/wps": 12883.2761025362, "speed/FLOPS": 202349554666887.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23296698927879333, "optim/lr": 0.0013164000000000001, "optim/total_tokens": 1150812160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.4649598598480225, "created_at": "2025-01-15T11:44:42.977907+00:00"} {"global_step": 2196, "acc_step": 0, "speed/wps": 12887.359145325328, "speed/FLOPS": 202413684464579.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1855054795742035, "optim/lr": 0.001317, "optim/total_tokens": 1151336448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 3.459723949432373, "created_at": "2025-01-15T11:44:53.152924+00:00"} {"global_step": 2197, "acc_step": 0, "speed/wps": 12885.727603285586, "speed/FLOPS": 202388058854871.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1823018193244934, "optim/lr": 0.0013176, "optim/total_tokens": 1151860736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.5323455333709717, "created_at": "2025-01-15T11:45:03.329398+00:00"} {"global_step": 2198, "acc_step": 0, "speed/wps": 12882.386524582646, "speed/FLOPS": 202335582622718.88, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.14638525247573853, "optim/lr": 0.0013182, "optim/total_tokens": 1152385024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.4763126373291016, "created_at": "2025-01-15T11:45:13.509758+00:00"} {"global_step": 2199, "acc_step": 0, "speed/wps": 12890.12721584599, "speed/FLOPS": 202457160815834.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1606767475605011, "optim/lr": 0.0013188, "optim/total_tokens": 1152909312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.5763299465179443, "created_at": "2025-01-15T11:45:23.686096+00:00"} {"global_step": 2200, "acc_step": 0, "speed/wps": 12878.518167525834, "speed/FLOPS": 202274824759462.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16448874771595, "optim/lr": 0.0013194, "optim/total_tokens": 1153433600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 3.5268807411193848, "created_at": "2025-01-15T11:45:33.868149+00:00"} {"global_step": 2201, "acc_step": 0, "speed/wps": 12889.376581328688, "speed/FLOPS": 202445371069260.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16827762126922607, "optim/lr": 0.00132, "optim/total_tokens": 1153957888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.558042526245117, "created_at": "2025-01-15T11:45:44.045539+00:00"} {"global_step": 2202, "acc_step": 0, "speed/wps": 12886.448300335833, "speed/FLOPS": 202399378392387.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15806874632835388, "optim/lr": 0.0013206, "optim/total_tokens": 1154482176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.574702024459839, "created_at": "2025-01-15T11:45:54.221795+00:00"} {"global_step": 2203, "acc_step": 0, "speed/wps": 12884.644950060663, "speed/FLOPS": 202371054298252.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16630128026008606, "optim/lr": 0.0013212, "optim/total_tokens": 1155006464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.3817925453186035, "created_at": "2025-01-15T11:46:04.402045+00:00"} {"global_step": 2204, "acc_step": 0, "speed/wps": 12889.151712098655, "speed/FLOPS": 202441839188999.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18766899406909943, "optim/lr": 0.0013218, "optim/total_tokens": 1155530752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.5079362392425537, "created_at": "2025-01-15T11:46:14.578739+00:00"} {"global_step": 2205, "acc_step": 0, "speed/wps": 12888.242491505598, "speed/FLOPS": 202427558630185.66, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1906491369009018, "optim/lr": 0.0013224, "optim/total_tokens": 1156055040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 3.349660873413086, "created_at": "2025-01-15T11:46:24.754230+00:00"} {"global_step": 2206, "acc_step": 0, "speed/wps": 12883.373210974209, "speed/FLOPS": 202351079888347.56, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18899370729923248, "optim/lr": 0.001323, "optim/total_tokens": 1156579328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.67873477935791, "created_at": "2025-01-15T11:46:34.932652+00:00"} {"global_step": 2207, "acc_step": 0, "speed/wps": 12886.57946164013, "speed/FLOPS": 202401438460906.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.15880794823169708, "optim/lr": 0.0013235999999999999, "optim/total_tokens": 1157103616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.5458314418792725, "created_at": "2025-01-15T11:46:45.108821+00:00"} {"global_step": 2208, "acc_step": 0, "speed/wps": 12883.829790033134, "speed/FLOPS": 202358251089873.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16410808265209198, "optim/lr": 0.0013242, "optim/total_tokens": 1157627904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.5555601119995117, "created_at": "2025-01-15T11:46:55.286201+00:00"} {"global_step": 2209, "acc_step": 0, "speed/wps": 12885.951745435987, "speed/FLOPS": 202391579315347.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18045592308044434, "optim/lr": 0.0013248000000000001, "optim/total_tokens": 1158152192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384815, "loss/out": 3.549445629119873, "created_at": "2025-01-15T11:47:05.464990+00:00"} {"global_step": 2210, "acc_step": 0, "speed/wps": 12887.112180332906, "speed/FLOPS": 202409805540008.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1767398566007614, "optim/lr": 0.0013254, "optim/total_tokens": 1158676480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 3.4922471046447754, "created_at": "2025-01-15T11:47:15.639440+00:00"} {"global_step": 2211, "acc_step": 0, "speed/wps": 12886.913372832594, "speed/FLOPS": 202406682995025.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13388171792030334, "optim/lr": 0.0013260000000000001, "optim/total_tokens": 1159200768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.462730884552002, "created_at": "2025-01-15T11:47:25.820079+00:00"} {"global_step": 2212, "acc_step": 0, "speed/wps": 12890.1850929038, "speed/FLOPS": 202458069854558.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16190892457962036, "optim/lr": 0.0013266, "optim/total_tokens": 1159725056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.4591410160064697, "created_at": "2025-01-15T11:47:35.992382+00:00"} {"global_step": 2213, "acc_step": 0, "speed/wps": 12878.53566443319, "speed/FLOPS": 202275099572435.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1481342762708664, "optim/lr": 0.0013272000000000002, "optim/total_tokens": 1160249344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 3.5304641723632812, "created_at": "2025-01-15T11:47:46.174065+00:00"} {"global_step": 2214, "acc_step": 0, "speed/wps": 12882.241875280133, "speed/FLOPS": 202333310706653.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.144102543592453, "optim/lr": 0.0013278, "optim/total_tokens": 1160773632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 3.5387539863586426, "created_at": "2025-01-15T11:47:56.353639+00:00"} {"global_step": 2215, "acc_step": 0, "speed/wps": 12882.710996799562, "speed/FLOPS": 202340678904757.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15648365020751953, "optim/lr": 0.0013284000000000002, "optim/total_tokens": 1161297920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.4061381816864014, "created_at": "2025-01-15T11:48:06.533078+00:00"} {"global_step": 2216, "acc_step": 0, "speed/wps": 12885.659867830303, "speed/FLOPS": 202386994976468.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15237456560134888, "optim/lr": 0.001329, "optim/total_tokens": 1161822208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.52650785446167, "created_at": "2025-01-15T11:48:16.708561+00:00"} {"global_step": 2217, "acc_step": 0, "speed/wps": 12881.354419678957, "speed/FLOPS": 202319371996945.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.36017322540283203, "optim/lr": 0.0013296, "optim/total_tokens": 1162346496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.4939792156219482, "created_at": "2025-01-15T11:48:26.889842+00:00"} {"global_step": 2218, "acc_step": 0, "speed/wps": 12882.778789700367, "speed/FLOPS": 202341743685421.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3794458508491516, "optim/lr": 0.0013302000000000001, "optim/total_tokens": 1162870784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475206, "loss/out": 3.602470874786377, "created_at": "2025-01-15T11:48:37.068948+00:00"} {"global_step": 2219, "acc_step": 0, "speed/wps": 12888.994149542594, "speed/FLOPS": 202439364452543.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.33553728461265564, "optim/lr": 0.0013308, "optim/total_tokens": 1163395072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.5797343254089355, "created_at": "2025-01-15T11:48:47.248824+00:00"} {"global_step": 2220, "acc_step": 0, "speed/wps": 12890.02618518608, "speed/FLOPS": 202455573990489.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.48354867100715637, "optim/lr": 0.0013314, "optim/total_tokens": 1163919360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.4860081672668457, "created_at": "2025-01-15T11:48:57.422848+00:00"} {"global_step": 2221, "acc_step": 0, "speed/wps": 12892.17457424532, "speed/FLOPS": 202489317392860.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5724344849586487, "optim/lr": 0.001332, "optim/total_tokens": 1164443648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.6988606452941895, "created_at": "2025-01-15T11:49:07.598904+00:00"} {"global_step": 2222, "acc_step": 0, "speed/wps": 12889.07804683635, "speed/FLOPS": 202440682174828.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.6785431504249573, "optim/lr": 0.0013326, "optim/total_tokens": 1164967936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423026, "loss/out": 3.66841721534729, "created_at": "2025-01-15T11:49:17.774670+00:00"} {"global_step": 2223, "acc_step": 0, "speed/wps": 12884.761177191063, "speed/FLOPS": 202372879805048.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5275205969810486, "optim/lr": 0.0013332, "optim/total_tokens": 1165492224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 3.762012481689453, "created_at": "2025-01-15T11:49:27.952417+00:00"} {"global_step": 2224, "acc_step": 0, "speed/wps": 12893.703992548812, "speed/FLOPS": 202513339008958.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5356765985488892, "optim/lr": 0.0013338, "optim/total_tokens": 1166016512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.751297950744629, "created_at": "2025-01-15T11:49:38.125311+00:00"} {"global_step": 2225, "acc_step": 0, "speed/wps": 12885.651350955064, "speed/FLOPS": 202386861207239.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4290969669818878, "optim/lr": 0.0013344, "optim/total_tokens": 1166540800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.8182621002197266, "created_at": "2025-01-15T11:49:48.301784+00:00"} {"global_step": 2226, "acc_step": 0, "speed/wps": 12888.592920156778, "speed/FLOPS": 202433062593691.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3953342139720917, "optim/lr": 0.001335, "optim/total_tokens": 1167065088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 3.4755125045776367, "created_at": "2025-01-15T11:49:58.478281+00:00"} {"global_step": 2227, "acc_step": 0, "speed/wps": 12886.792068294424, "speed/FLOPS": 202404777740563.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3065165877342224, "optim/lr": 0.0013356, "optim/total_tokens": 1167589376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.640110731124878, "created_at": "2025-01-15T11:50:08.652844+00:00"} {"global_step": 2228, "acc_step": 0, "speed/wps": 12880.40307939688, "speed/FLOPS": 202304429890537.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4266532361507416, "optim/lr": 0.0013362, "optim/total_tokens": 1168113664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.6244959831237793, "created_at": "2025-01-15T11:50:18.835436+00:00"} {"global_step": 2229, "acc_step": 0, "speed/wps": 12883.002068626067, "speed/FLOPS": 202345250587766.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3987167477607727, "optim/lr": 0.0013368, "optim/total_tokens": 1168637952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.699573516845703, "created_at": "2025-01-15T11:50:29.012948+00:00"} {"global_step": 2230, "acc_step": 0, "speed/wps": 12894.072889420635, "speed/FLOPS": 202519133041247.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.381594717502594, "optim/lr": 0.0013373999999999999, "optim/total_tokens": 1169162240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406246, "loss/out": 3.536224126815796, "created_at": "2025-01-15T11:50:39.183539+00:00"} {"global_step": 2231, "acc_step": 0, "speed/wps": 12884.528358177835, "speed/FLOPS": 202369223062519.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31964564323425293, "optim/lr": 0.001338, "optim/total_tokens": 1169686528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 3.4870879650115967, "created_at": "2025-01-15T11:50:49.362150+00:00"} {"global_step": 2232, "acc_step": 0, "speed/wps": 12884.861746669696, "speed/FLOPS": 202374459386907.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2318035513162613, "optim/lr": 0.0013386, "optim/total_tokens": 1170210816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.519606113433838, "created_at": "2025-01-15T11:50:59.540923+00:00"} {"global_step": 2233, "acc_step": 0, "speed/wps": 12887.912865187302, "speed/FLOPS": 202422381395902.78, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19397802650928497, "optim/lr": 0.0013392, "optim/total_tokens": 1170735104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.695981025695801, "created_at": "2025-01-15T11:51:09.714638+00:00"} {"global_step": 2234, "acc_step": 0, "speed/wps": 12891.043306448202, "speed/FLOPS": 202471549277582.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16660180687904358, "optim/lr": 0.0013398, "optim/total_tokens": 1171259392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 3.5797600746154785, "created_at": "2025-01-15T11:51:19.889142+00:00"} {"global_step": 2235, "acc_step": 0, "speed/wps": 12891.392426729935, "speed/FLOPS": 202477032691348.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16346941888332367, "optim/lr": 0.0013404, "optim/total_tokens": 1171783680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.5593957901000977, "created_at": "2025-01-15T11:51:30.060445+00:00"} {"global_step": 2236, "acc_step": 0, "speed/wps": 12890.301824965343, "speed/FLOPS": 202459903291991.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18482111394405365, "optim/lr": 0.001341, "optim/total_tokens": 1172307968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.6149115562438965, "created_at": "2025-01-15T11:51:40.235981+00:00"} {"global_step": 2237, "acc_step": 0, "speed/wps": 12892.258183450771, "speed/FLOPS": 202490630590323.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15646116435527802, "optim/lr": 0.0013416, "optim/total_tokens": 1172832256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.6511926651000977, "created_at": "2025-01-15T11:51:50.407828+00:00"} {"global_step": 2238, "acc_step": 0, "speed/wps": 12891.56379321825, "speed/FLOPS": 202479724237529.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.159749373793602, "optim/lr": 0.0013422, "optim/total_tokens": 1173356544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 3.5846261978149414, "created_at": "2025-01-15T11:52:00.578811+00:00"} {"global_step": 2239, "acc_step": 0, "speed/wps": 12892.262808230269, "speed/FLOPS": 202490703228840.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13435538113117218, "optim/lr": 0.0013428000000000001, "optim/total_tokens": 1173880832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.5761656761169434, "created_at": "2025-01-15T11:52:10.759099+00:00"} {"global_step": 2240, "acc_step": 0, "speed/wps": 12889.874365914926, "speed/FLOPS": 202453189460214.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17312732338905334, "optim/lr": 0.0013434, "optim/total_tokens": 1174405120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.521481990814209, "created_at": "2025-01-15T11:52:20.932580+00:00"} {"global_step": 2241, "acc_step": 0, "speed/wps": 12890.706882358789, "speed/FLOPS": 202466265274962.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14479021728038788, "optim/lr": 0.0013440000000000001, "optim/total_tokens": 1174929408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 3.615684747695923, "created_at": "2025-01-15T11:52:31.107441+00:00"} {"global_step": 2242, "acc_step": 0, "speed/wps": 12888.858126699355, "speed/FLOPS": 202437228026875.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12501677870750427, "optim/lr": 0.0013446, "optim/total_tokens": 1175453696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.503431797027588, "created_at": "2025-01-15T11:52:41.283015+00:00"} {"global_step": 2243, "acc_step": 0, "speed/wps": 12884.87177244338, "speed/FLOPS": 202374616855460.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14059478044509888, "optim/lr": 0.0013452000000000002, "optim/total_tokens": 1175977984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.521818161010742, "created_at": "2025-01-15T11:52:51.459256+00:00"} {"global_step": 2244, "acc_step": 0, "speed/wps": 12885.230823153946, "speed/FLOPS": 202380256240256.97, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15459559857845306, "optim/lr": 0.0013458, "optim/total_tokens": 1176502272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 3.4526071548461914, "created_at": "2025-01-15T11:53:01.637958+00:00"} {"global_step": 2245, "acc_step": 0, "speed/wps": 12885.968384744412, "speed/FLOPS": 202391840658551.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15808291733264923, "optim/lr": 0.0013464, "optim/total_tokens": 1177026560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 3.5557258129119873, "created_at": "2025-01-15T11:53:11.813824+00:00"} {"global_step": 2246, "acc_step": 0, "speed/wps": 12889.86494388312, "speed/FLOPS": 202453041474257.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1368314027786255, "optim/lr": 0.001347, "optim/total_tokens": 1177550848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.5146260261535645, "created_at": "2025-01-15T11:53:21.987419+00:00"} {"global_step": 2247, "acc_step": 0, "speed/wps": 12890.225467721935, "speed/FLOPS": 202458703996558.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16010421514511108, "optim/lr": 0.0013476, "optim/total_tokens": 1178075136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.495980739593506, "created_at": "2025-01-15T11:53:32.159695+00:00"} {"global_step": 2248, "acc_step": 0, "speed/wps": 12888.316894330257, "speed/FLOPS": 202428727228787.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18501490354537964, "optim/lr": 0.0013482000000000001, "optim/total_tokens": 1178599424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.6202917098999023, "created_at": "2025-01-15T11:53:42.333085+00:00"} {"global_step": 2249, "acc_step": 0, "speed/wps": 12885.778840988685, "speed/FLOPS": 202388863613406.97, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19052225351333618, "optim/lr": 0.0013488, "optim/total_tokens": 1179123712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 3.537829875946045, "created_at": "2025-01-15T11:53:52.508752+00:00"} {"global_step": 2250, "acc_step": 0, "speed/wps": 12887.901921207646, "speed/FLOPS": 202422209505663.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16781604290008545, "optim/lr": 0.0013494, "optim/total_tokens": 1179648000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497150, "loss/out": 3.541829824447632, "created_at": "2025-01-15T11:54:02.686185+00:00"} {"global_step": 2251, "acc_step": 0, "speed/wps": 12891.663338809936, "speed/FLOPS": 202481287737837.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15931132435798645, "optim/lr": 0.00135, "optim/total_tokens": 1180172288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.482375144958496, "created_at": "2025-01-15T11:54:12.857965+00:00"} {"global_step": 2252, "acc_step": 0, "speed/wps": 12884.974511865588, "speed/FLOPS": 202376230519264.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14211316406726837, "optim/lr": 0.0013506, "optim/total_tokens": 1180696576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.5509896278381348, "created_at": "2025-01-15T11:54:23.035692+00:00"} {"global_step": 2253, "acc_step": 0, "speed/wps": 12882.48691666928, "speed/FLOPS": 202337159418392.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15514828264713287, "optim/lr": 0.0013512, "optim/total_tokens": 1181220864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 3.575913906097412, "created_at": "2025-01-15T11:54:33.213629+00:00"} {"global_step": 2254, "acc_step": 0, "speed/wps": 12886.859893533865, "speed/FLOPS": 202405843029150.88, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14800883829593658, "optim/lr": 0.0013518, "optim/total_tokens": 1181745152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 3.5426366329193115, "created_at": "2025-01-15T11:54:43.392706+00:00"} {"global_step": 2255, "acc_step": 0, "speed/wps": 12892.629578218088, "speed/FLOPS": 202496463855494.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15312910079956055, "optim/lr": 0.0013524, "optim/total_tokens": 1182269440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 3.5136444568634033, "created_at": "2025-01-15T11:54:53.565176+00:00"} {"global_step": 2256, "acc_step": 0, "speed/wps": 12884.74410925955, "speed/FLOPS": 202372611729730.53, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13356231153011322, "optim/lr": 0.001353, "optim/total_tokens": 1182793728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.410785436630249, "created_at": "2025-01-15T11:55:03.743917+00:00"} {"global_step": 2257, "acc_step": 0, "speed/wps": 12888.38594187371, "speed/FLOPS": 202429811715338.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1646183282136917, "optim/lr": 0.0013536, "optim/total_tokens": 1183318016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 3.527524948120117, "created_at": "2025-01-15T11:55:13.917659+00:00"} {"global_step": 2258, "acc_step": 0, "speed/wps": 12890.77818944266, "speed/FLOPS": 202467385250701.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12695404887199402, "optim/lr": 0.0013542, "optim/total_tokens": 1183842304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.4112439155578613, "created_at": "2025-01-15T11:55:24.089860+00:00"} {"global_step": 2259, "acc_step": 0, "speed/wps": 12891.422965602373, "speed/FLOPS": 202477512346304.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14102773368358612, "optim/lr": 0.0013548, "optim/total_tokens": 1184366592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.715810775756836, "created_at": "2025-01-15T11:55:34.261830+00:00"} {"global_step": 2260, "acc_step": 0, "speed/wps": 12889.332515808663, "speed/FLOPS": 202444678959715.66, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15365512669086456, "optim/lr": 0.0013553999999999999, "optim/total_tokens": 1184890880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340518, "loss/out": 3.4939870834350586, "created_at": "2025-01-15T11:55:44.439749+00:00"} {"global_step": 2261, "acc_step": 0, "speed/wps": 12884.700556689604, "speed/FLOPS": 202371927676772.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1386220008134842, "optim/lr": 0.001356, "optim/total_tokens": 1185415168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.46047306060791, "created_at": "2025-01-15T11:55:54.619463+00:00"} {"global_step": 2262, "acc_step": 0, "speed/wps": 12883.659617451043, "speed/FLOPS": 202355578295630.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.13881221413612366, "optim/lr": 0.0013566, "optim/total_tokens": 1185939456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.4608237743377686, "created_at": "2025-01-15T11:56:04.800467+00:00"} {"global_step": 2263, "acc_step": 0, "speed/wps": 12879.713795188356, "speed/FLOPS": 202293603734866.5, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15677745640277863, "optim/lr": 0.0013572, "optim/total_tokens": 1186463744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.4090681076049805, "created_at": "2025-01-15T11:56:14.981569+00:00"} {"global_step": 2264, "acc_step": 0, "speed/wps": 12881.071790457987, "speed/FLOPS": 202314932916657.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14401574432849884, "optim/lr": 0.0013578, "optim/total_tokens": 1186988032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.6110715866088867, "created_at": "2025-01-15T11:56:25.161452+00:00"} {"global_step": 2265, "acc_step": 0, "speed/wps": 12887.625367947996, "speed/FLOPS": 202417865856696.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14542129635810852, "optim/lr": 0.0013584, "optim/total_tokens": 1187512320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305392, "loss/out": 3.4655284881591797, "created_at": "2025-01-15T11:56:35.335759+00:00"} {"global_step": 2266, "acc_step": 0, "speed/wps": 12887.159176105108, "speed/FLOPS": 202410543673189.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14147478342056274, "optim/lr": 0.001359, "optim/total_tokens": 1188036608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.341263771057129, "created_at": "2025-01-15T11:56:45.510727+00:00"} {"global_step": 2267, "acc_step": 0, "speed/wps": 12889.172500619654, "speed/FLOPS": 202442165701288.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15662114322185516, "optim/lr": 0.0013596, "optim/total_tokens": 1188560896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.5234642028808594, "created_at": "2025-01-15T11:56:55.686899+00:00"} {"global_step": 2268, "acc_step": 0, "speed/wps": 12887.229556528926, "speed/FLOPS": 202411649094458.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1681220680475235, "optim/lr": 0.0013602000000000002, "optim/total_tokens": 1189085184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.4733638763427734, "created_at": "2025-01-15T11:57:05.861886+00:00"} {"global_step": 2269, "acc_step": 0, "speed/wps": 12882.978180766353, "speed/FLOPS": 202344875396104.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16274195909500122, "optim/lr": 0.0013608000000000001, "optim/total_tokens": 1189609472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365596, "loss/out": 3.5034728050231934, "created_at": "2025-01-15T11:57:16.040299+00:00"} {"global_step": 2270, "acc_step": 0, "speed/wps": 12887.274596580135, "speed/FLOPS": 202412356510354.25, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16537787020206451, "optim/lr": 0.0013614, "optim/total_tokens": 1190133760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.430225372314453, "created_at": "2025-01-15T11:57:26.220100+00:00"} {"global_step": 2271, "acc_step": 0, "speed/wps": 12885.859257243372, "speed/FLOPS": 202390126661190.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1942380964756012, "optim/lr": 0.0013620000000000001, "optim/total_tokens": 1190658048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.4356136322021484, "created_at": "2025-01-15T11:57:36.396792+00:00"} {"global_step": 2272, "acc_step": 0, "speed/wps": 12888.871433052513, "speed/FLOPS": 202437437021436.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1755966693162918, "optim/lr": 0.0013626, "optim/total_tokens": 1191182336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 3.45705509185791, "created_at": "2025-01-15T11:57:46.569989+00:00"} {"global_step": 2273, "acc_step": 0, "speed/wps": 12888.993513308407, "speed/FLOPS": 202439354459610.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2081521898508072, "optim/lr": 0.0013632000000000002, "optim/total_tokens": 1191706624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 3.6546640396118164, "created_at": "2025-01-15T11:57:56.743218+00:00"} {"global_step": 2274, "acc_step": 0, "speed/wps": 12891.006291688898, "speed/FLOPS": 202470967909923.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23011969029903412, "optim/lr": 0.0013638, "optim/total_tokens": 1192230912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.5361247062683105, "created_at": "2025-01-15T11:58:06.914388+00:00"} {"global_step": 2275, "acc_step": 0, "speed/wps": 12887.830237597302, "speed/FLOPS": 202421083616059.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14542120695114136, "optim/lr": 0.0013644, "optim/total_tokens": 1192755200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.4701130390167236, "created_at": "2025-01-15T11:58:17.088476+00:00"} {"global_step": 2276, "acc_step": 0, "speed/wps": 12887.326850844931, "speed/FLOPS": 202413177235387.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20440617203712463, "optim/lr": 0.0013650000000000001, "optim/total_tokens": 1193279488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.5236926078796387, "created_at": "2025-01-15T11:58:27.266918+00:00"} {"global_step": 2277, "acc_step": 0, "speed/wps": 12885.87428349052, "speed/FLOPS": 202390362669050.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1883370727300644, "optim/lr": 0.0013656, "optim/total_tokens": 1193803776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 3.419315814971924, "created_at": "2025-01-15T11:58:37.445533+00:00"} {"global_step": 2278, "acc_step": 0, "speed/wps": 12888.879294354523, "speed/FLOPS": 202437560493986.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15311969816684723, "optim/lr": 0.0013662000000000001, "optim/total_tokens": 1194328064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 3.6252999305725098, "created_at": "2025-01-15T11:58:47.623901+00:00"} {"global_step": 2279, "acc_step": 0, "speed/wps": 12888.679707600197, "speed/FLOPS": 202434425709749.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16558445990085602, "optim/lr": 0.0013668, "optim/total_tokens": 1194852352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.335906982421875, "created_at": "2025-01-15T11:58:57.799655+00:00"} {"global_step": 2280, "acc_step": 0, "speed/wps": 12885.14935301866, "speed/FLOPS": 202378976639830.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17808133363723755, "optim/lr": 0.0013674, "optim/total_tokens": 1195376640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.47257661819458, "created_at": "2025-01-15T11:59:07.978126+00:00"} {"global_step": 2281, "acc_step": 0, "speed/wps": 12889.263318829899, "speed/FLOPS": 202443592126077.56, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.17465785145759583, "optim/lr": 0.001368, "optim/total_tokens": 1195900928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.410122871398926, "created_at": "2025-01-15T11:59:18.153976+00:00"} {"global_step": 2282, "acc_step": 0, "speed/wps": 12887.116735338137, "speed/FLOPS": 202409877082624.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18586543202400208, "optim/lr": 0.0013686, "optim/total_tokens": 1196425216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.5414199829101562, "created_at": "2025-01-15T11:59:28.329237+00:00"} {"global_step": 2283, "acc_step": 0, "speed/wps": 12887.199530789734, "speed/FLOPS": 202411177498964.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.21014808118343353, "optim/lr": 0.0013691999999999999, "optim/total_tokens": 1196949504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.39876389503479, "created_at": "2025-01-15T11:59:38.509428+00:00"} {"global_step": 2284, "acc_step": 0, "speed/wps": 12890.284310239764, "speed/FLOPS": 202459628199158.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1890196055173874, "optim/lr": 0.0013698, "optim/total_tokens": 1197473792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 3.4536237716674805, "created_at": "2025-01-15T11:59:48.681225+00:00"} {"global_step": 2285, "acc_step": 0, "speed/wps": 12891.398954398808, "speed/FLOPS": 202477135217358.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1432798057794571, "optim/lr": 0.0013704, "optim/total_tokens": 1197998080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 3.517512321472168, "created_at": "2025-01-15T11:59:58.852947+00:00"} {"global_step": 2286, "acc_step": 0, "speed/wps": 12887.178533860702, "speed/FLOPS": 202410847713341.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15765556693077087, "optim/lr": 0.001371, "optim/total_tokens": 1198522368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.4249563217163086, "created_at": "2025-01-15T12:00:09.027590+00:00"} {"global_step": 2287, "acc_step": 0, "speed/wps": 12888.363770238673, "speed/FLOPS": 202429463479343.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17556948959827423, "optim/lr": 0.0013716, "optim/total_tokens": 1199046656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.4991345405578613, "created_at": "2025-01-15T12:00:19.203496+00:00"} {"global_step": 2288, "acc_step": 0, "speed/wps": 12882.343544522015, "speed/FLOPS": 202334907561805.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17928212881088257, "optim/lr": 0.0013721999999999999, "optim/total_tokens": 1199570944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.5205440521240234, "created_at": "2025-01-15T12:00:29.382585+00:00"} {"global_step": 2289, "acc_step": 0, "speed/wps": 12885.537162913666, "speed/FLOPS": 202385067727138.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17683057487010956, "optim/lr": 0.0013728, "optim/total_tokens": 1200095232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 3.4122254848480225, "created_at": "2025-01-15T12:00:39.559594+00:00"} {"global_step": 2290, "acc_step": 0, "speed/wps": 12868.396583217052, "speed/FLOPS": 202115851369378.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19293101131916046, "optim/lr": 0.0013733999999999999, "optim/total_tokens": 1200619520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303224, "loss/out": 3.3988823890686035, "created_at": "2025-01-15T12:00:49.749173+00:00"} {"global_step": 2291, "acc_step": 0, "speed/wps": 12883.424644602079, "speed/FLOPS": 202351887724150.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19767315685749054, "optim/lr": 0.001374, "optim/total_tokens": 1201143808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.3959860801696777, "created_at": "2025-01-15T12:00:59.927316+00:00"} {"global_step": 2292, "acc_step": 0, "speed/wps": 12882.683823471254, "speed/FLOPS": 202340252110295.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19534918665885925, "optim/lr": 0.0013746, "optim/total_tokens": 1201668096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.4579343795776367, "created_at": "2025-01-15T12:01:10.105256+00:00"} {"global_step": 2293, "acc_step": 0, "speed/wps": 12886.774105268856, "speed/FLOPS": 202404495606563.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16940492391586304, "optim/lr": 0.0013752, "optim/total_tokens": 1202192384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478937, "loss/out": 3.554682731628418, "created_at": "2025-01-15T12:01:20.282124+00:00"} {"global_step": 2294, "acc_step": 0, "speed/wps": 12890.384657806462, "speed/FLOPS": 202461204295585.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14802207052707672, "optim/lr": 0.0013758000000000002, "optim/total_tokens": 1202716672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 3.4835562705993652, "created_at": "2025-01-15T12:01:30.455620+00:00"} {"global_step": 2295, "acc_step": 0, "speed/wps": 12888.388860721738, "speed/FLOPS": 202429857559857.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15398350358009338, "optim/lr": 0.0013764, "optim/total_tokens": 1203240960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.5800211429595947, "created_at": "2025-01-15T12:01:40.629412+00:00"} {"global_step": 2296, "acc_step": 0, "speed/wps": 12886.946592738306, "speed/FLOPS": 202407204759293.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13110046088695526, "optim/lr": 0.0013770000000000002, "optim/total_tokens": 1203765248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.4529435634613037, "created_at": "2025-01-15T12:01:50.805132+00:00"} {"global_step": 2297, "acc_step": 0, "speed/wps": 12886.526159744672, "speed/FLOPS": 202400601281393.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14384068548679352, "optim/lr": 0.0013776, "optim/total_tokens": 1204289536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.436692714691162, "created_at": "2025-01-15T12:02:00.981221+00:00"} {"global_step": 2298, "acc_step": 0, "speed/wps": 12887.243125036908, "speed/FLOPS": 202411862206521.47, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16307972371578217, "optim/lr": 0.0013782, "optim/total_tokens": 1204813824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305392, "loss/out": 3.402998685836792, "created_at": "2025-01-15T12:02:11.158012+00:00"} {"global_step": 2299, "acc_step": 0, "speed/wps": 12887.156255453909, "speed/FLOPS": 202410497800349.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20318356156349182, "optim/lr": 0.0013788000000000001, "optim/total_tokens": 1205338112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.4477574825286865, "created_at": "2025-01-15T12:02:21.343903+00:00"} {"global_step": 2300, "acc_step": 0, "speed/wps": 12894.91554311956, "speed/FLOPS": 202532368075516.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20857153832912445, "optim/lr": 0.0013794, "optim/total_tokens": 1205862400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472193, "loss/out": 3.5869345664978027, "created_at": "2025-01-15T12:02:31.512095+00:00"} {"global_step": 2301, "acc_step": 0, "speed/wps": 12887.024105337645, "speed/FLOPS": 202408422201179.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19607695937156677, "optim/lr": 0.0013800000000000002, "optim/total_tokens": 1206386688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.365670919418335, "created_at": "2025-01-15T12:02:41.689703+00:00"} {"global_step": 2302, "acc_step": 0, "speed/wps": 12881.773864421075, "speed/FLOPS": 202325959952998.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17629718780517578, "optim/lr": 0.0013806, "optim/total_tokens": 1206910976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 3.545480251312256, "created_at": "2025-01-15T12:02:51.870481+00:00"} {"global_step": 2303, "acc_step": 0, "speed/wps": 12884.43615581819, "speed/FLOPS": 202367774897761.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17230607569217682, "optim/lr": 0.0013812, "optim/total_tokens": 1207435264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.433901309967041, "created_at": "2025-01-15T12:03:02.047186+00:00"} {"global_step": 2304, "acc_step": 0, "speed/wps": 12888.816652083575, "speed/FLOPS": 202436576611042.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1608683466911316, "optim/lr": 0.0013818, "optim/total_tokens": 1207959552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.4655556678771973, "created_at": "2025-01-15T12:03:12.221279+00:00"} {"global_step": 2305, "acc_step": 0, "speed/wps": 12887.373844642536, "speed/FLOPS": 202413915337555.4, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1465507447719574, "optim/lr": 0.0013824, "optim/total_tokens": 1208483840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.463432788848877, "created_at": "2025-01-15T12:03:22.396529+00:00"} {"global_step": 2306, "acc_step": 0, "speed/wps": 12888.094105055452, "speed/FLOPS": 202425228017081.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14272713661193848, "optim/lr": 0.0013830000000000001, "optim/total_tokens": 1209008128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.4785072803497314, "created_at": "2025-01-15T12:03:32.574049+00:00"} {"global_step": 2307, "acc_step": 0, "speed/wps": 12897.511913475515, "speed/FLOPS": 202573147639743.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1574743241071701, "optim/lr": 0.0013836, "optim/total_tokens": 1209532416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.498790740966797, "created_at": "2025-01-15T12:03:42.740457+00:00"} {"global_step": 2308, "acc_step": 0, "speed/wps": 12890.245991704462, "speed/FLOPS": 202459026353906.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1509704738855362, "optim/lr": 0.0013842, "optim/total_tokens": 1210056704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.3909201622009277, "created_at": "2025-01-15T12:03:52.912397+00:00"} {"global_step": 2309, "acc_step": 0, "speed/wps": 12888.453299969504, "speed/FLOPS": 202430869666792.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17263640463352203, "optim/lr": 0.0013848, "optim/total_tokens": 1210580992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500338, "loss/out": 3.4064764976501465, "created_at": "2025-01-15T12:04:03.086015+00:00"} {"global_step": 2310, "acc_step": 0, "speed/wps": 12889.939300510216, "speed/FLOPS": 202454209347263.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17161181569099426, "optim/lr": 0.0013854, "optim/total_tokens": 1211105280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313436, "loss/out": 3.670358896255493, "created_at": "2025-01-15T12:04:13.259528+00:00"} {"global_step": 2311, "acc_step": 0, "speed/wps": 12890.704225759213, "speed/FLOPS": 202466223549416.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1322791874408722, "optim/lr": 0.001386, "optim/total_tokens": 1211629568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.4579474925994873, "created_at": "2025-01-15T12:04:23.431759+00:00"} {"global_step": 2312, "acc_step": 0, "speed/wps": 12888.962439423929, "speed/FLOPS": 202438866401555.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1395988017320633, "optim/lr": 0.0013866, "optim/total_tokens": 1212153856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.6039257049560547, "created_at": "2025-01-15T12:04:33.604896+00:00"} {"global_step": 2313, "acc_step": 0, "speed/wps": 12888.923635346431, "speed/FLOPS": 202438256930195.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14659743010997772, "optim/lr": 0.0013871999999999999, "optim/total_tokens": 1212678144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 3.454800844192505, "created_at": "2025-01-15T12:04:43.779634+00:00"} {"global_step": 2314, "acc_step": 0, "speed/wps": 12887.618801614899, "speed/FLOPS": 202417762723412.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15761014819145203, "optim/lr": 0.0013878, "optim/total_tokens": 1213202432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.65474271774292, "created_at": "2025-01-15T12:04:53.958255+00:00"} {"global_step": 2315, "acc_step": 0, "speed/wps": 12883.801302799184, "speed/FLOPS": 202357803658718.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15182410180568695, "optim/lr": 0.0013884, "optim/total_tokens": 1213726720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.459756374359131, "created_at": "2025-01-15T12:05:04.135413+00:00"} {"global_step": 2316, "acc_step": 0, "speed/wps": 12887.810051757375, "speed/FLOPS": 202420766569705.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1754606068134308, "optim/lr": 0.001389, "optim/total_tokens": 1214251008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.545992136001587, "created_at": "2025-01-15T12:05:14.312620+00:00"} {"global_step": 2317, "acc_step": 0, "speed/wps": 12890.809268540728, "speed/FLOPS": 202467873390643.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15999752283096313, "optim/lr": 0.0013896, "optim/total_tokens": 1214775296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.435683012008667, "created_at": "2025-01-15T12:05:24.484159+00:00"} {"global_step": 2318, "acc_step": 0, "speed/wps": 12885.907981770504, "speed/FLOPS": 202390891946843.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18400520086288452, "optim/lr": 0.0013902, "optim/total_tokens": 1215299584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470341, "loss/out": 3.4362807273864746, "created_at": "2025-01-15T12:05:34.662669+00:00"} {"global_step": 2319, "acc_step": 0, "speed/wps": 12885.623312240186, "speed/FLOPS": 202386420820692.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.255171537399292, "optim/lr": 0.0013908, "optim/total_tokens": 1215823872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.548891544342041, "created_at": "2025-01-15T12:05:44.838196+00:00"} {"global_step": 2320, "acc_step": 0, "speed/wps": 12890.831574929007, "speed/FLOPS": 202468223743123.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2934619188308716, "optim/lr": 0.0013914, "optim/total_tokens": 1216348160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.5442633628845215, "created_at": "2025-01-15T12:05:55.013275+00:00"} {"global_step": 2321, "acc_step": 0, "speed/wps": 12892.60554007939, "speed/FLOPS": 202496086303494.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24411800503730774, "optim/lr": 0.001392, "optim/total_tokens": 1216872448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 3.641899347305298, "created_at": "2025-01-15T12:06:05.186737+00:00"} {"global_step": 2322, "acc_step": 0, "speed/wps": 12888.68251208827, "speed/FLOPS": 202434469758088.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24937456846237183, "optim/lr": 0.0013926000000000001, "optim/total_tokens": 1217396736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.4836039543151855, "created_at": "2025-01-15T12:06:15.360411+00:00"} {"global_step": 2323, "acc_step": 0, "speed/wps": 12882.672086487364, "speed/FLOPS": 202340067764835.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21006320416927338, "optim/lr": 0.0013932, "optim/total_tokens": 1217921024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.5763490200042725, "created_at": "2025-01-15T12:06:25.539059+00:00"} {"global_step": 2324, "acc_step": 0, "speed/wps": 12887.948602904155, "speed/FLOPS": 202422942705854.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24283771216869354, "optim/lr": 0.0013938000000000002, "optim/total_tokens": 1218445312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.5119476318359375, "created_at": "2025-01-15T12:06:35.714235+00:00"} {"global_step": 2325, "acc_step": 0, "speed/wps": 12889.050565704512, "speed/FLOPS": 202440250545889.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.173437237739563, "optim/lr": 0.0013944, "optim/total_tokens": 1218969600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.5206544399261475, "created_at": "2025-01-15T12:06:45.889507+00:00"} {"global_step": 2326, "acc_step": 0, "speed/wps": 12881.914063896817, "speed/FLOPS": 202328161978418.94, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.15160197019577026, "optim/lr": 0.0013950000000000002, "optim/total_tokens": 1219493888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.463738441467285, "created_at": "2025-01-15T12:06:56.071173+00:00"} {"global_step": 2327, "acc_step": 0, "speed/wps": 12884.247913332996, "speed/FLOPS": 202364818290859.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13966140151023865, "optim/lr": 0.0013956, "optim/total_tokens": 1220018176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.456113338470459, "created_at": "2025-01-15T12:07:06.250704+00:00"} {"global_step": 2328, "acc_step": 0, "speed/wps": 12886.66467266783, "speed/FLOPS": 202402776817188.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14553333818912506, "optim/lr": 0.0013962, "optim/total_tokens": 1220542464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 3.4451968669891357, "created_at": "2025-01-15T12:07:16.428806+00:00"} {"global_step": 2329, "acc_step": 0, "speed/wps": 12886.961393776599, "speed/FLOPS": 202407437229938.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14458000659942627, "optim/lr": 0.0013968000000000001, "optim/total_tokens": 1221066752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317431, "loss/out": 3.5615222454071045, "created_at": "2025-01-15T12:07:26.607850+00:00"} {"global_step": 2330, "acc_step": 0, "speed/wps": 12892.050858325038, "speed/FLOPS": 202487374264329.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15147507190704346, "optim/lr": 0.0013974, "optim/total_tokens": 1221591040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.390470266342163, "created_at": "2025-01-15T12:07:36.779526+00:00"} {"global_step": 2331, "acc_step": 0, "speed/wps": 12887.608301868635, "speed/FLOPS": 202417597810469.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20006412267684937, "optim/lr": 0.0013980000000000002, "optim/total_tokens": 1222115328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.494580030441284, "created_at": "2025-01-15T12:07:46.959174+00:00"} {"global_step": 2332, "acc_step": 0, "speed/wps": 12892.196616094554, "speed/FLOPS": 202489663590391.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23957206308841705, "optim/lr": 0.0013986, "optim/total_tokens": 1222639616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 3.6357498168945312, "created_at": "2025-01-15T12:07:57.131209+00:00"} {"global_step": 2333, "acc_step": 0, "speed/wps": 12889.40874796741, "speed/FLOPS": 202445876290522.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18653751909732819, "optim/lr": 0.0013992, "optim/total_tokens": 1223163904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.4344632625579834, "created_at": "2025-01-15T12:08:07.307086+00:00"} {"global_step": 2334, "acc_step": 0, "speed/wps": 12888.73800248792, "speed/FLOPS": 202435341311066.47, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1439681053161621, "optim/lr": 0.0013998, "optim/total_tokens": 1223688192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.598729133605957, "created_at": "2025-01-15T12:08:17.483005+00:00"} {"global_step": 2335, "acc_step": 0, "speed/wps": 12888.488192044455, "speed/FLOPS": 202431417694776.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1564047783613205, "optim/lr": 0.0014004, "optim/total_tokens": 1224212480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.4454588890075684, "created_at": "2025-01-15T12:08:27.659645+00:00"} {"global_step": 2336, "acc_step": 0, "speed/wps": 12888.843459407111, "speed/FLOPS": 202436997656895.38, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14871302247047424, "optim/lr": 0.0014010000000000001, "optim/total_tokens": 1224736768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.482497215270996, "created_at": "2025-01-15T12:08:37.832762+00:00"} {"global_step": 2337, "acc_step": 0, "speed/wps": 12884.313158089379, "speed/FLOPS": 202365843049414.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13879482448101044, "optim/lr": 0.0014016, "optim/total_tokens": 1225261056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 3.438288927078247, "created_at": "2025-01-15T12:08:48.010493+00:00"} {"global_step": 2338, "acc_step": 0, "speed/wps": 12890.176683384494, "speed/FLOPS": 202457937771501.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16820040345191956, "optim/lr": 0.0014022, "optim/total_tokens": 1225785344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.446924924850464, "created_at": "2025-01-15T12:08:58.188324+00:00"} {"global_step": 2339, "acc_step": 0, "speed/wps": 12890.406987519043, "speed/FLOPS": 202461555014406.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14942459762096405, "optim/lr": 0.0014028, "optim/total_tokens": 1226309632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.4727349281311035, "created_at": "2025-01-15T12:09:08.362713+00:00"} {"global_step": 2340, "acc_step": 0, "speed/wps": 12883.984432160694, "speed/FLOPS": 202360679956987.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1403561383485794, "optim/lr": 0.0014034, "optim/total_tokens": 1226833920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407300, "loss/out": 3.5257391929626465, "created_at": "2025-01-15T12:09:18.539511+00:00"} {"global_step": 2341, "acc_step": 0, "speed/wps": 12880.57140121546, "speed/FLOPS": 202307073615996.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1437426656484604, "optim/lr": 0.001404, "optim/total_tokens": 1227358208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 3.4930906295776367, "created_at": "2025-01-15T12:09:28.719051+00:00"} {"global_step": 2342, "acc_step": 0, "speed/wps": 12887.469339096531, "speed/FLOPS": 202415415209178.53, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1939282864332199, "optim/lr": 0.0014046, "optim/total_tokens": 1227882496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.5388312339782715, "created_at": "2025-01-15T12:09:38.896097+00:00"} {"global_step": 2343, "acc_step": 0, "speed/wps": 12885.448488897922, "speed/FLOPS": 202383674979870.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18612176179885864, "optim/lr": 0.0014052, "optim/total_tokens": 1228406784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.574723243713379, "created_at": "2025-01-15T12:09:49.071902+00:00"} {"global_step": 2344, "acc_step": 0, "speed/wps": 12886.493029322015, "speed/FLOPS": 202400080922579.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2050442397594452, "optim/lr": 0.0014058, "optim/total_tokens": 1228931072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.4063637256622314, "created_at": "2025-01-15T12:09:59.249871+00:00"} {"global_step": 2345, "acc_step": 0, "speed/wps": 12887.405113845916, "speed/FLOPS": 202414406463362.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26493972539901733, "optim/lr": 0.0014064, "optim/total_tokens": 1229455360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.4951672554016113, "created_at": "2025-01-15T12:10:09.426556+00:00"} {"global_step": 2346, "acc_step": 0, "speed/wps": 12890.327974946535, "speed/FLOPS": 202460314013380.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.265994668006897, "optim/lr": 0.001407, "optim/total_tokens": 1229979648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 3.5072407722473145, "created_at": "2025-01-15T12:10:19.600580+00:00"} {"global_step": 2347, "acc_step": 0, "speed/wps": 12889.22150163991, "speed/FLOPS": 202442935329645.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21700024604797363, "optim/lr": 0.0014076, "optim/total_tokens": 1230503936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.4627363681793213, "created_at": "2025-01-15T12:10:29.773273+00:00"} {"global_step": 2348, "acc_step": 0, "speed/wps": 12887.942509503788, "speed/FLOPS": 202422847000628.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16853486001491547, "optim/lr": 0.0014082, "optim/total_tokens": 1231028224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429961, "loss/out": 3.5244300365448, "created_at": "2025-01-15T12:10:39.947650+00:00"} {"global_step": 2349, "acc_step": 0, "speed/wps": 12892.487900589806, "speed/FLOPS": 202494238613656.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18094958364963531, "optim/lr": 0.0014088, "optim/total_tokens": 1231552512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.4896798133850098, "created_at": "2025-01-15T12:10:50.122411+00:00"} {"global_step": 2350, "acc_step": 0, "speed/wps": 12890.31656995727, "speed/FLOPS": 202460134882351.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20036457479000092, "optim/lr": 0.0014094, "optim/total_tokens": 1232076800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.3710217475891113, "created_at": "2025-01-15T12:11:00.294332+00:00"} {"global_step": 2351, "acc_step": 0, "speed/wps": 12888.043469735805, "speed/FLOPS": 202424432719806.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18520012497901917, "optim/lr": 0.00141, "optim/total_tokens": 1232601088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.5060954093933105, "created_at": "2025-01-15T12:11:10.469165+00:00"} {"global_step": 2352, "acc_step": 0, "speed/wps": 12891.073130476465, "speed/FLOPS": 202472017704928.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18419018387794495, "optim/lr": 0.0014106000000000001, "optim/total_tokens": 1233125376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.376659393310547, "created_at": "2025-01-15T12:11:20.640920+00:00"} {"global_step": 2353, "acc_step": 0, "speed/wps": 12889.438361390534, "speed/FLOPS": 202446341410026.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20326846837997437, "optim/lr": 0.0014112, "optim/total_tokens": 1233649664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 3.517030954360962, "created_at": "2025-01-15T12:11:30.814197+00:00"} {"global_step": 2354, "acc_step": 0, "speed/wps": 12891.307633762142, "speed/FLOPS": 202475700901271.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20086956024169922, "optim/lr": 0.0014118000000000002, "optim/total_tokens": 1234173952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.59100079536438, "created_at": "2025-01-15T12:11:40.987981+00:00"} {"global_step": 2355, "acc_step": 0, "speed/wps": 12889.32309220807, "speed/FLOPS": 202444530949118.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16756275296211243, "optim/lr": 0.0014124, "optim/total_tokens": 1234698240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.476743459701538, "created_at": "2025-01-15T12:11:51.163941+00:00"} {"global_step": 2356, "acc_step": 0, "speed/wps": 12889.303447061207, "speed/FLOPS": 202444222395092.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.197641059756279, "optim/lr": 0.001413, "optim/total_tokens": 1235222528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 3.5001637935638428, "created_at": "2025-01-15T12:12:01.340130+00:00"} {"global_step": 2357, "acc_step": 0, "speed/wps": 12878.736744987473, "speed/FLOPS": 202278257818853.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1483827531337738, "optim/lr": 0.0014136, "optim/total_tokens": 1235746816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.4124512672424316, "created_at": "2025-01-15T12:12:11.526556+00:00"} {"global_step": 2358, "acc_step": 0, "speed/wps": 12878.462055652382, "speed/FLOPS": 202273943445383.66, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1451726108789444, "optim/lr": 0.0014142, "optim/total_tokens": 1236271104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.482055902481079, "created_at": "2025-01-15T12:12:21.708798+00:00"} {"global_step": 2359, "acc_step": 0, "speed/wps": 12879.737595506856, "speed/FLOPS": 202293977551573.62, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16585539281368256, "optim/lr": 0.0014148000000000001, "optim/total_tokens": 1236795392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.4019885063171387, "created_at": "2025-01-15T12:12:31.892874+00:00"} {"global_step": 2360, "acc_step": 0, "speed/wps": 12884.002273000104, "speed/FLOPS": 202360960171886.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13016937673091888, "optim/lr": 0.0014154, "optim/total_tokens": 1237319680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404041, "loss/out": 3.549126625061035, "created_at": "2025-01-15T12:12:42.069768+00:00"} {"global_step": 2361, "acc_step": 0, "speed/wps": 12893.495589712717, "speed/FLOPS": 202510065756041.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1300252228975296, "optim/lr": 0.001416, "optim/total_tokens": 1237843968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.594933032989502, "created_at": "2025-01-15T12:12:52.239266+00:00"} {"global_step": 2362, "acc_step": 0, "speed/wps": 12893.907875387938, "speed/FLOPS": 202516541269111.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12873072922229767, "optim/lr": 0.0014166, "optim/total_tokens": 1238368256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.5197319984436035, "created_at": "2025-01-15T12:13:02.409400+00:00"} {"global_step": 2363, "acc_step": 0, "speed/wps": 12894.126174615207, "speed/FLOPS": 202519969958449.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13266471028327942, "optim/lr": 0.0014172, "optim/total_tokens": 1238892544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.501701593399048, "created_at": "2025-01-15T12:13:12.578262+00:00"} {"global_step": 2364, "acc_step": 0, "speed/wps": 12887.040402844037, "speed/FLOPS": 202408678175912.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1677895337343216, "optim/lr": 0.0014178, "optim/total_tokens": 1239416832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 3.3734467029571533, "created_at": "2025-01-15T12:13:22.753356+00:00"} {"global_step": 2365, "acc_step": 0, "speed/wps": 12893.24371319968, "speed/FLOPS": 202506109689290.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17697928845882416, "optim/lr": 0.0014184, "optim/total_tokens": 1239941120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.4921810626983643, "created_at": "2025-01-15T12:13:32.927187+00:00"} {"global_step": 2366, "acc_step": 0, "speed/wps": 12891.59098984991, "speed/FLOPS": 202480151398002.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.15456129610538483, "optim/lr": 0.001419, "optim/total_tokens": 1240465408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.522210121154785, "created_at": "2025-01-15T12:13:43.102235+00:00"} {"global_step": 2367, "acc_step": 0, "speed/wps": 12893.79133079929, "speed/FLOPS": 202514710776197.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17934595048427582, "optim/lr": 0.0014196, "optim/total_tokens": 1240989696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 3.51517391204834, "created_at": "2025-01-15T12:13:53.271212+00:00"} {"global_step": 2368, "acc_step": 0, "speed/wps": 12890.412485837449, "speed/FLOPS": 202461641373052.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18468798696994781, "optim/lr": 0.0014202, "optim/total_tokens": 1241513984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.3677806854248047, "created_at": "2025-01-15T12:14:03.444938+00:00"} {"global_step": 2369, "acc_step": 0, "speed/wps": 12889.760192736761, "speed/FLOPS": 202451396213561.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16996018588542938, "optim/lr": 0.0014208, "optim/total_tokens": 1242038272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.484998941421509, "created_at": "2025-01-15T12:14:13.619811+00:00"} {"global_step": 2370, "acc_step": 0, "speed/wps": 12893.079788145003, "speed/FLOPS": 202503535021049.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1996719241142273, "optim/lr": 0.0014214, "optim/total_tokens": 1242562560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.434962272644043, "created_at": "2025-01-15T12:14:23.790350+00:00"} {"global_step": 2371, "acc_step": 0, "speed/wps": 12888.935200764325, "speed/FLOPS": 202438438580976.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19711792469024658, "optim/lr": 0.0014219999999999999, "optim/total_tokens": 1243086848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 3.436160087585449, "created_at": "2025-01-15T12:14:33.965910+00:00"} {"global_step": 2372, "acc_step": 0, "speed/wps": 12885.265025345752, "speed/FLOPS": 202380793432678.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14622114598751068, "optim/lr": 0.0014226, "optim/total_tokens": 1243611136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.427128791809082, "created_at": "2025-01-15T12:14:44.145114+00:00"} {"global_step": 2373, "acc_step": 0, "speed/wps": 12885.439122884662, "speed/FLOPS": 202383527873761.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1586744487285614, "optim/lr": 0.0014232, "optim/total_tokens": 1244135424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.5018675327301025, "created_at": "2025-01-15T12:14:54.323537+00:00"} {"global_step": 2374, "acc_step": 0, "speed/wps": 12892.788664697708, "speed/FLOPS": 202498962527264.38, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16658468544483185, "optim/lr": 0.0014238, "optim/total_tokens": 1244659712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.4158689975738525, "created_at": "2025-01-15T12:15:04.496358+00:00"} {"global_step": 2375, "acc_step": 0, "speed/wps": 12896.343817992372, "speed/FLOPS": 202554801095049.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17704641819000244, "optim/lr": 0.0014244, "optim/total_tokens": 1245184000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 3.4873294830322266, "created_at": "2025-01-15T12:15:14.667140+00:00"} {"global_step": 2376, "acc_step": 0, "speed/wps": 12889.759620175977, "speed/FLOPS": 202451387220707.22, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18402063846588135, "optim/lr": 0.001425, "optim/total_tokens": 1245708288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.4246959686279297, "created_at": "2025-01-15T12:15:24.840091+00:00"} {"global_step": 2377, "acc_step": 0, "speed/wps": 12871.285071214199, "speed/FLOPS": 202161219042571.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16932298243045807, "optim/lr": 0.0014256, "optim/total_tokens": 1246232576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.526671886444092, "created_at": "2025-01-15T12:15:35.032693+00:00"} {"global_step": 2378, "acc_step": 0, "speed/wps": 12880.734121832847, "speed/FLOPS": 202309629366893.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20772001147270203, "optim/lr": 0.0014262, "optim/total_tokens": 1246756864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499296, "loss/out": 3.3649892807006836, "created_at": "2025-01-15T12:15:45.213315+00:00"} {"global_step": 2379, "acc_step": 0, "speed/wps": 12888.923823027571, "speed/FLOPS": 202438259877985.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18083497881889343, "optim/lr": 0.0014268000000000002, "optim/total_tokens": 1247281152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.518181324005127, "created_at": "2025-01-15T12:15:55.388359+00:00"} {"global_step": 2380, "acc_step": 0, "speed/wps": 12889.94457057934, "speed/FLOPS": 202454292120940.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20631997287273407, "optim/lr": 0.0014274000000000001, "optim/total_tokens": 1247805440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.4160919189453125, "created_at": "2025-01-15T12:16:05.560409+00:00"} {"global_step": 2381, "acc_step": 0, "speed/wps": 12894.102551450062, "speed/FLOPS": 202519598924179.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23120678961277008, "optim/lr": 0.001428, "optim/total_tokens": 1248329728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.4013054370880127, "created_at": "2025-01-15T12:16:15.732351+00:00"} {"global_step": 2382, "acc_step": 0, "speed/wps": 12886.878347682707, "speed/FLOPS": 202406132876917.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14609216153621674, "optim/lr": 0.0014286000000000001, "optim/total_tokens": 1248854016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.448988199234009, "created_at": "2025-01-15T12:16:25.906948+00:00"} {"global_step": 2383, "acc_step": 0, "speed/wps": 12890.29314492822, "speed/FLOPS": 202459766960081.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16702501475811005, "optim/lr": 0.0014292, "optim/total_tokens": 1249378304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447029, "loss/out": 3.63165283203125, "created_at": "2025-01-15T12:16:36.085077+00:00"} {"global_step": 2384, "acc_step": 0, "speed/wps": 12888.78495582223, "speed/FLOPS": 202436078777702.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.19117861986160278, "optim/lr": 0.0014298000000000002, "optim/total_tokens": 1249902592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.5205955505371094, "created_at": "2025-01-15T12:16:46.260779+00:00"} {"global_step": 2385, "acc_step": 0, "speed/wps": 12889.319051502678, "speed/FLOPS": 202444467484288.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1790865808725357, "optim/lr": 0.0014304, "optim/total_tokens": 1250426880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.5352749824523926, "created_at": "2025-01-15T12:16:56.438460+00:00"} {"global_step": 2386, "acc_step": 0, "speed/wps": 12885.20729037714, "speed/FLOPS": 202379886625659.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15160709619522095, "optim/lr": 0.001431, "optim/total_tokens": 1250951168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.5091724395751953, "created_at": "2025-01-15T12:17:06.616618+00:00"} {"global_step": 2387, "acc_step": 0, "speed/wps": 12889.733963602408, "speed/FLOPS": 202450984248963.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17789579927921295, "optim/lr": 0.0014316, "optim/total_tokens": 1251475456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388560, "loss/out": 3.554513931274414, "created_at": "2025-01-15T12:17:16.792712+00:00"} {"global_step": 2388, "acc_step": 0, "speed/wps": 12887.709153010273, "speed/FLOPS": 202419181816232.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.20186610519886017, "optim/lr": 0.0014322, "optim/total_tokens": 1251999744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 3.4107096195220947, "created_at": "2025-01-15T12:17:26.967876+00:00"} {"global_step": 2389, "acc_step": 0, "speed/wps": 12893.537074365591, "speed/FLOPS": 202510717329521.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1751432865858078, "optim/lr": 0.0014328000000000001, "optim/total_tokens": 1252524032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.4333977699279785, "created_at": "2025-01-15T12:17:37.138795+00:00"} {"global_step": 2390, "acc_step": 0, "speed/wps": 12891.490122818188, "speed/FLOPS": 202478567142662.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17091549932956696, "optim/lr": 0.0014334, "optim/total_tokens": 1253048320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 3.4905552864074707, "created_at": "2025-01-15T12:17:47.309730+00:00"} {"global_step": 2391, "acc_step": 0, "speed/wps": 12894.271506770177, "speed/FLOPS": 202522252599651.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1674342304468155, "optim/lr": 0.001434, "optim/total_tokens": 1253572608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 3.4605712890625, "created_at": "2025-01-15T12:17:57.482329+00:00"} {"global_step": 2392, "acc_step": 0, "speed/wps": 12893.430134985967, "speed/FLOPS": 202509037699614.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17913927137851715, "optim/lr": 0.0014346, "optim/total_tokens": 1254096896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.5631155967712402, "created_at": "2025-01-15T12:18:07.654743+00:00"} {"global_step": 2393, "acc_step": 0, "speed/wps": 12888.52076971857, "speed/FLOPS": 202431929371914.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1777571588754654, "optim/lr": 0.0014352, "optim/total_tokens": 1254621184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.5211129188537598, "created_at": "2025-01-15T12:18:17.828452+00:00"} {"global_step": 2394, "acc_step": 0, "speed/wps": 12897.031270399499, "speed/FLOPS": 202565598479762.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18935054540634155, "optim/lr": 0.0014358, "optim/total_tokens": 1255145472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401645, "loss/out": 3.5321969985961914, "created_at": "2025-01-15T12:18:27.999583+00:00"} {"global_step": 2395, "acc_step": 0, "speed/wps": 12887.908569262952, "speed/FLOPS": 202422313922507.66, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19360429048538208, "optim/lr": 0.0014364, "optim/total_tokens": 1255669760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.519707441329956, "created_at": "2025-01-15T12:18:38.173344+00:00"} {"global_step": 2396, "acc_step": 0, "speed/wps": 12885.821036980737, "speed/FLOPS": 202389526359445.5, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19089575111865997, "optim/lr": 0.001437, "optim/total_tokens": 1256194048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.5203957557678223, "created_at": "2025-01-15T12:18:48.353065+00:00"} {"global_step": 2397, "acc_step": 0, "speed/wps": 12876.907640995367, "speed/FLOPS": 202249529227205.66, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20093877613544464, "optim/lr": 0.0014376, "optim/total_tokens": 1256718336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.4068803787231445, "created_at": "2025-01-15T12:18:58.536438+00:00"} {"global_step": 2398, "acc_step": 0, "speed/wps": 12872.649937615985, "speed/FLOPS": 202182656144934.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.17129170894622803, "optim/lr": 0.0014382, "optim/total_tokens": 1257242624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449664, "loss/out": 3.439852237701416, "created_at": "2025-01-15T12:19:08.722751+00:00"} {"global_step": 2399, "acc_step": 0, "speed/wps": 12880.4894866253, "speed/FLOPS": 202305787034795.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17833192646503448, "optim/lr": 0.0014388, "optim/total_tokens": 1257766912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.537935256958008, "created_at": "2025-01-15T12:19:18.905914+00:00"} {"global_step": 2400, "acc_step": 0, "speed/wps": 12890.969717565367, "speed/FLOPS": 202470393463055.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1679048240184784, "optim/lr": 0.0014394, "optim/total_tokens": 1258291200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.4125730991363525, "created_at": "2025-01-15T12:19:29.078835+00:00"} {"global_step": 2401, "acc_step": 0, "speed/wps": 12890.942903757872, "speed/FLOPS": 202469972315362.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13905854523181915, "optim/lr": 0.0014399999999999999, "optim/total_tokens": 1258815488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.418672800064087, "created_at": "2025-01-15T12:19:39.250439+00:00"} {"global_step": 2402, "acc_step": 0, "speed/wps": 12890.391007356793, "speed/FLOPS": 202461304023998.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16780586540699005, "optim/lr": 0.0014406, "optim/total_tokens": 1259339776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.3548479080200195, "created_at": "2025-01-15T12:19:49.423308+00:00"} {"global_step": 2403, "acc_step": 0, "speed/wps": 12893.569541726321, "speed/FLOPS": 202511227274036.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18444882333278656, "optim/lr": 0.0014412, "optim/total_tokens": 1259864064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.5535340309143066, "created_at": "2025-01-15T12:19:59.592635+00:00"} {"global_step": 2404, "acc_step": 0, "speed/wps": 12887.618773291406, "speed/FLOPS": 202417762278553.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1944434642791748, "optim/lr": 0.0014418, "optim/total_tokens": 1260388352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.460684299468994, "created_at": "2025-01-15T12:20:09.771988+00:00"} {"global_step": 2405, "acc_step": 0, "speed/wps": 12886.699245904956, "speed/FLOPS": 202403319837385.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22171592712402344, "optim/lr": 0.0014424000000000001, "optim/total_tokens": 1260912640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 3.426239490509033, "created_at": "2025-01-15T12:20:19.948799+00:00"} {"global_step": 2406, "acc_step": 0, "speed/wps": 12891.54535516938, "speed/FLOPS": 202479434642635.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22406572103500366, "optim/lr": 0.001443, "optim/total_tokens": 1261436928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492154, "loss/out": 3.4214987754821777, "created_at": "2025-01-15T12:20:30.121037+00:00"} {"global_step": 2407, "acc_step": 0, "speed/wps": 12886.84922510349, "speed/FLOPS": 202405675466791.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17638516426086426, "optim/lr": 0.0014436000000000002, "optim/total_tokens": 1261961216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 3.468371629714966, "created_at": "2025-01-15T12:20:40.296999+00:00"} {"global_step": 2408, "acc_step": 0, "speed/wps": 12891.15178141537, "speed/FLOPS": 202473253026001.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15480481088161469, "optim/lr": 0.0014442, "optim/total_tokens": 1262485504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.4301204681396484, "created_at": "2025-01-15T12:20:50.468669+00:00"} {"global_step": 2409, "acc_step": 0, "speed/wps": 12891.699410266334, "speed/FLOPS": 202481854289627.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1367388814687729, "optim/lr": 0.0014448, "optim/total_tokens": 1263009792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.435647487640381, "created_at": "2025-01-15T12:21:00.642567+00:00"} {"global_step": 2410, "acc_step": 0, "speed/wps": 12891.06382949283, "speed/FLOPS": 202471871620199.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1491738259792328, "optim/lr": 0.0014454000000000001, "optim/total_tokens": 1263534080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.344454288482666, "created_at": "2025-01-15T12:21:10.814223+00:00"} {"global_step": 2411, "acc_step": 0, "speed/wps": 12893.530887087498, "speed/FLOPS": 202510620149816.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1514529585838318, "optim/lr": 0.001446, "optim/total_tokens": 1264058368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.473755359649658, "created_at": "2025-01-15T12:21:20.986033+00:00"} {"global_step": 2412, "acc_step": 0, "speed/wps": 12883.132265858623, "speed/FLOPS": 202347295514214.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.16431288421154022, "optim/lr": 0.0014466000000000001, "optim/total_tokens": 1264582656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.4188101291656494, "created_at": "2025-01-15T12:21:31.166167+00:00"} {"global_step": 2413, "acc_step": 0, "speed/wps": 12891.711390436869, "speed/FLOPS": 202482042454667.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18426837027072906, "optim/lr": 0.0014472, "optim/total_tokens": 1265106944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.425175189971924, "created_at": "2025-01-15T12:21:41.337386+00:00"} {"global_step": 2414, "acc_step": 0, "speed/wps": 12893.404932736845, "speed/FLOPS": 202508641863660.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.18583570420742035, "optim/lr": 0.0014478, "optim/total_tokens": 1265631232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 3.424008846282959, "created_at": "2025-01-15T12:21:51.507233+00:00"} {"global_step": 2415, "acc_step": 0, "speed/wps": 12894.341824316665, "speed/FLOPS": 202523357033345.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1778804510831833, "optim/lr": 0.0014484, "optim/total_tokens": 1266155520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.4896841049194336, "created_at": "2025-01-15T12:22:01.676211+00:00"} {"global_step": 2416, "acc_step": 0, "speed/wps": 12894.428359591328, "speed/FLOPS": 202524716188745.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14605721831321716, "optim/lr": 0.001449, "optim/total_tokens": 1266679808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.435335636138916, "created_at": "2025-01-15T12:22:11.845919+00:00"} {"global_step": 2417, "acc_step": 0, "speed/wps": 12894.474468343364, "speed/FLOPS": 202525440390056.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15700440108776093, "optim/lr": 0.0014496, "optim/total_tokens": 1267204096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429413, "loss/out": 3.401386260986328, "created_at": "2025-01-15T12:22:22.017221+00:00"} {"global_step": 2418, "acc_step": 0, "speed/wps": 12891.491140714978, "speed/FLOPS": 202478583130130.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14627186954021454, "optim/lr": 0.0014502, "optim/total_tokens": 1267728384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.424891233444214, "created_at": "2025-01-15T12:22:32.196348+00:00"} {"global_step": 2419, "acc_step": 0, "speed/wps": 12897.94381660799, "speed/FLOPS": 202579931271936.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15871107578277588, "optim/lr": 0.0014508, "optim/total_tokens": 1268252672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.567382335662842, "created_at": "2025-01-15T12:22:42.371064+00:00"} {"global_step": 2420, "acc_step": 0, "speed/wps": 12890.96301203671, "speed/FLOPS": 202470288143513.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17191509902477264, "optim/lr": 0.0014514, "optim/total_tokens": 1268776960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.484013557434082, "created_at": "2025-01-15T12:22:52.542669+00:00"} {"global_step": 2421, "acc_step": 0, "speed/wps": 12891.965036361731, "speed/FLOPS": 202486026312463.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1718905121088028, "optim/lr": 0.001452, "optim/total_tokens": 1269301248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 3.568173885345459, "created_at": "2025-01-15T12:23:02.715422+00:00"} {"global_step": 2422, "acc_step": 0, "speed/wps": 12889.110893086667, "speed/FLOPS": 202441198070325.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.16767965257167816, "optim/lr": 0.0014526, "optim/total_tokens": 1269825536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.462736129760742, "created_at": "2025-01-15T12:23:12.891023+00:00"} {"global_step": 2423, "acc_step": 0, "speed/wps": 12891.631130214297, "speed/FLOPS": 202480781857583.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16204676032066345, "optim/lr": 0.0014532, "optim/total_tokens": 1270349824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.376314640045166, "created_at": "2025-01-15T12:23:23.063077+00:00"} {"global_step": 2424, "acc_step": 0, "speed/wps": 12893.582466819527, "speed/FLOPS": 202511430280385.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22136284410953522, "optim/lr": 0.0014537999999999999, "optim/total_tokens": 1270874112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.5243144035339355, "created_at": "2025-01-15T12:23:33.235053+00:00"} {"global_step": 2425, "acc_step": 0, "speed/wps": 12894.106143885607, "speed/FLOPS": 202519655348316.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24307572841644287, "optim/lr": 0.0014544, "optim/total_tokens": 1271398400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470104, "loss/out": 3.5327649116516113, "created_at": "2025-01-15T12:23:43.407674+00:00"} {"global_step": 2426, "acc_step": 0, "speed/wps": 12882.479122123348, "speed/FLOPS": 202337036994337.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19312764704227448, "optim/lr": 0.001455, "optim/total_tokens": 1271922688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.5061941146850586, "created_at": "2025-01-15T12:23:53.585830+00:00"} {"global_step": 2427, "acc_step": 0, "speed/wps": 12895.091639548908, "speed/FLOPS": 202535133911921.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1667216420173645, "optim/lr": 0.0014556, "optim/total_tokens": 1272446976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.442991256713867, "created_at": "2025-01-15T12:24:03.755726+00:00"} {"global_step": 2428, "acc_step": 0, "speed/wps": 12887.25276863118, "speed/FLOPS": 202412013672421.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2535831034183502, "optim/lr": 0.0014562, "optim/total_tokens": 1272971264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.3940963745117188, "created_at": "2025-01-15T12:24:13.930010+00:00"} {"global_step": 2429, "acc_step": 0, "speed/wps": 12891.234525992635, "speed/FLOPS": 202474552643290.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16435189545154572, "optim/lr": 0.0014568, "optim/total_tokens": 1273495552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 3.5144505500793457, "created_at": "2025-01-15T12:24:24.108899+00:00"} {"global_step": 2430, "acc_step": 0, "speed/wps": 12888.828192521767, "speed/FLOPS": 202436757869482.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17691905796527863, "optim/lr": 0.0014574, "optim/total_tokens": 1274019840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.3981950283050537, "created_at": "2025-01-15T12:24:34.288464+00:00"} {"global_step": 2431, "acc_step": 0, "speed/wps": 12890.47543170472, "speed/FLOPS": 202462630024394.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14370527863502502, "optim/lr": 0.001458, "optim/total_tokens": 1274544128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.4695258140563965, "created_at": "2025-01-15T12:24:44.462847+00:00"} {"global_step": 2432, "acc_step": 0, "speed/wps": 12890.438533550032, "speed/FLOPS": 202462050488171.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16457584500312805, "optim/lr": 0.0014586, "optim/total_tokens": 1275068416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 3.407813310623169, "created_at": "2025-01-15T12:24:54.634652+00:00"} {"global_step": 2433, "acc_step": 0, "speed/wps": 12892.023463714937, "speed/FLOPS": 202486943994333.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14736884832382202, "optim/lr": 0.0014592000000000001, "optim/total_tokens": 1275592704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.411086082458496, "created_at": "2025-01-15T12:25:04.807974+00:00"} {"global_step": 2434, "acc_step": 0, "speed/wps": 12882.014978044424, "speed/FLOPS": 202329746973777.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.14194904267787933, "optim/lr": 0.0014598, "optim/total_tokens": 1276116992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.502232074737549, "created_at": "2025-01-15T12:25:14.986936+00:00"} {"global_step": 2435, "acc_step": 0, "speed/wps": 12883.30659824918, "speed/FLOPS": 202350033643965.53, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1581316590309143, "optim/lr": 0.0014604000000000002, "optim/total_tokens": 1276641280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.5085091590881348, "created_at": "2025-01-15T12:25:25.166391+00:00"} {"global_step": 2436, "acc_step": 0, "speed/wps": 12882.790545746868, "speed/FLOPS": 202341928330286.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.17640598118305206, "optim/lr": 0.001461, "optim/total_tokens": 1277165568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.4687211513519287, "created_at": "2025-01-15T12:25:35.345442+00:00"} {"global_step": 2437, "acc_step": 0, "speed/wps": 12883.132837830786, "speed/FLOPS": 202347304497822.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14796999096870422, "optim/lr": 0.0014616000000000002, "optim/total_tokens": 1277689856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.5587515830993652, "created_at": "2025-01-15T12:25:45.528096+00:00"} {"global_step": 2438, "acc_step": 0, "speed/wps": 12893.826032064302, "speed/FLOPS": 202515255807248.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1617351621389389, "optim/lr": 0.0014622, "optim/total_tokens": 1278214144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.5563182830810547, "created_at": "2025-01-15T12:25:55.697536+00:00"} {"global_step": 2439, "acc_step": 0, "speed/wps": 12891.955306590775, "speed/FLOPS": 202485873493040.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1691252738237381, "optim/lr": 0.0014628, "optim/total_tokens": 1278738432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.4012720584869385, "created_at": "2025-01-15T12:26:05.868213+00:00"} {"global_step": 2440, "acc_step": 0, "speed/wps": 12888.33630513565, "speed/FLOPS": 202429032102159.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16493506729602814, "optim/lr": 0.0014634000000000001, "optim/total_tokens": 1279262720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.4719293117523193, "created_at": "2025-01-15T12:26:16.043674+00:00"} {"global_step": 2441, "acc_step": 0, "speed/wps": 12891.788946194818, "speed/FLOPS": 202483260574419.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1513577550649643, "optim/lr": 0.001464, "optim/total_tokens": 1279787008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 3.4885106086730957, "created_at": "2025-01-15T12:26:26.221129+00:00"} {"global_step": 2442, "acc_step": 0, "speed/wps": 12893.679982297395, "speed/FLOPS": 202512961894966.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.16553230583667755, "optim/lr": 0.0014646000000000001, "optim/total_tokens": 1280311296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.4934210777282715, "created_at": "2025-01-15T12:26:36.396491+00:00"} {"global_step": 2443, "acc_step": 0, "speed/wps": 12886.784118678888, "speed/FLOPS": 202404652880927.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1272120475769043, "optim/lr": 0.0014652, "optim/total_tokens": 1280835584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.3882570266723633, "created_at": "2025-01-15T12:26:46.571419+00:00"} {"global_step": 2444, "acc_step": 0, "speed/wps": 12892.064512568713, "speed/FLOPS": 202487588722989.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15793024003505707, "optim/lr": 0.0014658, "optim/total_tokens": 1281359872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419553, "loss/out": 3.454058885574341, "created_at": "2025-01-15T12:26:56.741926+00:00"} {"global_step": 2445, "acc_step": 0, "speed/wps": 12892.395100059177, "speed/FLOPS": 202492781053801.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17890287935733795, "optim/lr": 0.0014664, "optim/total_tokens": 1281884160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.389860153198242, "created_at": "2025-01-15T12:27:06.916156+00:00"} {"global_step": 2446, "acc_step": 0, "speed/wps": 12885.635722359015, "speed/FLOPS": 202386615738662.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.20470787584781647, "optim/lr": 0.001467, "optim/total_tokens": 1282408448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.5278007984161377, "created_at": "2025-01-15T12:27:17.095896+00:00"} {"global_step": 2447, "acc_step": 0, "speed/wps": 12883.753110796051, "speed/FLOPS": 202357046737086.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18152569234371185, "optim/lr": 0.0014676000000000001, "optim/total_tokens": 1282932736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.4573287963867188, "created_at": "2025-01-15T12:27:27.276277+00:00"} {"global_step": 2448, "acc_step": 0, "speed/wps": 12880.861878865557, "speed/FLOPS": 202311635966649.25, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17781436443328857, "optim/lr": 0.0014682, "optim/total_tokens": 1283457024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.563082218170166, "created_at": "2025-01-15T12:27:37.459270+00:00"} {"global_step": 2449, "acc_step": 0, "speed/wps": 12883.03971671434, "speed/FLOPS": 202345841902727.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14418034255504608, "optim/lr": 0.0014688, "optim/total_tokens": 1283981312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.620914936065674, "created_at": "2025-01-15T12:27:47.638197+00:00"} {"global_step": 2450, "acc_step": 0, "speed/wps": 12892.53159203727, "speed/FLOPS": 202494924847877.34, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1513475924730301, "optim/lr": 0.0014694, "optim/total_tokens": 1284505600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290845, "loss/out": 3.4115071296691895, "created_at": "2025-01-15T12:27:57.809349+00:00"} {"global_step": 2451, "acc_step": 0, "speed/wps": 12874.675037144627, "speed/FLOPS": 202214463115810.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.14619779586791992, "optim/lr": 0.00147, "optim/total_tokens": 1285029888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.534788131713867, "created_at": "2025-01-15T12:28:07.994874+00:00"} {"global_step": 2452, "acc_step": 0, "speed/wps": 12877.611557623817, "speed/FLOPS": 202260585205138.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1369975358247757, "optim/lr": 0.0014706, "optim/total_tokens": 1285554176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.490750551223755, "created_at": "2025-01-15T12:28:18.180101+00:00"} {"global_step": 2453, "acc_step": 0, "speed/wps": 12878.32811577994, "speed/FLOPS": 202271839735634.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19139884412288666, "optim/lr": 0.0014712, "optim/total_tokens": 1286078464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.4954426288604736, "created_at": "2025-01-15T12:28:28.364566+00:00"} {"global_step": 2454, "acc_step": 0, "speed/wps": 12884.732353186226, "speed/FLOPS": 202372427084444.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2161925733089447, "optim/lr": 0.0014717999999999999, "optim/total_tokens": 1286602752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.4043707847595215, "created_at": "2025-01-15T12:28:38.543077+00:00"} {"global_step": 2455, "acc_step": 0, "speed/wps": 12891.383759393999, "speed/FLOPS": 202476896558927.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20244963467121124, "optim/lr": 0.0014724, "optim/total_tokens": 1287127040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 3.4471359252929688, "created_at": "2025-01-15T12:28:48.715269+00:00"} {"global_step": 2456, "acc_step": 0, "speed/wps": 12886.49890306285, "speed/FLOPS": 202400173177750.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23101995885372162, "optim/lr": 0.001473, "optim/total_tokens": 1287651328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.494610548019409, "created_at": "2025-01-15T12:28:58.890804+00:00"} {"global_step": 2457, "acc_step": 0, "speed/wps": 12892.973902828691, "speed/FLOPS": 202501871946655.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20754516124725342, "optim/lr": 0.0014736, "optim/total_tokens": 1288175616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.388319253921509, "created_at": "2025-01-15T12:29:09.061798+00:00"} {"global_step": 2458, "acc_step": 0, "speed/wps": 12889.053737460965, "speed/FLOPS": 202440300362683.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20245584845542908, "optim/lr": 0.0014742, "optim/total_tokens": 1288699904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 3.5155811309814453, "created_at": "2025-01-15T12:29:19.234706+00:00"} {"global_step": 2459, "acc_step": 0, "speed/wps": 12874.078465813245, "speed/FLOPS": 202205093143276.34, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.20708440244197845, "optim/lr": 0.0014748, "optim/total_tokens": 1289224192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.60621976852417, "created_at": "2025-01-15T12:29:29.421162+00:00"} {"global_step": 2460, "acc_step": 0, "speed/wps": 12883.926404364278, "speed/FLOPS": 202359768550705.97, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1889008730649948, "optim/lr": 0.0014754, "optim/total_tokens": 1289748480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.3817086219787598, "created_at": "2025-01-15T12:29:39.600982+00:00"} {"global_step": 2461, "acc_step": 0, "speed/wps": 12881.001295720334, "speed/FLOPS": 202313825699932.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19513307511806488, "optim/lr": 0.001476, "optim/total_tokens": 1290272768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.543490409851074, "created_at": "2025-01-15T12:29:49.786739+00:00"} {"global_step": 2462, "acc_step": 0, "speed/wps": 12878.042654506022, "speed/FLOPS": 202267356174062.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20867063105106354, "optim/lr": 0.0014766, "optim/total_tokens": 1290797056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.4971461296081543, "created_at": "2025-01-15T12:29:59.971165+00:00"} {"global_step": 2463, "acc_step": 0, "speed/wps": 12877.642753548726, "speed/FLOPS": 202261075180006.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17385487258434296, "optim/lr": 0.0014772000000000001, "optim/total_tokens": 1291321344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.4000442028045654, "created_at": "2025-01-15T12:30:10.153874+00:00"} {"global_step": 2464, "acc_step": 0, "speed/wps": 12881.654037290913, "speed/FLOPS": 202324077903259.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21053245663642883, "optim/lr": 0.0014778, "optim/total_tokens": 1291845632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 3.535278081893921, "created_at": "2025-01-15T12:30:20.337184+00:00"} {"global_step": 2465, "acc_step": 0, "speed/wps": 12889.946688525002, "speed/FLOPS": 202454325386187.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17718344926834106, "optim/lr": 0.0014784000000000002, "optim/total_tokens": 1292369920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 3.4742307662963867, "created_at": "2025-01-15T12:30:30.509684+00:00"} {"global_step": 2466, "acc_step": 0, "speed/wps": 12889.422725967803, "speed/FLOPS": 202446095834226.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17250774800777435, "optim/lr": 0.001479, "optim/total_tokens": 1292894208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.5011141300201416, "created_at": "2025-01-15T12:30:40.689135+00:00"} {"global_step": 2467, "acc_step": 0, "speed/wps": 12885.081144692083, "speed/FLOPS": 202377905334333.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1722250133752823, "optim/lr": 0.0014796000000000002, "optim/total_tokens": 1293418496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.3965885639190674, "created_at": "2025-01-15T12:30:50.866229+00:00"} {"global_step": 2468, "acc_step": 0, "speed/wps": 12882.33521951942, "speed/FLOPS": 202334776806200.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2042088359594345, "optim/lr": 0.0014802, "optim/total_tokens": 1293942784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.530900478363037, "created_at": "2025-01-15T12:31:01.046568+00:00"} {"global_step": 2469, "acc_step": 0, "speed/wps": 12884.643030850199, "speed/FLOPS": 202371024154415.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1857410967350006, "optim/lr": 0.0014808, "optim/total_tokens": 1294467072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 3.4158272743225098, "created_at": "2025-01-15T12:31:11.223905+00:00"} {"global_step": 2470, "acc_step": 0, "speed/wps": 12888.112561559044, "speed/FLOPS": 202425517901832.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.18604633212089539, "optim/lr": 0.0014814000000000001, "optim/total_tokens": 1294991360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.3580737113952637, "created_at": "2025-01-15T12:31:21.398145+00:00"} {"global_step": 2471, "acc_step": 0, "speed/wps": 12891.293174622539, "speed/FLOPS": 202475473800615.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20968596637248993, "optim/lr": 0.001482, "optim/total_tokens": 1295515648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.396085262298584, "created_at": "2025-01-15T12:31:31.571792+00:00"} {"global_step": 2472, "acc_step": 0, "speed/wps": 12891.463217101138, "speed/FLOPS": 202478144551404.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14416924118995667, "optim/lr": 0.0014826, "optim/total_tokens": 1296039936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 3.4737353324890137, "created_at": "2025-01-15T12:31:41.743197+00:00"} {"global_step": 2473, "acc_step": 0, "speed/wps": 12887.43559134585, "speed/FLOPS": 202414885154379.8, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15069162845611572, "optim/lr": 0.0014832, "optim/total_tokens": 1296564224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.517273426055908, "created_at": "2025-01-15T12:31:51.919070+00:00"} {"global_step": 2474, "acc_step": 0, "speed/wps": 12877.92282136571, "speed/FLOPS": 202265474029925.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15137822926044464, "optim/lr": 0.0014838, "optim/total_tokens": 1297088512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.4036028385162354, "created_at": "2025-01-15T12:32:02.103312+00:00"} {"global_step": 2475, "acc_step": 0, "speed/wps": 12879.613136325488, "speed/FLOPS": 202292022749105.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13634157180786133, "optim/lr": 0.0014844, "optim/total_tokens": 1297612800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.4536995887756348, "created_at": "2025-01-15T12:32:12.283800+00:00"} {"global_step": 2476, "acc_step": 0, "speed/wps": 12873.980814405928, "speed/FLOPS": 202203559393737.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15956832468509674, "optim/lr": 0.001485, "optim/total_tokens": 1298137088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.5070347785949707, "created_at": "2025-01-15T12:32:22.470331+00:00"} {"global_step": 2477, "acc_step": 0, "speed/wps": 12888.7870627636, "speed/FLOPS": 202436111870111.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17752030491828918, "optim/lr": 0.0014856, "optim/total_tokens": 1298661376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 3.6050875186920166, "created_at": "2025-01-15T12:32:32.645342+00:00"} {"global_step": 2478, "acc_step": 0, "speed/wps": 12893.336126704331, "speed/FLOPS": 202507561170370.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1844778209924698, "optim/lr": 0.0014862, "optim/total_tokens": 1299185664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.3726413249969482, "created_at": "2025-01-15T12:32:42.817066+00:00"} {"global_step": 2479, "acc_step": 0, "speed/wps": 12886.17059771944, "speed/FLOPS": 202395016691193.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.173723503947258, "optim/lr": 0.0014868, "optim/total_tokens": 1299709952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.4221105575561523, "created_at": "2025-01-15T12:32:52.993065+00:00"} {"global_step": 2480, "acc_step": 0, "speed/wps": 12882.733395453553, "speed/FLOPS": 202341030706397.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.14374300837516785, "optim/lr": 0.0014874, "optim/total_tokens": 1300234240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.3662147521972656, "created_at": "2025-01-15T12:33:03.171319+00:00"} {"global_step": 2481, "acc_step": 0, "speed/wps": 12885.50238473499, "speed/FLOPS": 202384521488051.97, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.14357507228851318, "optim/lr": 0.001488, "optim/total_tokens": 1300758528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.4577035903930664, "created_at": "2025-01-15T12:33:13.347195+00:00"} {"global_step": 2482, "acc_step": 0, "speed/wps": 12889.285892552385, "speed/FLOPS": 202443946677408.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12648627161979675, "optim/lr": 0.0014885999999999999, "optim/total_tokens": 1301282816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 3.472903251647949, "created_at": "2025-01-15T12:33:23.522547+00:00"} {"global_step": 2483, "acc_step": 0, "speed/wps": 12888.201882435918, "speed/FLOPS": 202426920808943.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13743118941783905, "optim/lr": 0.0014892, "optim/total_tokens": 1301807104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.3882532119750977, "created_at": "2025-01-15T12:33:33.699078+00:00"} {"global_step": 2484, "acc_step": 0, "speed/wps": 12894.823324362165, "speed/FLOPS": 202530919653210.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1357688456773758, "optim/lr": 0.0014898, "optim/total_tokens": 1302331392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.4085516929626465, "created_at": "2025-01-15T12:33:43.868165+00:00"} {"global_step": 2485, "acc_step": 0, "speed/wps": 12889.287911123693, "speed/FLOPS": 202443978381844.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13417024910449982, "optim/lr": 0.0014904, "optim/total_tokens": 1302855680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 3.473479986190796, "created_at": "2025-01-15T12:33:54.042968+00:00"} {"global_step": 2486, "acc_step": 0, "speed/wps": 12892.22194352895, "speed/FLOPS": 202490061392552.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15536215901374817, "optim/lr": 0.001491, "optim/total_tokens": 1303379968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.3604722023010254, "created_at": "2025-01-15T12:34:04.218693+00:00"} {"global_step": 2487, "acc_step": 0, "speed/wps": 12892.273004959608, "speed/FLOPS": 202490863382486.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15338170528411865, "optim/lr": 0.0014916, "optim/total_tokens": 1303904256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.433741569519043, "created_at": "2025-01-15T12:34:14.391240+00:00"} {"global_step": 2488, "acc_step": 0, "speed/wps": 12893.151026990892, "speed/FLOPS": 202504653925016.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1647786945104599, "optim/lr": 0.0014922, "optim/total_tokens": 1304428544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.5290660858154297, "created_at": "2025-01-15T12:34:24.560973+00:00"} {"global_step": 2489, "acc_step": 0, "speed/wps": 12894.724879511898, "speed/FLOPS": 202529373441559.2, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17120659351348877, "optim/lr": 0.0014928, "optim/total_tokens": 1304952832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 3.4977524280548096, "created_at": "2025-01-15T12:34:34.729719+00:00"} {"global_step": 2490, "acc_step": 0, "speed/wps": 12891.95426264001, "speed/FLOPS": 202485857096359.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16047710180282593, "optim/lr": 0.0014934000000000002, "optim/total_tokens": 1305477120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.516446113586426, "created_at": "2025-01-15T12:34:44.900327+00:00"} {"global_step": 2491, "acc_step": 0, "speed/wps": 12887.264156398753, "speed/FLOPS": 202412192532959.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1576368659734726, "optim/lr": 0.001494, "optim/total_tokens": 1306001408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.472270965576172, "created_at": "2025-01-15T12:34:55.078946+00:00"} {"global_step": 2492, "acc_step": 0, "speed/wps": 12885.510165243413, "speed/FLOPS": 202384643691628.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16446322202682495, "optim/lr": 0.0014946, "optim/total_tokens": 1306525696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446221, "loss/out": 3.361593723297119, "created_at": "2025-01-15T12:35:05.260134+00:00"} {"global_step": 2493, "acc_step": 0, "speed/wps": 12886.972739728279, "speed/FLOPS": 202407615433701.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.19656817615032196, "optim/lr": 0.0014952000000000001, "optim/total_tokens": 1307049984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 3.54327392578125, "created_at": "2025-01-15T12:35:15.436082+00:00"} {"global_step": 2494, "acc_step": 0, "speed/wps": 12889.373907570394, "speed/FLOPS": 202445329074211.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18615403771400452, "optim/lr": 0.0014958, "optim/total_tokens": 1307574272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.4436330795288086, "created_at": "2025-01-15T12:35:25.610933+00:00"} {"global_step": 2495, "acc_step": 0, "speed/wps": 12890.52981707959, "speed/FLOPS": 202463484221439.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17279255390167236, "optim/lr": 0.0014964000000000002, "optim/total_tokens": 1308098560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 3.340552806854248, "created_at": "2025-01-15T12:35:35.782539+00:00"} {"global_step": 2496, "acc_step": 0, "speed/wps": 12890.294934771699, "speed/FLOPS": 202459795072032.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15469177067279816, "optim/lr": 0.001497, "optim/total_tokens": 1308622848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.37847900390625, "created_at": "2025-01-15T12:35:45.959626+00:00"} {"global_step": 2497, "acc_step": 0, "speed/wps": 12891.14742311111, "speed/FLOPS": 202473184572844.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13670603930950165, "optim/lr": 0.0014976, "optim/total_tokens": 1309147136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.4844517707824707, "created_at": "2025-01-15T12:35:56.135144+00:00"} {"global_step": 2498, "acc_step": 0, "speed/wps": 12889.443818746333, "speed/FLOPS": 202446427125298.03, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.14790736138820648, "optim/lr": 0.0014982, "optim/total_tokens": 1309671424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 3.381640911102295, "created_at": "2025-01-15T12:36:06.314471+00:00"} {"global_step": 2499, "acc_step": 0, "speed/wps": 12885.548898040095, "speed/FLOPS": 202385252043424.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16476356983184814, "optim/lr": 0.0014988, "optim/total_tokens": 1310195712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.4335150718688965, "created_at": "2025-01-15T12:36:16.492829+00:00"} {"global_step": 2500, "acc_step": 0, "speed/wps": 12888.78164019432, "speed/FLOPS": 202436026701209.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1628313958644867, "optim/lr": 0.0014994000000000001, "optim/total_tokens": 1310720000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.2167809009552, "created_at": "2025-01-15T12:36:26.668562+00:00"} {"global_step": 2501, "acc_step": 0, "speed/wps": 5401.626620550772, "speed/FLOPS": 84839968688560.73, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16284209489822388, "optim/lr": 0.0015, "optim/total_tokens": 1311244288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.4413256645202637, "created_at": "2025-01-15T12:36:50.939268+00:00"} {"global_step": 2502, "acc_step": 0, "speed/wps": 12932.93986437738, "speed/FLOPS": 203129592291763.75, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15875285863876343, "optim/lr": 0.0015006, "optim/total_tokens": 1311768576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.402259349822998, "created_at": "2025-01-15T12:37:01.083103+00:00"} {"global_step": 2503, "acc_step": 0, "speed/wps": 12906.650685451812, "speed/FLOPS": 202716684611620.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1484057456254959, "optim/lr": 0.0015011999999999998, "optim/total_tokens": 1312292864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 3.4477405548095703, "created_at": "2025-01-15T12:37:11.242294+00:00"} {"global_step": 2504, "acc_step": 0, "speed/wps": 12899.553613001484, "speed/FLOPS": 202605215336388.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15465505421161652, "optim/lr": 0.0015018000000000002, "optim/total_tokens": 1312817152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.506988048553467, "created_at": "2025-01-15T12:37:21.407702+00:00"} {"global_step": 2505, "acc_step": 0, "speed/wps": 12894.313786754034, "speed/FLOPS": 202522916664896.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16477049887180328, "optim/lr": 0.0015024, "optim/total_tokens": 1313341440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.6007046699523926, "created_at": "2025-01-15T12:37:31.576535+00:00"} {"global_step": 2506, "acc_step": 0, "speed/wps": 12892.66608971564, "speed/FLOPS": 202497037318735.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15054337680339813, "optim/lr": 0.001503, "optim/total_tokens": 1313865728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.4356725215911865, "created_at": "2025-01-15T12:37:41.746547+00:00"} {"global_step": 2507, "acc_step": 0, "speed/wps": 12895.54086229872, "speed/FLOPS": 202542189572504.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15957166254520416, "optim/lr": 0.0015036, "optim/total_tokens": 1314390016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 3.322604179382324, "created_at": "2025-01-15T12:37:51.915363+00:00"} {"global_step": 2508, "acc_step": 0, "speed/wps": 12894.601750247011, "speed/FLOPS": 202527439527256.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18792352080345154, "optim/lr": 0.0015041999999999998, "optim/total_tokens": 1314914304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.411890983581543, "created_at": "2025-01-15T12:38:02.083697+00:00"} {"global_step": 2509, "acc_step": 0, "speed/wps": 12892.93495530024, "speed/FLOPS": 202501260222201.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1604275107383728, "optim/lr": 0.0015048000000000001, "optim/total_tokens": 1315438592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.4470019340515137, "created_at": "2025-01-15T12:38:12.256022+00:00"} {"global_step": 2510, "acc_step": 0, "speed/wps": 12891.74744326774, "speed/FLOPS": 202482608713918.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13920174539089203, "optim/lr": 0.0015054, "optim/total_tokens": 1315962880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286322, "loss/out": 3.38748836517334, "created_at": "2025-01-15T12:38:22.427171+00:00"} {"global_step": 2511, "acc_step": 0, "speed/wps": 12890.910181594181, "speed/FLOPS": 202469458368817.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14435772597789764, "optim/lr": 0.001506, "optim/total_tokens": 1316487168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.416994571685791, "created_at": "2025-01-15T12:38:32.599158+00:00"} {"global_step": 2512, "acc_step": 0, "speed/wps": 12891.113875720628, "speed/FLOPS": 202472657664978.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.1591629534959793, "optim/lr": 0.0015065999999999999, "optim/total_tokens": 1317011456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.35233473777771, "created_at": "2025-01-15T12:38:42.770561+00:00"} {"global_step": 2513, "acc_step": 0, "speed/wps": 12895.893798397225, "speed/FLOPS": 202547732918916.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16131818294525146, "optim/lr": 0.0015072, "optim/total_tokens": 1317535744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.468402147293091, "created_at": "2025-01-15T12:38:52.939666+00:00"} {"global_step": 2514, "acc_step": 0, "speed/wps": 12891.8555392005, "speed/FLOPS": 202484306509081.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1690603792667389, "optim/lr": 0.0015078000000000001, "optim/total_tokens": 1318060032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377784, "loss/out": 3.508803367614746, "created_at": "2025-01-15T12:39:03.110272+00:00"} {"global_step": 2515, "acc_step": 0, "speed/wps": 12894.527147248915, "speed/FLOPS": 202526267784656.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17752133309841156, "optim/lr": 0.0015084, "optim/total_tokens": 1318584320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.546180248260498, "created_at": "2025-01-15T12:39:13.281697+00:00"} {"global_step": 2516, "acc_step": 0, "speed/wps": 12887.660731142, "speed/FLOPS": 202418421284252.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17721186578273773, "optim/lr": 0.0015090000000000001, "optim/total_tokens": 1319108608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.371889591217041, "created_at": "2025-01-15T12:39:23.461646+00:00"} {"global_step": 2517, "acc_step": 0, "speed/wps": 12877.61320490627, "speed/FLOPS": 202260611077972.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16176466643810272, "optim/lr": 0.0015096, "optim/total_tokens": 1319632896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.44795298576355, "created_at": "2025-01-15T12:39:33.645274+00:00"} {"global_step": 2518, "acc_step": 0, "speed/wps": 12890.706939032925, "speed/FLOPS": 202466266165108.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1603807657957077, "optim/lr": 0.0015102, "optim/total_tokens": 1320157184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350240, "loss/out": 3.4989380836486816, "created_at": "2025-01-15T12:39:43.817132+00:00"} {"global_step": 2519, "acc_step": 0, "speed/wps": 12889.96588937517, "speed/FLOPS": 202454626961924.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14096753299236298, "optim/lr": 0.0015108, "optim/total_tokens": 1320681472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.3701541423797607, "created_at": "2025-01-15T12:39:53.991939+00:00"} {"global_step": 2520, "acc_step": 0, "speed/wps": 12883.873227026828, "speed/FLOPS": 202358933327548.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15546099841594696, "optim/lr": 0.0015114000000000002, "optim/total_tokens": 1321205760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.4213757514953613, "created_at": "2025-01-15T12:40:04.168921+00:00"} {"global_step": 2521, "acc_step": 0, "speed/wps": 12888.40173057935, "speed/FLOPS": 202430059698656.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16917851567268372, "optim/lr": 0.001512, "optim/total_tokens": 1321730048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.4483590126037598, "created_at": "2025-01-15T12:40:14.343475+00:00"} {"global_step": 2522, "acc_step": 0, "speed/wps": 12891.447553148619, "speed/FLOPS": 202477898527505.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18063944578170776, "optim/lr": 0.0015126, "optim/total_tokens": 1322254336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 3.4308085441589355, "created_at": "2025-01-15T12:40:24.519677+00:00"} {"global_step": 2523, "acc_step": 0, "speed/wps": 12893.256515978279, "speed/FLOPS": 202506310774520.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18359346687793732, "optim/lr": 0.0015132, "optim/total_tokens": 1322778624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 3.4507970809936523, "created_at": "2025-01-15T12:40:34.693248+00:00"} {"global_step": 2524, "acc_step": 0, "speed/wps": 12892.5255935102, "speed/FLOPS": 202494830632766.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17111079394817352, "optim/lr": 0.0015138000000000003, "optim/total_tokens": 1323302912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.3356287479400635, "created_at": "2025-01-15T12:40:44.864778+00:00"} {"global_step": 2525, "acc_step": 0, "speed/wps": 12892.714105118148, "speed/FLOPS": 202497791466611.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20315693318843842, "optim/lr": 0.0015144000000000002, "optim/total_tokens": 1323827200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413426, "loss/out": 3.5016090869903564, "created_at": "2025-01-15T12:40:55.036513+00:00"} {"global_step": 2526, "acc_step": 0, "speed/wps": 12889.634856262606, "speed/FLOPS": 202449427632004.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1886410415172577, "optim/lr": 0.001515, "optim/total_tokens": 1324351488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.4905693531036377, "created_at": "2025-01-15T12:41:05.208812+00:00"} {"global_step": 2527, "acc_step": 0, "speed/wps": 12889.290730041874, "speed/FLOPS": 202444022656828.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1882062405347824, "optim/lr": 0.0015156, "optim/total_tokens": 1324875776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.5184946060180664, "created_at": "2025-01-15T12:41:15.383521+00:00"} {"global_step": 2528, "acc_step": 0, "speed/wps": 12890.319664414577, "speed/FLOPS": 202460183485055.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16712254285812378, "optim/lr": 0.0015161999999999999, "optim/total_tokens": 1325400064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.519516944885254, "created_at": "2025-01-15T12:41:25.555907+00:00"} {"global_step": 2529, "acc_step": 0, "speed/wps": 12892.885949663265, "speed/FLOPS": 202500490521332.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1300145983695984, "optim/lr": 0.0015168000000000002, "optim/total_tokens": 1325924352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.4729397296905518, "created_at": "2025-01-15T12:41:35.730213+00:00"} {"global_step": 2530, "acc_step": 0, "speed/wps": 12891.743084560703, "speed/FLOPS": 202482540254435.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16323938965797424, "optim/lr": 0.0015174000000000001, "optim/total_tokens": 1326448640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.51135516166687, "created_at": "2025-01-15T12:41:45.901408+00:00"} {"global_step": 2531, "acc_step": 0, "speed/wps": 12894.55725439958, "speed/FLOPS": 202526740658829.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17357012629508972, "optim/lr": 0.001518, "optim/total_tokens": 1326972928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.4973745346069336, "created_at": "2025-01-15T12:41:56.070040+00:00"} {"global_step": 2532, "acc_step": 0, "speed/wps": 12887.024428667688, "speed/FLOPS": 202408427279522.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1410602182149887, "optim/lr": 0.0015186, "optim/total_tokens": 1327497216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.3373918533325195, "created_at": "2025-01-15T12:42:06.247105+00:00"} {"global_step": 2533, "acc_step": 0, "speed/wps": 12888.452989551039, "speed/FLOPS": 202430864791243.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17258405685424805, "optim/lr": 0.0015191999999999998, "optim/total_tokens": 1328021504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318539, "loss/out": 3.465785264968872, "created_at": "2025-01-15T12:42:16.423839+00:00"} {"global_step": 2534, "acc_step": 0, "speed/wps": 12890.912493499978, "speed/FLOPS": 202469494680474.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18374194204807281, "optim/lr": 0.0015198000000000002, "optim/total_tokens": 1328545792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.4001080989837646, "created_at": "2025-01-15T12:42:26.595207+00:00"} {"global_step": 2535, "acc_step": 0, "speed/wps": 12887.961929803367, "speed/FLOPS": 202423152023119.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11984092742204666, "optim/lr": 0.0015204, "optim/total_tokens": 1329070080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.4639241695404053, "created_at": "2025-01-15T12:42:36.768974+00:00"} {"global_step": 2536, "acc_step": 0, "speed/wps": 12892.277260097087, "speed/FLOPS": 202490930215268.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1414809226989746, "optim/lr": 0.001521, "optim/total_tokens": 1329594368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 3.448390483856201, "created_at": "2025-01-15T12:42:46.939167+00:00"} {"global_step": 2537, "acc_step": 0, "speed/wps": 12890.804551526417, "speed/FLOPS": 202467799303452.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1441856473684311, "optim/lr": 0.0015216, "optim/total_tokens": 1330118656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.479112148284912, "created_at": "2025-01-15T12:42:57.113411+00:00"} {"global_step": 2538, "acc_step": 0, "speed/wps": 12887.242155013908, "speed/FLOPS": 202411846970977.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15156319737434387, "optim/lr": 0.0015221999999999998, "optim/total_tokens": 1330642944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.4217350482940674, "created_at": "2025-01-15T12:43:07.288610+00:00"} {"global_step": 2539, "acc_step": 0, "speed/wps": 12887.510910072971, "speed/FLOPS": 202416068138488.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1368337869644165, "optim/lr": 0.0015228000000000002, "optim/total_tokens": 1331167232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 3.4182419776916504, "created_at": "2025-01-15T12:43:17.465245+00:00"} {"global_step": 2540, "acc_step": 0, "speed/wps": 12891.400285203314, "speed/FLOPS": 202477156119471.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17567741870880127, "optim/lr": 0.0015234, "optim/total_tokens": 1331691520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 3.4306845664978027, "created_at": "2025-01-15T12:43:27.639180+00:00"} {"global_step": 2541, "acc_step": 0, "speed/wps": 12893.187546166622, "speed/FLOPS": 202505227508854.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23464210331439972, "optim/lr": 0.001524, "optim/total_tokens": 1332215808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 3.538619041442871, "created_at": "2025-01-15T12:43:37.811814+00:00"} {"global_step": 2542, "acc_step": 0, "speed/wps": 12889.875238347837, "speed/FLOPS": 202453203162972.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24147312343120575, "optim/lr": 0.0015246, "optim/total_tokens": 1332740096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.5084478855133057, "created_at": "2025-01-15T12:43:47.986293+00:00"} {"global_step": 2543, "acc_step": 0, "speed/wps": 12866.826212191605, "speed/FLOPS": 202091186534507.7, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27391576766967773, "optim/lr": 0.0015252, "optim/total_tokens": 1333264384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.4490528106689453, "created_at": "2025-01-15T12:43:58.182367+00:00"} {"global_step": 2544, "acc_step": 0, "speed/wps": 12885.990959905463, "speed/FLOPS": 202392195232476.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29207083582878113, "optim/lr": 0.0015258000000000001, "optim/total_tokens": 1333788672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.4418559074401855, "created_at": "2025-01-15T12:44:08.359381+00:00"} {"global_step": 2545, "acc_step": 0, "speed/wps": 12886.702443643768, "speed/FLOPS": 202403370062268.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23566266894340515, "optim/lr": 0.0015264, "optim/total_tokens": 1334312960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.601289749145508, "created_at": "2025-01-15T12:44:18.536568+00:00"} {"global_step": 2546, "acc_step": 0, "speed/wps": 12883.57445323591, "speed/FLOPS": 202354240674603.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22218599915504456, "optim/lr": 0.0015270000000000001, "optim/total_tokens": 1334837248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346379, "loss/out": 3.486168384552002, "created_at": "2025-01-15T12:44:28.717540+00:00"} {"global_step": 2547, "acc_step": 0, "speed/wps": 12884.510740000936, "speed/FLOPS": 202368946344841.0, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19948948919773102, "optim/lr": 0.0015276, "optim/total_tokens": 1335361536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.353914976119995, "created_at": "2025-01-15T12:44:38.897476+00:00"} {"global_step": 2548, "acc_step": 0, "speed/wps": 12890.142541297457, "speed/FLOPS": 202457401523108.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.21073894202709198, "optim/lr": 0.0015282, "optim/total_tokens": 1335885824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.444375991821289, "created_at": "2025-01-15T12:44:49.070705+00:00"} {"global_step": 2549, "acc_step": 0, "speed/wps": 12881.796084133643, "speed/FLOPS": 202326308944117.1, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.274056077003479, "optim/lr": 0.0015288000000000003, "optim/total_tokens": 1336410112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.3546524047851562, "created_at": "2025-01-15T12:44:59.249639+00:00"} {"global_step": 2550, "acc_step": 0, "speed/wps": 12877.213356540746, "speed/FLOPS": 202254330909943.03, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21027448773384094, "optim/lr": 0.0015294000000000002, "optim/total_tokens": 1336934400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391766, "loss/out": 3.5004220008850098, "created_at": "2025-01-15T12:45:09.432093+00:00"} {"global_step": 2551, "acc_step": 0, "speed/wps": 12886.487999242467, "speed/FLOPS": 202400001918268.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12736651301383972, "optim/lr": 0.0015300000000000001, "optim/total_tokens": 1337458688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 3.4060022830963135, "created_at": "2025-01-15T12:45:19.607184+00:00"} {"global_step": 2552, "acc_step": 0, "speed/wps": 12883.791641960266, "speed/FLOPS": 202357651921968.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1367354840040207, "optim/lr": 0.0015306, "optim/total_tokens": 1337982976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.37544584274292, "created_at": "2025-01-15T12:45:29.784257+00:00"} {"global_step": 2553, "acc_step": 0, "speed/wps": 12887.64558625361, "speed/FLOPS": 202418183412969.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13951192796230316, "optim/lr": 0.0015312, "optim/total_tokens": 1338507264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.420286178588867, "created_at": "2025-01-15T12:45:39.960445+00:00"} {"global_step": 2554, "acc_step": 0, "speed/wps": 12889.399460009588, "speed/FLOPS": 202445730410382.16, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1263759732246399, "optim/lr": 0.0015318000000000003, "optim/total_tokens": 1339031552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.461374521255493, "created_at": "2025-01-15T12:45:50.133367+00:00"} {"global_step": 2555, "acc_step": 0, "speed/wps": 12886.6386083456, "speed/FLOPS": 202402367441191.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.123467817902565, "optim/lr": 0.0015324000000000002, "optim/total_tokens": 1339555840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.484105348587036, "created_at": "2025-01-15T12:46:00.313500+00:00"} {"global_step": 2556, "acc_step": 0, "speed/wps": 12889.585607734702, "speed/FLOPS": 202448654116198.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14790984988212585, "optim/lr": 0.001533, "optim/total_tokens": 1340080128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.433499336242676, "created_at": "2025-01-15T12:46:10.493007+00:00"} {"global_step": 2557, "acc_step": 0, "speed/wps": 12883.080769702909, "speed/FLOPS": 202346486696325.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21318210661411285, "optim/lr": 0.0015336, "optim/total_tokens": 1340604416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.405229091644287, "created_at": "2025-01-15T12:46:20.670674+00:00"} {"global_step": 2558, "acc_step": 0, "speed/wps": 12887.99732553001, "speed/FLOPS": 202423707961645.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14693017303943634, "optim/lr": 0.0015341999999999999, "optim/total_tokens": 1341128704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.3407387733459473, "created_at": "2025-01-15T12:46:30.848211+00:00"} {"global_step": 2559, "acc_step": 0, "speed/wps": 12891.264504411427, "speed/FLOPS": 202475023495552.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1585073322057724, "optim/lr": 0.0015348000000000002, "optim/total_tokens": 1341652992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 3.332972764968872, "created_at": "2025-01-15T12:46:41.020827+00:00"} {"global_step": 2560, "acc_step": 0, "speed/wps": 12889.040084872464, "speed/FLOPS": 202440085930020.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17643482983112335, "optim/lr": 0.0015354000000000001, "optim/total_tokens": 1342177280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 3.5073795318603516, "created_at": "2025-01-15T12:46:51.193748+00:00"} {"global_step": 2561, "acc_step": 0, "speed/wps": 12887.7629060768, "speed/FLOPS": 202420026082006.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17193201184272766, "optim/lr": 0.001536, "optim/total_tokens": 1342701568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.4837305545806885, "created_at": "2025-01-15T12:47:01.373972+00:00"} {"global_step": 2562, "acc_step": 0, "speed/wps": 12878.093663778636, "speed/FLOPS": 202268157344783.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17048007249832153, "optim/lr": 0.0015366, "optim/total_tokens": 1343225856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.490288257598877, "created_at": "2025-01-15T12:47:11.555876+00:00"} {"global_step": 2563, "acc_step": 0, "speed/wps": 12881.951707984597, "speed/FLOPS": 202328753230545.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16985473036766052, "optim/lr": 0.0015371999999999999, "optim/total_tokens": 1343750144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.468733787536621, "created_at": "2025-01-15T12:47:21.736596+00:00"} {"global_step": 2564, "acc_step": 0, "speed/wps": 12887.55873484746, "speed/FLOPS": 202416819292289.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14459308981895447, "optim/lr": 0.0015377999999999998, "optim/total_tokens": 1344274432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 3.495492696762085, "created_at": "2025-01-15T12:47:31.911953+00:00"} {"global_step": 2565, "acc_step": 0, "speed/wps": 12882.304636429453, "speed/FLOPS": 202334296456746.88, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1787235289812088, "optim/lr": 0.0015384, "optim/total_tokens": 1344798720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 3.4066224098205566, "created_at": "2025-01-15T12:47:42.092579+00:00"} {"global_step": 2566, "acc_step": 0, "speed/wps": 12869.152202606594, "speed/FLOPS": 202127719410220.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14636574685573578, "optim/lr": 0.001539, "optim/total_tokens": 1345323008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 3.489168405532837, "created_at": "2025-01-15T12:47:52.285078+00:00"} {"global_step": 2567, "acc_step": 0, "speed/wps": 12879.098346104858, "speed/FLOPS": 202283937261294.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15905466675758362, "optim/lr": 0.0015396, "optim/total_tokens": 1345847296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.386751174926758, "created_at": "2025-01-15T12:48:02.473396+00:00"} {"global_step": 2568, "acc_step": 0, "speed/wps": 12876.87608104013, "speed/FLOPS": 202249033534740.75, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19042018055915833, "optim/lr": 0.0015402, "optim/total_tokens": 1346371584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.4194419384002686, "created_at": "2025-01-15T12:48:12.658309+00:00"} {"global_step": 2569, "acc_step": 0, "speed/wps": 12882.904365528224, "speed/FLOPS": 202343716026357.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1490769237279892, "optim/lr": 0.0015408, "optim/total_tokens": 1346895872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 3.476221799850464, "created_at": "2025-01-15T12:48:22.837859+00:00"} {"global_step": 2570, "acc_step": 0, "speed/wps": 12873.800377533797, "speed/FLOPS": 202200725384714.25, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0026, "optim/grad_norm": 0.33596089482307434, "optim/lr": 0.0015414, "optim/total_tokens": 1347420160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348373, "loss/out": 3.4351420402526855, "created_at": "2025-01-15T12:48:33.024830+00:00"} {"global_step": 2571, "acc_step": 0, "speed/wps": 12877.058733169477, "speed/FLOPS": 202251902337422.47, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1773034632205963, "optim/lr": 0.001542, "optim/total_tokens": 1347944448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 3.3927011489868164, "created_at": "2025-01-15T12:48:43.210310+00:00"} {"global_step": 2572, "acc_step": 0, "speed/wps": 12884.773125595764, "speed/FLOPS": 202373067471163.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1584506332874298, "optim/lr": 0.0015426, "optim/total_tokens": 1348468736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.353590965270996, "created_at": "2025-01-15T12:48:53.387060+00:00"} {"global_step": 2573, "acc_step": 0, "speed/wps": 12886.544712053405, "speed/FLOPS": 202400892670895.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14543746411800385, "optim/lr": 0.0015432, "optim/total_tokens": 1348993024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.4127566814422607, "created_at": "2025-01-15T12:49:03.563654+00:00"} {"global_step": 2574, "acc_step": 0, "speed/wps": 12886.27147945637, "speed/FLOPS": 202396601177498.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20516689121723175, "optim/lr": 0.0015438, "optim/total_tokens": 1349517312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.4339470863342285, "created_at": "2025-01-15T12:49:13.743328+00:00"} {"global_step": 2575, "acc_step": 0, "speed/wps": 12887.184775226997, "speed/FLOPS": 202410945742575.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3511841297149658, "optim/lr": 0.0015444000000000002, "optim/total_tokens": 1350041600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 3.627124309539795, "created_at": "2025-01-15T12:49:23.921997+00:00"} {"global_step": 2576, "acc_step": 0, "speed/wps": 12886.84038923514, "speed/FLOPS": 202405536687337.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23954454064369202, "optim/lr": 0.0015450000000000001, "optim/total_tokens": 1350565888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.393099308013916, "created_at": "2025-01-15T12:49:34.128362+00:00"} {"global_step": 2577, "acc_step": 0, "speed/wps": 12890.150096011117, "speed/FLOPS": 202457520180268.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2904678285121918, "optim/lr": 0.0015456, "optim/total_tokens": 1351090176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.384977340698242, "created_at": "2025-01-15T12:49:44.300438+00:00"} {"global_step": 2578, "acc_step": 0, "speed/wps": 12887.477713220444, "speed/FLOPS": 202415546736301.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45677074790000916, "optim/lr": 0.0015462, "optim/total_tokens": 1351614464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.5700924396514893, "created_at": "2025-01-15T12:49:54.477034+00:00"} {"global_step": 2579, "acc_step": 0, "speed/wps": 12876.448561213972, "speed/FLOPS": 202242318748397.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3957894444465637, "optim/lr": 0.0015467999999999999, "optim/total_tokens": 1352138752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.4942657947540283, "created_at": "2025-01-15T12:50:04.660524+00:00"} {"global_step": 2580, "acc_step": 0, "speed/wps": 12888.5691201487, "speed/FLOPS": 202432688781860.38, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.38605356216430664, "optim/lr": 0.0015474000000000002, "optim/total_tokens": 1352663040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.483031749725342, "created_at": "2025-01-15T12:50:14.838200+00:00"} {"global_step": 2581, "acc_step": 0, "speed/wps": 12891.19660564167, "speed/FLOPS": 202473957052070.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.32378262281417847, "optim/lr": 0.0015480000000000001, "optim/total_tokens": 1353187328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402750, "loss/out": 3.646463394165039, "created_at": "2025-01-15T12:50:25.010852+00:00"} {"global_step": 2582, "acc_step": 0, "speed/wps": 12890.30343299309, "speed/FLOPS": 202459928548277.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.32941797375679016, "optim/lr": 0.0015486, "optim/total_tokens": 1353711616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.577141284942627, "created_at": "2025-01-15T12:50:35.188100+00:00"} {"global_step": 2583, "acc_step": 0, "speed/wps": 12888.67012796917, "speed/FLOPS": 202434275248480.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3007736802101135, "optim/lr": 0.0015492, "optim/total_tokens": 1354235904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.479433059692383, "created_at": "2025-01-15T12:50:45.364245+00:00"} {"global_step": 2584, "acc_step": 0, "speed/wps": 12889.72964993647, "speed/FLOPS": 202450916496912.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29504111409187317, "optim/lr": 0.0015497999999999998, "optim/total_tokens": 1354760192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.5880074501037598, "created_at": "2025-01-15T12:50:55.542846+00:00"} {"global_step": 2585, "acc_step": 0, "speed/wps": 12891.757327423315, "speed/FLOPS": 202482763958164.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22362764179706573, "optim/lr": 0.0015504000000000002, "optim/total_tokens": 1355284480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 3.443202495574951, "created_at": "2025-01-15T12:51:05.716375+00:00"} {"global_step": 2586, "acc_step": 0, "speed/wps": 12886.632794655621, "speed/FLOPS": 202402276129201.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3485119044780731, "optim/lr": 0.001551, "optim/total_tokens": 1355808768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.5167388916015625, "created_at": "2025-01-15T12:51:15.891087+00:00"} {"global_step": 2587, "acc_step": 0, "speed/wps": 12881.210232130892, "speed/FLOPS": 202317107333368.88, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3789674639701843, "optim/lr": 0.0015516, "optim/total_tokens": 1356333056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.5252628326416016, "created_at": "2025-01-15T12:51:26.072508+00:00"} {"global_step": 2588, "acc_step": 0, "speed/wps": 12883.956134138332, "speed/FLOPS": 202360235497660.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2717567980289459, "optim/lr": 0.0015521999999999999, "optim/total_tokens": 1356857344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.5429494380950928, "created_at": "2025-01-15T12:51:36.249409+00:00"} {"global_step": 2589, "acc_step": 0, "speed/wps": 12887.707616439426, "speed/FLOPS": 202419157682276.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.24356913566589355, "optim/lr": 0.0015527999999999998, "optim/total_tokens": 1357381632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 3.4801130294799805, "created_at": "2025-01-15T12:51:46.423192+00:00"} {"global_step": 2590, "acc_step": 0, "speed/wps": 12886.981764582739, "speed/FLOPS": 202407757181441.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24148204922676086, "optim/lr": 0.0015534000000000001, "optim/total_tokens": 1357905920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.504049777984619, "created_at": "2025-01-15T12:51:56.600216+00:00"} {"global_step": 2591, "acc_step": 0, "speed/wps": 12888.589614070255, "speed/FLOPS": 202433010667060.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.17649458348751068, "optim/lr": 0.001554, "optim/total_tokens": 1358430208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.435403347015381, "created_at": "2025-01-15T12:52:06.777985+00:00"} {"global_step": 2592, "acc_step": 0, "speed/wps": 12887.199516628909, "speed/FLOPS": 202411177276549.6, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.7869341373443604, "optim/lr": 0.0015546, "optim/total_tokens": 1358954496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.3645009994506836, "created_at": "2025-01-15T12:52:16.952520+00:00"} {"global_step": 2593, "acc_step": 0, "speed/wps": 12887.685794104531, "speed/FLOPS": 202418814932519.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1571306735277176, "optim/lr": 0.0015551999999999999, "optim/total_tokens": 1359478784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.5135984420776367, "created_at": "2025-01-15T12:52:27.128372+00:00"} {"global_step": 2594, "acc_step": 0, "speed/wps": 12887.684905445765, "speed/FLOPS": 202418800974912.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18023476004600525, "optim/lr": 0.0015558, "optim/total_tokens": 1360003072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 3.430551290512085, "created_at": "2025-01-15T12:52:37.302168+00:00"} {"global_step": 2595, "acc_step": 0, "speed/wps": 12886.620025087479, "speed/FLOPS": 202402075565586.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1505631059408188, "optim/lr": 0.0015564, "optim/total_tokens": 1360527360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.423335552215576, "created_at": "2025-01-15T12:52:47.481638+00:00"} {"global_step": 2596, "acc_step": 0, "speed/wps": 12887.848021818423, "speed/FLOPS": 202421362941690.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1523534655570984, "optim/lr": 0.001557, "optim/total_tokens": 1361051648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.412179470062256, "created_at": "2025-01-15T12:52:57.658094+00:00"} {"global_step": 2597, "acc_step": 0, "speed/wps": 12887.234170620886, "speed/FLOPS": 202411721565113.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1222093254327774, "optim/lr": 0.0015576, "optim/total_tokens": 1361575936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 3.445570945739746, "created_at": "2025-01-15T12:53:07.832375+00:00"} {"global_step": 2598, "acc_step": 0, "speed/wps": 12887.53884253047, "speed/FLOPS": 202416506856116.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14194123446941376, "optim/lr": 0.0015582, "optim/total_tokens": 1362100224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 3.4712815284729004, "created_at": "2025-01-15T12:53:18.006893+00:00"} {"global_step": 2599, "acc_step": 0, "speed/wps": 12888.896837132417, "speed/FLOPS": 202437836027419.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13204975426197052, "optim/lr": 0.0015588, "optim/total_tokens": 1362624512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.3592371940612793, "created_at": "2025-01-15T12:53:28.182045+00:00"} {"global_step": 2600, "acc_step": 0, "speed/wps": 12891.063753923132, "speed/FLOPS": 202471870433273.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13631822168827057, "optim/lr": 0.0015594, "optim/total_tokens": 1363148800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 3.339153528213501, "created_at": "2025-01-15T12:53:38.355256+00:00"} {"global_step": 2601, "acc_step": 0, "speed/wps": 12889.189980452311, "speed/FLOPS": 202442440246080.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.180971145629883, "optim/lr": 0.0015600000000000002, "optim/total_tokens": 1363673088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 3.7752647399902344, "created_at": "2025-01-15T12:53:48.531770+00:00"} {"global_step": 2602, "acc_step": 0, "speed/wps": 12886.618362520361, "speed/FLOPS": 202402049452685.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6446548700332642, "optim/lr": 0.0015606, "optim/total_tokens": 1364197376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.5431301593780518, "created_at": "2025-01-15T12:53:58.706562+00:00"} {"global_step": 2603, "acc_step": 0, "speed/wps": 12888.365253853286, "speed/FLOPS": 202429486781549.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1969655603170395, "optim/lr": 0.0015612, "optim/total_tokens": 1364721664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.4000070095062256, "created_at": "2025-01-15T12:54:08.879920+00:00"} {"global_step": 2604, "acc_step": 0, "speed/wps": 12888.247767268893, "speed/FLOPS": 202427641493297.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16778255999088287, "optim/lr": 0.0015618, "optim/total_tokens": 1365245952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.648937225341797, "created_at": "2025-01-15T12:54:19.057665+00:00"} {"global_step": 2605, "acc_step": 0, "speed/wps": 12880.522661631752, "speed/FLOPS": 202306308093859.0, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.003981590270996, "optim/lr": 0.0015624000000000002, "optim/total_tokens": 1365770240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.5422167778015137, "created_at": "2025-01-15T12:54:29.238640+00:00"} {"global_step": 2606, "acc_step": 0, "speed/wps": 12896.22030092796, "speed/FLOPS": 202552861089822.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 101.1973648071289, "optim/lr": 0.0015630000000000002, "optim/total_tokens": 1366294528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 6.082343101501465, "created_at": "2025-01-15T12:54:39.407917+00:00"} {"global_step": 2607, "acc_step": 0, "speed/wps": 12952.731437905266, "speed/FLOPS": 203440446150491.22, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.299631953239441, "optim/lr": 0.0015636, "optim/total_tokens": 1366818816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 10.12051010131836, "created_at": "2025-01-15T12:54:49.538749+00:00"} {"global_step": 2608, "acc_step": 0, "speed/wps": 12925.922054308014, "speed/FLOPS": 203019367941142.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.579559564590454, "optim/lr": 0.0015642, "optim/total_tokens": 1367343104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 9.862543106079102, "created_at": "2025-01-15T12:54:59.682483+00:00"} {"global_step": 2609, "acc_step": 0, "speed/wps": 12896.781049737487, "speed/FLOPS": 202561668420426.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.597702503204346, "optim/lr": 0.0015647999999999999, "optim/total_tokens": 1367867392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 9.012588500976562, "created_at": "2025-01-15T12:55:09.859011+00:00"} {"global_step": 2610, "acc_step": 0, "speed/wps": 12870.367539988723, "speed/FLOPS": 202146807953855.78, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.588657855987549, "optim/lr": 0.0015654000000000002, "optim/total_tokens": 1368391680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 9.14312744140625, "created_at": "2025-01-15T12:55:20.046565+00:00"} {"global_step": 2611, "acc_step": 0, "speed/wps": 12893.695155546833, "speed/FLOPS": 202513200211698.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 11.761312484741211, "optim/lr": 0.0015660000000000001, "optim/total_tokens": 1368915968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 6.4028425216674805, "created_at": "2025-01-15T12:55:30.215789+00:00"} {"global_step": 2612, "acc_step": 0, "speed/wps": 12888.127060774188, "speed/FLOPS": 202425745631930.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.29885196685791, "optim/lr": 0.0015666, "optim/total_tokens": 1369440256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314573, "loss/out": 6.674660682678223, "created_at": "2025-01-15T12:55:40.390367+00:00"} {"global_step": 2613, "acc_step": 0, "speed/wps": 12883.085768819297, "speed/FLOPS": 202346565214317.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8995473384857178, "optim/lr": 0.0015672, "optim/total_tokens": 1369964544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 6.638798713684082, "created_at": "2025-01-15T12:55:50.568324+00:00"} {"global_step": 2614, "acc_step": 0, "speed/wps": 12890.366848874206, "speed/FLOPS": 202460924581832.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.290830373764038, "optim/lr": 0.0015677999999999998, "optim/total_tokens": 1370488832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.588249683380127, "created_at": "2025-01-15T12:56:00.740042+00:00"} {"global_step": 2615, "acc_step": 0, "speed/wps": 12887.986499428196, "speed/FLOPS": 202423537922840.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.0630364418029785, "optim/lr": 0.0015684000000000002, "optim/total_tokens": 1371013120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 7.428754806518555, "created_at": "2025-01-15T12:56:10.913863+00:00"} {"global_step": 2616, "acc_step": 0, "speed/wps": 12881.36486443539, "speed/FLOPS": 202319536046198.1, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.240962028503418, "optim/lr": 0.001569, "optim/total_tokens": 1371537408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 7.077343463897705, "created_at": "2025-01-15T12:56:21.092897+00:00"} {"global_step": 2617, "acc_step": 0, "speed/wps": 12885.979605063865, "speed/FLOPS": 202392016889086.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.3272473812103271, "optim/lr": 0.0015696, "optim/total_tokens": 1372061696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 7.019208908081055, "created_at": "2025-01-15T12:56:31.268064+00:00"} {"global_step": 2618, "acc_step": 0, "speed/wps": 12884.330171799731, "speed/FLOPS": 202366110273114.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4934558868408203, "optim/lr": 0.0015702, "optim/total_tokens": 1372585984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 8.209552764892578, "created_at": "2025-01-15T12:56:41.445278+00:00"} {"global_step": 2619, "acc_step": 0, "speed/wps": 12879.02456589169, "speed/FLOPS": 202282778441662.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2557523250579834, "optim/lr": 0.0015707999999999998, "optim/total_tokens": 1373110272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 7.005101203918457, "created_at": "2025-01-15T12:56:51.629033+00:00"} {"global_step": 2620, "acc_step": 0, "speed/wps": 12880.81465196948, "speed/FLOPS": 202310894203349.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2327810525894165, "optim/lr": 0.0015714000000000001, "optim/total_tokens": 1373634560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383767, "loss/out": 6.592592239379883, "created_at": "2025-01-15T12:57:01.808276+00:00"} {"global_step": 2621, "acc_step": 0, "speed/wps": 12883.065738176978, "speed/FLOPS": 202346250605555.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3769358396530151, "optim/lr": 0.001572, "optim/total_tokens": 1374158848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 6.859285354614258, "created_at": "2025-01-15T12:57:11.988008+00:00"} {"global_step": 2622, "acc_step": 0, "speed/wps": 12882.263308925063, "speed/FLOPS": 202333647351500.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2485437393188477, "optim/lr": 0.0015726, "optim/total_tokens": 1374683136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 6.342504024505615, "created_at": "2025-01-15T12:57:22.166201+00:00"} {"global_step": 2623, "acc_step": 0, "speed/wps": 12879.24715721807, "speed/FLOPS": 202286274544315.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1566582918167114, "optim/lr": 0.0015731999999999999, "optim/total_tokens": 1375207424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 6.084018707275391, "created_at": "2025-01-15T12:57:32.349674+00:00"} {"global_step": 2624, "acc_step": 0, "speed/wps": 12880.953850879689, "speed/FLOPS": 202313080513513.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1420056819915771, "optim/lr": 0.0015738, "optim/total_tokens": 1375731712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408639, "loss/out": 5.869872093200684, "created_at": "2025-01-15T12:57:42.528835+00:00"} {"global_step": 2625, "acc_step": 0, "speed/wps": 12877.186284163214, "speed/FLOPS": 202253905701052.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1628363132476807, "optim/lr": 0.0015744, "optim/total_tokens": 1376256000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311795, "loss/out": 5.703019618988037, "created_at": "2025-01-15T12:57:52.715745+00:00"} {"global_step": 2626, "acc_step": 0, "speed/wps": 12891.715646384528, "speed/FLOPS": 202482109300174.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2926480770111084, "optim/lr": 0.001575, "optim/total_tokens": 1376780288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 5.709079265594482, "created_at": "2025-01-15T12:58:02.891051+00:00"} {"global_step": 2627, "acc_step": 0, "speed/wps": 12887.493506781675, "speed/FLOPS": 202415794795883.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.942266583442688, "optim/lr": 0.0015756000000000001, "optim/total_tokens": 1377304576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.447352409362793, "created_at": "2025-01-15T12:58:13.065297+00:00"} {"global_step": 2628, "acc_step": 0, "speed/wps": 12884.458917880276, "speed/FLOPS": 202368132407225.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1439311504364014, "optim/lr": 0.0015762, "optim/total_tokens": 1377828864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 5.282017707824707, "created_at": "2025-01-15T12:58:23.249727+00:00"} {"global_step": 2629, "acc_step": 0, "speed/wps": 12888.557287455717, "speed/FLOPS": 202432502933157.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8603079319000244, "optim/lr": 0.0015768, "optim/total_tokens": 1378353152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 5.047414779663086, "created_at": "2025-01-15T12:58:33.422854+00:00"} {"global_step": 2630, "acc_step": 0, "speed/wps": 12887.516383498638, "speed/FLOPS": 202416154106159.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9852790832519531, "optim/lr": 0.0015774, "optim/total_tokens": 1378877440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.9771647453308105, "created_at": "2025-01-15T12:58:43.597977+00:00"} {"global_step": 2631, "acc_step": 0, "speed/wps": 12887.566733804291, "speed/FLOPS": 202416944926898.25, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1576879024505615, "optim/lr": 0.0015780000000000002, "optim/total_tokens": 1379401728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 5.134726047515869, "created_at": "2025-01-15T12:58:53.775141+00:00"} {"global_step": 2632, "acc_step": 0, "speed/wps": 12887.37089557163, "speed/FLOPS": 202413869018344.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9456276893615723, "optim/lr": 0.0015786, "optim/total_tokens": 1379926016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 4.902555465698242, "created_at": "2025-01-15T12:59:03.951388+00:00"} {"global_step": 2633, "acc_step": 0, "speed/wps": 12887.271659355472, "speed/FLOPS": 202412310377204.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5788230895996094, "optim/lr": 0.0015792, "optim/total_tokens": 1380450304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 4.639863014221191, "created_at": "2025-01-15T12:59:14.125524+00:00"} {"global_step": 2634, "acc_step": 0, "speed/wps": 12888.960367836957, "speed/FLOPS": 202438833864435.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6121317744255066, "optim/lr": 0.0015798, "optim/total_tokens": 1380974592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.501555442810059, "created_at": "2025-01-15T12:59:24.298311+00:00"} {"global_step": 2635, "acc_step": 0, "speed/wps": 12886.731564379594, "speed/FLOPS": 202403827443440.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5778740048408508, "optim/lr": 0.0015804000000000003, "optim/total_tokens": 1381498880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 4.451676845550537, "created_at": "2025-01-15T12:59:34.473024+00:00"} {"global_step": 2636, "acc_step": 0, "speed/wps": 12888.974372013927, "speed/FLOPS": 202439053819278.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5206647515296936, "optim/lr": 0.0015810000000000002, "optim/total_tokens": 1382023168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.259506702423096, "created_at": "2025-01-15T12:59:44.647017+00:00"} {"global_step": 2637, "acc_step": 0, "speed/wps": 12892.045377504173, "speed/FLOPS": 202487288180506.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.402278333902359, "optim/lr": 0.0015816, "optim/total_tokens": 1382547456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 4.132748603820801, "created_at": "2025-01-15T12:59:54.820540+00:00"} {"global_step": 2638, "acc_step": 0, "speed/wps": 12889.982922739962, "speed/FLOPS": 202454894494324.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3335627317428589, "optim/lr": 0.0015822, "optim/total_tokens": 1383071744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.019270896911621, "created_at": "2025-01-15T13:00:04.992787+00:00"} {"global_step": 2639, "acc_step": 0, "speed/wps": 12888.556474217066, "speed/FLOPS": 202432490160126.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2603156268596649, "optim/lr": 0.0015827999999999999, "optim/total_tokens": 1383596032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.946220874786377, "created_at": "2025-01-15T13:00:15.165931+00:00"} {"global_step": 2640, "acc_step": 0, "speed/wps": 12890.71905786383, "speed/FLOPS": 202466456508000.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25114214420318604, "optim/lr": 0.0015834000000000002, "optim/total_tokens": 1384120320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.9752755165100098, "created_at": "2025-01-15T13:00:25.342104+00:00"} {"global_step": 2641, "acc_step": 0, "speed/wps": 12881.827160024719, "speed/FLOPS": 202326797033689.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23870187997817993, "optim/lr": 0.0015840000000000001, "optim/total_tokens": 1384644608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280160, "loss/out": 3.787724018096924, "created_at": "2025-01-15T13:00:35.521595+00:00"} {"global_step": 2642, "acc_step": 0, "speed/wps": 12885.474615652047, "speed/FLOPS": 202384085336445.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24844306707382202, "optim/lr": 0.0015846, "optim/total_tokens": 1385168896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.8043107986450195, "created_at": "2025-01-15T13:00:45.698318+00:00"} {"global_step": 2643, "acc_step": 0, "speed/wps": 12893.94538169653, "speed/FLOPS": 202517130357226.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3461713492870331, "optim/lr": 0.0015852, "optim/total_tokens": 1385693184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.7752904891967773, "created_at": "2025-01-15T13:00:55.869488+00:00"} {"global_step": 2644, "acc_step": 0, "speed/wps": 12885.451149230408, "speed/FLOPS": 202383716764047.56, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.42303064465522766, "optim/lr": 0.0015857999999999998, "optim/total_tokens": 1386217472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.8217720985412598, "created_at": "2025-01-15T13:01:06.045387+00:00"} {"global_step": 2645, "acc_step": 0, "speed/wps": 12884.491885679672, "speed/FLOPS": 202368650211817.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21023719012737274, "optim/lr": 0.0015864000000000002, "optim/total_tokens": 1386741760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460613, "loss/out": 3.763821601867676, "created_at": "2025-01-15T13:01:16.224041+00:00"} {"global_step": 2646, "acc_step": 0, "speed/wps": 12868.895884561549, "speed/FLOPS": 202123693583105.53, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2060829997062683, "optim/lr": 0.001587, "optim/total_tokens": 1387266048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 3.6597540378570557, "created_at": "2025-01-15T13:01:26.412966+00:00"} {"global_step": 2647, "acc_step": 0, "speed/wps": 12881.30300847431, "speed/FLOPS": 202318564513330.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17758123576641083, "optim/lr": 0.0015876, "optim/total_tokens": 1387790336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.6542704105377197, "created_at": "2025-01-15T13:01:36.596102+00:00"} {"global_step": 2648, "acc_step": 0, "speed/wps": 12872.149076959678, "speed/FLOPS": 202174789440072.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19619245827198029, "optim/lr": 0.0015882, "optim/total_tokens": 1388314624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.627959966659546, "created_at": "2025-01-15T13:01:46.782644+00:00"} {"global_step": 2649, "acc_step": 0, "speed/wps": 12888.517365703325, "speed/FLOPS": 202431875907177.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1698397994041443, "optim/lr": 0.0015887999999999998, "optim/total_tokens": 1388838912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.5951004028320312, "created_at": "2025-01-15T13:01:56.960444+00:00"} {"global_step": 2650, "acc_step": 0, "speed/wps": 12890.775212837569, "speed/FLOPS": 202467338499028.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16018979251384735, "optim/lr": 0.0015894000000000001, "optim/total_tokens": 1389363200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 3.7317960262298584, "created_at": "2025-01-15T13:02:07.132252+00:00"} {"global_step": 2651, "acc_step": 0, "speed/wps": 12889.051171253726, "speed/FLOPS": 202440260056872.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14831691980361938, "optim/lr": 0.00159, "optim/total_tokens": 1389887488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 3.6401596069335938, "created_at": "2025-01-15T13:02:17.305119+00:00"} {"global_step": 2652, "acc_step": 0, "speed/wps": 12877.623853336703, "speed/FLOPS": 202260778326205.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.14927473664283752, "optim/lr": 0.0015906, "optim/total_tokens": 1390411776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.6011548042297363, "created_at": "2025-01-15T13:02:27.492428+00:00"} {"global_step": 2653, "acc_step": 0, "speed/wps": 12895.84804935144, "speed/FLOPS": 202547014367286.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14450591802597046, "optim/lr": 0.0015912, "optim/total_tokens": 1390936064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.6624956130981445, "created_at": "2025-01-15T13:02:37.662020+00:00"} {"global_step": 2654, "acc_step": 0, "speed/wps": 12890.748346706878, "speed/FLOPS": 202466916529527.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14467960596084595, "optim/lr": 0.0015918, "optim/total_tokens": 1391460352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.679763078689575, "created_at": "2025-01-15T13:02:47.835920+00:00"} {"global_step": 2655, "acc_step": 0, "speed/wps": 12897.119862801506, "speed/FLOPS": 202566989945174.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14516843855381012, "optim/lr": 0.0015924000000000001, "optim/total_tokens": 1391984640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.53340482711792, "created_at": "2025-01-15T13:02:58.003490+00:00"} {"global_step": 2656, "acc_step": 0, "speed/wps": 12900.116064740732, "speed/FLOPS": 202614049413842.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13102881610393524, "optim/lr": 0.001593, "optim/total_tokens": 1392508928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.573317050933838, "created_at": "2025-01-15T13:03:08.167609+00:00"} {"global_step": 2657, "acc_step": 0, "speed/wps": 12891.240559923448, "speed/FLOPS": 202474647414465.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12914203107357025, "optim/lr": 0.0015936000000000001, "optim/total_tokens": 1393033216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410217, "loss/out": 3.5336124897003174, "created_at": "2025-01-15T13:03:18.338749+00:00"} {"global_step": 2658, "acc_step": 0, "speed/wps": 12886.588392727175, "speed/FLOPS": 202401578735900.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13035501539707184, "optim/lr": 0.0015942, "optim/total_tokens": 1393557504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 3.54543399810791, "created_at": "2025-01-15T13:03:28.518657+00:00"} {"global_step": 2659, "acc_step": 0, "speed/wps": 12896.16888916284, "speed/FLOPS": 202552053597404.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1271541714668274, "optim/lr": 0.0015948, "optim/total_tokens": 1394081792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.504915714263916, "created_at": "2025-01-15T13:03:38.685848+00:00"} {"global_step": 2660, "acc_step": 0, "speed/wps": 12890.616898627568, "speed/FLOPS": 202464851956813.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.14355017244815826, "optim/lr": 0.0015954000000000003, "optim/total_tokens": 1394606080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.473155975341797, "created_at": "2025-01-15T13:03:48.862436+00:00"} {"global_step": 2661, "acc_step": 0, "speed/wps": 12891.084115265969, "speed/FLOPS": 202472190236142.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0014513731002808, "optim/lr": 0.0015960000000000002, "optim/total_tokens": 1395130368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 3.590616226196289, "created_at": "2025-01-15T13:03:59.036160+00:00"} {"global_step": 2662, "acc_step": 0, "speed/wps": 12895.757705510869, "speed/FLOPS": 202545595393124.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16478651762008667, "optim/lr": 0.0015966, "optim/total_tokens": 1395654656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.5600671768188477, "created_at": "2025-01-15T13:04:09.203738+00:00"} {"global_step": 2663, "acc_step": 0, "speed/wps": 12895.506720147885, "speed/FLOPS": 202541653323109.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15515339374542236, "optim/lr": 0.0015972, "optim/total_tokens": 1396178944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.5019121170043945, "created_at": "2025-01-15T13:04:19.373590+00:00"} {"global_step": 2664, "acc_step": 0, "speed/wps": 12894.763530677123, "speed/FLOPS": 202529980511221.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13443408906459808, "optim/lr": 0.0015978, "optim/total_tokens": 1396703232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 3.4383363723754883, "created_at": "2025-01-15T13:04:29.542249+00:00"} {"global_step": 2665, "acc_step": 0, "speed/wps": 12889.273600529179, "speed/FLOPS": 202443753614293.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14782249927520752, "optim/lr": 0.0015984000000000003, "optim/total_tokens": 1397227520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.4637670516967773, "created_at": "2025-01-15T13:04:39.715215+00:00"} {"global_step": 2666, "acc_step": 0, "speed/wps": 12895.309014713155, "speed/FLOPS": 202538548087580.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1122601181268692, "optim/lr": 0.0015990000000000002, "optim/total_tokens": 1397751808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.50527024269104, "created_at": "2025-01-15T13:04:49.883375+00:00"} {"global_step": 2667, "acc_step": 0, "speed/wps": 12891.50262217678, "speed/FLOPS": 202478763462265.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0019222497940063, "optim/lr": 0.0015996, "optim/total_tokens": 1398276096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.4080545902252197, "created_at": "2025-01-15T13:05:00.054206+00:00"} {"global_step": 2668, "acc_step": 0, "speed/wps": 12892.556318452322, "speed/FLOPS": 202495313210203.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.025867462158203, "optim/lr": 0.0016002, "optim/total_tokens": 1398800384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.689539909362793, "created_at": "2025-01-15T13:05:10.224960+00:00"} {"global_step": 2669, "acc_step": 0, "speed/wps": 12891.06635753593, "speed/FLOPS": 202471911326590.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3322241306304932, "optim/lr": 0.0016007999999999999, "optim/total_tokens": 1399324672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.4853129386901855, "created_at": "2025-01-15T13:05:20.397855+00:00"} {"global_step": 2670, "acc_step": 0, "speed/wps": 12889.636608146795, "speed/FLOPS": 202449455147752.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12813061475753784, "optim/lr": 0.0016014000000000002, "optim/total_tokens": 1399848960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.451561212539673, "created_at": "2025-01-15T13:05:30.571677+00:00"} {"global_step": 2671, "acc_step": 0, "speed/wps": 12890.956062098363, "speed/FLOPS": 202470178985181.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15578751266002655, "optim/lr": 0.0016020000000000001, "optim/total_tokens": 1400373248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.4644031524658203, "created_at": "2025-01-15T13:05:40.743420+00:00"} {"global_step": 2672, "acc_step": 0, "speed/wps": 12895.662636457693, "speed/FLOPS": 202544102203006.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15693286061286926, "optim/lr": 0.0016026, "optim/total_tokens": 1400897536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 3.5583858489990234, "created_at": "2025-01-15T13:05:50.911143+00:00"} {"global_step": 2673, "acc_step": 0, "speed/wps": 12892.815330279447, "speed/FLOPS": 202499381346867.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1450606882572174, "optim/lr": 0.0016032, "optim/total_tokens": 1401421824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.5402884483337402, "created_at": "2025-01-15T13:06:01.082147+00:00"} {"global_step": 2674, "acc_step": 0, "speed/wps": 12893.094929271992, "speed/FLOPS": 202503772833255.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.129109725356102, "optim/lr": 0.0016037999999999998, "optim/total_tokens": 1401946112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.5219855308532715, "created_at": "2025-01-15T13:06:11.252223+00:00"} {"global_step": 2675, "acc_step": 0, "speed/wps": 12896.14523133297, "speed/FLOPS": 202551682018677.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1594046801328659, "optim/lr": 0.0016044000000000002, "optim/total_tokens": 1402470400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 3.3829843997955322, "created_at": "2025-01-15T13:06:21.421012+00:00"} {"global_step": 2676, "acc_step": 0, "speed/wps": 12893.537330692436, "speed/FLOPS": 202510721355486.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7488757967948914, "optim/lr": 0.001605, "optim/total_tokens": 1402994688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.4642088413238525, "created_at": "2025-01-15T13:06:31.592843+00:00"} {"global_step": 2677, "acc_step": 0, "speed/wps": 12895.82310824657, "speed/FLOPS": 202546622632961.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0011693239212036, "optim/lr": 0.0016056, "optim/total_tokens": 1403518976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.595576286315918, "created_at": "2025-01-15T13:06:41.761569+00:00"} {"global_step": 2678, "acc_step": 0, "speed/wps": 12891.501694023293, "speed/FLOPS": 202478748884339.66, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9811811447143555, "optim/lr": 0.0016062, "optim/total_tokens": 1404043264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.4927639961242676, "created_at": "2025-01-15T13:06:51.934238+00:00"} {"global_step": 2679, "acc_step": 0, "speed/wps": 12893.53453354105, "speed/FLOPS": 202510677422380.25, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20418991148471832, "optim/lr": 0.0016068, "optim/total_tokens": 1404567552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297240, "loss/out": 3.4608166217803955, "created_at": "2025-01-15T13:07:02.104868+00:00"} {"global_step": 2680, "acc_step": 0, "speed/wps": 12892.090237469116, "speed/FLOPS": 202487992767899.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16494035720825195, "optim/lr": 0.0016074000000000001, "optim/total_tokens": 1405091840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.635303258895874, "created_at": "2025-01-15T13:07:12.275511+00:00"} {"global_step": 2681, "acc_step": 0, "speed/wps": 12889.960275741452, "speed/FLOPS": 202454538792092.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1848801076412201, "optim/lr": 0.001608, "optim/total_tokens": 1405616128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.4774622917175293, "created_at": "2025-01-15T13:07:22.449121+00:00"} {"global_step": 2682, "acc_step": 0, "speed/wps": 12892.478852668257, "speed/FLOPS": 202494096503615.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16434116661548615, "optim/lr": 0.0016086, "optim/total_tokens": 1406140416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280980, "loss/out": 3.5759072303771973, "created_at": "2025-01-15T13:07:32.631539+00:00"} {"global_step": 2683, "acc_step": 0, "speed/wps": 12897.795944260444, "speed/FLOPS": 202577608733520.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1396574229001999, "optim/lr": 0.0016092, "optim/total_tokens": 1406664704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.484726905822754, "created_at": "2025-01-15T13:07:42.801749+00:00"} {"global_step": 2684, "acc_step": 0, "speed/wps": 12893.318248193984, "speed/FLOPS": 202507280363797.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.15187707543373108, "optim/lr": 0.0016098, "optim/total_tokens": 1407188992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 3.6447079181671143, "created_at": "2025-01-15T13:07:52.974040+00:00"} {"global_step": 2685, "acc_step": 0, "speed/wps": 12895.72405258841, "speed/FLOPS": 202545066827733.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15768371522426605, "optim/lr": 0.0016104000000000001, "optim/total_tokens": 1407713280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.4812188148498535, "created_at": "2025-01-15T13:08:03.142017+00:00"} {"global_step": 2686, "acc_step": 0, "speed/wps": 12898.928325475184, "speed/FLOPS": 202595394336552.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1337575763463974, "optim/lr": 0.0016110000000000002, "optim/total_tokens": 1408237568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 3.4380064010620117, "created_at": "2025-01-15T13:08:13.307113+00:00"} {"global_step": 2687, "acc_step": 0, "speed/wps": 12892.747983274145, "speed/FLOPS": 202498323569605.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.13060231506824493, "optim/lr": 0.0016116000000000001, "optim/total_tokens": 1408761856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.5752148628234863, "created_at": "2025-01-15T13:08:23.480129+00:00"} {"global_step": 2688, "acc_step": 0, "speed/wps": 12896.807052284688, "speed/FLOPS": 202562076826161.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.154092937707901, "optim/lr": 0.0016122, "optim/total_tokens": 1409286144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 3.4717917442321777, "created_at": "2025-01-15T13:08:33.646731+00:00"} {"global_step": 2689, "acc_step": 0, "speed/wps": 12892.813531468299, "speed/FLOPS": 202499353094067.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14179553091526031, "optim/lr": 0.0016128, "optim/total_tokens": 1409810432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 3.46946120262146, "created_at": "2025-01-15T13:08:43.820687+00:00"} {"global_step": 2690, "acc_step": 0, "speed/wps": 12891.35412758449, "speed/FLOPS": 202476431150640.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13063502311706543, "optim/lr": 0.0016133999999999999, "optim/total_tokens": 1410334720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.422865867614746, "created_at": "2025-01-15T13:08:53.991697+00:00"} {"global_step": 2691, "acc_step": 0, "speed/wps": 12890.74954631725, "speed/FLOPS": 202466935371057.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13901878893375397, "optim/lr": 0.0016140000000000002, "optim/total_tokens": 1410859008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 3.530186414718628, "created_at": "2025-01-15T13:09:04.163223+00:00"} {"global_step": 2692, "acc_step": 0, "speed/wps": 12892.497253232152, "speed/FLOPS": 202494385509756.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1356407105922699, "optim/lr": 0.0016146, "optim/total_tokens": 1411383296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.3953962326049805, "created_at": "2025-01-15T13:09:14.333208+00:00"} {"global_step": 2693, "acc_step": 0, "speed/wps": 12892.161734643063, "speed/FLOPS": 202489115729264.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11692308634519577, "optim/lr": 0.0016152, "optim/total_tokens": 1411907584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.437446355819702, "created_at": "2025-01-15T13:09:24.503351+00:00"} {"global_step": 2694, "acc_step": 0, "speed/wps": 12891.019360420396, "speed/FLOPS": 202471173172310.3, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12231232970952988, "optim/lr": 0.0016158, "optim/total_tokens": 1412431872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.420145034790039, "created_at": "2025-01-15T13:09:34.674641+00:00"} {"global_step": 2695, "acc_step": 0, "speed/wps": 12891.859331140859, "speed/FLOPS": 202484366066715.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12838329374790192, "optim/lr": 0.0016163999999999998, "optim/total_tokens": 1412956160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478894, "loss/out": 3.6140358448028564, "created_at": "2025-01-15T13:09:44.849572+00:00"} {"global_step": 2696, "acc_step": 0, "speed/wps": 12894.083589897255, "speed/FLOPS": 202519301106936.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15094152092933655, "optim/lr": 0.0016170000000000002, "optim/total_tokens": 1413480448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 3.4277172088623047, "created_at": "2025-01-15T13:09:55.019127+00:00"} {"global_step": 2697, "acc_step": 0, "speed/wps": 12893.537421647126, "speed/FLOPS": 202510722784054.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15103232860565186, "optim/lr": 0.0016176, "optim/total_tokens": 1414004736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.4608700275421143, "created_at": "2025-01-15T13:10:05.190716+00:00"} {"global_step": 2698, "acc_step": 0, "speed/wps": 12895.545391361069, "speed/FLOPS": 202542260707651.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1221708208322525, "optim/lr": 0.0016182, "optim/total_tokens": 1414529024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.400205612182617, "created_at": "2025-01-15T13:10:15.360042+00:00"} {"global_step": 2699, "acc_step": 0, "speed/wps": 12892.581438253383, "speed/FLOPS": 202495707751197.25, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13748718798160553, "optim/lr": 0.0016187999999999999, "optim/total_tokens": 1415053312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.4798426628112793, "created_at": "2025-01-15T13:10:25.533328+00:00"} {"global_step": 2700, "acc_step": 0, "speed/wps": 12874.422543378674, "speed/FLOPS": 202210497354253.53, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15262627601623535, "optim/lr": 0.0016193999999999998, "optim/total_tokens": 1415577600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412896, "loss/out": 3.6053359508514404, "created_at": "2025-01-15T13:10:35.718098+00:00"} {"global_step": 2701, "acc_step": 0, "speed/wps": 12889.267193057254, "speed/FLOPS": 202443652976141.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1306166797876358, "optim/lr": 0.0016200000000000001, "optim/total_tokens": 1416101888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507696, "loss/out": 3.483856678009033, "created_at": "2025-01-15T13:10:45.891470+00:00"} {"global_step": 2702, "acc_step": 0, "speed/wps": 12889.96870268703, "speed/FLOPS": 202454671148852.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15091322362422943, "optim/lr": 0.0016206, "optim/total_tokens": 1416626176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.4637553691864014, "created_at": "2025-01-15T13:10:56.067485+00:00"} {"global_step": 2703, "acc_step": 0, "speed/wps": 12890.190972372078, "speed/FLOPS": 202458162199687.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1387399286031723, "optim/lr": 0.0016212, "optim/total_tokens": 1417150464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.329766035079956, "created_at": "2025-01-15T13:11:06.244142+00:00"} {"global_step": 2704, "acc_step": 0, "speed/wps": 12892.826262526618, "speed/FLOPS": 202499553052832.22, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1702726036310196, "optim/lr": 0.0016217999999999998, "optim/total_tokens": 1417674752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306453, "loss/out": 3.5081839561462402, "created_at": "2025-01-15T13:11:16.416540+00:00"} {"global_step": 2705, "acc_step": 0, "speed/wps": 12893.673328286544, "speed/FLOPS": 202512857384582.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15083883702754974, "optim/lr": 0.0016224, "optim/total_tokens": 1418199040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.474429130554199, "created_at": "2025-01-15T13:11:26.589306+00:00"} {"global_step": 2706, "acc_step": 0, "speed/wps": 12891.880699965111, "speed/FLOPS": 202484701693464.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14525166153907776, "optim/lr": 0.001623, "optim/total_tokens": 1418723328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.526616096496582, "created_at": "2025-01-15T13:11:36.765358+00:00"} {"global_step": 2707, "acc_step": 0, "speed/wps": 12891.457530131389, "speed/FLOPS": 202478055229729.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1872425526380539, "optim/lr": 0.0016236, "optim/total_tokens": 1419247616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.515119791030884, "created_at": "2025-01-15T13:11:46.936673+00:00"} {"global_step": 2708, "acc_step": 0, "speed/wps": 12891.631515181347, "speed/FLOPS": 202480787904019.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16652996838092804, "optim/lr": 0.0016242, "optim/total_tokens": 1419771904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 3.5239973068237305, "created_at": "2025-01-15T13:11:57.108666+00:00"} {"global_step": 2709, "acc_step": 0, "speed/wps": 12888.906607141858, "speed/FLOPS": 202437989478843.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1839205026626587, "optim/lr": 0.0016248, "optim/total_tokens": 1420296192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.4829397201538086, "created_at": "2025-01-15T13:12:07.288663+00:00"} {"global_step": 2710, "acc_step": 0, "speed/wps": 12889.516096388277, "speed/FLOPS": 202447562344984.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2037803679704666, "optim/lr": 0.0016254, "optim/total_tokens": 1420820480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.5557072162628174, "created_at": "2025-01-15T13:12:17.468682+00:00"} {"global_step": 2711, "acc_step": 0, "speed/wps": 12892.631676968102, "speed/FLOPS": 202496496819247.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17876982688903809, "optim/lr": 0.001626, "optim/total_tokens": 1421344768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.3796048164367676, "created_at": "2025-01-15T13:12:27.643728+00:00"} {"global_step": 2712, "acc_step": 0, "speed/wps": 12892.310236031532, "speed/FLOPS": 202491448147632.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24266210198402405, "optim/lr": 0.0016266000000000002, "optim/total_tokens": 1421869056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.4403278827667236, "created_at": "2025-01-15T13:12:37.814054+00:00"} {"global_step": 2713, "acc_step": 0, "speed/wps": 12886.114728268798, "speed/FLOPS": 202394139184698.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22767819464206696, "optim/lr": 0.0016272, "optim/total_tokens": 1422393344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286083, "loss/out": 3.5015828609466553, "created_at": "2025-01-15T13:12:47.989447+00:00"} {"global_step": 2714, "acc_step": 0, "speed/wps": 12895.333410312414, "speed/FLOPS": 202538931253990.25, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1622019112110138, "optim/lr": 0.0016278, "optim/total_tokens": 1422917632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.437138319015503, "created_at": "2025-01-15T13:12:58.162346+00:00"} {"global_step": 2715, "acc_step": 0, "speed/wps": 12903.258716021564, "speed/FLOPS": 202663409070663.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16824528574943542, "optim/lr": 0.0016284, "optim/total_tokens": 1423441920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 3.4932892322540283, "created_at": "2025-01-15T13:13:08.323937+00:00"} {"global_step": 2716, "acc_step": 0, "speed/wps": 12891.227163660047, "speed/FLOPS": 202474437007740.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1440257430076599, "optim/lr": 0.0016290000000000002, "optim/total_tokens": 1423966208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 3.5154521465301514, "created_at": "2025-01-15T13:13:18.495080+00:00"} {"global_step": 2717, "acc_step": 0, "speed/wps": 12887.883981052146, "speed/FLOPS": 202421927730869.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12902870774269104, "optim/lr": 0.0016296000000000001, "optim/total_tokens": 1424490496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.5039401054382324, "created_at": "2025-01-15T13:13:28.669075+00:00"} {"global_step": 2718, "acc_step": 0, "speed/wps": 12893.736477287914, "speed/FLOPS": 202513849226424.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15551716089248657, "optim/lr": 0.0016302, "optim/total_tokens": 1425014784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.472972869873047, "created_at": "2025-01-15T13:13:38.840479+00:00"} {"global_step": 2719, "acc_step": 0, "speed/wps": 12893.15334442516, "speed/FLOPS": 202504690323506.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14185313880443573, "optim/lr": 0.0016308, "optim/total_tokens": 1425539072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.4514970779418945, "created_at": "2025-01-15T13:13:49.012438+00:00"} {"global_step": 2720, "acc_step": 0, "speed/wps": 12892.337004744153, "speed/FLOPS": 202491868587047.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13862182199954987, "optim/lr": 0.0016313999999999999, "optim/total_tokens": 1426063360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 3.377958297729492, "created_at": "2025-01-15T13:13:59.184456+00:00"} {"global_step": 2721, "acc_step": 0, "speed/wps": 12887.56425079863, "speed/FLOPS": 202416905927882.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.13371644914150238, "optim/lr": 0.0016320000000000002, "optim/total_tokens": 1426587648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.4554240703582764, "created_at": "2025-01-15T13:14:09.365442+00:00"} {"global_step": 2722, "acc_step": 0, "speed/wps": 12894.214819255752, "speed/FLOPS": 202521362244340.2, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.11452816426753998, "optim/lr": 0.0016326000000000001, "optim/total_tokens": 1427111936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.4478044509887695, "created_at": "2025-01-15T13:14:19.536596+00:00"} {"global_step": 2723, "acc_step": 0, "speed/wps": 12892.168133183375, "speed/FLOPS": 202489216227132.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11899159103631973, "optim/lr": 0.0016332, "optim/total_tokens": 1427636224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 3.389859914779663, "created_at": "2025-01-15T13:14:29.707917+00:00"} {"global_step": 2724, "acc_step": 0, "speed/wps": 12900.864316632697, "speed/FLOPS": 202625801738012.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1676315814256668, "optim/lr": 0.0016338, "optim/total_tokens": 1428160512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.3442811965942383, "created_at": "2025-01-15T13:14:39.871313+00:00"} {"global_step": 2725, "acc_step": 0, "speed/wps": 12895.041246969318, "speed/FLOPS": 202534342427212.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2004317194223404, "optim/lr": 0.0016343999999999998, "optim/total_tokens": 1428684800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.3843395709991455, "created_at": "2025-01-15T13:14:50.039406+00:00"} {"global_step": 2726, "acc_step": 0, "speed/wps": 12894.201664893886, "speed/FLOPS": 202521155637011.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17789797484874725, "optim/lr": 0.0016350000000000002, "optim/total_tokens": 1429209088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453916, "loss/out": 3.330634593963623, "created_at": "2025-01-15T13:15:00.209266+00:00"} {"global_step": 2727, "acc_step": 0, "speed/wps": 12894.41499568763, "speed/FLOPS": 202524506290274.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13851860165596008, "optim/lr": 0.0016356, "optim/total_tokens": 1429733376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.4877939224243164, "created_at": "2025-01-15T13:15:10.378731+00:00"} {"global_step": 2728, "acc_step": 0, "speed/wps": 12891.808426505635, "speed/FLOPS": 202483566539469.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14651213586330414, "optim/lr": 0.0016362, "optim/total_tokens": 1430257664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.4282703399658203, "created_at": "2025-01-15T13:15:20.550401+00:00"} {"global_step": 2729, "acc_step": 0, "speed/wps": 12894.897368403455, "speed/FLOPS": 202532082616625.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1565941423177719, "optim/lr": 0.0016367999999999999, "optim/total_tokens": 1430781952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 3.537752628326416, "created_at": "2025-01-15T13:15:30.718649+00:00"} {"global_step": 2730, "acc_step": 0, "speed/wps": 12887.574089596168, "speed/FLOPS": 202417060459717.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1776491105556488, "optim/lr": 0.0016373999999999998, "optim/total_tokens": 1431306240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.547921657562256, "created_at": "2025-01-15T13:15:40.892968+00:00"} {"global_step": 2731, "acc_step": 0, "speed/wps": 12891.351638395887, "speed/FLOPS": 202476392054513.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1888902336359024, "optim/lr": 0.0016380000000000001, "optim/total_tokens": 1431830528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 3.582674503326416, "created_at": "2025-01-15T13:15:51.064541+00:00"} {"global_step": 2732, "acc_step": 0, "speed/wps": 12895.668515015766, "speed/FLOPS": 202544194533838.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13122381269931793, "optim/lr": 0.0016386, "optim/total_tokens": 1432354816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.357931137084961, "created_at": "2025-01-15T13:16:01.235227+00:00"} {"global_step": 2733, "acc_step": 0, "speed/wps": 12892.945050335287, "speed/FLOPS": 202501418778598.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1686590015888214, "optim/lr": 0.0016392, "optim/total_tokens": 1432879104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 3.595487594604492, "created_at": "2025-01-15T13:16:11.414175+00:00"} {"global_step": 2734, "acc_step": 0, "speed/wps": 12893.291109239206, "speed/FLOPS": 202506854109219.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1435922235250473, "optim/lr": 0.0016397999999999999, "optim/total_tokens": 1433403392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.384568929672241, "created_at": "2025-01-15T13:16:21.588457+00:00"} {"global_step": 2735, "acc_step": 0, "speed/wps": 12894.159510867747, "speed/FLOPS": 202520493550104.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16553010046482086, "optim/lr": 0.0016404, "optim/total_tokens": 1433927680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.575930118560791, "created_at": "2025-01-15T13:16:31.757204+00:00"} {"global_step": 2736, "acc_step": 0, "speed/wps": 12897.05302984481, "speed/FLOPS": 202565940241751.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16761130094528198, "optim/lr": 0.001641, "optim/total_tokens": 1434451968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428908, "loss/out": 3.3661630153656006, "created_at": "2025-01-15T13:16:41.929027+00:00"} {"global_step": 2737, "acc_step": 0, "speed/wps": 12891.194109429096, "speed/FLOPS": 202473917845622.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14409147202968597, "optim/lr": 0.0016416, "optim/total_tokens": 1434976256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 3.4180939197540283, "created_at": "2025-01-15T13:16:52.101268+00:00"} {"global_step": 2738, "acc_step": 0, "speed/wps": 12899.222294156198, "speed/FLOPS": 202600011518643.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18002475798130035, "optim/lr": 0.0016422000000000001, "optim/total_tokens": 1435500544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.5935916900634766, "created_at": "2025-01-15T13:17:02.270659+00:00"} {"global_step": 2739, "acc_step": 0, "speed/wps": 12896.142873826708, "speed/FLOPS": 202551644990801.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16695700585842133, "optim/lr": 0.0016428, "optim/total_tokens": 1436024832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.5344231128692627, "created_at": "2025-01-15T13:17:12.441334+00:00"} {"global_step": 2740, "acc_step": 0, "speed/wps": 12890.461209384444, "speed/FLOPS": 202462406643310.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23100242018699646, "optim/lr": 0.0016434, "optim/total_tokens": 1436549120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.5607547760009766, "created_at": "2025-01-15T13:17:22.612962+00:00"} {"global_step": 2741, "acc_step": 0, "speed/wps": 12897.349339552815, "speed/FLOPS": 202570594192887.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2175849974155426, "optim/lr": 0.001644, "optim/total_tokens": 1437073408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.4338626861572266, "created_at": "2025-01-15T13:17:32.779135+00:00"} {"global_step": 2742, "acc_step": 0, "speed/wps": 12895.809714273797, "speed/FLOPS": 202546412262214.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20275703072547913, "optim/lr": 0.0016446000000000002, "optim/total_tokens": 1437597696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.4082131385803223, "created_at": "2025-01-15T13:17:42.946634+00:00"} {"global_step": 2743, "acc_step": 0, "speed/wps": 12892.061593236549, "speed/FLOPS": 202487542870865.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17126885056495667, "optim/lr": 0.0016452, "optim/total_tokens": 1438121984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.3689537048339844, "created_at": "2025-01-15T13:17:53.119502+00:00"} {"global_step": 2744, "acc_step": 0, "speed/wps": 12894.443541690567, "speed/FLOPS": 202524954644476.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1439541131258011, "optim/lr": 0.0016458, "optim/total_tokens": 1438646272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.4497690200805664, "created_at": "2025-01-15T13:18:03.287953+00:00"} {"global_step": 2745, "acc_step": 0, "speed/wps": 12890.718720179806, "speed/FLOPS": 202466451204208.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1627887487411499, "optim/lr": 0.0016464, "optim/total_tokens": 1439170560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.4861040115356445, "created_at": "2025-01-15T13:18:13.460281+00:00"} {"global_step": 2746, "acc_step": 0, "speed/wps": 12894.235456359447, "speed/FLOPS": 202521686378412.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13251934945583344, "optim/lr": 0.0016470000000000002, "optim/total_tokens": 1439694848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.481518268585205, "created_at": "2025-01-15T13:18:23.630125+00:00"} {"global_step": 2747, "acc_step": 0, "speed/wps": 12891.766923453131, "speed/FLOPS": 202482914676998.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13233856856822968, "optim/lr": 0.0016476000000000001, "optim/total_tokens": 1440219136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 3.3237929344177246, "created_at": "2025-01-15T13:18:33.800927+00:00"} {"global_step": 2748, "acc_step": 0, "speed/wps": 12893.91059355434, "speed/FLOPS": 202516583961649.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.127602681517601, "optim/lr": 0.0016482, "optim/total_tokens": 1440743424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.368567705154419, "created_at": "2025-01-15T13:18:43.979224+00:00"} {"global_step": 2749, "acc_step": 0, "speed/wps": 12895.204664220162, "speed/FLOPS": 202536909119695.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1402817964553833, "optim/lr": 0.0016488, "optim/total_tokens": 1441267712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.547093391418457, "created_at": "2025-01-15T13:18:54.152478+00:00"} {"global_step": 2750, "acc_step": 0, "speed/wps": 12890.583395322355, "speed/FLOPS": 202464325741366.78, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13745927810668945, "optim/lr": 0.0016493999999999999, "optim/total_tokens": 1441792000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.4514319896698, "created_at": "2025-01-15T13:19:04.324445+00:00"} {"global_step": 2751, "acc_step": 0, "speed/wps": 12887.097976053596, "speed/FLOPS": 202409582442282.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17159704864025116, "optim/lr": 0.0016500000000000002, "optim/total_tokens": 1442316288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442492, "loss/out": 3.4437756538391113, "created_at": "2025-01-15T13:19:14.499282+00:00"} {"global_step": 2752, "acc_step": 0, "speed/wps": 12878.18682064259, "speed/FLOPS": 202269620501342.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.13734398782253265, "optim/lr": 0.0016506000000000001, "optim/total_tokens": 1442840576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 3.3537676334381104, "created_at": "2025-01-15T13:19:24.686183+00:00"} {"global_step": 2753, "acc_step": 0, "speed/wps": 12878.374213162886, "speed/FLOPS": 202272563758377.9, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.15085765719413757, "optim/lr": 0.0016512, "optim/total_tokens": 1443364864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.50620174407959, "created_at": "2025-01-15T13:19:34.867483+00:00"} {"global_step": 2754, "acc_step": 0, "speed/wps": 12885.919049786145, "speed/FLOPS": 202391065785238.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0032, "optim/grad_norm": 0.14966046810150146, "optim/lr": 0.0016518, "optim/total_tokens": 1443889152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.369539260864258, "created_at": "2025-01-15T13:19:45.047499+00:00"} {"global_step": 2755, "acc_step": 0, "speed/wps": 12884.660952547078, "speed/FLOPS": 202371305639291.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12939390540122986, "optim/lr": 0.0016523999999999998, "optim/total_tokens": 1444413440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.4505133628845215, "created_at": "2025-01-15T13:19:55.224496+00:00"} {"global_step": 2756, "acc_step": 0, "speed/wps": 12893.06582347223, "speed/FLOPS": 202503315686674.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1317538172006607, "optim/lr": 0.0016530000000000002, "optim/total_tokens": 1444937728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 3.2869462966918945, "created_at": "2025-01-15T13:20:05.397203+00:00"} {"global_step": 2757, "acc_step": 0, "speed/wps": 12888.75978117213, "speed/FLOPS": 202435683375229.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.11674795299768448, "optim/lr": 0.0016536, "optim/total_tokens": 1445462016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.31518292427063, "created_at": "2025-01-15T13:20:15.571508+00:00"} {"global_step": 2758, "acc_step": 0, "speed/wps": 12890.738926939483, "speed/FLOPS": 202466768579136.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1551954746246338, "optim/lr": 0.0016542, "optim/total_tokens": 1445986304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.420158863067627, "created_at": "2025-01-15T13:20:25.744164+00:00"} {"global_step": 2759, "acc_step": 0, "speed/wps": 12894.436409608214, "speed/FLOPS": 202524842625322.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18563194572925568, "optim/lr": 0.0016547999999999999, "optim/total_tokens": 1446510592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.6097917556762695, "created_at": "2025-01-15T13:20:35.912667+00:00"} {"global_step": 2760, "acc_step": 0, "speed/wps": 12891.525161232603, "speed/FLOPS": 202479117469109.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17994271218776703, "optim/lr": 0.0016554, "optim/total_tokens": 1447034880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 3.4633522033691406, "created_at": "2025-01-15T13:20:46.085698+00:00"} {"global_step": 2761, "acc_step": 0, "speed/wps": 12892.539599530424, "speed/FLOPS": 202495050616560.78, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1498376727104187, "optim/lr": 0.0016560000000000001, "optim/total_tokens": 1447559168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.4974288940429688, "created_at": "2025-01-15T13:20:56.258015+00:00"} {"global_step": 2762, "acc_step": 0, "speed/wps": 12893.126774299875, "speed/FLOPS": 202504273003177.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19363851845264435, "optim/lr": 0.0016566, "optim/total_tokens": 1448083456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.4996933937072754, "created_at": "2025-01-15T13:21:06.427919+00:00"} {"global_step": 2763, "acc_step": 0, "speed/wps": 12890.229379127048, "speed/FLOPS": 202458765430550.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18922924995422363, "optim/lr": 0.0016572, "optim/total_tokens": 1448607744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 3.529325246810913, "created_at": "2025-01-15T13:21:16.599923+00:00"} {"global_step": 2764, "acc_step": 0, "speed/wps": 12889.923565885023, "speed/FLOPS": 202453962213352.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.150923490524292, "optim/lr": 0.0016578, "optim/total_tokens": 1449132032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.506253480911255, "created_at": "2025-01-15T13:21:26.775675+00:00"} {"global_step": 2765, "acc_step": 0, "speed/wps": 12889.859430684579, "speed/FLOPS": 202452954881898.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18596896529197693, "optim/lr": 0.0016584, "optim/total_tokens": 1449656320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.398937225341797, "created_at": "2025-01-15T13:21:36.950198+00:00"} {"global_step": 2766, "acc_step": 0, "speed/wps": 12894.756023720161, "speed/FLOPS": 202529862604146.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21935176849365234, "optim/lr": 0.001659, "optim/total_tokens": 1450180608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.35113525390625, "created_at": "2025-01-15T13:21:47.121488+00:00"} {"global_step": 2767, "acc_step": 0, "speed/wps": 12878.736556424372, "speed/FLOPS": 202278254857211.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24232995510101318, "optim/lr": 0.0016596, "optim/total_tokens": 1450704896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 3.453761577606201, "created_at": "2025-01-15T13:21:57.302783+00:00"} {"global_step": 2768, "acc_step": 0, "speed/wps": 12891.715066564633, "speed/FLOPS": 202482100193306.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1755971759557724, "optim/lr": 0.0016602000000000001, "optim/total_tokens": 1451229184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 3.482823610305786, "created_at": "2025-01-15T13:22:07.479775+00:00"} {"global_step": 2769, "acc_step": 0, "speed/wps": 12890.212771314937, "speed/FLOPS": 202458504582039.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15831272304058075, "optim/lr": 0.0016608, "optim/total_tokens": 1451753472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.387120246887207, "created_at": "2025-01-15T13:22:17.651733+00:00"} {"global_step": 2770, "acc_step": 0, "speed/wps": 12888.857136366887, "speed/FLOPS": 202437212472342.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15472066402435303, "optim/lr": 0.0016614, "optim/total_tokens": 1452277760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.4299020767211914, "created_at": "2025-01-15T13:22:27.824668+00:00"} {"global_step": 2771, "acc_step": 0, "speed/wps": 12886.518455885553, "speed/FLOPS": 202400480281699.66, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15308640897274017, "optim/lr": 0.0016620000000000003, "optim/total_tokens": 1452802048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394163, "loss/out": 3.5133490562438965, "created_at": "2025-01-15T13:22:38.000681+00:00"} {"global_step": 2772, "acc_step": 0, "speed/wps": 12875.552061938866, "speed/FLOPS": 202228237995362.5, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1438397616147995, "optim/lr": 0.0016626000000000002, "optim/total_tokens": 1453326336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316658, "loss/out": 3.301544666290283, "created_at": "2025-01-15T13:22:48.184262+00:00"} {"global_step": 2773, "acc_step": 0, "speed/wps": 12875.264087613683, "speed/FLOPS": 202223714962867.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12934571504592896, "optim/lr": 0.0016632, "optim/total_tokens": 1453850624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.3795084953308105, "created_at": "2025-01-15T13:22:58.369803+00:00"} {"global_step": 2774, "acc_step": 0, "speed/wps": 12879.080813489652, "speed/FLOPS": 202283661887480.38, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15439806878566742, "optim/lr": 0.0016638, "optim/total_tokens": 1454374912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457889, "loss/out": 3.4451708793640137, "created_at": "2025-01-15T13:23:08.554707+00:00"} {"global_step": 2775, "acc_step": 0, "speed/wps": 12887.498856279584, "speed/FLOPS": 202415878817099.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13904352486133575, "optim/lr": 0.0016644, "optim/total_tokens": 1454899200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.4550323486328125, "created_at": "2025-01-15T13:23:18.731133+00:00"} {"global_step": 2776, "acc_step": 0, "speed/wps": 12884.380768509047, "speed/FLOPS": 202366904963961.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12424586713314056, "optim/lr": 0.0016650000000000002, "optim/total_tokens": 1455423488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.3621866703033447, "created_at": "2025-01-15T13:23:28.909794+00:00"} {"global_step": 2777, "acc_step": 0, "speed/wps": 12888.917121987946, "speed/FLOPS": 202438154628950.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1219264566898346, "optim/lr": 0.0016656000000000002, "optim/total_tokens": 1455947776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 3.5549874305725098, "created_at": "2025-01-15T13:23:39.088059+00:00"} {"global_step": 2778, "acc_step": 0, "speed/wps": 12888.854188977832, "speed/FLOPS": 202437166179547.5, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12091223150491714, "optim/lr": 0.0016662, "optim/total_tokens": 1456472064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.504859685897827, "created_at": "2025-01-15T13:23:49.262558+00:00"} {"global_step": 2779, "acc_step": 0, "speed/wps": 12889.508652203478, "speed/FLOPS": 202447445423833.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1391788125038147, "optim/lr": 0.0016668, "optim/total_tokens": 1456996352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.38486647605896, "created_at": "2025-01-15T13:23:59.436008+00:00"} {"global_step": 2780, "acc_step": 0, "speed/wps": 12881.326884227183, "speed/FLOPS": 202318939514837.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18780462443828583, "optim/lr": 0.0016673999999999999, "optim/total_tokens": 1457520640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.484262228012085, "created_at": "2025-01-15T13:24:09.621371+00:00"} {"global_step": 2781, "acc_step": 0, "speed/wps": 12884.225136682624, "speed/FLOPS": 202364460552266.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2074756771326065, "optim/lr": 0.0016680000000000002, "optim/total_tokens": 1458044928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.379547595977783, "created_at": "2025-01-15T13:24:19.799526+00:00"} {"global_step": 2782, "acc_step": 0, "speed/wps": 12880.135837201344, "speed/FLOPS": 202300232484626.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19994902610778809, "optim/lr": 0.0016686000000000001, "optim/total_tokens": 1458569216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.347463369369507, "created_at": "2025-01-15T13:24:29.986971+00:00"} {"global_step": 2783, "acc_step": 0, "speed/wps": 12875.113074064662, "speed/FLOPS": 202221343087565.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.22052645683288574, "optim/lr": 0.0016692, "optim/total_tokens": 1459093504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.3080735206604004, "created_at": "2025-01-15T13:24:40.172480+00:00"} {"global_step": 2784, "acc_step": 0, "speed/wps": 12886.502360289267, "speed/FLOPS": 202400227478242.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24726152420043945, "optim/lr": 0.0016698, "optim/total_tokens": 1459617792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.411299228668213, "created_at": "2025-01-15T13:24:50.350781+00:00"} {"global_step": 2785, "acc_step": 0, "speed/wps": 12886.365220737594, "speed/FLOPS": 202398073513133.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16982612013816833, "optim/lr": 0.0016703999999999998, "optim/total_tokens": 1460142080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.364941120147705, "created_at": "2025-01-15T13:25:00.527820+00:00"} {"global_step": 2786, "acc_step": 0, "speed/wps": 12887.15105728808, "speed/FLOPS": 202410416156011.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1845196783542633, "optim/lr": 0.0016710000000000002, "optim/total_tokens": 1460666368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 3.48952054977417, "created_at": "2025-01-15T13:25:10.706037+00:00"} {"global_step": 2787, "acc_step": 0, "speed/wps": 12873.104803158893, "speed/FLOPS": 202189800433337.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20105870068073273, "optim/lr": 0.0016716, "optim/total_tokens": 1461190656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.5373878479003906, "created_at": "2025-01-15T13:25:20.891698+00:00"} {"global_step": 2788, "acc_step": 0, "speed/wps": 12890.622623818623, "speed/FLOPS": 202464941878806.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1572447568178177, "optim/lr": 0.0016722, "optim/total_tokens": 1461714944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.389692783355713, "created_at": "2025-01-15T13:25:31.068553+00:00"} {"global_step": 2789, "acc_step": 0, "speed/wps": 12889.521344862462, "speed/FLOPS": 202447644779483.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14953847229480743, "optim/lr": 0.0016728, "optim/total_tokens": 1462239232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.2840189933776855, "created_at": "2025-01-15T13:25:41.244648+00:00"} {"global_step": 2790, "acc_step": 0, "speed/wps": 12882.34869989729, "speed/FLOPS": 202334988534058.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1656855344772339, "optim/lr": 0.0016734, "optim/total_tokens": 1462763520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.4541537761688232, "created_at": "2025-01-15T13:25:51.423968+00:00"} {"global_step": 2791, "acc_step": 0, "speed/wps": 12884.146369725198, "speed/FLOPS": 202363223408963.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14784768223762512, "optim/lr": 0.0016740000000000001, "optim/total_tokens": 1463287808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 3.4086458683013916, "created_at": "2025-01-15T13:26:01.604199+00:00"} {"global_step": 2792, "acc_step": 0, "speed/wps": 12882.489977888388, "speed/FLOPS": 202337207499045.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.12814341485500336, "optim/lr": 0.0016746, "optim/total_tokens": 1463812096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.5029919147491455, "created_at": "2025-01-15T13:26:11.783176+00:00"} {"global_step": 2793, "acc_step": 0, "speed/wps": 12887.106408697486, "speed/FLOPS": 202409714888542.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13395266234874725, "optim/lr": 0.0016752, "optim/total_tokens": 1464336384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.516657829284668, "created_at": "2025-01-15T13:26:21.959270+00:00"} {"global_step": 2794, "acc_step": 0, "speed/wps": 12885.832152039755, "speed/FLOPS": 202389700936721.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15918676555156708, "optim/lr": 0.0016758, "optim/total_tokens": 1464860672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 3.415947437286377, "created_at": "2025-01-15T13:26:32.136179+00:00"} {"global_step": 2795, "acc_step": 0, "speed/wps": 12887.638664674998, "speed/FLOPS": 202418074700066.06, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16512924432754517, "optim/lr": 0.0016764, "optim/total_tokens": 1465384960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 3.3846018314361572, "created_at": "2025-01-15T13:26:42.311197+00:00"} {"global_step": 2796, "acc_step": 0, "speed/wps": 12890.045480720284, "speed/FLOPS": 202455877053369.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1442793905735016, "optim/lr": 0.001677, "optim/total_tokens": 1465909248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.3269824981689453, "created_at": "2025-01-15T13:26:52.484325+00:00"} {"global_step": 2797, "acc_step": 0, "speed/wps": 12885.605995451144, "speed/FLOPS": 202386148836725.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1220528855919838, "optim/lr": 0.0016776000000000002, "optim/total_tokens": 1466433536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.3936383724212646, "created_at": "2025-01-15T13:27:02.660475+00:00"} {"global_step": 2798, "acc_step": 0, "speed/wps": 12876.414481208356, "speed/FLOPS": 202241783475078.1, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1159304529428482, "optim/lr": 0.0016782000000000001, "optim/total_tokens": 1466957824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.540262222290039, "created_at": "2025-01-15T13:27:12.846811+00:00"} {"global_step": 2799, "acc_step": 0, "speed/wps": 12890.396059374252, "speed/FLOPS": 202461383372875.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1122383400797844, "optim/lr": 0.0016788, "optim/total_tokens": 1467482112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.373619556427002, "created_at": "2025-01-15T13:27:23.018762+00:00"} {"global_step": 2800, "acc_step": 0, "speed/wps": 12888.331299583073, "speed/FLOPS": 202428953483077.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13865746557712555, "optim/lr": 0.0016794, "optim/total_tokens": 1468006400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 3.4063587188720703, "created_at": "2025-01-15T13:27:33.194411+00:00"} {"global_step": 2801, "acc_step": 0, "speed/wps": 12892.562550856102, "speed/FLOPS": 202495411098668.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11914872378110886, "optim/lr": 0.0016800000000000003, "optim/total_tokens": 1468530688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.380455255508423, "created_at": "2025-01-15T13:27:43.367635+00:00"} {"global_step": 2802, "acc_step": 0, "speed/wps": 12888.401361148926, "speed/FLOPS": 202430053896243.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11307571828365326, "optim/lr": 0.0016806000000000002, "optim/total_tokens": 1469054976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.4503068923950195, "created_at": "2025-01-15T13:27:53.543126+00:00"} {"global_step": 2803, "acc_step": 0, "speed/wps": 12888.841967419814, "speed/FLOPS": 202436974223184.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13792982697486877, "optim/lr": 0.0016812, "optim/total_tokens": 1469579264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.395888328552246, "created_at": "2025-01-15T13:28:03.716803+00:00"} {"global_step": 2804, "acc_step": 0, "speed/wps": 12888.540777196815, "speed/FLOPS": 202432243616853.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12358751893043518, "optim/lr": 0.0016818, "optim/total_tokens": 1470103552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 3.55883526802063, "created_at": "2025-01-15T13:28:13.891364+00:00"} {"global_step": 2805, "acc_step": 0, "speed/wps": 12886.331566082577, "speed/FLOPS": 202397544920530.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1357778012752533, "optim/lr": 0.0016824, "optim/total_tokens": 1470627840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.441087484359741, "created_at": "2025-01-15T13:28:24.067520+00:00"} {"global_step": 2806, "acc_step": 0, "speed/wps": 12888.850957120656, "speed/FLOPS": 202437115418789.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15316630899906158, "optim/lr": 0.0016830000000000003, "optim/total_tokens": 1471152128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.3364505767822266, "created_at": "2025-01-15T13:28:34.240360+00:00"} {"global_step": 2807, "acc_step": 0, "speed/wps": 12889.851917631033, "speed/FLOPS": 202452836879068.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1654636263847351, "optim/lr": 0.0016836000000000002, "optim/total_tokens": 1471676416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365072, "loss/out": 3.4244484901428223, "created_at": "2025-01-15T13:28:44.412466+00:00"} {"global_step": 2808, "acc_step": 0, "speed/wps": 12890.507842324536, "speed/FLOPS": 202463139077715.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2031196653842926, "optim/lr": 0.0016842, "optim/total_tokens": 1472200704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.3212039470672607, "created_at": "2025-01-15T13:28:54.585202+00:00"} {"global_step": 2809, "acc_step": 0, "speed/wps": 12888.764661939866, "speed/FLOPS": 202435760034393.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21939833462238312, "optim/lr": 0.0016848, "optim/total_tokens": 1472724992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.414196014404297, "created_at": "2025-01-15T13:29:04.758984+00:00"} {"global_step": 2810, "acc_step": 0, "speed/wps": 12889.623193967533, "speed/FLOPS": 202449244459634.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17983204126358032, "optim/lr": 0.0016853999999999999, "optim/total_tokens": 1473249280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.4982759952545166, "created_at": "2025-01-15T13:29:14.934639+00:00"} {"global_step": 2811, "acc_step": 0, "speed/wps": 12892.290510917888, "speed/FLOPS": 202491138337617.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1402416229248047, "optim/lr": 0.0016860000000000002, "optim/total_tokens": 1473773568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493489, "loss/out": 3.4816341400146484, "created_at": "2025-01-15T13:29:25.105967+00:00"} {"global_step": 2812, "acc_step": 0, "speed/wps": 12894.953564512518, "speed/FLOPS": 202532965253740.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19667454063892365, "optim/lr": 0.0016866000000000001, "optim/total_tokens": 1474297856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 3.3632476329803467, "created_at": "2025-01-15T13:29:35.274269+00:00"} {"global_step": 2813, "acc_step": 0, "speed/wps": 12887.792155524201, "speed/FLOPS": 202420485484771.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2333773821592331, "optim/lr": 0.0016872, "optim/total_tokens": 1474822144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.470116138458252, "created_at": "2025-01-15T13:29:45.450365+00:00"} {"global_step": 2814, "acc_step": 0, "speed/wps": 12891.984002312622, "speed/FLOPS": 202486324198784.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20082366466522217, "optim/lr": 0.0016878, "optim/total_tokens": 1475346432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.489304542541504, "created_at": "2025-01-15T13:29:55.620860+00:00"} {"global_step": 2815, "acc_step": 0, "speed/wps": 12888.585637562152, "speed/FLOPS": 202432948210535.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16779908537864685, "optim/lr": 0.0016883999999999998, "optim/total_tokens": 1475870720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 3.320201873779297, "created_at": "2025-01-15T13:30:05.794031+00:00"} {"global_step": 2816, "acc_step": 0, "speed/wps": 12886.564437288176, "speed/FLOPS": 202401202482812.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17194399237632751, "optim/lr": 0.001689, "optim/total_tokens": 1476395008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.5755558013916016, "created_at": "2025-01-15T13:30:15.971668+00:00"} {"global_step": 2817, "acc_step": 0, "speed/wps": 12891.18398053104, "speed/FLOPS": 202473758757359.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15207073092460632, "optim/lr": 0.0016896, "optim/total_tokens": 1476919296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.4490036964416504, "created_at": "2025-01-15T13:30:26.145561+00:00"} {"global_step": 2818, "acc_step": 0, "speed/wps": 12891.489700072238, "speed/FLOPS": 202478560502856.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12479925900697708, "optim/lr": 0.0016902, "optim/total_tokens": 1477443584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.41505765914917, "created_at": "2025-01-15T13:30:36.319551+00:00"} {"global_step": 2819, "acc_step": 0, "speed/wps": 12892.426414892392, "speed/FLOPS": 202493272896288.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12856973707675934, "optim/lr": 0.0016908, "optim/total_tokens": 1477967872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 3.393237590789795, "created_at": "2025-01-15T13:30:46.489681+00:00"} {"global_step": 2820, "acc_step": 0, "speed/wps": 12889.411939968208, "speed/FLOPS": 202445926425281.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1292230635881424, "optim/lr": 0.0016914, "optim/total_tokens": 1478492160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.495558738708496, "created_at": "2025-01-15T13:30:56.662221+00:00"} {"global_step": 2821, "acc_step": 0, "speed/wps": 12889.870968267987, "speed/FLOPS": 202453136095500.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13148514926433563, "optim/lr": 0.001692, "optim/total_tokens": 1479016448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.446394920349121, "created_at": "2025-01-15T13:31:06.834498+00:00"} {"global_step": 2822, "acc_step": 0, "speed/wps": 12888.671878410789, "speed/FLOPS": 202434302741571.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16281995177268982, "optim/lr": 0.0016926, "optim/total_tokens": 1479540736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.3711037635803223, "created_at": "2025-01-15T13:31:17.007886+00:00"} {"global_step": 2823, "acc_step": 0, "speed/wps": 12891.135099198156, "speed/FLOPS": 202472991008856.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16107943654060364, "optim/lr": 0.0016932000000000002, "optim/total_tokens": 1480065024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302869, "loss/out": 3.498253583908081, "created_at": "2025-01-15T13:31:27.178963+00:00"} {"global_step": 2824, "acc_step": 0, "speed/wps": 12891.895377713621, "speed/FLOPS": 202484932227674.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14689883589744568, "optim/lr": 0.0016938, "optim/total_tokens": 1480589312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 3.4722330570220947, "created_at": "2025-01-15T13:31:37.349657+00:00"} {"global_step": 2825, "acc_step": 0, "speed/wps": 12889.311923907582, "speed/FLOPS": 202444355535612.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15526950359344482, "optim/lr": 0.0016944, "optim/total_tokens": 1481113600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.3362932205200195, "created_at": "2025-01-15T13:31:47.527153+00:00"} {"global_step": 2826, "acc_step": 0, "speed/wps": 12891.777917691412, "speed/FLOPS": 202483087356618.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1329103708267212, "optim/lr": 0.001695, "optim/total_tokens": 1481637888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 3.371075391769409, "created_at": "2025-01-15T13:31:57.698020+00:00"} {"global_step": 2827, "acc_step": 0, "speed/wps": 12894.319441984277, "speed/FLOPS": 202523005488058.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13542218506336212, "optim/lr": 0.0016956000000000002, "optim/total_tokens": 1482162176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.216279983520508, "created_at": "2025-01-15T13:32:07.870771+00:00"} {"global_step": 2828, "acc_step": 0, "speed/wps": 12893.668872595563, "speed/FLOPS": 202512787401832.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14738206565380096, "optim/lr": 0.0016962000000000001, "optim/total_tokens": 1482686464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 3.34267520904541, "created_at": "2025-01-15T13:32:18.041113+00:00"} {"global_step": 2829, "acc_step": 0, "speed/wps": 12895.282650744368, "speed/FLOPS": 202538134005223.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14453119039535522, "optim/lr": 0.0016968, "optim/total_tokens": 1483210752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.361398220062256, "created_at": "2025-01-15T13:32:28.211912+00:00"} {"global_step": 2830, "acc_step": 0, "speed/wps": 12891.718387244648, "speed/FLOPS": 202482152349148.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14128008484840393, "optim/lr": 0.0016974, "optim/total_tokens": 1483735040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 3.4659264087677, "created_at": "2025-01-15T13:32:38.382809+00:00"} {"global_step": 2831, "acc_step": 0, "speed/wps": 12891.371155171775, "speed/FLOPS": 202476698592297.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14359953999519348, "optim/lr": 0.0016979999999999999, "optim/total_tokens": 1484259328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.4373064041137695, "created_at": "2025-01-15T13:32:48.555615+00:00"} {"global_step": 2832, "acc_step": 0, "speed/wps": 12883.734328278666, "speed/FLOPS": 202356751731841.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11789696663618088, "optim/lr": 0.0016986000000000002, "optim/total_tokens": 1484783616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.3676748275756836, "created_at": "2025-01-15T13:32:58.732708+00:00"} {"global_step": 2833, "acc_step": 0, "speed/wps": 12892.598176180716, "speed/FLOPS": 202495970643346.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.15777051448822021, "optim/lr": 0.0016992, "optim/total_tokens": 1485307904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.3286242485046387, "created_at": "2025-01-15T13:33:08.903853+00:00"} {"global_step": 2834, "acc_step": 0, "speed/wps": 12888.073692965227, "speed/FLOPS": 202424907417155.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13170446455478668, "optim/lr": 0.0016998, "optim/total_tokens": 1485832192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 3.366868019104004, "created_at": "2025-01-15T13:33:19.081103+00:00"} {"global_step": 2835, "acc_step": 0, "speed/wps": 12896.699335388574, "speed/FLOPS": 202560384984289.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13319967687129974, "optim/lr": 0.0017004, "optim/total_tokens": 1486356480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.5020534992218018, "created_at": "2025-01-15T13:33:29.248828+00:00"} {"global_step": 2836, "acc_step": 0, "speed/wps": 12887.233939326172, "speed/FLOPS": 202411717932311.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15125125646591187, "optim/lr": 0.0017009999999999998, "optim/total_tokens": 1486880768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.3807270526885986, "created_at": "2025-01-15T13:33:39.426007+00:00"} {"global_step": 2837, "acc_step": 0, "speed/wps": 12887.99302010764, "speed/FLOPS": 202423640339071.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14650782942771912, "optim/lr": 0.0017016000000000002, "optim/total_tokens": 1487405056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.5403523445129395, "created_at": "2025-01-15T13:33:49.599661+00:00"} {"global_step": 2838, "acc_step": 0, "speed/wps": 12888.437751917798, "speed/FLOPS": 202430625463275.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18333248794078827, "optim/lr": 0.0017022, "optim/total_tokens": 1487929344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 3.4657046794891357, "created_at": "2025-01-15T13:33:59.773294+00:00"} {"global_step": 2839, "acc_step": 0, "speed/wps": 12890.591379143238, "speed/FLOPS": 202464451138244.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16847938299179077, "optim/lr": 0.0017028, "optim/total_tokens": 1488453632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 3.4520788192749023, "created_at": "2025-01-15T13:34:09.944800+00:00"} {"global_step": 2840, "acc_step": 0, "speed/wps": 12889.537969780898, "speed/FLOPS": 202447905896672.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16211456060409546, "optim/lr": 0.0017033999999999999, "optim/total_tokens": 1488977920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.352058172225952, "created_at": "2025-01-15T13:34:20.119554+00:00"} {"global_step": 2841, "acc_step": 0, "speed/wps": 12896.425911907852, "speed/FLOPS": 202556090492801.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1918300837278366, "optim/lr": 0.0017039999999999998, "optim/total_tokens": 1489502208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 3.5196571350097656, "created_at": "2025-01-15T13:34:30.286585+00:00"} {"global_step": 2842, "acc_step": 0, "speed/wps": 12886.245393126763, "speed/FLOPS": 202396191455845.3, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17323465645313263, "optim/lr": 0.0017046000000000001, "optim/total_tokens": 1490026496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.4760379791259766, "created_at": "2025-01-15T13:34:40.462951+00:00"} {"global_step": 2843, "acc_step": 0, "speed/wps": 12883.893451253263, "speed/FLOPS": 202359250976815.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1394607275724411, "optim/lr": 0.0017052, "optim/total_tokens": 1490550784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.33072829246521, "created_at": "2025-01-15T13:34:50.640555+00:00"} {"global_step": 2844, "acc_step": 0, "speed/wps": 12888.831734804218, "speed/FLOPS": 202436813505895.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.16264165937900543, "optim/lr": 0.0017058, "optim/total_tokens": 1491075072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.441079616546631, "created_at": "2025-01-15T13:35:00.813850+00:00"} {"global_step": 2845, "acc_step": 0, "speed/wps": 12892.831159366766, "speed/FLOPS": 202499629964435.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17306268215179443, "optim/lr": 0.0017063999999999998, "optim/total_tokens": 1491599360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.47564697265625, "created_at": "2025-01-15T13:35:10.983885+00:00"} {"global_step": 2846, "acc_step": 0, "speed/wps": 12894.154598842386, "speed/FLOPS": 202520416399996.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13441401720046997, "optim/lr": 0.001707, "optim/total_tokens": 1492123648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.468687057495117, "created_at": "2025-01-15T13:35:21.152598+00:00"} {"global_step": 2847, "acc_step": 0, "speed/wps": 12889.316356520034, "speed/FLOPS": 202444425155882.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1498742401599884, "optim/lr": 0.0017076, "optim/total_tokens": 1492647936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.3230412006378174, "created_at": "2025-01-15T13:35:31.332463+00:00"} {"global_step": 2848, "acc_step": 0, "speed/wps": 12895.008905705785, "speed/FLOPS": 202533834463227.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1378195732831955, "optim/lr": 0.0017082, "optim/total_tokens": 1493172224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.441833257675171, "created_at": "2025-01-15T13:35:41.500658+00:00"} {"global_step": 2849, "acc_step": 0, "speed/wps": 12891.461474167185, "speed/FLOPS": 202478117176231.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11878857761621475, "optim/lr": 0.0017088000000000001, "optim/total_tokens": 1493696512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339038, "loss/out": 3.355818748474121, "created_at": "2025-01-15T13:35:51.671554+00:00"} {"global_step": 2850, "acc_step": 0, "speed/wps": 12891.26733482157, "speed/FLOPS": 202475067951033.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1746789664030075, "optim/lr": 0.0017094, "optim/total_tokens": 1494220800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.4507243633270264, "created_at": "2025-01-15T13:36:01.844209+00:00"} {"global_step": 2851, "acc_step": 0, "speed/wps": 12887.037745395392, "speed/FLOPS": 202408636437030.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19671696424484253, "optim/lr": 0.00171, "optim/total_tokens": 1494745088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.3474555015563965, "created_at": "2025-01-15T13:36:12.018894+00:00"} {"global_step": 2852, "acc_step": 0, "speed/wps": 12890.95607390593, "speed/FLOPS": 202470179170635.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1820923238992691, "optim/lr": 0.0017106, "optim/total_tokens": 1495269376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.463449239730835, "created_at": "2025-01-15T13:36:22.191291+00:00"} {"global_step": 2853, "acc_step": 0, "speed/wps": 12889.126445123356, "speed/FLOPS": 202441442336432.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17305313050746918, "optim/lr": 0.0017112000000000002, "optim/total_tokens": 1495793664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 3.4769530296325684, "created_at": "2025-01-15T13:36:32.369238+00:00"} {"global_step": 2854, "acc_step": 0, "speed/wps": 12890.323649076596, "speed/FLOPS": 202460246069648.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.183674156665802, "optim/lr": 0.0017118, "optim/total_tokens": 1496317952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.5143449306488037, "created_at": "2025-01-15T13:36:42.544159+00:00"} {"global_step": 2855, "acc_step": 0, "speed/wps": 12889.872047298128, "speed/FLOPS": 202453153043151.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17995613813400269, "optim/lr": 0.0017124, "optim/total_tokens": 1496842240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.4054012298583984, "created_at": "2025-01-15T13:36:52.716371+00:00"} {"global_step": 2856, "acc_step": 0, "speed/wps": 12895.57492664454, "speed/FLOPS": 202542724599864.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2031441628932953, "optim/lr": 0.001713, "optim/total_tokens": 1497366528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456592, "loss/out": 3.454646348953247, "created_at": "2025-01-15T13:37:02.883961+00:00"} {"global_step": 2857, "acc_step": 0, "speed/wps": 12891.232732345798, "speed/FLOPS": 202474524471602.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1591588407754898, "optim/lr": 0.0017136000000000002, "optim/total_tokens": 1497890816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329132, "loss/out": 3.445000171661377, "created_at": "2025-01-15T13:37:13.055482+00:00"} {"global_step": 2858, "acc_step": 0, "speed/wps": 12886.918291142587, "speed/FLOPS": 202406760243842.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18409258127212524, "optim/lr": 0.0017142000000000001, "optim/total_tokens": 1498415104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.4379959106445312, "created_at": "2025-01-15T13:37:23.230181+00:00"} {"global_step": 2859, "acc_step": 0, "speed/wps": 12887.109354107188, "speed/FLOPS": 202409761150249.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17818084359169006, "optim/lr": 0.0017148, "optim/total_tokens": 1498939392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.423530101776123, "created_at": "2025-01-15T13:37:33.405882+00:00"} {"global_step": 2860, "acc_step": 0, "speed/wps": 12882.42955799386, "speed/FLOPS": 202336258521576.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16409353911876678, "optim/lr": 0.0017154, "optim/total_tokens": 1499463680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 3.5463290214538574, "created_at": "2025-01-15T13:37:43.588685+00:00"} {"global_step": 2861, "acc_step": 0, "speed/wps": 12889.363848818994, "speed/FLOPS": 202445171087698.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1393217146396637, "optim/lr": 0.0017159999999999999, "optim/total_tokens": 1499987968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.2486908435821533, "created_at": "2025-01-15T13:37:53.761211+00:00"} {"global_step": 2862, "acc_step": 0, "speed/wps": 12887.851915258212, "speed/FLOPS": 202421424093512.47, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13729926943778992, "optim/lr": 0.0017166000000000002, "optim/total_tokens": 1500512256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.505061388015747, "created_at": "2025-01-15T13:38:03.944311+00:00"} {"global_step": 2863, "acc_step": 0, "speed/wps": 12888.299099372938, "speed/FLOPS": 202428447734529.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1234082281589508, "optim/lr": 0.0017172, "optim/total_tokens": 1501036544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 3.4604854583740234, "created_at": "2025-01-15T13:38:14.119330+00:00"} {"global_step": 2864, "acc_step": 0, "speed/wps": 12888.14715790876, "speed/FLOPS": 202426061285045.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1367960274219513, "optim/lr": 0.0017178, "optim/total_tokens": 1501560832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.336817741394043, "created_at": "2025-01-15T13:38:24.293655+00:00"} {"global_step": 2865, "acc_step": 0, "speed/wps": 12892.644196256446, "speed/FLOPS": 202496693451873.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13043473660945892, "optim/lr": 0.0017184, "optim/total_tokens": 1502085120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 3.3685944080352783, "created_at": "2025-01-15T13:38:34.466451+00:00"} {"global_step": 2866, "acc_step": 0, "speed/wps": 12885.380850755855, "speed/FLOPS": 202382612629904.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.11840809136629105, "optim/lr": 0.0017189999999999998, "optim/total_tokens": 1502609408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.3838229179382324, "created_at": "2025-01-15T13:38:44.642833+00:00"} {"global_step": 2867, "acc_step": 0, "speed/wps": 12896.800790958176, "speed/FLOPS": 202561978483424.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15636126697063446, "optim/lr": 0.0017196000000000002, "optim/total_tokens": 1503133696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.425922155380249, "created_at": "2025-01-15T13:38:54.812206+00:00"} {"global_step": 2868, "acc_step": 0, "speed/wps": 12890.839279223013, "speed/FLOPS": 202468344749647.94, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1505478173494339, "optim/lr": 0.0017202, "optim/total_tokens": 1503657984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417683, "loss/out": 3.3924214839935303, "created_at": "2025-01-15T13:39:04.988585+00:00"} {"global_step": 2869, "acc_step": 0, "speed/wps": 12887.386025653188, "speed/FLOPS": 202414106657066.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10762333124876022, "optim/lr": 0.0017208, "optim/total_tokens": 1504182272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 3.289376735687256, "created_at": "2025-01-15T13:39:15.167182+00:00"} {"global_step": 2870, "acc_step": 0, "speed/wps": 12890.989875519905, "speed/FLOPS": 202470710071431.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14166171848773956, "optim/lr": 0.0017213999999999999, "optim/total_tokens": 1504706560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.443910598754883, "created_at": "2025-01-15T13:39:25.340848+00:00"} {"global_step": 2871, "acc_step": 0, "speed/wps": 12888.613711596608, "speed/FLOPS": 202433389151824.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11477477103471756, "optim/lr": 0.001722, "optim/total_tokens": 1505230848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.3701891899108887, "created_at": "2025-01-15T13:39:35.516550+00:00"} {"global_step": 2872, "acc_step": 0, "speed/wps": 12897.557362288317, "speed/FLOPS": 202573861475802.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16400963068008423, "optim/lr": 0.0017226000000000001, "optim/total_tokens": 1505755136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.403965473175049, "created_at": "2025-01-15T13:39:45.683264+00:00"} {"global_step": 2873, "acc_step": 0, "speed/wps": 12889.31645803929, "speed/FLOPS": 202444426750381.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1879226714372635, "optim/lr": 0.0017232, "optim/total_tokens": 1506279424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 3.459540843963623, "created_at": "2025-01-15T13:39:55.858791+00:00"} {"global_step": 2874, "acc_step": 0, "speed/wps": 12890.669657004655, "speed/FLOPS": 202465680599623.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16102176904678345, "optim/lr": 0.0017238, "optim/total_tokens": 1506803712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 3.3551974296569824, "created_at": "2025-01-15T13:40:06.033064+00:00"} {"global_step": 2875, "acc_step": 0, "speed/wps": 12888.340973132656, "speed/FLOPS": 202429105419466.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16824723780155182, "optim/lr": 0.0017244, "optim/total_tokens": 1507328000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.3844592571258545, "created_at": "2025-01-15T13:40:16.208561+00:00"} {"global_step": 2876, "acc_step": 0, "speed/wps": 12888.387496310606, "speed/FLOPS": 202429836129905.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1577945500612259, "optim/lr": 0.001725, "optim/total_tokens": 1507852288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.4115233421325684, "created_at": "2025-01-15T13:40:26.382373+00:00"} {"global_step": 2877, "acc_step": 0, "speed/wps": 12890.013869319757, "speed/FLOPS": 202455380552885.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1358645111322403, "optim/lr": 0.0017256, "optim/total_tokens": 1508376576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 3.3589816093444824, "created_at": "2025-01-15T13:40:36.555293+00:00"} {"global_step": 2878, "acc_step": 0, "speed/wps": 12890.029440063809, "speed/FLOPS": 202455625112817.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17123694717884064, "optim/lr": 0.0017262, "optim/total_tokens": 1508900864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.327479362487793, "created_at": "2025-01-15T13:40:46.727972+00:00"} {"global_step": 2879, "acc_step": 0, "speed/wps": 12892.843839672485, "speed/FLOPS": 202499829126061.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13751527667045593, "optim/lr": 0.0017268000000000001, "optim/total_tokens": 1509425152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.440412998199463, "created_at": "2025-01-15T13:40:56.897717+00:00"} {"global_step": 2880, "acc_step": 0, "speed/wps": 12893.717277015427, "speed/FLOPS": 202513547659761.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2015516310930252, "optim/lr": 0.0017274, "optim/total_tokens": 1509949440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 3.4240307807922363, "created_at": "2025-01-15T13:41:07.068279+00:00"} {"global_step": 2881, "acc_step": 0, "speed/wps": 12891.928191087381, "speed/FLOPS": 202485447606799.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17920249700546265, "optim/lr": 0.001728, "optim/total_tokens": 1510473728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.457536220550537, "created_at": "2025-01-15T13:41:17.240114+00:00"} {"global_step": 2882, "acc_step": 0, "speed/wps": 12893.634650581715, "speed/FLOPS": 202512249898079.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1568012684583664, "optim/lr": 0.0017286000000000003, "optim/total_tokens": 1510998016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.489811897277832, "created_at": "2025-01-15T13:41:27.413066+00:00"} {"global_step": 2883, "acc_step": 0, "speed/wps": 12888.51529307911, "speed/FLOPS": 202431843353766.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1254381239414215, "optim/lr": 0.0017292000000000002, "optim/total_tokens": 1511522304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.350571632385254, "created_at": "2025-01-15T13:41:37.592480+00:00"} {"global_step": 2884, "acc_step": 0, "speed/wps": 12892.932909601837, "speed/FLOPS": 202501228091697.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17756401002407074, "optim/lr": 0.0017298, "optim/total_tokens": 1512046592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365310, "loss/out": 3.3287510871887207, "created_at": "2025-01-15T13:41:47.763759+00:00"} {"global_step": 2885, "acc_step": 0, "speed/wps": 12891.03683463576, "speed/FLOPS": 202471447628874.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15675176680088043, "optim/lr": 0.0017304, "optim/total_tokens": 1512570880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.4791111946105957, "created_at": "2025-01-15T13:41:57.936343+00:00"} {"global_step": 2886, "acc_step": 0, "speed/wps": 12891.31230155011, "speed/FLOPS": 202475774215295.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1521630734205246, "optim/lr": 0.001731, "optim/total_tokens": 1513095168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 3.462010383605957, "created_at": "2025-01-15T13:42:08.113999+00:00"} {"global_step": 2887, "acc_step": 0, "speed/wps": 12895.357318018347, "speed/FLOPS": 202539306757364.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.156458780169487, "optim/lr": 0.0017316000000000002, "optim/total_tokens": 1513619456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.4141507148742676, "created_at": "2025-01-15T13:42:18.286436+00:00"} {"global_step": 2888, "acc_step": 0, "speed/wps": 12892.316076103403, "speed/FLOPS": 202491539873986.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15213266015052795, "optim/lr": 0.0017322000000000001, "optim/total_tokens": 1514143744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 3.4852871894836426, "created_at": "2025-01-15T13:42:28.460079+00:00"} {"global_step": 2889, "acc_step": 0, "speed/wps": 12897.793161809144, "speed/FLOPS": 202577565031299.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.140095055103302, "optim/lr": 0.0017328, "optim/total_tokens": 1514668032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.45872163772583, "created_at": "2025-01-15T13:42:38.625982+00:00"} {"global_step": 2890, "acc_step": 0, "speed/wps": 12890.050551356364, "speed/FLOPS": 202455956694676.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1302923858165741, "optim/lr": 0.0017334, "optim/total_tokens": 1515192320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.3735134601593018, "created_at": "2025-01-15T13:42:48.799162+00:00"} {"global_step": 2891, "acc_step": 0, "speed/wps": 12893.628785690255, "speed/FLOPS": 202512157781899.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13928519189357758, "optim/lr": 0.0017339999999999999, "optim/total_tokens": 1515716608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.4015185832977295, "created_at": "2025-01-15T13:42:58.968428+00:00"} {"global_step": 2892, "acc_step": 0, "speed/wps": 12892.505635072539, "speed/FLOPS": 202494517158077.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14532078802585602, "optim/lr": 0.0017346000000000002, "optim/total_tokens": 1516240896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.281371593475342, "created_at": "2025-01-15T13:43:09.139647+00:00"} {"global_step": 2893, "acc_step": 0, "speed/wps": 12893.805332482087, "speed/FLOPS": 202514930691865.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14429955184459686, "optim/lr": 0.0017352000000000001, "optim/total_tokens": 1516765184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 3.407741069793701, "created_at": "2025-01-15T13:43:19.309034+00:00"} {"global_step": 2894, "acc_step": 0, "speed/wps": 12891.20932642209, "speed/FLOPS": 202474156849407.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15899762511253357, "optim/lr": 0.0017358, "optim/total_tokens": 1517289472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.428957462310791, "created_at": "2025-01-15T13:43:29.483415+00:00"} {"global_step": 2895, "acc_step": 0, "speed/wps": 12897.03859563393, "speed/FLOPS": 202565713532634.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.14560456573963165, "optim/lr": 0.0017364, "optim/total_tokens": 1517813760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 3.3687350749969482, "created_at": "2025-01-15T13:43:39.651333+00:00"} {"global_step": 2896, "acc_step": 0, "speed/wps": 12894.1195260505, "speed/FLOPS": 202519865533605.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14601477980613708, "optim/lr": 0.0017369999999999998, "optim/total_tokens": 1518338048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.367314577102661, "created_at": "2025-01-15T13:43:49.822696+00:00"} {"global_step": 2897, "acc_step": 0, "speed/wps": 12894.095850931104, "speed/FLOPS": 202519493683321.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15317945182323456, "optim/lr": 0.0017376000000000002, "optim/total_tokens": 1518862336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.4269158840179443, "created_at": "2025-01-15T13:43:59.993036+00:00"} {"global_step": 2898, "acc_step": 0, "speed/wps": 12893.391580404066, "speed/FLOPS": 202508432146925.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17964792251586914, "optim/lr": 0.0017382, "optim/total_tokens": 1519386624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.3135786056518555, "created_at": "2025-01-15T13:44:10.179726+00:00"} {"global_step": 2899, "acc_step": 0, "speed/wps": 12888.347439877964, "speed/FLOPS": 202429206988587.84, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1822660267353058, "optim/lr": 0.0017388, "optim/total_tokens": 1519910912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349184, "loss/out": 3.440037488937378, "created_at": "2025-01-15T13:44:20.352997+00:00"} {"global_step": 2900, "acc_step": 0, "speed/wps": 12894.256919300893, "speed/FLOPS": 202522023483400.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1701623648405075, "optim/lr": 0.0017393999999999999, "optim/total_tokens": 1520435200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 3.346743583679199, "created_at": "2025-01-15T13:44:30.526184+00:00"} {"global_step": 2901, "acc_step": 0, "speed/wps": 12886.12076684674, "speed/FLOPS": 202394234028863.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16666507720947266, "optim/lr": 0.00174, "optim/total_tokens": 1520959488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.435945510864258, "created_at": "2025-01-15T13:44:40.703080+00:00"} {"global_step": 2902, "acc_step": 0, "speed/wps": 12885.981187236666, "speed/FLOPS": 202392041739284.4, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20394733548164368, "optim/lr": 0.0017406000000000001, "optim/total_tokens": 1521483776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.4317688941955566, "created_at": "2025-01-15T13:44:50.879403+00:00"} {"global_step": 2903, "acc_step": 0, "speed/wps": 12872.140781610255, "speed/FLOPS": 202174659150211.25, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20605053007602692, "optim/lr": 0.0017412, "optim/total_tokens": 1522008064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472295, "loss/out": 3.52382230758667, "created_at": "2025-01-15T13:45:01.067771+00:00"} {"global_step": 2904, "acc_step": 0, "speed/wps": 12884.339640000902, "speed/FLOPS": 202366258984224.44, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1783250868320465, "optim/lr": 0.0017418, "optim/total_tokens": 1522532352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290845, "loss/out": 3.5291531085968018, "created_at": "2025-01-15T13:45:11.245023+00:00"} {"global_step": 2905, "acc_step": 0, "speed/wps": 12886.271989172668, "speed/FLOPS": 202396609183293.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15981212258338928, "optim/lr": 0.0017424, "optim/total_tokens": 1523056640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.380889892578125, "created_at": "2025-01-15T13:45:21.422985+00:00"} {"global_step": 2906, "acc_step": 0, "speed/wps": 12894.764334066173, "speed/FLOPS": 202529993129550.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16538779437541962, "optim/lr": 0.001743, "optim/total_tokens": 1523580928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.46785306930542, "created_at": "2025-01-15T13:45:31.592712+00:00"} {"global_step": 2907, "acc_step": 0, "speed/wps": 12889.101865249871, "speed/FLOPS": 202441056275742.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13481828570365906, "optim/lr": 0.0017436, "optim/total_tokens": 1524105216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 3.4837546348571777, "created_at": "2025-01-15T13:45:41.769726+00:00"} {"global_step": 2908, "acc_step": 0, "speed/wps": 12887.953577474096, "speed/FLOPS": 202423020838311.44, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1523096263408661, "optim/lr": 0.0017442000000000002, "optim/total_tokens": 1524629504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 3.3174729347229004, "created_at": "2025-01-15T13:45:51.948379+00:00"} {"global_step": 2909, "acc_step": 0, "speed/wps": 12886.710990228024, "speed/FLOPS": 202403504298118.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14293041825294495, "optim/lr": 0.0017448000000000001, "optim/total_tokens": 1525153792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.4272067546844482, "created_at": "2025-01-15T13:46:02.125451+00:00"} {"global_step": 2910, "acc_step": 0, "speed/wps": 12884.047631368854, "speed/FLOPS": 202361672587396.22, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12376078963279724, "optim/lr": 0.0017454, "optim/total_tokens": 1525678080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.3130037784576416, "created_at": "2025-01-15T13:46:12.307802+00:00"} {"global_step": 2911, "acc_step": 0, "speed/wps": 12894.849754978104, "speed/FLOPS": 202531334782352.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13071247935295105, "optim/lr": 0.001746, "optim/total_tokens": 1526202368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.4839279651641846, "created_at": "2025-01-15T13:46:22.485663+00:00"} {"global_step": 2912, "acc_step": 0, "speed/wps": 12887.770145255461, "speed/FLOPS": 202420139783255.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1478111445903778, "optim/lr": 0.0017466000000000003, "optim/total_tokens": 1526726656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399303, "loss/out": 3.39923357963562, "created_at": "2025-01-15T13:46:32.660830+00:00"} {"global_step": 2913, "acc_step": 0, "speed/wps": 12882.781464260972, "speed/FLOPS": 202341785693071.0, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17229273915290833, "optim/lr": 0.0017472000000000002, "optim/total_tokens": 1527250944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.510857343673706, "created_at": "2025-01-15T13:46:42.840943+00:00"} {"global_step": 2914, "acc_step": 0, "speed/wps": 12887.9717208195, "speed/FLOPS": 202423305804482.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17575936019420624, "optim/lr": 0.0017478, "optim/total_tokens": 1527775232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.493567943572998, "created_at": "2025-01-15T13:46:53.016323+00:00"} {"global_step": 2915, "acc_step": 0, "speed/wps": 12893.398581398174, "speed/FLOPS": 202508542107158.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1695406287908554, "optim/lr": 0.0017484, "optim/total_tokens": 1528299520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.431422710418701, "created_at": "2025-01-15T13:47:03.188335+00:00"} {"global_step": 2916, "acc_step": 0, "speed/wps": 12891.894784887345, "speed/FLOPS": 202484922916523.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14470578730106354, "optim/lr": 0.001749, "optim/total_tokens": 1528823808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 3.438732147216797, "created_at": "2025-01-15T13:47:13.359165+00:00"} {"global_step": 2917, "acc_step": 0, "speed/wps": 12892.570278478253, "speed/FLOPS": 202495532471593.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14822322130203247, "optim/lr": 0.0017496000000000002, "optim/total_tokens": 1529348096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.355034351348877, "created_at": "2025-01-15T13:47:23.532010+00:00"} {"global_step": 2918, "acc_step": 0, "speed/wps": 12888.350895735754, "speed/FLOPS": 202429261267583.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14642153680324554, "optim/lr": 0.0017502000000000001, "optim/total_tokens": 1529872384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.384063720703125, "created_at": "2025-01-15T13:47:33.712187+00:00"} {"global_step": 2919, "acc_step": 0, "speed/wps": 12863.157433626187, "speed/FLOPS": 202033563325707.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11463715881109238, "optim/lr": 0.0017508, "optim/total_tokens": 1530396672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.3924577236175537, "created_at": "2025-01-15T13:47:43.908712+00:00"} {"global_step": 2920, "acc_step": 0, "speed/wps": 12887.761325826785, "speed/FLOPS": 202420001262008.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5538851022720337, "optim/lr": 0.0017514, "optim/total_tokens": 1530920960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341575, "loss/out": 3.4470295906066895, "created_at": "2025-01-15T13:47:54.085170+00:00"} {"global_step": 2921, "acc_step": 0, "speed/wps": 12887.960485228474, "speed/FLOPS": 202423129334086.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11364468932151794, "optim/lr": 0.0017519999999999999, "optim/total_tokens": 1531445248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.4449286460876465, "created_at": "2025-01-15T13:48:04.260030+00:00"} {"global_step": 2922, "acc_step": 0, "speed/wps": 12875.423264103782, "speed/FLOPS": 202226215048375.72, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1336328238248825, "optim/lr": 0.0017526000000000002, "optim/total_tokens": 1531969536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.447174549102783, "created_at": "2025-01-15T13:48:14.444922+00:00"} {"global_step": 2923, "acc_step": 0, "speed/wps": 12887.073952637425, "speed/FLOPS": 202409205121520.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1387239694595337, "optim/lr": 0.0017532000000000001, "optim/total_tokens": 1532493824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.2645866870880127, "created_at": "2025-01-15T13:48:24.620398+00:00"} {"global_step": 2924, "acc_step": 0, "speed/wps": 12889.734054503446, "speed/FLOPS": 202450985676689.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16048242151737213, "optim/lr": 0.0017538, "optim/total_tokens": 1533018112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.3358778953552246, "created_at": "2025-01-15T13:48:34.792719+00:00"} {"global_step": 2925, "acc_step": 0, "speed/wps": 12896.256721683549, "speed/FLOPS": 202553433127836.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17715580761432648, "optim/lr": 0.0017544, "optim/total_tokens": 1533542400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 3.2551138401031494, "created_at": "2025-01-15T13:48:44.960527+00:00"} {"global_step": 2926, "acc_step": 0, "speed/wps": 12891.15782943825, "speed/FLOPS": 202473348018512.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15485629439353943, "optim/lr": 0.0017549999999999998, "optim/total_tokens": 1534066688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.4733457565307617, "created_at": "2025-01-15T13:48:55.142517+00:00"} {"global_step": 2927, "acc_step": 0, "speed/wps": 12891.957886944912, "speed/FLOPS": 202485914021048.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19804157316684723, "optim/lr": 0.0017556000000000002, "optim/total_tokens": 1534590976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 3.479665756225586, "created_at": "2025-01-15T13:49:05.317762+00:00"} {"global_step": 2928, "acc_step": 0, "speed/wps": 12891.624714493828, "speed/FLOPS": 202480681089877.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.6223861575126648, "optim/lr": 0.0017562, "optim/total_tokens": 1535115264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350240, "loss/out": 3.41528582572937, "created_at": "2025-01-15T13:49:15.490972+00:00"} {"global_step": 2929, "acc_step": 0, "speed/wps": 12892.213066042832, "speed/FLOPS": 202489921959434.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16424807906150818, "optim/lr": 0.0017568, "optim/total_tokens": 1535639552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.379207134246826, "created_at": "2025-01-15T13:49:25.662889+00:00"} {"global_step": 2930, "acc_step": 0, "speed/wps": 12892.154548397813, "speed/FLOPS": 202489002859407.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23359191417694092, "optim/lr": 0.0017573999999999999, "optim/total_tokens": 1536163840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.3704049587249756, "created_at": "2025-01-15T13:49:35.833772+00:00"} {"global_step": 2931, "acc_step": 0, "speed/wps": 12885.970119114281, "speed/FLOPS": 202391867899213.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.151670902967453, "optim/lr": 0.001758, "optim/total_tokens": 1536688128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.4648618698120117, "created_at": "2025-01-15T13:49:46.015381+00:00"} {"global_step": 2932, "acc_step": 0, "speed/wps": 12893.927303037899, "speed/FLOPS": 202516846407050.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7680716514587402, "optim/lr": 0.0017586000000000001, "optim/total_tokens": 1537212416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 3.399120569229126, "created_at": "2025-01-15T13:49:56.185603+00:00"} {"global_step": 2933, "acc_step": 0, "speed/wps": 12890.916382887803, "speed/FLOPS": 202469555768655.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2039501965045929, "optim/lr": 0.0017592, "optim/total_tokens": 1537736704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.6074271202087402, "created_at": "2025-01-15T13:50:06.359729+00:00"} {"global_step": 2934, "acc_step": 0, "speed/wps": 12898.38206353704, "speed/FLOPS": 202586814542171.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15388502180576324, "optim/lr": 0.0017598000000000002, "optim/total_tokens": 1538260992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.3672146797180176, "created_at": "2025-01-15T13:50:16.525339+00:00"} {"global_step": 2935, "acc_step": 0, "speed/wps": 12884.376425397904, "speed/FLOPS": 202366836749432.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17383335530757904, "optim/lr": 0.0017604, "optim/total_tokens": 1538785280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.481367588043213, "created_at": "2025-01-15T13:50:26.706877+00:00"} {"global_step": 2936, "acc_step": 0, "speed/wps": 12885.336320605926, "speed/FLOPS": 202381913222709.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.15256474912166595, "optim/lr": 0.001761, "optim/total_tokens": 1539309568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364017, "loss/out": 3.344883441925049, "created_at": "2025-01-15T13:50:36.883116+00:00"} {"global_step": 2937, "acc_step": 0, "speed/wps": 12892.402062178471, "speed/FLOPS": 202492890403452.22, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14188748598098755, "optim/lr": 0.0017616, "optim/total_tokens": 1539833856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.3715262413024902, "created_at": "2025-01-15T13:50:47.054856+00:00"} {"global_step": 2938, "acc_step": 0, "speed/wps": 12896.962614754717, "speed/FLOPS": 202564520148518.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.15723499655723572, "optim/lr": 0.0017622000000000002, "optim/total_tokens": 1540358144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.515533924102783, "created_at": "2025-01-15T13:50:57.221743+00:00"} {"global_step": 2939, "acc_step": 0, "speed/wps": 12891.954951127882, "speed/FLOPS": 202485867910007.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 4.191669940948486, "optim/lr": 0.0017628000000000001, "optim/total_tokens": 1540882432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.425123691558838, "created_at": "2025-01-15T13:51:07.392648+00:00"} {"global_step": 2940, "acc_step": 0, "speed/wps": 12894.227143151049, "speed/FLOPS": 202521555808050.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2690894305706024, "optim/lr": 0.0017634, "optim/total_tokens": 1541406720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 3.4513754844665527, "created_at": "2025-01-15T13:51:17.562565+00:00"} {"global_step": 2941, "acc_step": 0, "speed/wps": 12888.740336041317, "speed/FLOPS": 202435377962729.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.24828900396823883, "optim/lr": 0.001764, "optim/total_tokens": 1541931008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.4716713428497314, "created_at": "2025-01-15T13:51:27.736495+00:00"} {"global_step": 2942, "acc_step": 0, "speed/wps": 12885.908460782408, "speed/FLOPS": 202390899470383.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15979917347431183, "optim/lr": 0.0017645999999999998, "optim/total_tokens": 1542455296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.5656259059906006, "created_at": "2025-01-15T13:51:37.914920+00:00"} {"global_step": 2943, "acc_step": 0, "speed/wps": 12891.854042973013, "speed/FLOPS": 202484283008773.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18503348529338837, "optim/lr": 0.0017652000000000002, "optim/total_tokens": 1542979584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 3.3717522621154785, "created_at": "2025-01-15T13:51:48.088260+00:00"} {"global_step": 2944, "acc_step": 0, "speed/wps": 12888.880292953034, "speed/FLOPS": 202437576178348.7, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.19587236642837524, "optim/lr": 0.0017658, "optim/total_tokens": 1543503872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372174, "loss/out": 3.399132490158081, "created_at": "2025-01-15T13:51:58.266280+00:00"} {"global_step": 2945, "acc_step": 0, "speed/wps": 12896.713604568231, "speed/FLOPS": 202560609101363.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23289912939071655, "optim/lr": 0.0017664, "optim/total_tokens": 1544028160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.470439910888672, "created_at": "2025-01-15T13:52:08.436095+00:00"} {"global_step": 2946, "acc_step": 0, "speed/wps": 12893.042871535807, "speed/FLOPS": 202502955194973.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6823886632919312, "optim/lr": 0.001767, "optim/total_tokens": 1544552448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.6061558723449707, "created_at": "2025-01-15T13:52:18.609935+00:00"} {"global_step": 2947, "acc_step": 0, "speed/wps": 12893.729874013146, "speed/FLOPS": 202513745512920.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4359906017780304, "optim/lr": 0.0017675999999999998, "optim/total_tokens": 1545076736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474683, "loss/out": 3.6085610389709473, "created_at": "2025-01-15T13:52:28.779675+00:00"} {"global_step": 2948, "acc_step": 0, "speed/wps": 12888.744603011399, "speed/FLOPS": 202435444981357.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3072918653488159, "optim/lr": 0.0017682000000000002, "optim/total_tokens": 1545601024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 3.605607032775879, "created_at": "2025-01-15T13:52:38.954592+00:00"} {"global_step": 2949, "acc_step": 0, "speed/wps": 12891.1865877348, "speed/FLOPS": 202473799707077.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28222084045410156, "optim/lr": 0.0017688, "optim/total_tokens": 1546125312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.571094036102295, "created_at": "2025-01-15T13:52:49.125680+00:00"} {"global_step": 2950, "acc_step": 0, "speed/wps": 12895.176329885617, "speed/FLOPS": 202536464090036.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2461780309677124, "optim/lr": 0.0017694, "optim/total_tokens": 1546649600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.6958541870117188, "created_at": "2025-01-15T13:52:59.293949+00:00"} {"global_step": 2951, "acc_step": 0, "speed/wps": 12893.353751307099, "speed/FLOPS": 202507837988975.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19977794587612152, "optim/lr": 0.0017699999999999999, "optim/total_tokens": 1547173888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.471545696258545, "created_at": "2025-01-15T13:53:09.463267+00:00"} {"global_step": 2952, "acc_step": 0, "speed/wps": 12890.386867987645, "speed/FLOPS": 202461239009518.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18008124828338623, "optim/lr": 0.0017705999999999998, "optim/total_tokens": 1547698176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.47835373878479, "created_at": "2025-01-15T13:53:19.639568+00:00"} {"global_step": 2953, "acc_step": 0, "speed/wps": 12889.886147901158, "speed/FLOPS": 202453374512498.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1484047770500183, "optim/lr": 0.0017712000000000001, "optim/total_tokens": 1548222464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.3861145973205566, "created_at": "2025-01-15T13:53:29.814404+00:00"} {"global_step": 2954, "acc_step": 0, "speed/wps": 12894.569789242027, "speed/FLOPS": 202526937535754.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.15343770384788513, "optim/lr": 0.0017718, "optim/total_tokens": 1548746752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.4062976837158203, "created_at": "2025-01-15T13:53:39.982873+00:00"} {"global_step": 2955, "acc_step": 0, "speed/wps": 12892.246173920117, "speed/FLOPS": 202490441964141.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12173415720462799, "optim/lr": 0.0017724, "optim/total_tokens": 1549271040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 3.3427295684814453, "created_at": "2025-01-15T13:53:50.154305+00:00"} {"global_step": 2956, "acc_step": 0, "speed/wps": 12891.601112302322, "speed/FLOPS": 202480310385026.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13317817449569702, "optim/lr": 0.001773, "optim/total_tokens": 1549795328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.447706699371338, "created_at": "2025-01-15T13:54:00.325471+00:00"} {"global_step": 2957, "acc_step": 0, "speed/wps": 12895.184133896364, "speed/FLOPS": 202536586662748.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1334514170885086, "optim/lr": 0.0017736, "optim/total_tokens": 1550319616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.379150867462158, "created_at": "2025-01-15T13:54:10.494475+00:00"} {"global_step": 2958, "acc_step": 0, "speed/wps": 12891.877393372588, "speed/FLOPS": 202484649758885.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18339255452156067, "optim/lr": 0.0017742, "optim/total_tokens": 1550843904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.49299955368042, "created_at": "2025-01-15T13:54:20.670547+00:00"} {"global_step": 2959, "acc_step": 0, "speed/wps": 12895.227744371023, "speed/FLOPS": 202537271625180.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22822214663028717, "optim/lr": 0.0017748, "optim/total_tokens": 1551368192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.4672482013702393, "created_at": "2025-01-15T13:54:30.843836+00:00"} {"global_step": 2960, "acc_step": 0, "speed/wps": 12898.072110297784, "speed/FLOPS": 202581946300628.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23870237171649933, "optim/lr": 0.0017754000000000001, "optim/total_tokens": 1551892480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 3.4427695274353027, "created_at": "2025-01-15T13:54:41.009461+00:00"} {"global_step": 2961, "acc_step": 0, "speed/wps": 12895.168589686862, "speed/FLOPS": 202536342519578.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14934873580932617, "optim/lr": 0.001776, "optim/total_tokens": 1552416768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 3.502347469329834, "created_at": "2025-01-15T13:54:51.181021+00:00"} {"global_step": 2962, "acc_step": 0, "speed/wps": 12897.025703794194, "speed/FLOPS": 202565511048576.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15579113364219666, "optim/lr": 0.0017766, "optim/total_tokens": 1552941056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 3.424160957336426, "created_at": "2025-01-15T13:55:01.349515+00:00"} {"global_step": 2963, "acc_step": 0, "speed/wps": 12889.255215068144, "speed/FLOPS": 202443464845363.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.16286635398864746, "optim/lr": 0.0017772, "optim/total_tokens": 1553465344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.487044095993042, "created_at": "2025-01-15T13:55:11.526695+00:00"} {"global_step": 2964, "acc_step": 0, "speed/wps": 12894.836498895953, "speed/FLOPS": 202531126577366.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16452768445014954, "optim/lr": 0.0017778000000000002, "optim/total_tokens": 1553989632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.444983720779419, "created_at": "2025-01-15T13:55:21.697809+00:00"} {"global_step": 2965, "acc_step": 0, "speed/wps": 12891.167939402481, "speed/FLOPS": 202473506809392.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.12175619602203369, "optim/lr": 0.0017784, "optim/total_tokens": 1554513920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.449284553527832, "created_at": "2025-01-15T13:55:31.872479+00:00"} {"global_step": 2966, "acc_step": 0, "speed/wps": 12895.497670351815, "speed/FLOPS": 202541511183625.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13630710542201996, "optim/lr": 0.001779, "optim/total_tokens": 1555038208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 3.3498001098632812, "created_at": "2025-01-15T13:55:42.040585+00:00"} {"global_step": 2967, "acc_step": 0, "speed/wps": 12899.072061106675, "speed/FLOPS": 202597651902158.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12579724192619324, "optim/lr": 0.0017795999999999999, "optim/total_tokens": 1555562496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 3.438716411590576, "created_at": "2025-01-15T13:55:52.210155+00:00"} {"global_step": 2968, "acc_step": 0, "speed/wps": 12877.757676044635, "speed/FLOPS": 202262880195725.0, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.002, "optim/grad_norm": 0.12782041728496552, "optim/lr": 0.0017802000000000002, "optim/total_tokens": 1556086784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.3783748149871826, "created_at": "2025-01-15T13:56:02.392312+00:00"} {"global_step": 2969, "acc_step": 0, "speed/wps": 12871.926110492, "speed/FLOPS": 202171287445310.03, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13288211822509766, "optim/lr": 0.0017808000000000001, "optim/total_tokens": 1556611072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.514012336730957, "created_at": "2025-01-15T13:56:12.580806+00:00"} {"global_step": 2970, "acc_step": 0, "speed/wps": 12875.890151582624, "speed/FLOPS": 202233548157801.56, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.13905265927314758, "optim/lr": 0.0017814, "optim/total_tokens": 1557135360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.3389532566070557, "created_at": "2025-01-15T13:56:22.766957+00:00"} {"global_step": 2971, "acc_step": 0, "speed/wps": 12886.415511311881, "speed/FLOPS": 202398863395709.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13921812176704407, "optim/lr": 0.001782, "optim/total_tokens": 1557659648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 3.4914560317993164, "created_at": "2025-01-15T13:56:32.941806+00:00"} {"global_step": 2972, "acc_step": 0, "speed/wps": 12884.530132265805, "speed/FLOPS": 202369250927008.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11835837364196777, "optim/lr": 0.0017825999999999999, "optim/total_tokens": 1558183936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.3304989337921143, "created_at": "2025-01-15T13:56:43.118383+00:00"} {"global_step": 2973, "acc_step": 0, "speed/wps": 12889.289044356983, "speed/FLOPS": 202443996180831.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.11534711718559265, "optim/lr": 0.0017832000000000002, "optim/total_tokens": 1558708224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.4448885917663574, "created_at": "2025-01-15T13:56:53.291255+00:00"} {"global_step": 2974, "acc_step": 0, "speed/wps": 12880.512924368873, "speed/FLOPS": 202306155156765.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.13123570382595062, "optim/lr": 0.0017838, "optim/total_tokens": 1559232512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461929, "loss/out": 3.3973193168640137, "created_at": "2025-01-15T13:57:03.474007+00:00"} {"global_step": 2975, "acc_step": 0, "speed/wps": 12883.756105388971, "speed/FLOPS": 202357093771283.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11378543078899384, "optim/lr": 0.0017844, "optim/total_tokens": 1559756800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288006, "loss/out": 3.3817667961120605, "created_at": "2025-01-15T13:57:13.656605+00:00"} {"global_step": 2976, "acc_step": 0, "speed/wps": 12884.484691468404, "speed/FLOPS": 202368537216843.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12406110018491745, "optim/lr": 0.001785, "optim/total_tokens": 1560281088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326440, "loss/out": 3.4178662300109863, "created_at": "2025-01-15T13:57:23.833824+00:00"} {"global_step": 2977, "acc_step": 0, "speed/wps": 12873.398765259939, "speed/FLOPS": 202194417512080.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.14460571110248566, "optim/lr": 0.0017855999999999998, "optim/total_tokens": 1560805376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.3965225219726562, "created_at": "2025-01-15T13:57:34.022736+00:00"} {"global_step": 2978, "acc_step": 0, "speed/wps": 12901.590826093996, "speed/FLOPS": 202637212567430.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12651380896568298, "optim/lr": 0.0017862000000000002, "optim/total_tokens": 1561329664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.150871753692627, "created_at": "2025-01-15T13:57:44.185910+00:00"} {"global_step": 2979, "acc_step": 0, "speed/wps": 12890.562588081722, "speed/FLOPS": 202463998935060.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13246239721775055, "optim/lr": 0.0017868, "optim/total_tokens": 1561853952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343977, "loss/out": 3.37608003616333, "created_at": "2025-01-15T13:57:54.358194+00:00"} {"global_step": 2980, "acc_step": 0, "speed/wps": 12877.454570476286, "speed/FLOPS": 202258119506263.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.12396415323019028, "optim/lr": 0.0017874, "optim/total_tokens": 1562378240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.3689050674438477, "created_at": "2025-01-15T13:58:04.541167+00:00"} {"global_step": 2981, "acc_step": 0, "speed/wps": 12883.176595620796, "speed/FLOPS": 202347991774044.4, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12472381442785263, "optim/lr": 0.0017879999999999999, "optim/total_tokens": 1562902528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.4279391765594482, "created_at": "2025-01-15T13:58:14.720110+00:00"} {"global_step": 2982, "acc_step": 0, "speed/wps": 12876.886400717285, "speed/FLOPS": 202249195619451.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13353990018367767, "optim/lr": 0.0017886, "optim/total_tokens": 1563426816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.4328339099884033, "created_at": "2025-01-15T13:58:24.902497+00:00"} {"global_step": 2983, "acc_step": 0, "speed/wps": 12889.272142675298, "speed/FLOPS": 202443730716694.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13842995464801788, "optim/lr": 0.0017892000000000001, "optim/total_tokens": 1563951104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.4194250106811523, "created_at": "2025-01-15T13:58:35.075166+00:00"} {"global_step": 2984, "acc_step": 0, "speed/wps": 12897.013941844818, "speed/FLOPS": 202565326310998.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3804302513599396, "optim/lr": 0.0017898, "optim/total_tokens": 1564475392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374658, "loss/out": 3.3799211978912354, "created_at": "2025-01-15T13:58:45.241800+00:00"} {"global_step": 2985, "acc_step": 0, "speed/wps": 12898.689166083497, "speed/FLOPS": 202591638009666.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15317711234092712, "optim/lr": 0.0017904, "optim/total_tokens": 1564999680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.3752644062042236, "created_at": "2025-01-15T13:58:55.410478+00:00"} {"global_step": 2986, "acc_step": 0, "speed/wps": 12898.195203006899, "speed/FLOPS": 202583879640772.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13775065541267395, "optim/lr": 0.001791, "optim/total_tokens": 1565523968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.4152274131774902, "created_at": "2025-01-15T13:59:05.577755+00:00"} {"global_step": 2987, "acc_step": 0, "speed/wps": 12897.866098126648, "speed/FLOPS": 202578710596390.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1348879486322403, "optim/lr": 0.0017916, "optim/total_tokens": 1566048256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.3550567626953125, "created_at": "2025-01-15T13:59:15.745324+00:00"} {"global_step": 2988, "acc_step": 0, "speed/wps": 12899.193094478274, "speed/FLOPS": 202599552897576.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11247836798429489, "optim/lr": 0.0017922, "optim/total_tokens": 1566572544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2785797119140625, "created_at": "2025-01-15T13:59:25.911963+00:00"} {"global_step": 2989, "acc_step": 0, "speed/wps": 12891.59637463363, "speed/FLOPS": 202480235973429.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1121520921587944, "optim/lr": 0.0017928, "optim/total_tokens": 1567096832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.2795658111572266, "created_at": "2025-01-15T13:59:36.085175+00:00"} {"global_step": 2990, "acc_step": 0, "speed/wps": 12891.183360611822, "speed/FLOPS": 202473749020676.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11968640238046646, "optim/lr": 0.0017934000000000001, "optim/total_tokens": 1567621120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.215909719467163, "created_at": "2025-01-15T13:59:46.257862+00:00"} {"global_step": 2991, "acc_step": 0, "speed/wps": 12895.330244917053, "speed/FLOPS": 202538881537106.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13715597987174988, "optim/lr": 0.001794, "optim/total_tokens": 1568145408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277752, "loss/out": 3.434969186782837, "created_at": "2025-01-15T13:59:56.428350+00:00"} {"global_step": 2992, "acc_step": 0, "speed/wps": 12896.582834875191, "speed/FLOPS": 202558555183637.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14551371335983276, "optim/lr": 0.0017946, "optim/total_tokens": 1568669696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.4035515785217285, "created_at": "2025-01-15T14:00:06.601480+00:00"} {"global_step": 2993, "acc_step": 0, "speed/wps": 12889.139746128418, "speed/FLOPS": 202441651246994.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16520126163959503, "optim/lr": 0.0017952000000000003, "optim/total_tokens": 1569193984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497719, "loss/out": 3.4583683013916016, "created_at": "2025-01-15T14:00:16.774274+00:00"} {"global_step": 2994, "acc_step": 0, "speed/wps": 12893.034627189954, "speed/FLOPS": 202502825706192.97, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12207746505737305, "optim/lr": 0.0017958000000000002, "optim/total_tokens": 1569718272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 3.2470734119415283, "created_at": "2025-01-15T14:00:26.944162+00:00"} {"global_step": 2995, "acc_step": 0, "speed/wps": 12896.01192072978, "speed/FLOPS": 202549588192466.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11810129135847092, "optim/lr": 0.0017964, "optim/total_tokens": 1570242560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.4463796615600586, "created_at": "2025-01-15T14:00:37.115175+00:00"} {"global_step": 2996, "acc_step": 0, "speed/wps": 12889.728417461016, "speed/FLOPS": 202450897139191.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.17023617029190063, "optim/lr": 0.001797, "optim/total_tokens": 1570766848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.389141321182251, "created_at": "2025-01-15T14:00:47.287408+00:00"} {"global_step": 2997, "acc_step": 0, "speed/wps": 12899.028020301212, "speed/FLOPS": 202596960180790.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15087303519248962, "optim/lr": 0.0017976, "optim/total_tokens": 1571291136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.4491372108459473, "created_at": "2025-01-15T14:00:57.454300+00:00"} {"global_step": 2998, "acc_step": 0, "speed/wps": 12893.430616919699, "speed/FLOPS": 202509045269045.53, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11310921609401703, "optim/lr": 0.0017982000000000002, "optim/total_tokens": 1571815424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 3.3748795986175537, "created_at": "2025-01-15T14:01:07.624905+00:00"} {"global_step": 2999, "acc_step": 0, "speed/wps": 12896.117457718688, "speed/FLOPS": 202551245795899.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12884216010570526, "optim/lr": 0.0017988000000000001, "optim/total_tokens": 1572339712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 3.412355422973633, "created_at": "2025-01-15T14:01:17.794455+00:00"} {"global_step": 3000, "acc_step": 0, "speed/wps": 12890.357478070886, "speed/FLOPS": 202460777400489.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12183988839387894, "optim/lr": 0.0017994, "optim/total_tokens": 1572864000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.4593005180358887, "created_at": "2025-01-15T14:01:27.969514+00:00"} {"global_step": 3001, "acc_step": 0, "speed/wps": 12590.752208564973, "speed/FLOPS": 197755064942115.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.2493, "optim/grad_norm": 0.11488629132509232, "optim/lr": 0.0018, "optim/total_tokens": 1573388288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 3.453697443008423, "created_at": "2025-01-15T14:01:38.386309+00:00"} {"global_step": 3002, "acc_step": 0, "speed/wps": 12898.519812441833, "speed/FLOPS": 202588978078008.12, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16451594233512878, "optim/lr": 0.0018005999999999999, "optim/total_tokens": 1573912576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.471971035003662, "created_at": "2025-01-15T14:01:48.553847+00:00"} {"global_step": 3003, "acc_step": 0, "speed/wps": 12890.371817046593, "speed/FLOPS": 202461002613806.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15828561782836914, "optim/lr": 0.0018012000000000002, "optim/total_tokens": 1574436864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 3.3411991596221924, "created_at": "2025-01-15T14:01:58.727592+00:00"} {"global_step": 3004, "acc_step": 0, "speed/wps": 12896.12883634235, "speed/FLOPS": 202551424512820.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14442655444145203, "optim/lr": 0.0018018, "optim/total_tokens": 1574961152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.3718557357788086, "created_at": "2025-01-15T14:02:08.896796+00:00"} {"global_step": 3005, "acc_step": 0, "speed/wps": 12904.80462420657, "speed/FLOPS": 202687689683009.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15211531519889832, "optim/lr": 0.0018024, "optim/total_tokens": 1575485440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.377023220062256, "created_at": "2025-01-15T14:02:19.060091+00:00"} {"global_step": 3006, "acc_step": 0, "speed/wps": 12904.91967464966, "speed/FLOPS": 202689496708314.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1641295701265335, "optim/lr": 0.001803, "optim/total_tokens": 1576009728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 3.216505289077759, "created_at": "2025-01-15T14:02:29.225225+00:00"} {"global_step": 3007, "acc_step": 0, "speed/wps": 12901.229372938857, "speed/FLOPS": 202631535448940.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13627073168754578, "optim/lr": 0.0018035999999999998, "optim/total_tokens": 1576534016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.340573787689209, "created_at": "2025-01-15T14:02:39.391102+00:00"} {"global_step": 3008, "acc_step": 0, "speed/wps": 12892.10922032194, "speed/FLOPS": 202488290919688.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13805381953716278, "optim/lr": 0.0018042000000000002, "optim/total_tokens": 1577058304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.3062500953674316, "created_at": "2025-01-15T14:02:49.596627+00:00"} {"global_step": 3009, "acc_step": 0, "speed/wps": 12895.411149081945, "speed/FLOPS": 202540152248190.0, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14786353707313538, "optim/lr": 0.0018048, "optim/total_tokens": 1577582592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.2234654426574707, "created_at": "2025-01-15T14:02:59.765008+00:00"} {"global_step": 3010, "acc_step": 0, "speed/wps": 12906.01011812423, "speed/FLOPS": 202706623621508.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12352693825960159, "optim/lr": 0.0018054, "optim/total_tokens": 1578106880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423603, "loss/out": 3.4763102531433105, "created_at": "2025-01-15T14:03:09.926301+00:00"} {"global_step": 3011, "acc_step": 0, "speed/wps": 12900.010630198829, "speed/FLOPS": 202612393419480.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14697447419166565, "optim/lr": 0.0018059999999999999, "optim/total_tokens": 1578631168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.3672215938568115, "created_at": "2025-01-15T14:03:20.092063+00:00"} {"global_step": 3012, "acc_step": 0, "speed/wps": 12899.312797682303, "speed/FLOPS": 202601433000885.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14685440063476562, "optim/lr": 0.0018066, "optim/total_tokens": 1579155456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 3.367655038833618, "created_at": "2025-01-15T14:03:30.259079+00:00"} {"global_step": 3013, "acc_step": 0, "speed/wps": 12899.010178020331, "speed/FLOPS": 202596679943251.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12319991737604141, "optim/lr": 0.0018072000000000001, "optim/total_tokens": 1579679744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 3.353407859802246, "created_at": "2025-01-15T14:03:40.425459+00:00"} {"global_step": 3014, "acc_step": 0, "speed/wps": 12901.898532091, "speed/FLOPS": 202642045512945.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15541306138038635, "optim/lr": 0.0018078, "optim/total_tokens": 1580204032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.3786959648132324, "created_at": "2025-01-15T14:03:50.592359+00:00"} {"global_step": 3015, "acc_step": 0, "speed/wps": 12901.388444649676, "speed/FLOPS": 202634033888746.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12271122634410858, "optim/lr": 0.0018084, "optim/total_tokens": 1580728320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.262866973876953, "created_at": "2025-01-15T14:04:00.755594+00:00"} {"global_step": 3016, "acc_step": 0, "speed/wps": 12905.631573286992, "speed/FLOPS": 202700678054664.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1308385282754898, "optim/lr": 0.001809, "optim/total_tokens": 1581252608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308327, "loss/out": 3.3287110328674316, "created_at": "2025-01-15T14:04:10.917303+00:00"} {"global_step": 3017, "acc_step": 0, "speed/wps": 12902.94594479237, "speed/FLOPS": 202658496568716.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11823686212301254, "optim/lr": 0.0018096, "optim/total_tokens": 1581776896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.3321731090545654, "created_at": "2025-01-15T14:04:21.079840+00:00"} {"global_step": 3018, "acc_step": 0, "speed/wps": 12902.170417008925, "speed/FLOPS": 202646315839190.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12891115248203278, "optim/lr": 0.0018102, "optim/total_tokens": 1582301184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.396059513092041, "created_at": "2025-01-15T14:04:31.243490+00:00"} {"global_step": 3019, "acc_step": 0, "speed/wps": 12902.147272966453, "speed/FLOPS": 202645952330200.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1313370019197464, "optim/lr": 0.0018108000000000002, "optim/total_tokens": 1582825472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 3.3147406578063965, "created_at": "2025-01-15T14:04:41.411705+00:00"} {"global_step": 3020, "acc_step": 0, "speed/wps": 12894.996671291354, "speed/FLOPS": 202533642304936.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.11431735754013062, "optim/lr": 0.0018114000000000001, "optim/total_tokens": 1583349760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.342033624649048, "created_at": "2025-01-15T14:04:51.580260+00:00"} {"global_step": 3021, "acc_step": 0, "speed/wps": 12901.133371705446, "speed/FLOPS": 202630027617649.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1102684736251831, "optim/lr": 0.001812, "optim/total_tokens": 1583874048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.2903895378112793, "created_at": "2025-01-15T14:05:01.747018+00:00"} {"global_step": 3022, "acc_step": 0, "speed/wps": 12896.442877263169, "speed/FLOPS": 202556356957019.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12963847815990448, "optim/lr": 0.0018126, "optim/total_tokens": 1584398336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392297, "loss/out": 3.418658971786499, "created_at": "2025-01-15T14:05:11.915382+00:00"} {"global_step": 3023, "acc_step": 0, "speed/wps": 12907.667493866618, "speed/FLOPS": 202732654985015.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17492008209228516, "optim/lr": 0.0018132000000000003, "optim/total_tokens": 1584922624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.3627963066101074, "created_at": "2025-01-15T14:05:22.074165+00:00"} {"global_step": 3024, "acc_step": 0, "speed/wps": 12904.49027886702, "speed/FLOPS": 202682752457496.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14283427596092224, "optim/lr": 0.0018138000000000002, "optim/total_tokens": 1585446912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.4727988243103027, "created_at": "2025-01-15T14:05:32.237331+00:00"} {"global_step": 3025, "acc_step": 0, "speed/wps": 12904.272300626577, "speed/FLOPS": 202679328809696.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13816548883914948, "optim/lr": 0.0018144, "optim/total_tokens": 1585971200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 3.3766965866088867, "created_at": "2025-01-15T14:05:42.398954+00:00"} {"global_step": 3026, "acc_step": 0, "speed/wps": 12901.644329561528, "speed/FLOPS": 202638052912909.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1998082548379898, "optim/lr": 0.001815, "optim/total_tokens": 1586495488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 3.3598687648773193, "created_at": "2025-01-15T14:05:52.564458+00:00"} {"global_step": 3027, "acc_step": 0, "speed/wps": 12908.112438551589, "speed/FLOPS": 202739643452711.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1621202975511551, "optim/lr": 0.0018156, "optim/total_tokens": 1587019776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.2723422050476074, "created_at": "2025-01-15T14:06:02.723183+00:00"} {"global_step": 3028, "acc_step": 0, "speed/wps": 12909.35389555798, "speed/FLOPS": 202759142240937.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13670478761196136, "optim/lr": 0.0018162000000000002, "optim/total_tokens": 1587544064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.373116970062256, "created_at": "2025-01-15T14:06:12.882129+00:00"} {"global_step": 3029, "acc_step": 0, "speed/wps": 12902.989366458256, "speed/FLOPS": 202659178565646.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15583565831184387, "optim/lr": 0.0018168000000000001, "optim/total_tokens": 1588068352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.2984652519226074, "created_at": "2025-01-15T14:06:23.045863+00:00"} {"global_step": 3030, "acc_step": 0, "speed/wps": 12904.518090923644, "speed/FLOPS": 202683189284063.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12526777386665344, "optim/lr": 0.0018174, "optim/total_tokens": 1588592640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.28126859664917, "created_at": "2025-01-15T14:06:33.207708+00:00"} {"global_step": 3031, "acc_step": 0, "speed/wps": 12898.773804377572, "speed/FLOPS": 202592967370380.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15219134092330933, "optim/lr": 0.001818, "optim/total_tokens": 1589116928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 3.395589828491211, "created_at": "2025-01-15T14:06:43.377397+00:00"} {"global_step": 3032, "acc_step": 0, "speed/wps": 12907.240844967368, "speed/FLOPS": 202725953877776.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15139828622341156, "optim/lr": 0.0018185999999999999, "optim/total_tokens": 1589641216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.4116530418395996, "created_at": "2025-01-15T14:06:53.536481+00:00"} {"global_step": 3033, "acc_step": 0, "speed/wps": 12906.182039160833, "speed/FLOPS": 202709323877632.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1414613276720047, "optim/lr": 0.0018192000000000002, "optim/total_tokens": 1590165504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 3.4081833362579346, "created_at": "2025-01-15T14:07:03.697053+00:00"} {"global_step": 3034, "acc_step": 0, "speed/wps": 12906.036943758136, "speed/FLOPS": 202707044954951.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14652161300182343, "optim/lr": 0.0018198, "optim/total_tokens": 1590689792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.39751935005188, "created_at": "2025-01-15T14:07:13.862601+00:00"} {"global_step": 3035, "acc_step": 0, "speed/wps": 12900.208378136409, "speed/FLOPS": 202615499322573.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13482245802879333, "optim/lr": 0.0018204, "optim/total_tokens": 1591214080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.373060703277588, "created_at": "2025-01-15T14:07:24.031808+00:00"} {"global_step": 3036, "acc_step": 0, "speed/wps": 12906.468110171396, "speed/FLOPS": 202713817015956.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15696285665035248, "optim/lr": 0.001821, "optim/total_tokens": 1591738368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 3.2914974689483643, "created_at": "2025-01-15T14:07:34.191565+00:00"} {"global_step": 3037, "acc_step": 0, "speed/wps": 12902.382561392214, "speed/FLOPS": 202649647858247.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13759928941726685, "optim/lr": 0.0018215999999999998, "optim/total_tokens": 1592262656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.4351069927215576, "created_at": "2025-01-15T14:07:44.359979+00:00"} {"global_step": 3038, "acc_step": 0, "speed/wps": 12908.798256028645, "speed/FLOPS": 202750415158600.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11680655926465988, "optim/lr": 0.0018222000000000002, "optim/total_tokens": 1592786944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.2981462478637695, "created_at": "2025-01-15T14:07:54.520662+00:00"} {"global_step": 3039, "acc_step": 0, "speed/wps": 12906.217101857581, "speed/FLOPS": 202709874585466.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.11696653068065643, "optim/lr": 0.0018228, "optim/total_tokens": 1593311232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.299610137939453, "created_at": "2025-01-15T14:08:04.679904+00:00"} {"global_step": 3040, "acc_step": 0, "speed/wps": 12907.083274267601, "speed/FLOPS": 202723479013411.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12242978066205978, "optim/lr": 0.0018234, "optim/total_tokens": 1593835520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.455615758895874, "created_at": "2025-01-15T14:08:14.839300+00:00"} {"global_step": 3041, "acc_step": 0, "speed/wps": 12897.045855873685, "speed/FLOPS": 202565827564676.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1518244445323944, "optim/lr": 0.001824, "optim/total_tokens": 1594359808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 3.3410677909851074, "created_at": "2025-01-15T14:08:25.006397+00:00"} {"global_step": 3042, "acc_step": 0, "speed/wps": 12906.577270990112, "speed/FLOPS": 202715531536617.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14503352344036102, "optim/lr": 0.0018246, "optim/total_tokens": 1594884096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.397944450378418, "created_at": "2025-01-15T14:08:35.166750+00:00"} {"global_step": 3043, "acc_step": 0, "speed/wps": 12908.144952128801, "speed/FLOPS": 202740154123119.94, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12563063204288483, "optim/lr": 0.0018252000000000001, "optim/total_tokens": 1595408384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.33502197265625, "created_at": "2025-01-15T14:08:45.325073+00:00"} {"global_step": 3044, "acc_step": 0, "speed/wps": 12900.092738878911, "speed/FLOPS": 202613683049129.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14924056828022003, "optim/lr": 0.0018258, "optim/total_tokens": 1595932672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349355, "loss/out": 3.295954704284668, "created_at": "2025-01-15T14:08:55.494215+00:00"} {"global_step": 3045, "acc_step": 0, "speed/wps": 12905.124105230454, "speed/FLOPS": 202692707571502.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18067046999931335, "optim/lr": 0.0018264000000000002, "optim/total_tokens": 1596456960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 3.358429193496704, "created_at": "2025-01-15T14:09:05.657641+00:00"} {"global_step": 3046, "acc_step": 0, "speed/wps": 12898.004793470207, "speed/FLOPS": 202580888997348.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17193885147571564, "optim/lr": 0.001827, "optim/total_tokens": 1596981248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480757, "loss/out": 3.2964961528778076, "created_at": "2025-01-15T14:09:15.826169+00:00"} {"global_step": 3047, "acc_step": 0, "speed/wps": 12904.435456158328, "speed/FLOPS": 202681891391522.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18375495076179504, "optim/lr": 0.0018276, "optim/total_tokens": 1597505536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.4083352088928223, "created_at": "2025-01-15T14:09:25.987025+00:00"} {"global_step": 3048, "acc_step": 0, "speed/wps": 12905.709117919216, "speed/FLOPS": 202701895999671.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17132268846035004, "optim/lr": 0.0018282, "optim/total_tokens": 1598029824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.3574893474578857, "created_at": "2025-01-15T14:09:36.152766+00:00"} {"global_step": 3049, "acc_step": 0, "speed/wps": 12902.750695228457, "speed/FLOPS": 202655429906014.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16359557211399078, "optim/lr": 0.0018288000000000002, "optim/total_tokens": 1598554112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.3078956604003906, "created_at": "2025-01-15T14:09:46.317900+00:00"} {"global_step": 3050, "acc_step": 0, "speed/wps": 12902.507491773627, "speed/FLOPS": 202651610061559.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.184526264667511, "optim/lr": 0.0018294000000000001, "optim/total_tokens": 1599078400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 3.455115795135498, "created_at": "2025-01-15T14:09:56.483134+00:00"} {"global_step": 3051, "acc_step": 0, "speed/wps": 12906.154281475629, "speed/FLOPS": 202708887905043.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15176096558570862, "optim/lr": 0.00183, "optim/total_tokens": 1599602688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.377483606338501, "created_at": "2025-01-15T14:10:06.645874+00:00"} {"global_step": 3052, "acc_step": 0, "speed/wps": 12907.19387899341, "speed/FLOPS": 202725216212617.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15235216915607452, "optim/lr": 0.0018306, "optim/total_tokens": 1600126976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 3.340099811553955, "created_at": "2025-01-15T14:10:16.804463+00:00"} {"global_step": 3053, "acc_step": 0, "speed/wps": 12910.258135344877, "speed/FLOPS": 202773344569346.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11863814294338226, "optim/lr": 0.0018312000000000003, "optim/total_tokens": 1600651264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.273739814758301, "created_at": "2025-01-15T14:10:26.961670+00:00"} {"global_step": 3054, "acc_step": 0, "speed/wps": 12904.887214000943, "speed/FLOPS": 202688986869220.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13522790372371674, "optim/lr": 0.0018318000000000002, "optim/total_tokens": 1601175552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379944, "loss/out": 3.30678129196167, "created_at": "2025-01-15T14:10:37.123372+00:00"} {"global_step": 3055, "acc_step": 0, "speed/wps": 12902.149028253523, "speed/FLOPS": 202645979899395.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.135222926735878, "optim/lr": 0.0018324, "optim/total_tokens": 1601699840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.2459287643432617, "created_at": "2025-01-15T14:10:47.285921+00:00"} {"global_step": 3056, "acc_step": 0, "speed/wps": 12902.753011383253, "speed/FLOPS": 202655466284408.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14100946485996246, "optim/lr": 0.001833, "optim/total_tokens": 1602224128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.407252550125122, "created_at": "2025-01-15T14:10:57.450630+00:00"} {"global_step": 3057, "acc_step": 0, "speed/wps": 12900.587660884126, "speed/FLOPS": 202621456479315.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14554361999034882, "optim/lr": 0.0018336, "optim/total_tokens": 1602748416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.366633415222168, "created_at": "2025-01-15T14:11:07.617771+00:00"} {"global_step": 3058, "acc_step": 0, "speed/wps": 12902.241863985504, "speed/FLOPS": 202647438012136.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.1281505674123764, "optim/lr": 0.0018342000000000002, "optim/total_tokens": 1603272704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.181957721710205, "created_at": "2025-01-15T14:11:17.780100+00:00"} {"global_step": 3059, "acc_step": 0, "speed/wps": 12905.25732619901, "speed/FLOPS": 202694799989878.7, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1219247579574585, "optim/lr": 0.0018348000000000001, "optim/total_tokens": 1603796992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.4075350761413574, "created_at": "2025-01-15T14:11:27.943477+00:00"} {"global_step": 3060, "acc_step": 0, "speed/wps": 12904.162393164217, "speed/FLOPS": 202677602561971.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13513420522212982, "optim/lr": 0.0018354, "optim/total_tokens": 1604321280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 3.453521728515625, "created_at": "2025-01-15T14:11:38.107650+00:00"} {"global_step": 3061, "acc_step": 0, "speed/wps": 12907.112013700225, "speed/FLOPS": 202723930405693.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13181523978710175, "optim/lr": 0.001836, "optim/total_tokens": 1604845568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.340346097946167, "created_at": "2025-01-15T14:11:48.266334+00:00"} {"global_step": 3062, "acc_step": 0, "speed/wps": 12906.31721592994, "speed/FLOPS": 202711447014544.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14802920818328857, "optim/lr": 0.0018365999999999999, "optim/total_tokens": 1605369856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.2663393020629883, "created_at": "2025-01-15T14:11:58.425802+00:00"} {"global_step": 3063, "acc_step": 0, "speed/wps": 12901.814146106031, "speed/FLOPS": 202640720115093.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.13412560522556305, "optim/lr": 0.0018372000000000002, "optim/total_tokens": 1605894144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.24257755279541, "created_at": "2025-01-15T14:12:08.590166+00:00"} {"global_step": 3064, "acc_step": 0, "speed/wps": 12904.489222236049, "speed/FLOPS": 202682735861654.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14780177175998688, "optim/lr": 0.0018378000000000001, "optim/total_tokens": 1606418432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.2594339847564697, "created_at": "2025-01-15T14:12:18.750833+00:00"} {"global_step": 3065, "acc_step": 0, "speed/wps": 12896.184426330297, "speed/FLOPS": 202552297629969.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.1434415876865387, "optim/lr": 0.0018384, "optim/total_tokens": 1606942720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 3.3926243782043457, "created_at": "2025-01-15T14:12:28.921682+00:00"} {"global_step": 3066, "acc_step": 0, "speed/wps": 12889.182984027917, "speed/FLOPS": 202442330357620.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1880670040845871, "optim/lr": 0.001839, "optim/total_tokens": 1607467008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 3.2926292419433594, "created_at": "2025-01-15T14:12:39.094550+00:00"} {"global_step": 3067, "acc_step": 0, "speed/wps": 12900.549434794486, "speed/FLOPS": 202620856086049.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22568398714065552, "optim/lr": 0.0018396, "optim/total_tokens": 1607991296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.3392958641052246, "created_at": "2025-01-15T14:12:49.261982+00:00"} {"global_step": 3068, "acc_step": 0, "speed/wps": 12891.232114783506, "speed/FLOPS": 202474514771938.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2118186056613922, "optim/lr": 0.0018402, "optim/total_tokens": 1608515584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.352802276611328, "created_at": "2025-01-15T14:12:59.433474+00:00"} {"global_step": 3069, "acc_step": 0, "speed/wps": 12898.250826178068, "speed/FLOPS": 202584753279109.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.1655472368001938, "optim/lr": 0.0018408, "optim/total_tokens": 1609039872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368810, "loss/out": 3.357274055480957, "created_at": "2025-01-15T14:13:09.601155+00:00"} {"global_step": 3070, "acc_step": 0, "speed/wps": 12894.60672286577, "speed/FLOPS": 202527517629067.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16039429605007172, "optim/lr": 0.0018414, "optim/total_tokens": 1609564160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 3.332777738571167, "created_at": "2025-01-15T14:13:19.771921+00:00"} {"global_step": 3071, "acc_step": 0, "speed/wps": 12902.070304571207, "speed/FLOPS": 202644743435787.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13726112246513367, "optim/lr": 0.001842, "optim/total_tokens": 1610088448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 3.409128427505493, "created_at": "2025-01-15T14:13:29.939662+00:00"} {"global_step": 3072, "acc_step": 0, "speed/wps": 12898.052261163844, "speed/FLOPS": 202581634542704.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15365199744701385, "optim/lr": 0.0018426, "optim/total_tokens": 1610612736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.3167107105255127, "created_at": "2025-01-15T14:13:40.105849+00:00"} {"global_step": 3073, "acc_step": 0, "speed/wps": 12902.918826805473, "speed/FLOPS": 202658070643468.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20023782551288605, "optim/lr": 0.0018432, "optim/total_tokens": 1611137024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.440704822540283, "created_at": "2025-01-15T14:13:50.269376+00:00"} {"global_step": 3074, "acc_step": 0, "speed/wps": 12898.550080097984, "speed/FLOPS": 202589453473140.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.20914798974990845, "optim/lr": 0.0018438, "optim/total_tokens": 1611661312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.3696680068969727, "created_at": "2025-01-15T14:14:00.434701+00:00"} {"global_step": 3075, "acc_step": 0, "speed/wps": 12899.54655684596, "speed/FLOPS": 202605104509769.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14613047242164612, "optim/lr": 0.0018444000000000002, "optim/total_tokens": 1612185600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402407, "loss/out": 3.4139580726623535, "created_at": "2025-01-15T14:14:10.599767+00:00"} {"global_step": 3076, "acc_step": 0, "speed/wps": 12900.060784938192, "speed/FLOPS": 202613181168581.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1404200792312622, "optim/lr": 0.001845, "optim/total_tokens": 1612709888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.414602756500244, "created_at": "2025-01-15T14:14:20.765133+00:00"} {"global_step": 3077, "acc_step": 0, "speed/wps": 12904.68401442939, "speed/FLOPS": 202685795340723.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1284853219985962, "optim/lr": 0.0018456, "optim/total_tokens": 1613234176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.3243560791015625, "created_at": "2025-01-15T14:14:30.928083+00:00"} {"global_step": 3078, "acc_step": 0, "speed/wps": 12896.383711430075, "speed/FLOPS": 202555427676308.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11064542084932327, "optim/lr": 0.0018461999999999999, "optim/total_tokens": 1613758464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287430, "loss/out": 3.3989763259887695, "created_at": "2025-01-15T14:14:41.096178+00:00"} {"global_step": 3079, "acc_step": 0, "speed/wps": 12902.943959795395, "speed/FLOPS": 202658465391611.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13939397037029266, "optim/lr": 0.0018468000000000002, "optim/total_tokens": 1614282752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 3.4005086421966553, "created_at": "2025-01-15T14:14:51.258222+00:00"} {"global_step": 3080, "acc_step": 0, "speed/wps": 12904.417545669818, "speed/FLOPS": 202681610082688.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.1462458074092865, "optim/lr": 0.0018474000000000001, "optim/total_tokens": 1614807040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.318265438079834, "created_at": "2025-01-15T14:15:01.419196+00:00"} {"global_step": 3081, "acc_step": 0, "speed/wps": 12903.926886012947, "speed/FLOPS": 202673903598540.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.13041655719280243, "optim/lr": 0.001848, "optim/total_tokens": 1615331328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 3.2303578853607178, "created_at": "2025-01-15T14:15:11.584651+00:00"} {"global_step": 3082, "acc_step": 0, "speed/wps": 12904.877286035495, "speed/FLOPS": 202688830936880.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.13808688521385193, "optim/lr": 0.0018486, "optim/total_tokens": 1615855616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 3.3451156616210938, "created_at": "2025-01-15T14:15:21.746395+00:00"} {"global_step": 3083, "acc_step": 0, "speed/wps": 12903.628414199002, "speed/FLOPS": 202669215688558.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12487112730741501, "optim/lr": 0.0018491999999999998, "optim/total_tokens": 1616379904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.375849723815918, "created_at": "2025-01-15T14:15:31.907639+00:00"} {"global_step": 3084, "acc_step": 0, "speed/wps": 12908.79126316535, "speed/FLOPS": 202750305326073.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14865145087242126, "optim/lr": 0.0018498000000000002, "optim/total_tokens": 1616904192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.393064260482788, "created_at": "2025-01-15T14:15:42.070459+00:00"} {"global_step": 3085, "acc_step": 0, "speed/wps": 12905.882507935446, "speed/FLOPS": 202704619328139.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17306369543075562, "optim/lr": 0.0018504, "optim/total_tokens": 1617428480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.493746519088745, "created_at": "2025-01-15T14:15:52.232069+00:00"} {"global_step": 3086, "acc_step": 0, "speed/wps": 12905.741774424514, "speed/FLOPS": 202702408914961.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1825937032699585, "optim/lr": 0.001851, "optim/total_tokens": 1617952768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 3.466190814971924, "created_at": "2025-01-15T14:16:02.393948+00:00"} {"global_step": 3087, "acc_step": 0, "speed/wps": 12903.426527220672, "speed/FLOPS": 202666044776142.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18416284024715424, "optim/lr": 0.0018516, "optim/total_tokens": 1618477056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.396249771118164, "created_at": "2025-01-15T14:16:12.557679+00:00"} {"global_step": 3088, "acc_step": 0, "speed/wps": 12903.231009948682, "speed/FLOPS": 202662973908717.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.16832302510738373, "optim/lr": 0.0018521999999999998, "optim/total_tokens": 1619001344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.3217570781707764, "created_at": "2025-01-15T14:16:22.720865+00:00"} {"global_step": 3089, "acc_step": 0, "speed/wps": 12899.758384389363, "speed/FLOPS": 202608431552420.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15374597907066345, "optim/lr": 0.0018528000000000001, "optim/total_tokens": 1619525632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350002, "loss/out": 3.393981695175171, "created_at": "2025-01-15T14:16:32.885424+00:00"} {"global_step": 3090, "acc_step": 0, "speed/wps": 12897.79857186955, "speed/FLOPS": 202577650003732.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16537441313266754, "optim/lr": 0.0018534, "optim/total_tokens": 1620049920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.374241590499878, "created_at": "2025-01-15T14:16:43.053322+00:00"} {"global_step": 3091, "acc_step": 0, "speed/wps": 12903.192992962215, "speed/FLOPS": 202662376799703.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15249653160572052, "optim/lr": 0.001854, "optim/total_tokens": 1620574208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.4425222873687744, "created_at": "2025-01-15T14:16:53.216176+00:00"} {"global_step": 3092, "acc_step": 0, "speed/wps": 12900.917577558172, "speed/FLOPS": 202626638274034.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17630897462368011, "optim/lr": 0.0018545999999999999, "optim/total_tokens": 1621098496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.335017681121826, "created_at": "2025-01-15T14:17:03.382381+00:00"} {"global_step": 3093, "acc_step": 0, "speed/wps": 12899.888885054927, "speed/FLOPS": 202610481244699.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15022806823253632, "optim/lr": 0.0018552, "optim/total_tokens": 1621622784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.3369202613830566, "created_at": "2025-01-15T14:17:13.548918+00:00"} {"global_step": 3094, "acc_step": 0, "speed/wps": 12893.357245286079, "speed/FLOPS": 202507892866716.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13406939804553986, "optim/lr": 0.0018558000000000001, "optim/total_tokens": 1622147072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.3507649898529053, "created_at": "2025-01-15T14:17:23.725924+00:00"} {"global_step": 3095, "acc_step": 0, "speed/wps": 12868.738614454565, "speed/FLOPS": 202121223439963.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1342756301164627, "optim/lr": 0.0018564, "optim/total_tokens": 1622671360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.4832682609558105, "created_at": "2025-01-15T14:17:33.915109+00:00"} {"global_step": 3096, "acc_step": 0, "speed/wps": 12883.619092688605, "speed/FLOPS": 202354941798549.03, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.11617029458284378, "optim/lr": 0.001857, "optim/total_tokens": 1623195648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366944, "loss/out": 3.230696678161621, "created_at": "2025-01-15T14:17:44.092893+00:00"} {"global_step": 3097, "acc_step": 0, "speed/wps": 12878.786149888709, "speed/FLOPS": 202279033790717.3, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.112189382314682, "optim/lr": 0.0018576, "optim/total_tokens": 1623719936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 3.288043260574341, "created_at": "2025-01-15T14:17:54.275872+00:00"} {"global_step": 3098, "acc_step": 0, "speed/wps": 12894.651875931853, "speed/FLOPS": 202528226820017.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13651630282402039, "optim/lr": 0.0018582, "optim/total_tokens": 1624244224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 3.3118162155151367, "created_at": "2025-01-15T14:18:04.449383+00:00"} {"global_step": 3099, "acc_step": 0, "speed/wps": 12892.05851565723, "speed/FLOPS": 202487494533253.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11562319844961166, "optim/lr": 0.0018588, "optim/total_tokens": 1624768512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.205942153930664, "created_at": "2025-01-15T14:18:14.622428+00:00"} {"global_step": 3100, "acc_step": 0, "speed/wps": 12888.04537107584, "speed/FLOPS": 202424462582964.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11933750659227371, "optim/lr": 0.0018594, "optim/total_tokens": 1625292800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330478, "loss/out": 3.248739719390869, "created_at": "2025-01-15T14:18:24.798988+00:00"} {"global_step": 3101, "acc_step": 0, "speed/wps": 12888.476684083287, "speed/FLOPS": 202431236946432.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14261266589164734, "optim/lr": 0.00186, "optim/total_tokens": 1625817088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 3.3445892333984375, "created_at": "2025-01-15T14:18:34.972226+00:00"} {"global_step": 3102, "acc_step": 0, "speed/wps": 12890.591355529437, "speed/FLOPS": 202464450767357.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14165113866329193, "optim/lr": 0.0018606, "optim/total_tokens": 1626341376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.3167824745178223, "created_at": "2025-01-15T14:18:45.146720+00:00"} {"global_step": 3103, "acc_step": 0, "speed/wps": 12893.335224273049, "speed/FLOPS": 202507546996446.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12664613127708435, "optim/lr": 0.0018612, "optim/total_tokens": 1626865664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.4398257732391357, "created_at": "2025-01-15T14:18:55.318615+00:00"} {"global_step": 3104, "acc_step": 0, "speed/wps": 12889.2858441539, "speed/FLOPS": 202443945917244.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12995734810829163, "optim/lr": 0.0018618000000000003, "optim/total_tokens": 1627389952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 3.3055315017700195, "created_at": "2025-01-15T14:19:05.494737+00:00"} {"global_step": 3105, "acc_step": 0, "speed/wps": 12889.76516752476, "speed/FLOPS": 202451474349442.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15606440603733063, "optim/lr": 0.0018624000000000002, "optim/total_tokens": 1627914240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.378091335296631, "created_at": "2025-01-15T14:19:15.668977+00:00"} {"global_step": 3106, "acc_step": 0, "speed/wps": 12883.552834779417, "speed/FLOPS": 202353901127037.34, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13925689458847046, "optim/lr": 0.001863, "optim/total_tokens": 1628438528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.4887757301330566, "created_at": "2025-01-15T14:19:25.850773+00:00"} {"global_step": 3107, "acc_step": 0, "speed/wps": 12890.262257219549, "speed/FLOPS": 202459281826172.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11539099365472794, "optim/lr": 0.0018636, "optim/total_tokens": 1628962816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.37863826751709, "created_at": "2025-01-15T14:19:36.022561+00:00"} {"global_step": 3108, "acc_step": 0, "speed/wps": 12886.137723977852, "speed/FLOPS": 202394500363908.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.140987828373909, "optim/lr": 0.0018641999999999999, "optim/total_tokens": 1629487104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373224, "loss/out": 3.4116337299346924, "created_at": "2025-01-15T14:19:46.201326+00:00"} {"global_step": 3109, "acc_step": 0, "speed/wps": 12889.032061658909, "speed/FLOPS": 202439959914426.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1483781337738037, "optim/lr": 0.0018648000000000002, "optim/total_tokens": 1630011392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 3.391951084136963, "created_at": "2025-01-15T14:19:56.379450+00:00"} {"global_step": 3110, "acc_step": 0, "speed/wps": 12884.81051154108, "speed/FLOPS": 202373654668808.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.132017582654953, "optim/lr": 0.0018654000000000001, "optim/total_tokens": 1630535680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.3793911933898926, "created_at": "2025-01-15T14:20:06.555586+00:00"} {"global_step": 3111, "acc_step": 0, "speed/wps": 12888.700820373515, "speed/FLOPS": 202434757314866.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15826989710330963, "optim/lr": 0.001866, "optim/total_tokens": 1631059968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.4413211345672607, "created_at": "2025-01-15T14:20:16.729011+00:00"} {"global_step": 3112, "acc_step": 0, "speed/wps": 12892.590670563699, "speed/FLOPS": 202495852757317.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11601024866104126, "optim/lr": 0.0018666, "optim/total_tokens": 1631584256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 3.445709705352783, "created_at": "2025-01-15T14:20:26.903222+00:00"} {"global_step": 3113, "acc_step": 0, "speed/wps": 12894.611169753554, "speed/FLOPS": 202527587473550.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1298256814479828, "optim/lr": 0.0018671999999999998, "optim/total_tokens": 1632108544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.376051902770996, "created_at": "2025-01-15T14:20:37.071712+00:00"} {"global_step": 3114, "acc_step": 0, "speed/wps": 12893.05434395038, "speed/FLOPS": 202503135385009.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11434673517942429, "optim/lr": 0.0018678000000000002, "optim/total_tokens": 1632632832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 3.435526132583618, "created_at": "2025-01-15T14:20:47.244037+00:00"} {"global_step": 3115, "acc_step": 0, "speed/wps": 12890.875113461865, "speed/FLOPS": 202468907575610.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15853101015090942, "optim/lr": 0.0018684, "optim/total_tokens": 1633157120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.4448161125183105, "created_at": "2025-01-15T14:20:57.415355+00:00"} {"global_step": 3116, "acc_step": 0, "speed/wps": 12891.692729958142, "speed/FLOPS": 202481749366207.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1609746515750885, "optim/lr": 0.001869, "optim/total_tokens": 1633681408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 3.4539003372192383, "created_at": "2025-01-15T14:21:07.591671+00:00"} {"global_step": 3117, "acc_step": 0, "speed/wps": 12878.156352697766, "speed/FLOPS": 202269141960401.53, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13197050988674164, "optim/lr": 0.0018696, "optim/total_tokens": 1634205696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 3.3737568855285645, "created_at": "2025-01-15T14:21:17.777402+00:00"} {"global_step": 3118, "acc_step": 0, "speed/wps": 12869.839510612863, "speed/FLOPS": 202138514526919.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11749386787414551, "optim/lr": 0.0018701999999999998, "optim/total_tokens": 1634729984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.305464744567871, "created_at": "2025-01-15T14:21:27.968259+00:00"} {"global_step": 3119, "acc_step": 0, "speed/wps": 12863.002010776812, "speed/FLOPS": 202031122196284.94, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13748842477798462, "optim/lr": 0.0018708000000000002, "optim/total_tokens": 1635254272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.354461669921875, "created_at": "2025-01-15T14:21:38.162850+00:00"} {"global_step": 3120, "acc_step": 0, "speed/wps": 12873.89786251271, "speed/FLOPS": 202202256520266.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11506536602973938, "optim/lr": 0.0018714, "optim/total_tokens": 1635778560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.3532865047454834, "created_at": "2025-01-15T14:21:48.354404+00:00"} {"global_step": 3121, "acc_step": 0, "speed/wps": 12887.989828809581, "speed/FLOPS": 202423590215349.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12025416642427444, "optim/lr": 0.001872, "optim/total_tokens": 1636302848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.377492904663086, "created_at": "2025-01-15T14:21:58.530614+00:00"} {"global_step": 3122, "acc_step": 0, "speed/wps": 12887.125873679583, "speed/FLOPS": 202410020612834.4, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11576695740222931, "optim/lr": 0.0018725999999999999, "optim/total_tokens": 1636827136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418007, "loss/out": 3.4140396118164062, "created_at": "2025-01-15T14:22:08.707561+00:00"} {"global_step": 3123, "acc_step": 0, "speed/wps": 12890.180097721206, "speed/FLOPS": 202457991398351.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09885739535093307, "optim/lr": 0.0018732, "optim/total_tokens": 1637351424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 3.4552173614501953, "created_at": "2025-01-15T14:22:18.880278+00:00"} {"global_step": 3124, "acc_step": 0, "speed/wps": 12889.36585915146, "speed/FLOPS": 202445202662732.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11230208724737167, "optim/lr": 0.0018738000000000001, "optim/total_tokens": 1637875712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.364995002746582, "created_at": "2025-01-15T14:22:29.053306+00:00"} {"global_step": 3125, "acc_step": 0, "speed/wps": 12888.192550187547, "speed/FLOPS": 202426774233159.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.13769516348838806, "optim/lr": 0.0018744, "optim/total_tokens": 1638400000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482132, "loss/out": 3.2585957050323486, "created_at": "2025-01-15T14:22:39.228343+00:00"} {"global_step": 3126, "acc_step": 0, "speed/wps": 12883.225032152019, "speed/FLOPS": 202348752536326.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12661612033843994, "optim/lr": 0.001875, "optim/total_tokens": 1638924288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 3.4223833084106445, "created_at": "2025-01-15T14:22:49.409077+00:00"} {"global_step": 3127, "acc_step": 0, "speed/wps": 12887.917685123417, "speed/FLOPS": 202422457099622.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1337750405073166, "optim/lr": 0.0018756, "optim/total_tokens": 1639448576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.258878231048584, "created_at": "2025-01-15T14:22:59.584020+00:00"} {"global_step": 3128, "acc_step": 0, "speed/wps": 12891.120532993973, "speed/FLOPS": 202472762226604.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12189119309186935, "optim/lr": 0.0018762, "optim/total_tokens": 1639972864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 3.2011804580688477, "created_at": "2025-01-15T14:23:09.758196+00:00"} {"global_step": 3129, "acc_step": 0, "speed/wps": 12890.797924097624, "speed/FLOPS": 202467695210575.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13326166570186615, "optim/lr": 0.0018768, "optim/total_tokens": 1640497152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.3612513542175293, "created_at": "2025-01-15T14:23:19.930128+00:00"} {"global_step": 3130, "acc_step": 0, "speed/wps": 12892.126279428297, "speed/FLOPS": 202488558856395.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14746500551700592, "optim/lr": 0.0018774000000000002, "optim/total_tokens": 1641021440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435320, "loss/out": 3.2658514976501465, "created_at": "2025-01-15T14:23:30.101083+00:00"} {"global_step": 3131, "acc_step": 0, "speed/wps": 12888.577259637705, "speed/FLOPS": 202432816623720.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13854555785655975, "optim/lr": 0.0018780000000000001, "optim/total_tokens": 1641545728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2635440826416016, "created_at": "2025-01-15T14:23:40.279761+00:00"} {"global_step": 3132, "acc_step": 0, "speed/wps": 12885.560728749158, "speed/FLOPS": 202385437860967.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16020330786705017, "optim/lr": 0.0018786, "optim/total_tokens": 1642070016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 3.302041530609131, "created_at": "2025-01-15T14:23:50.457643+00:00"} {"global_step": 3133, "acc_step": 0, "speed/wps": 12894.338750374087, "speed/FLOPS": 202523308752852.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2008524388074875, "optim/lr": 0.0018792, "optim/total_tokens": 1642594304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 3.4524497985839844, "created_at": "2025-01-15T14:24:00.628123+00:00"} {"global_step": 3134, "acc_step": 0, "speed/wps": 12892.675467419023, "speed/FLOPS": 202497184608453.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19268500804901123, "optim/lr": 0.0018798000000000003, "optim/total_tokens": 1643118592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 3.4242348670959473, "created_at": "2025-01-15T14:24:10.801355+00:00"} {"global_step": 3135, "acc_step": 0, "speed/wps": 12887.486131019725, "speed/FLOPS": 202415678949406.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1442413479089737, "optim/lr": 0.0018804000000000002, "optim/total_tokens": 1643642880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453930, "loss/out": 3.432687282562256, "created_at": "2025-01-15T14:24:20.978833+00:00"} {"global_step": 3136, "acc_step": 0, "speed/wps": 12886.479991005226, "speed/FLOPS": 202399876137897.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14419984817504883, "optim/lr": 0.001881, "optim/total_tokens": 1644167168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.37007999420166, "created_at": "2025-01-15T14:24:31.155033+00:00"} {"global_step": 3137, "acc_step": 0, "speed/wps": 12886.356053977026, "speed/FLOPS": 202397929536562.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15804825723171234, "optim/lr": 0.0018816, "optim/total_tokens": 1644691456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.3461356163024902, "created_at": "2025-01-15T14:24:41.333661+00:00"} {"global_step": 3138, "acc_step": 0, "speed/wps": 12886.676816938723, "speed/FLOPS": 202402967559650.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1307581663131714, "optim/lr": 0.0018821999999999999, "optim/total_tokens": 1645215744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398779, "loss/out": 3.290541648864746, "created_at": "2025-01-15T14:24:51.511085+00:00"} {"global_step": 3139, "acc_step": 0, "speed/wps": 12889.953368204338, "speed/FLOPS": 202454430299730.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1444505900144577, "optim/lr": 0.0018828000000000002, "optim/total_tokens": 1645740032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 3.360995292663574, "created_at": "2025-01-15T14:25:01.701699+00:00"} {"global_step": 3140, "acc_step": 0, "speed/wps": 12888.750950946112, "speed/FLOPS": 202435544684395.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13770200312137604, "optim/lr": 0.0018834000000000001, "optim/total_tokens": 1646264320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.388096570968628, "created_at": "2025-01-15T14:25:11.874701+00:00"} {"global_step": 3141, "acc_step": 0, "speed/wps": 12890.333865170944, "speed/FLOPS": 202460406527448.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1329299807548523, "optim/lr": 0.001884, "optim/total_tokens": 1646788608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.3835067749023438, "created_at": "2025-01-15T14:25:22.051625+00:00"} {"global_step": 3142, "acc_step": 0, "speed/wps": 12890.287963121642, "speed/FLOPS": 202459685572688.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13730330765247345, "optim/lr": 0.0018846, "optim/total_tokens": 1647312896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.21232008934021, "created_at": "2025-01-15T14:25:32.226803+00:00"} {"global_step": 3143, "acc_step": 0, "speed/wps": 12889.176558939624, "speed/FLOPS": 202442229442780.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14632157981395721, "optim/lr": 0.0018851999999999999, "optim/total_tokens": 1647837184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.4307186603546143, "created_at": "2025-01-15T14:25:42.402097+00:00"} {"global_step": 3144, "acc_step": 0, "speed/wps": 12892.946075548576, "speed/FLOPS": 202501434880982.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17892508208751678, "optim/lr": 0.0018858000000000002, "optim/total_tokens": 1648361472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 3.252570867538452, "created_at": "2025-01-15T14:25:52.573740+00:00"} {"global_step": 3145, "acc_step": 0, "speed/wps": 12892.11726153855, "speed/FLOPS": 202488417218045.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16347000002861023, "optim/lr": 0.0018864, "optim/total_tokens": 1648885760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 3.390587329864502, "created_at": "2025-01-15T14:26:02.747433+00:00"} {"global_step": 3146, "acc_step": 0, "speed/wps": 12893.278272137619, "speed/FLOPS": 202506652484899.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12174960970878601, "optim/lr": 0.001887, "optim/total_tokens": 1649410048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.3059988021850586, "created_at": "2025-01-15T14:26:12.917258+00:00"} {"global_step": 3147, "acc_step": 0, "speed/wps": 12886.754275553334, "speed/FLOPS": 202404184153632.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.15949466824531555, "optim/lr": 0.0018876, "optim/total_tokens": 1649934336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380023, "loss/out": 3.294665813446045, "created_at": "2025-01-15T14:26:23.091886+00:00"} {"global_step": 3148, "acc_step": 0, "speed/wps": 12892.510676937254, "speed/FLOPS": 202494596347491.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1231016218662262, "optim/lr": 0.0018881999999999998, "optim/total_tokens": 1650458624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 3.3851499557495117, "created_at": "2025-01-15T14:26:33.261877+00:00"} {"global_step": 3149, "acc_step": 0, "speed/wps": 12890.941924912546, "speed/FLOPS": 202469956941252.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12554942071437836, "optim/lr": 0.0018888000000000002, "optim/total_tokens": 1650982912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.326610803604126, "created_at": "2025-01-15T14:26:43.435916+00:00"} {"global_step": 3150, "acc_step": 0, "speed/wps": 12894.245774345387, "speed/FLOPS": 202521848436559.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1296306550502777, "optim/lr": 0.0018894, "optim/total_tokens": 1651507200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.3368773460388184, "created_at": "2025-01-15T14:26:53.606793+00:00"} {"global_step": 3151, "acc_step": 0, "speed/wps": 12895.48494583304, "speed/FLOPS": 202541311327573.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13761469721794128, "optim/lr": 0.00189, "optim/total_tokens": 1652031488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.272435188293457, "created_at": "2025-01-15T14:27:03.775573+00:00"} {"global_step": 3152, "acc_step": 0, "speed/wps": 12891.289935638015, "speed/FLOPS": 202475422927912.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1509515941143036, "optim/lr": 0.0018906, "optim/total_tokens": 1652555776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.332148313522339, "created_at": "2025-01-15T14:27:13.947128+00:00"} {"global_step": 3153, "acc_step": 0, "speed/wps": 12885.19791173779, "speed/FLOPS": 202379739321241.22, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13923625648021698, "optim/lr": 0.0018912, "optim/total_tokens": 1653080064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.3186583518981934, "created_at": "2025-01-15T14:27:24.123334+00:00"} {"global_step": 3154, "acc_step": 0, "speed/wps": 12874.828106286359, "speed/FLOPS": 202216867277020.56, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11462201178073883, "optim/lr": 0.0018918000000000001, "optim/total_tokens": 1653604352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.295714855194092, "created_at": "2025-01-15T14:27:34.310726+00:00"} {"global_step": 3155, "acc_step": 0, "speed/wps": 12885.183342432043, "speed/FLOPS": 202379510490274.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11334926635026932, "optim/lr": 0.0018924, "optim/total_tokens": 1654128640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 3.4018609523773193, "created_at": "2025-01-15T14:27:44.491512+00:00"} {"global_step": 3156, "acc_step": 0, "speed/wps": 12885.927637805415, "speed/FLOPS": 202391200671881.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1085590273141861, "optim/lr": 0.0018930000000000002, "optim/total_tokens": 1654652928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 3.364542007446289, "created_at": "2025-01-15T14:27:54.667173+00:00"} {"global_step": 3157, "acc_step": 0, "speed/wps": 12887.75077038031, "speed/FLOPS": 202419835474217.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1110691949725151, "optim/lr": 0.0018936, "optim/total_tokens": 1655177216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.374542236328125, "created_at": "2025-01-15T14:28:04.841110+00:00"} {"global_step": 3158, "acc_step": 0, "speed/wps": 12887.517693439493, "speed/FLOPS": 202416174680580.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13212460279464722, "optim/lr": 0.0018942, "optim/total_tokens": 1655701504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.351077079772949, "created_at": "2025-01-15T14:28:15.023589+00:00"} {"global_step": 3159, "acc_step": 0, "speed/wps": 12891.525005358943, "speed/FLOPS": 202479115020899.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11466939747333527, "optim/lr": 0.0018948, "optim/total_tokens": 1656225792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429723, "loss/out": 3.3136541843414307, "created_at": "2025-01-15T14:28:25.197489+00:00"} {"global_step": 3160, "acc_step": 0, "speed/wps": 12889.74976269418, "speed/FLOPS": 202451232395410.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12776398658752441, "optim/lr": 0.0018954000000000002, "optim/total_tokens": 1656750080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.394803524017334, "created_at": "2025-01-15T14:28:35.370537+00:00"} {"global_step": 3161, "acc_step": 0, "speed/wps": 12885.565296814955, "speed/FLOPS": 202385509608718.1, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11112534999847412, "optim/lr": 0.0018960000000000001, "optim/total_tokens": 1657274368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.3376946449279785, "created_at": "2025-01-15T14:28:45.548088+00:00"} {"global_step": 3162, "acc_step": 0, "speed/wps": 12886.472641181988, "speed/FLOPS": 202399760698824.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.11582684516906738, "optim/lr": 0.0018966, "optim/total_tokens": 1657798656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.437060832977295, "created_at": "2025-01-15T14:28:55.723816+00:00"} {"global_step": 3163, "acc_step": 0, "speed/wps": 12887.109435530792, "speed/FLOPS": 202409762429119.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11708366870880127, "optim/lr": 0.0018972, "optim/total_tokens": 1658322944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.365224838256836, "created_at": "2025-01-15T14:29:05.898434+00:00"} {"global_step": 3164, "acc_step": 0, "speed/wps": 12883.672161780167, "speed/FLOPS": 202355775321556.6, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1294596642255783, "optim/lr": 0.0018978000000000003, "optim/total_tokens": 1658847232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359993, "loss/out": 3.1981546878814697, "created_at": "2025-01-15T14:29:16.076722+00:00"} {"global_step": 3165, "acc_step": 0, "speed/wps": 12871.049445387584, "speed/FLOPS": 202157518215181.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13418883085250854, "optim/lr": 0.0018984000000000002, "optim/total_tokens": 1659371520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 3.283494472503662, "created_at": "2025-01-15T14:29:26.266804+00:00"} {"global_step": 3166, "acc_step": 0, "speed/wps": 12891.081777321455, "speed/FLOPS": 202472153515511.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14131507277488708, "optim/lr": 0.001899, "optim/total_tokens": 1659895808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.344651699066162, "created_at": "2025-01-15T14:29:36.438017+00:00"} {"global_step": 3167, "acc_step": 0, "speed/wps": 12889.476734148622, "speed/FLOPS": 202446944106921.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13570664823055267, "optim/lr": 0.0018996, "optim/total_tokens": 1660420096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.3297009468078613, "created_at": "2025-01-15T14:29:46.611351+00:00"} {"global_step": 3168, "acc_step": 0, "speed/wps": 12898.158579805026, "speed/FLOPS": 202583304423061.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10676949471235275, "optim/lr": 0.0019002, "optim/total_tokens": 1660944384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.268104314804077, "created_at": "2025-01-15T14:29:56.780032+00:00"} {"global_step": 3169, "acc_step": 0, "speed/wps": 12889.319013728056, "speed/FLOPS": 202444466890985.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11597801744937897, "optim/lr": 0.0019008000000000002, "optim/total_tokens": 1661468672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 3.3986282348632812, "created_at": "2025-01-15T14:30:06.952478+00:00"} {"global_step": 3170, "acc_step": 0, "speed/wps": 12892.094644830484, "speed/FLOPS": 202488061991566.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14621946215629578, "optim/lr": 0.0019014000000000001, "optim/total_tokens": 1661992960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.310457706451416, "created_at": "2025-01-15T14:30:17.122995+00:00"} {"global_step": 3171, "acc_step": 0, "speed/wps": 12892.169461784735, "speed/FLOPS": 202489237094642.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17598293721675873, "optim/lr": 0.001902, "optim/total_tokens": 1662517248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.3863301277160645, "created_at": "2025-01-15T14:30:27.293436+00:00"} {"global_step": 3172, "acc_step": 0, "speed/wps": 12889.00745263541, "speed/FLOPS": 202439573395896.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16627895832061768, "optim/lr": 0.0019026, "optim/total_tokens": 1663041536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291904, "loss/out": 3.3684167861938477, "created_at": "2025-01-15T14:30:37.466264+00:00"} {"global_step": 3173, "acc_step": 0, "speed/wps": 12891.86598327955, "speed/FLOPS": 202484470547695.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12946534156799316, "optim/lr": 0.0019031999999999999, "optim/total_tokens": 1663565824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.2879090309143066, "created_at": "2025-01-15T14:30:47.639276+00:00"} {"global_step": 3174, "acc_step": 0, "speed/wps": 12890.037126870802, "speed/FLOPS": 202455745844683.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14688962697982788, "optim/lr": 0.0019038000000000002, "optim/total_tokens": 1664090112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 3.4268269538879395, "created_at": "2025-01-15T14:30:57.811774+00:00"} {"global_step": 3175, "acc_step": 0, "speed/wps": 12890.840080942413, "speed/FLOPS": 202468357341752.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14398708939552307, "optim/lr": 0.0019044, "optim/total_tokens": 1664614400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.247976303100586, "created_at": "2025-01-15T14:31:07.983166+00:00"} {"global_step": 3176, "acc_step": 0, "speed/wps": 12894.28412611547, "speed/FLOPS": 202522450803809.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12163233011960983, "optim/lr": 0.001905, "optim/total_tokens": 1665138688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371356, "loss/out": 3.224109411239624, "created_at": "2025-01-15T14:31:18.153616+00:00"} {"global_step": 3177, "acc_step": 0, "speed/wps": 12890.202462143647, "speed/FLOPS": 202458342662338.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.145615816116333, "optim/lr": 0.0019056, "optim/total_tokens": 1665662976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.31538724899292, "created_at": "2025-01-15T14:31:28.328482+00:00"} {"global_step": 3178, "acc_step": 0, "speed/wps": 12894.540226664425, "speed/FLOPS": 202526473214850.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.121174156665802, "optim/lr": 0.0019062, "optim/total_tokens": 1666187264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.3063693046569824, "created_at": "2025-01-15T14:31:38.497802+00:00"} {"global_step": 3179, "acc_step": 0, "speed/wps": 12892.089754454615, "speed/FLOPS": 202487985181493.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11894715577363968, "optim/lr": 0.0019068000000000002, "optim/total_tokens": 1666711552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 3.3092432022094727, "created_at": "2025-01-15T14:31:48.675591+00:00"} {"global_step": 3180, "acc_step": 0, "speed/wps": 12892.940179393272, "speed/FLOPS": 202501342273761.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12542316317558289, "optim/lr": 0.0019074, "optim/total_tokens": 1667235840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 3.3490307331085205, "created_at": "2025-01-15T14:31:58.848625+00:00"} {"global_step": 3181, "acc_step": 0, "speed/wps": 12891.37461147047, "speed/FLOPS": 202476752878217.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09992873668670654, "optim/lr": 0.001908, "optim/total_tokens": 1667760128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.304766893386841, "created_at": "2025-01-15T14:32:09.023126+00:00"} {"global_step": 3182, "acc_step": 0, "speed/wps": 12892.223380792386, "speed/FLOPS": 202490083966749.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12202160060405731, "optim/lr": 0.0019086, "optim/total_tokens": 1668284416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 3.3358688354492188, "created_at": "2025-01-15T14:32:19.193496+00:00"} {"global_step": 3183, "acc_step": 0, "speed/wps": 12889.855121656881, "speed/FLOPS": 202452887202697.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.13575899600982666, "optim/lr": 0.0019092, "optim/total_tokens": 1668808704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2743449211120605, "created_at": "2025-01-15T14:32:29.365755+00:00"} {"global_step": 3184, "acc_step": 0, "speed/wps": 12894.328760661518, "speed/FLOPS": 202523151850689.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15207001566886902, "optim/lr": 0.0019098000000000001, "optim/total_tokens": 1669332992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 3.3416476249694824, "created_at": "2025-01-15T14:32:39.536549+00:00"} {"global_step": 3185, "acc_step": 0, "speed/wps": 12897.192728863569, "speed/FLOPS": 202568134406807.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1620890200138092, "optim/lr": 0.0019104, "optim/total_tokens": 1669857280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 3.370709180831909, "created_at": "2025-01-15T14:32:49.706311+00:00"} {"global_step": 3186, "acc_step": 0, "speed/wps": 12896.930690501149, "speed/FLOPS": 202564018734247.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16239061951637268, "optim/lr": 0.0019110000000000002, "optim/total_tokens": 1670381568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.3899085521698, "created_at": "2025-01-15T14:32:59.873519+00:00"} {"global_step": 3187, "acc_step": 0, "speed/wps": 12891.540579754024, "speed/FLOPS": 202479359638175.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15507706999778748, "optim/lr": 0.0019116, "optim/total_tokens": 1670905856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361335, "loss/out": 3.265033721923828, "created_at": "2025-01-15T14:33:10.044271+00:00"} {"global_step": 3188, "acc_step": 0, "speed/wps": 12892.949544503046, "speed/FLOPS": 202501489365679.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19039331376552582, "optim/lr": 0.0019122, "optim/total_tokens": 1671430144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.2378268241882324, "created_at": "2025-01-15T14:33:20.219165+00:00"} {"global_step": 3189, "acc_step": 0, "speed/wps": 12892.433964016183, "speed/FLOPS": 202493391465651.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2138628512620926, "optim/lr": 0.0019127999999999999, "optim/total_tokens": 1671954432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 3.300942897796631, "created_at": "2025-01-15T14:33:30.389171+00:00"} {"global_step": 3190, "acc_step": 0, "speed/wps": 12892.138552092654, "speed/FLOPS": 202488751615453.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1988164335489273, "optim/lr": 0.0019134000000000002, "optim/total_tokens": 1672478720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.2864677906036377, "created_at": "2025-01-15T14:33:40.560687+00:00"} {"global_step": 3191, "acc_step": 0, "speed/wps": 12888.51569438443, "speed/FLOPS": 202431849656817.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1673736274242401, "optim/lr": 0.0019140000000000001, "optim/total_tokens": 1673003008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 3.421098470687866, "created_at": "2025-01-15T14:33:50.734819+00:00"} {"global_step": 3192, "acc_step": 0, "speed/wps": 12895.015338973097, "speed/FLOPS": 202533935506530.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14876028895378113, "optim/lr": 0.0019146, "optim/total_tokens": 1673527296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 3.345996856689453, "created_at": "2025-01-15T14:34:00.902945+00:00"} {"global_step": 3193, "acc_step": 0, "speed/wps": 12892.245331872047, "speed/FLOPS": 202490428738619.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13736939430236816, "optim/lr": 0.0019152, "optim/total_tokens": 1674051584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.2900097370147705, "created_at": "2025-01-15T14:34:11.073687+00:00"} {"global_step": 3194, "acc_step": 0, "speed/wps": 12894.615399261784, "speed/FLOPS": 202527653903789.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1805219203233719, "optim/lr": 0.0019157999999999998, "optim/total_tokens": 1674575872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 3.4345905780792236, "created_at": "2025-01-15T14:34:21.245627+00:00"} {"global_step": 3195, "acc_step": 0, "speed/wps": 12894.508146642258, "speed/FLOPS": 202525969354019.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13460473716259003, "optim/lr": 0.0019164000000000002, "optim/total_tokens": 1675100160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 3.367241859436035, "created_at": "2025-01-15T14:34:31.416368+00:00"} {"global_step": 3196, "acc_step": 0, "speed/wps": 12887.315340237203, "speed/FLOPS": 202412996445476.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12360282242298126, "optim/lr": 0.001917, "optim/total_tokens": 1675624448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.261475086212158, "created_at": "2025-01-15T14:34:41.595007+00:00"} {"global_step": 3197, "acc_step": 0, "speed/wps": 12892.596405775537, "speed/FLOPS": 202495942836700.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15737460553646088, "optim/lr": 0.0019176, "optim/total_tokens": 1676148736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.235029697418213, "created_at": "2025-01-15T14:34:51.765542+00:00"} {"global_step": 3198, "acc_step": 0, "speed/wps": 12891.497551579074, "speed/FLOPS": 202478683821560.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1358380913734436, "optim/lr": 0.0019182, "optim/total_tokens": 1676673024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315630, "loss/out": 3.3487696647644043, "created_at": "2025-01-15T14:35:01.940389+00:00"} {"global_step": 3199, "acc_step": 0, "speed/wps": 12894.086394374854, "speed/FLOPS": 202519345155111.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13576620817184448, "optim/lr": 0.0019187999999999998, "optim/total_tokens": 1677197312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.2569727897644043, "created_at": "2025-01-15T14:35:12.109413+00:00"} {"global_step": 3200, "acc_step": 0, "speed/wps": 12888.801016988764, "speed/FLOPS": 202436331040393.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13424932956695557, "optim/lr": 0.0019194000000000001, "optim/total_tokens": 1677721600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.287438154220581, "created_at": "2025-01-15T14:35:22.286034+00:00"} {"global_step": 3201, "acc_step": 0, "speed/wps": 12890.180476698526, "speed/FLOPS": 202457997350711.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1393689215183258, "optim/lr": 0.00192, "optim/total_tokens": 1678245888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.291206121444702, "created_at": "2025-01-15T14:35:32.457877+00:00"} {"global_step": 3202, "acc_step": 0, "speed/wps": 12890.00947636738, "speed/FLOPS": 202455311555531.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13015533983707428, "optim/lr": 0.0019206, "optim/total_tokens": 1678770176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373078, "loss/out": 3.4220776557922363, "created_at": "2025-01-15T14:35:42.629880+00:00"} {"global_step": 3203, "acc_step": 0, "speed/wps": 12895.329819556093, "speed/FLOPS": 202538874856228.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13024231791496277, "optim/lr": 0.0019211999999999999, "optim/total_tokens": 1679294464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.3843507766723633, "created_at": "2025-01-15T14:35:52.797937+00:00"} {"global_step": 3204, "acc_step": 0, "speed/wps": 12891.589664910396, "speed/FLOPS": 202480130588006.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13283950090408325, "optim/lr": 0.0019218, "optim/total_tokens": 1679818752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.426771402359009, "created_at": "2025-01-15T14:36:02.972117+00:00"} {"global_step": 3205, "acc_step": 0, "speed/wps": 12893.678138357256, "speed/FLOPS": 202512932933352.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11102112382650375, "optim/lr": 0.0019224, "optim/total_tokens": 1680343040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 3.3536252975463867, "created_at": "2025-01-15T14:36:13.141850+00:00"} {"global_step": 3206, "acc_step": 0, "speed/wps": 12892.347568885501, "speed/FLOPS": 202492034511404.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12466058880090714, "optim/lr": 0.001923, "optim/total_tokens": 1680867328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.345872402191162, "created_at": "2025-01-15T14:36:23.311979+00:00"} {"global_step": 3207, "acc_step": 0, "speed/wps": 12887.255458030639, "speed/FLOPS": 202412055913135.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13341262936592102, "optim/lr": 0.0019236, "optim/total_tokens": 1681391616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 3.3559255599975586, "created_at": "2025-01-15T14:36:33.487818+00:00"} {"global_step": 3208, "acc_step": 0, "speed/wps": 12894.944022812519, "speed/FLOPS": 202532815388229.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14998629689216614, "optim/lr": 0.0019242, "optim/total_tokens": 1681915904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.3538520336151123, "created_at": "2025-01-15T14:36:43.659202+00:00"} {"global_step": 3209, "acc_step": 0, "speed/wps": 12894.920088283445, "speed/FLOPS": 202532439463561.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.171478733420372, "optim/lr": 0.0019248, "optim/total_tokens": 1682440192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414236, "loss/out": 3.371720790863037, "created_at": "2025-01-15T14:36:53.830228+00:00"} {"global_step": 3210, "acc_step": 0, "speed/wps": 12895.483279796132, "speed/FLOPS": 202541285160174.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14925935864448547, "optim/lr": 0.0019254, "optim/total_tokens": 1682964480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.421039581298828, "created_at": "2025-01-15T14:37:04.000773+00:00"} {"global_step": 3211, "acc_step": 0, "speed/wps": 12891.152465094361, "speed/FLOPS": 202473263764119.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11765215545892715, "optim/lr": 0.001926, "optim/total_tokens": 1683488768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.3950703144073486, "created_at": "2025-01-15T14:37:14.172732+00:00"} {"global_step": 3212, "acc_step": 0, "speed/wps": 12893.62875734034, "speed/FLOPS": 202512157336625.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12285874783992767, "optim/lr": 0.0019266, "optim/total_tokens": 1684013056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.3678035736083984, "created_at": "2025-01-15T14:37:24.343852+00:00"} {"global_step": 3213, "acc_step": 0, "speed/wps": 12896.087593801609, "speed/FLOPS": 202550776742044.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12070465832948685, "optim/lr": 0.0019272, "optim/total_tokens": 1684537344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 3.3260061740875244, "created_at": "2025-01-15T14:37:34.514253+00:00"} {"global_step": 3214, "acc_step": 0, "speed/wps": 12893.897105506516, "speed/FLOPS": 202516372113323.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12659327685832977, "optim/lr": 0.0019278, "optim/total_tokens": 1685061632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.343167543411255, "created_at": "2025-01-15T14:37:44.683178+00:00"} {"global_step": 3215, "acc_step": 0, "speed/wps": 12892.812952730741, "speed/FLOPS": 202499344004198.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1356806606054306, "optim/lr": 0.0019284000000000003, "optim/total_tokens": 1685585920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.2674379348754883, "created_at": "2025-01-15T14:37:54.853429+00:00"} {"global_step": 3216, "acc_step": 0, "speed/wps": 12895.690118615024, "speed/FLOPS": 202544533848051.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16391929984092712, "optim/lr": 0.0019290000000000002, "optim/total_tokens": 1686110208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.242013931274414, "created_at": "2025-01-15T14:38:05.021047+00:00"} {"global_step": 3217, "acc_step": 0, "speed/wps": 12888.413996153557, "speed/FLOPS": 202430252346353.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16448456048965454, "optim/lr": 0.0019296, "optim/total_tokens": 1686634496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 3.3413946628570557, "created_at": "2025-01-15T14:38:15.196196+00:00"} {"global_step": 3218, "acc_step": 0, "speed/wps": 12892.672124986655, "speed/FLOPS": 202497132110959.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.11311519145965576, "optim/lr": 0.0019302, "optim/total_tokens": 1687158784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 3.3426175117492676, "created_at": "2025-01-15T14:38:25.368879+00:00"} {"global_step": 3219, "acc_step": 0, "speed/wps": 12893.50170962721, "speed/FLOPS": 202510161877708.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12404866516590118, "optim/lr": 0.0019307999999999999, "optim/total_tokens": 1687683072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.41119384765625, "created_at": "2025-01-15T14:38:35.539205+00:00"} {"global_step": 3220, "acc_step": 0, "speed/wps": 12890.05707884696, "speed/FLOPS": 202456059217886.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13317008316516876, "optim/lr": 0.0019314000000000002, "optim/total_tokens": 1688207360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431584, "loss/out": 3.2977888584136963, "created_at": "2025-01-15T14:38:45.711883+00:00"} {"global_step": 3221, "acc_step": 0, "speed/wps": 12893.87899626966, "speed/FLOPS": 202516087682874.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12277564406394958, "optim/lr": 0.0019320000000000001, "optim/total_tokens": 1688731648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.389694929122925, "created_at": "2025-01-15T14:38:55.881136+00:00"} {"global_step": 3222, "acc_step": 0, "speed/wps": 12894.883091427788, "speed/FLOPS": 202531858377104.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1412048637866974, "optim/lr": 0.0019326, "optim/total_tokens": 1689255936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.362489700317383, "created_at": "2025-01-15T14:39:06.050715+00:00"} {"global_step": 3223, "acc_step": 0, "speed/wps": 12892.27490990979, "speed/FLOPS": 202490893302347.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12247414141893387, "optim/lr": 0.0019332, "optim/total_tokens": 1689780224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.4006903171539307, "created_at": "2025-01-15T14:39:16.222167+00:00"} {"global_step": 3224, "acc_step": 0, "speed/wps": 12895.909383350076, "speed/FLOPS": 202547977702016.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15766143798828125, "optim/lr": 0.0019337999999999998, "optim/total_tokens": 1690304512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.297635793685913, "created_at": "2025-01-15T14:39:26.389474+00:00"} {"global_step": 3225, "acc_step": 0, "speed/wps": 12893.637692294897, "speed/FLOPS": 202512297672364.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.14993442595005035, "optim/lr": 0.0019344000000000002, "optim/total_tokens": 1690828800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.3540453910827637, "created_at": "2025-01-15T14:39:36.559651+00:00"} {"global_step": 3226, "acc_step": 0, "speed/wps": 12894.009941491697, "speed/FLOPS": 202518144357524.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11526117473840714, "optim/lr": 0.001935, "optim/total_tokens": 1691353088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.353717803955078, "created_at": "2025-01-15T14:39:46.731880+00:00"} {"global_step": 3227, "acc_step": 0, "speed/wps": 12895.3770418834, "speed/FLOPS": 202539616547768.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12728892266750336, "optim/lr": 0.0019356, "optim/total_tokens": 1691877376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 3.4613122940063477, "created_at": "2025-01-15T14:39:56.900608+00:00"} {"global_step": 3228, "acc_step": 0, "speed/wps": 12891.926884973509, "speed/FLOPS": 202485427092486.38, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1330118477344513, "optim/lr": 0.0019362, "optim/total_tokens": 1692401664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.3008365631103516, "created_at": "2025-01-15T14:40:07.072417+00:00"} {"global_step": 3229, "acc_step": 0, "speed/wps": 12894.495425257619, "speed/FLOPS": 202525769547192.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13904358446598053, "optim/lr": 0.0019367999999999998, "optim/total_tokens": 1692925952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.2928848266601562, "created_at": "2025-01-15T14:40:17.242797+00:00"} {"global_step": 3230, "acc_step": 0, "speed/wps": 12889.251998355634, "speed/FLOPS": 202443414322473.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10947974026203156, "optim/lr": 0.0019374000000000001, "optim/total_tokens": 1693450240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.2437760829925537, "created_at": "2025-01-15T14:40:27.419247+00:00"} {"global_step": 3231, "acc_step": 0, "speed/wps": 12897.184057272081, "speed/FLOPS": 202567998207547.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1304723173379898, "optim/lr": 0.001938, "optim/total_tokens": 1693974528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 3.4603583812713623, "created_at": "2025-01-15T14:40:37.587172+00:00"} {"global_step": 3232, "acc_step": 0, "speed/wps": 12899.627333026288, "speed/FLOPS": 202606373210679.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14117327332496643, "optim/lr": 0.0019386, "optim/total_tokens": 1694498816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.319669723510742, "created_at": "2025-01-15T14:40:47.759049+00:00"} {"global_step": 3233, "acc_step": 0, "speed/wps": 12892.289456283826, "speed/FLOPS": 202491121773140.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1521173119544983, "optim/lr": 0.0019391999999999999, "optim/total_tokens": 1695023104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.3491649627685547, "created_at": "2025-01-15T14:40:57.929543+00:00"} {"global_step": 3234, "acc_step": 0, "speed/wps": 12891.418399286633, "speed/FLOPS": 202477440626041.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14252084493637085, "optim/lr": 0.0019398, "optim/total_tokens": 1695547392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.2747504711151123, "created_at": "2025-01-15T14:41:08.102575+00:00"} {"global_step": 3235, "acc_step": 0, "speed/wps": 12899.635970068508, "speed/FLOPS": 202606508867296.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1178920641541481, "optim/lr": 0.0019404000000000001, "optim/total_tokens": 1696071680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 3.2535650730133057, "created_at": "2025-01-15T14:41:18.267236+00:00"} {"global_step": 3236, "acc_step": 0, "speed/wps": 12898.555225991944, "speed/FLOPS": 202589534296475.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14338253438472748, "optim/lr": 0.001941, "optim/total_tokens": 1696595968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.3493127822875977, "created_at": "2025-01-15T14:41:28.435303+00:00"} {"global_step": 3237, "acc_step": 0, "speed/wps": 12894.838694063466, "speed/FLOPS": 202531161055489.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15866945683956146, "optim/lr": 0.0019416000000000001, "optim/total_tokens": 1697120256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.286818742752075, "created_at": "2025-01-15T14:41:38.607440+00:00"} {"global_step": 3238, "acc_step": 0, "speed/wps": 12896.839299748897, "speed/FLOPS": 202562583316899.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19107450544834137, "optim/lr": 0.0019422, "optim/total_tokens": 1697644544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.275512218475342, "created_at": "2025-01-15T14:41:48.774081+00:00"} {"global_step": 3239, "acc_step": 0, "speed/wps": 12893.642710238844, "speed/FLOPS": 202512376486069.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12572112679481506, "optim/lr": 0.0019428, "optim/total_tokens": 1698168832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.3815813064575195, "created_at": "2025-01-15T14:41:58.943778+00:00"} {"global_step": 3240, "acc_step": 0, "speed/wps": 12892.34806491045, "speed/FLOPS": 202492042302157.9, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13960394263267517, "optim/lr": 0.0019434, "optim/total_tokens": 1698693120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.2924342155456543, "created_at": "2025-01-15T14:42:09.116348+00:00"} {"global_step": 3241, "acc_step": 0, "speed/wps": 12896.787556864496, "speed/FLOPS": 202561770623797.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1355935037136078, "optim/lr": 0.0019440000000000002, "optim/total_tokens": 1699217408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.3618898391723633, "created_at": "2025-01-15T14:42:19.283045+00:00"} {"global_step": 3242, "acc_step": 0, "speed/wps": 12896.715738920662, "speed/FLOPS": 202560642624301.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12979120016098022, "optim/lr": 0.0019446, "optim/total_tokens": 1699741696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.3683319091796875, "created_at": "2025-01-15T14:42:29.450761+00:00"} {"global_step": 3243, "acc_step": 0, "speed/wps": 12895.702907346542, "speed/FLOPS": 202544734712653.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1353786587715149, "optim/lr": 0.0019452, "optim/total_tokens": 1700265984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 3.195538282394409, "created_at": "2025-01-15T14:42:39.619706+00:00"} {"global_step": 3244, "acc_step": 0, "speed/wps": 12894.219871908212, "speed/FLOPS": 202521441603190.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1137162521481514, "optim/lr": 0.0019458, "optim/total_tokens": 1700790272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417722, "loss/out": 3.3021109104156494, "created_at": "2025-01-15T14:42:49.788820+00:00"} {"global_step": 3245, "acc_step": 0, "speed/wps": 12896.126504839574, "speed/FLOPS": 202551387893365.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12104363739490509, "optim/lr": 0.0019464000000000003, "optim/total_tokens": 1701314560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.422267436981201, "created_at": "2025-01-15T14:42:59.962208+00:00"} {"global_step": 3246, "acc_step": 0, "speed/wps": 12893.445035968416, "speed/FLOPS": 202509271740019.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10758896917104721, "optim/lr": 0.0019470000000000002, "optim/total_tokens": 1701838848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.313615083694458, "created_at": "2025-01-15T14:43:10.132063+00:00"} {"global_step": 3247, "acc_step": 0, "speed/wps": 12892.391730608722, "speed/FLOPS": 202492728131952.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12043917179107666, "optim/lr": 0.0019476, "optim/total_tokens": 1702363136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.3847951889038086, "created_at": "2025-01-15T14:43:20.305672+00:00"} {"global_step": 3248, "acc_step": 0, "speed/wps": 12891.663192380027, "speed/FLOPS": 202481285437954.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13732342422008514, "optim/lr": 0.0019482, "optim/total_tokens": 1702887424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.305072069168091, "created_at": "2025-01-15T14:43:30.477542+00:00"} {"global_step": 3249, "acc_step": 0, "speed/wps": 12894.974359938078, "speed/FLOPS": 202533291874474.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11401262879371643, "optim/lr": 0.0019487999999999999, "optim/total_tokens": 1703411712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 3.406881809234619, "created_at": "2025-01-15T14:43:40.646869+00:00"} {"global_step": 3250, "acc_step": 0, "speed/wps": 12896.540764548014, "speed/FLOPS": 202557894411338.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13680781424045563, "optim/lr": 0.0019494000000000002, "optim/total_tokens": 1703936000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 3.282292127609253, "created_at": "2025-01-15T14:43:50.815645+00:00"} {"global_step": 3251, "acc_step": 0, "speed/wps": 12891.852083824839, "speed/FLOPS": 202484252237658.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13124187290668488, "optim/lr": 0.0019500000000000001, "optim/total_tokens": 1704460288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.242185115814209, "created_at": "2025-01-15T14:44:00.989704+00:00"} {"global_step": 3252, "acc_step": 0, "speed/wps": 12898.048505769422, "speed/FLOPS": 202581575559074.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14601732790470123, "optim/lr": 0.0019506, "optim/total_tokens": 1704984576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 3.2718300819396973, "created_at": "2025-01-15T14:44:11.157195+00:00"} {"global_step": 3253, "acc_step": 0, "speed/wps": 12895.321164648205, "speed/FLOPS": 202538738919006.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13554663956165314, "optim/lr": 0.0019512, "optim/total_tokens": 1705508864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.264051675796509, "created_at": "2025-01-15T14:44:21.325158+00:00"} {"global_step": 3254, "acc_step": 0, "speed/wps": 12893.253785094688, "speed/FLOPS": 202506267882241.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15079206228256226, "optim/lr": 0.0019517999999999998, "optim/total_tokens": 1706033152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.440566062927246, "created_at": "2025-01-15T14:44:31.494639+00:00"} {"global_step": 3255, "acc_step": 0, "speed/wps": 12892.338001516504, "speed/FLOPS": 202491884242727.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13316775858402252, "optim/lr": 0.0019524000000000002, "optim/total_tokens": 1706557440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 3.3071937561035156, "created_at": "2025-01-15T14:44:41.666771+00:00"} {"global_step": 3256, "acc_step": 0, "speed/wps": 12892.016233928643, "speed/FLOPS": 202486830440605.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13968515396118164, "optim/lr": 0.001953, "optim/total_tokens": 1707081728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 3.297842502593994, "created_at": "2025-01-15T14:44:51.838300+00:00"} {"global_step": 3257, "acc_step": 0, "speed/wps": 12893.820274495358, "speed/FLOPS": 202515165376716.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1500897854566574, "optim/lr": 0.0019536, "optim/total_tokens": 1707606016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.2106151580810547, "created_at": "2025-01-15T14:45:02.007505+00:00"} {"global_step": 3258, "acc_step": 0, "speed/wps": 12894.575912539143, "speed/FLOPS": 202527033710550.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16577717661857605, "optim/lr": 0.0019542, "optim/total_tokens": 1708130304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 3.2698934078216553, "created_at": "2025-01-15T14:45:12.177379+00:00"} {"global_step": 3259, "acc_step": 0, "speed/wps": 12892.183500098676, "speed/FLOPS": 202489457585653.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1460476815700531, "optim/lr": 0.0019548, "optim/total_tokens": 1708654592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439064, "loss/out": 3.447195529937744, "created_at": "2025-01-15T14:45:22.348486+00:00"} {"global_step": 3260, "acc_step": 0, "speed/wps": 12894.6941359628, "speed/FLOPS": 202528890571874.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11857052147388458, "optim/lr": 0.0019554000000000004, "optim/total_tokens": 1709178880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.28085994720459, "created_at": "2025-01-15T14:45:32.517834+00:00"} {"global_step": 3261, "acc_step": 0, "speed/wps": 12895.17992528258, "speed/FLOPS": 202536520560686.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12637829780578613, "optim/lr": 0.0019560000000000003, "optim/total_tokens": 1709703168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.388225793838501, "created_at": "2025-01-15T14:45:42.689538+00:00"} {"global_step": 3262, "acc_step": 0, "speed/wps": 12895.61405082352, "speed/FLOPS": 202543339098857.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1161600649356842, "optim/lr": 0.0019566, "optim/total_tokens": 1710227456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.290511131286621, "created_at": "2025-01-15T14:45:52.860826+00:00"} {"global_step": 3263, "acc_step": 0, "speed/wps": 12894.75719926604, "speed/FLOPS": 202529881067709.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10233401507139206, "optim/lr": 0.0019572, "optim/total_tokens": 1710751744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 3.2978906631469727, "created_at": "2025-01-15T14:46:03.032974+00:00"} {"global_step": 3264, "acc_step": 0, "speed/wps": 12892.752551712932, "speed/FLOPS": 202498395323214.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13090866804122925, "optim/lr": 0.0019578, "optim/total_tokens": 1711276032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 3.305799961090088, "created_at": "2025-01-15T14:46:13.208607+00:00"} {"global_step": 3265, "acc_step": 0, "speed/wps": 12898.467060715746, "speed/FLOPS": 202588149539661.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13090667128562927, "optim/lr": 0.0019584000000000003, "optim/total_tokens": 1711800320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.3586997985839844, "created_at": "2025-01-15T14:46:23.377583+00:00"} {"global_step": 3266, "acc_step": 0, "speed/wps": 12893.017253860127, "speed/FLOPS": 202502552834174.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.12308759987354279, "optim/lr": 0.0019590000000000002, "optim/total_tokens": 1712324608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347602, "loss/out": 3.3624460697174072, "created_at": "2025-01-15T14:46:33.548145+00:00"} {"global_step": 3267, "acc_step": 0, "speed/wps": 12894.571082895864, "speed/FLOPS": 202526957854365.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10767341405153275, "optim/lr": 0.0019596, "optim/total_tokens": 1712848896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.253274440765381, "created_at": "2025-01-15T14:46:43.717591+00:00"} {"global_step": 3268, "acc_step": 0, "speed/wps": 12895.854720968597, "speed/FLOPS": 202547119154201.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10663697123527527, "optim/lr": 0.0019602, "optim/total_tokens": 1713373184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.314936876296997, "created_at": "2025-01-15T14:46:53.887620+00:00"} {"global_step": 3269, "acc_step": 0, "speed/wps": 12890.523376480338, "speed/FLOPS": 202463383062978.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1349819153547287, "optim/lr": 0.0019608, "optim/total_tokens": 1713897472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 3.35300350189209, "created_at": "2025-01-15T14:47:04.064477+00:00"} {"global_step": 3270, "acc_step": 0, "speed/wps": 12893.84190261591, "speed/FLOPS": 202515505076070.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12216927856206894, "optim/lr": 0.0019614000000000003, "optim/total_tokens": 1714421760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 3.161160469055176, "created_at": "2025-01-15T14:47:14.234756+00:00"} {"global_step": 3271, "acc_step": 0, "speed/wps": 12895.623052336849, "speed/FLOPS": 202543480479993.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16581781208515167, "optim/lr": 0.001962, "optim/total_tokens": 1714946048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.3002357482910156, "created_at": "2025-01-15T14:47:24.402350+00:00"} {"global_step": 3272, "acc_step": 0, "speed/wps": 12890.38403206086, "speed/FLOPS": 202461194467391.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17547400295734406, "optim/lr": 0.0019626, "optim/total_tokens": 1715470336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.360092878341675, "created_at": "2025-01-15T14:47:34.574857+00:00"} {"global_step": 3273, "acc_step": 0, "speed/wps": 12886.258052222096, "speed/FLOPS": 202396390284332.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1376456320285797, "optim/lr": 0.0019632, "optim/total_tokens": 1715994624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.3531150817871094, "created_at": "2025-01-15T14:47:44.749802+00:00"} {"global_step": 3274, "acc_step": 0, "speed/wps": 12892.414206636286, "speed/FLOPS": 202493081148850.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12063529342412949, "optim/lr": 0.0019638, "optim/total_tokens": 1716518912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296240, "loss/out": 3.2904059886932373, "created_at": "2025-01-15T14:47:54.921223+00:00"} {"global_step": 3275, "acc_step": 0, "speed/wps": 12894.813756862708, "speed/FLOPS": 202530769382483.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.1459648162126541, "optim/lr": 0.0019644000000000003, "optim/total_tokens": 1717043200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.3059091567993164, "created_at": "2025-01-15T14:48:05.089485+00:00"} {"global_step": 3276, "acc_step": 0, "speed/wps": 12896.441678956446, "speed/FLOPS": 202556338135965.47, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14327074587345123, "optim/lr": 0.001965, "optim/total_tokens": 1717567488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.333433151245117, "created_at": "2025-01-15T14:48:15.259454+00:00"} {"global_step": 3277, "acc_step": 0, "speed/wps": 12894.402483592508, "speed/FLOPS": 202524309770626.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1720171421766281, "optim/lr": 0.0019656, "optim/total_tokens": 1718091776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 3.304873466491699, "created_at": "2025-01-15T14:48:25.429405+00:00"} {"global_step": 3278, "acc_step": 0, "speed/wps": 12893.550162403726, "speed/FLOPS": 202510922895144.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15476074814796448, "optim/lr": 0.0019662, "optim/total_tokens": 1718616064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.248131275177002, "created_at": "2025-01-15T14:48:35.600003+00:00"} {"global_step": 3279, "acc_step": 0, "speed/wps": 12894.871230597913, "speed/FLOPS": 202531672086472.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14938507974147797, "optim/lr": 0.0019668, "optim/total_tokens": 1719140352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.406186580657959, "created_at": "2025-01-15T14:48:45.768184+00:00"} {"global_step": 3280, "acc_step": 0, "speed/wps": 12893.865286205451, "speed/FLOPS": 202515872347476.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18733134865760803, "optim/lr": 0.0019674000000000002, "optim/total_tokens": 1719664640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 3.260451078414917, "created_at": "2025-01-15T14:48:55.939162+00:00"} {"global_step": 3281, "acc_step": 0, "speed/wps": 12888.503348355744, "speed/FLOPS": 202431655745472.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10882802307605743, "optim/lr": 0.001968, "optim/total_tokens": 1720188928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 3.2951998710632324, "created_at": "2025-01-15T14:49:06.112343+00:00"} {"global_step": 3282, "acc_step": 0, "speed/wps": 12896.616941361724, "speed/FLOPS": 202559090872876.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13923245668411255, "optim/lr": 0.0019686, "optim/total_tokens": 1720713216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.3589892387390137, "created_at": "2025-01-15T14:49:16.279376+00:00"} {"global_step": 3283, "acc_step": 0, "speed/wps": 12896.569483050283, "speed/FLOPS": 202558345474880.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1421346664428711, "optim/lr": 0.0019692, "optim/total_tokens": 1721237504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.315518856048584, "created_at": "2025-01-15T14:49:26.447098+00:00"} {"global_step": 3284, "acc_step": 0, "speed/wps": 12891.218099795904, "speed/FLOPS": 202474294647299.2, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09948140382766724, "optim/lr": 0.0019698, "optim/total_tokens": 1721761792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.2561464309692383, "created_at": "2025-01-15T14:49:36.624850+00:00"} {"global_step": 3285, "acc_step": 0, "speed/wps": 12882.989681269115, "speed/FLOPS": 202345056027304.03, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11038802564144135, "optim/lr": 0.0019704, "optim/total_tokens": 1722286080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 3.3318402767181396, "created_at": "2025-01-15T14:49:46.803160+00:00"} {"global_step": 3286, "acc_step": 0, "speed/wps": 12883.67487681212, "speed/FLOPS": 202355817964864.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12569968402385712, "optim/lr": 0.001971, "optim/total_tokens": 1722810368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.386615037918091, "created_at": "2025-01-15T14:49:56.980574+00:00"} {"global_step": 3287, "acc_step": 0, "speed/wps": 12883.29002714058, "speed/FLOPS": 202349773371933.84, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.12719286978244781, "optim/lr": 0.0019716, "optim/total_tokens": 1723334656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.3402023315429688, "created_at": "2025-01-15T14:50:07.159665+00:00"} {"global_step": 3288, "acc_step": 0, "speed/wps": 12880.315394320582, "speed/FLOPS": 202303052675921.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12507861852645874, "optim/lr": 0.0019722, "optim/total_tokens": 1723858944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.3310916423797607, "created_at": "2025-01-15T14:50:17.343656+00:00"} {"global_step": 3289, "acc_step": 0, "speed/wps": 12890.041036981027, "speed/FLOPS": 202455807258337.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13754339516162872, "optim/lr": 0.0019728, "optim/total_tokens": 1724383232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.3144030570983887, "created_at": "2025-01-15T14:50:27.515860+00:00"} {"global_step": 3290, "acc_step": 0, "speed/wps": 12895.459498052405, "speed/FLOPS": 202540911635209.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11540025472640991, "optim/lr": 0.0019734, "optim/total_tokens": 1724907520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.3739516735076904, "created_at": "2025-01-15T14:50:37.687507+00:00"} {"global_step": 3291, "acc_step": 0, "speed/wps": 12885.437670617039, "speed/FLOPS": 202383505063903.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.10235729068517685, "optim/lr": 0.001974, "optim/total_tokens": 1725431808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.252634048461914, "created_at": "2025-01-15T14:50:47.863146+00:00"} {"global_step": 3292, "acc_step": 0, "speed/wps": 12888.129192285489, "speed/FLOPS": 202425779110244.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12073159962892532, "optim/lr": 0.0019746, "optim/total_tokens": 1725956096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.275222063064575, "created_at": "2025-01-15T14:50:58.040091+00:00"} {"global_step": 3293, "acc_step": 0, "speed/wps": 12887.576021481293, "speed/FLOPS": 202417090802628.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13793981075286865, "optim/lr": 0.0019752, "optim/total_tokens": 1726480384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338116, "loss/out": 3.274352788925171, "created_at": "2025-01-15T14:51:08.214054+00:00"} {"global_step": 3294, "acc_step": 0, "speed/wps": 12883.989553464955, "speed/FLOPS": 202360760394108.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15443682670593262, "optim/lr": 0.0019757999999999998, "optim/total_tokens": 1727004672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453409, "loss/out": 3.4273481369018555, "created_at": "2025-01-15T14:51:18.391020+00:00"} {"global_step": 3295, "acc_step": 0, "speed/wps": 12894.46084207352, "speed/FLOPS": 202525226370764.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12941370904445648, "optim/lr": 0.0019764, "optim/total_tokens": 1727528960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.3711347579956055, "created_at": "2025-01-15T14:51:28.560921+00:00"} {"global_step": 3296, "acc_step": 0, "speed/wps": 12896.97881094912, "speed/FLOPS": 202564774532007.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11178003251552582, "optim/lr": 0.001977, "optim/total_tokens": 1728053248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.2534167766571045, "created_at": "2025-01-15T14:51:38.728126+00:00"} {"global_step": 3297, "acc_step": 0, "speed/wps": 12891.247444043192, "speed/FLOPS": 202474755539026.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.11966649442911148, "optim/lr": 0.0019776, "optim/total_tokens": 1728577536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 3.294137716293335, "created_at": "2025-01-15T14:51:48.909360+00:00"} {"global_step": 3298, "acc_step": 0, "speed/wps": 12884.133212308327, "speed/FLOPS": 202363016753651.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0036, "optim/grad_norm": 0.10886368155479431, "optim/lr": 0.0019782, "optim/total_tokens": 1729101824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.2696475982666016, "created_at": "2025-01-15T14:51:59.087587+00:00"} {"global_step": 3299, "acc_step": 0, "speed/wps": 12884.72832951404, "speed/FLOPS": 202372363887143.62, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10118205100297928, "optim/lr": 0.0019787999999999997, "optim/total_tokens": 1729626112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.2694509029388428, "created_at": "2025-01-15T14:52:09.265123+00:00"} {"global_step": 3300, "acc_step": 0, "speed/wps": 12881.760642284105, "speed/FLOPS": 202325752281167.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0029, "optim/grad_norm": 0.10467881709337234, "optim/lr": 0.0019794, "optim/total_tokens": 1730150400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 3.2037925720214844, "created_at": "2025-01-15T14:52:19.444997+00:00"} {"global_step": 3301, "acc_step": 0, "speed/wps": 12896.547755962902, "speed/FLOPS": 202558004221116.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12651345133781433, "optim/lr": 0.00198, "optim/total_tokens": 1730674688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.355698585510254, "created_at": "2025-01-15T14:52:29.613027+00:00"} {"global_step": 3302, "acc_step": 0, "speed/wps": 12896.914287604104, "speed/FLOPS": 202563761104209.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17449072003364563, "optim/lr": 0.0019806, "optim/total_tokens": 1731198976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 3.280991792678833, "created_at": "2025-01-15T14:52:39.779925+00:00"} {"global_step": 3303, "acc_step": 0, "speed/wps": 12894.096141538797, "speed/FLOPS": 202519498247715.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16297729313373566, "optim/lr": 0.0019812, "optim/total_tokens": 1731723264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 3.4362800121307373, "created_at": "2025-01-15T14:52:49.949365+00:00"} {"global_step": 3304, "acc_step": 0, "speed/wps": 12891.901551612835, "speed/FLOPS": 202485029197245.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17263126373291016, "optim/lr": 0.0019818, "optim/total_tokens": 1732247552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.4259071350097656, "created_at": "2025-01-15T14:53:00.123187+00:00"} {"global_step": 3305, "acc_step": 0, "speed/wps": 12891.264374521963, "speed/FLOPS": 202475021455460.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13865487277507782, "optim/lr": 0.0019824, "optim/total_tokens": 1732771840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 3.2433905601501465, "created_at": "2025-01-15T14:53:10.295172+00:00"} {"global_step": 3306, "acc_step": 0, "speed/wps": 12897.637736506056, "speed/FLOPS": 202575123863337.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1800910085439682, "optim/lr": 0.001983, "optim/total_tokens": 1733296128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.3161587715148926, "created_at": "2025-01-15T14:53:20.461169+00:00"} {"global_step": 3307, "acc_step": 0, "speed/wps": 12894.282141419946, "speed/FLOPS": 202522419631439.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14273405075073242, "optim/lr": 0.0019836, "optim/total_tokens": 1733820416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 3.544945240020752, "created_at": "2025-01-15T14:53:30.632325+00:00"} {"global_step": 3308, "acc_step": 0, "speed/wps": 12897.786762412108, "speed/FLOPS": 202577464519975.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13108354806900024, "optim/lr": 0.0019842, "optim/total_tokens": 1734344704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.3407397270202637, "created_at": "2025-01-15T14:53:40.801099+00:00"} {"global_step": 3309, "acc_step": 0, "speed/wps": 12897.191570602072, "speed/FLOPS": 202568116214719.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15017834305763245, "optim/lr": 0.0019848, "optim/total_tokens": 1734868992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.4043972492218018, "created_at": "2025-01-15T14:53:50.970301+00:00"} {"global_step": 3310, "acc_step": 0, "speed/wps": 12894.764888168385, "speed/FLOPS": 202530001832486.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10846797376871109, "optim/lr": 0.0019854, "optim/total_tokens": 1735393280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.327915668487549, "created_at": "2025-01-15T14:54:01.142199+00:00"} {"global_step": 3311, "acc_step": 0, "speed/wps": 12894.496162454558, "speed/FLOPS": 202525781125883.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1265215128660202, "optim/lr": 0.0019860000000000004, "optim/total_tokens": 1735917568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.38331937789917, "created_at": "2025-01-15T14:54:11.313452+00:00"} {"global_step": 3312, "acc_step": 0, "speed/wps": 12890.195943950748, "speed/FLOPS": 202458240285161.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1136384904384613, "optim/lr": 0.0019866000000000003, "optim/total_tokens": 1736441856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 3.3186841011047363, "created_at": "2025-01-15T14:54:21.487189+00:00"} {"global_step": 3313, "acc_step": 0, "speed/wps": 12895.32301969233, "speed/FLOPS": 202538768055023.62, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11402800679206848, "optim/lr": 0.0019872, "optim/total_tokens": 1736966144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.340758800506592, "created_at": "2025-01-15T14:54:31.655041+00:00"} {"global_step": 3314, "acc_step": 0, "speed/wps": 12893.225930620605, "speed/FLOPS": 202505830389449.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12777073681354523, "optim/lr": 0.0019878, "optim/total_tokens": 1737490432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.340578079223633, "created_at": "2025-01-15T14:54:41.826619+00:00"} {"global_step": 3315, "acc_step": 0, "speed/wps": 12895.57468559751, "speed/FLOPS": 202542720813889.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17803524434566498, "optim/lr": 0.0019884, "optim/total_tokens": 1738014720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.304872989654541, "created_at": "2025-01-15T14:54:51.995956+00:00"} {"global_step": 3316, "acc_step": 0, "speed/wps": 12897.833008659341, "speed/FLOPS": 202578190880836.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17022764682769775, "optim/lr": 0.0019890000000000003, "optim/total_tokens": 1738539008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 3.478292942047119, "created_at": "2025-01-15T14:55:02.162414+00:00"} {"global_step": 3317, "acc_step": 0, "speed/wps": 12898.30496983529, "speed/FLOPS": 202585603679647.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1519336700439453, "optim/lr": 0.0019896000000000002, "optim/total_tokens": 1739063296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.3336567878723145, "created_at": "2025-01-15T14:55:12.328535+00:00"} {"global_step": 3318, "acc_step": 0, "speed/wps": 12895.678563522219, "speed/FLOPS": 202544352359440.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1482529491186142, "optim/lr": 0.0019902, "optim/total_tokens": 1739587584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 3.2985963821411133, "created_at": "2025-01-15T14:55:22.499706+00:00"} {"global_step": 3319, "acc_step": 0, "speed/wps": 12897.563236664122, "speed/FLOPS": 202573953740946.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15197046101093292, "optim/lr": 0.0019908, "optim/total_tokens": 1740111872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 3.2623395919799805, "created_at": "2025-01-15T14:55:32.665843+00:00"} {"global_step": 3320, "acc_step": 0, "speed/wps": 12889.76051738465, "speed/FLOPS": 202451401312602.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11507190763950348, "optim/lr": 0.0019914, "optim/total_tokens": 1740636160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.3438796997070312, "created_at": "2025-01-15T14:55:42.843779+00:00"} {"global_step": 3321, "acc_step": 0, "speed/wps": 12894.420224524303, "speed/FLOPS": 202524588416339.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11168621480464935, "optim/lr": 0.0019920000000000003, "optim/total_tokens": 1741160448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.275169849395752, "created_at": "2025-01-15T14:55:53.015055+00:00"} {"global_step": 3322, "acc_step": 0, "speed/wps": 12894.00855462648, "speed/FLOPS": 202518122574900.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13129876554012299, "optim/lr": 0.0019926, "optim/total_tokens": 1741684736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.3572235107421875, "created_at": "2025-01-15T14:56:03.184858+00:00"} {"global_step": 3323, "acc_step": 0, "speed/wps": 12887.943519759192, "speed/FLOPS": 202422862868078.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12245190143585205, "optim/lr": 0.0019932, "optim/total_tokens": 1742209024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 3.3186254501342773, "created_at": "2025-01-15T14:56:13.363525+00:00"} {"global_step": 3324, "acc_step": 0, "speed/wps": 12889.943662719903, "speed/FLOPS": 202454277861760.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10826298594474792, "optim/lr": 0.0019938, "optim/total_tokens": 1742733312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.3648955821990967, "created_at": "2025-01-15T14:56:23.535985+00:00"} {"global_step": 3325, "acc_step": 0, "speed/wps": 12895.263826324574, "speed/FLOPS": 202537838341842.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11635314673185349, "optim/lr": 0.0019944, "optim/total_tokens": 1743257600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.251796245574951, "created_at": "2025-01-15T14:56:33.707770+00:00"} {"global_step": 3326, "acc_step": 0, "speed/wps": 12893.019183834085, "speed/FLOPS": 202502583147067.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15518823266029358, "optim/lr": 0.0019950000000000002, "optim/total_tokens": 1743781888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429723, "loss/out": 3.1869475841522217, "created_at": "2025-01-15T14:56:43.877410+00:00"} {"global_step": 3327, "acc_step": 0, "speed/wps": 12894.305934179136, "speed/FLOPS": 202522793329416.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14848844707012177, "optim/lr": 0.0019956, "optim/total_tokens": 1744306176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 3.3190810680389404, "created_at": "2025-01-15T14:56:54.046596+00:00"} {"global_step": 3328, "acc_step": 0, "speed/wps": 12896.012502117932, "speed/FLOPS": 202549597323965.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12472284585237503, "optim/lr": 0.0019962, "optim/total_tokens": 1744830464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.379460334777832, "created_at": "2025-01-15T14:57:04.219507+00:00"} {"global_step": 3329, "acc_step": 0, "speed/wps": 12895.28294258711, "speed/FLOPS": 202538138589014.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1556658297777176, "optim/lr": 0.0019968, "optim/total_tokens": 1745354752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.3504257202148438, "created_at": "2025-01-15T14:57:14.390023+00:00"} {"global_step": 3330, "acc_step": 0, "speed/wps": 12894.251286571343, "speed/FLOPS": 202521935013642.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1360112726688385, "optim/lr": 0.0019974, "optim/total_tokens": 1745879040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 3.3706140518188477, "created_at": "2025-01-15T14:57:24.561224+00:00"} {"global_step": 3331, "acc_step": 0, "speed/wps": 12894.875069208561, "speed/FLOPS": 202531732377127.56, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11846692860126495, "optim/lr": 0.001998, "optim/total_tokens": 1746403328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.3732106685638428, "created_at": "2025-01-15T14:57:34.731318+00:00"} {"global_step": 3332, "acc_step": 0, "speed/wps": 12894.510535448897, "speed/FLOPS": 202526006873510.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1394084244966507, "optim/lr": 0.0019986, "optim/total_tokens": 1746927616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.3396008014678955, "created_at": "2025-01-15T14:57:44.909702+00:00"} {"global_step": 3333, "acc_step": 0, "speed/wps": 12892.695869285828, "speed/FLOPS": 202497505047806.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11835363507270813, "optim/lr": 0.0019992, "optim/total_tokens": 1747451904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.2604594230651855, "created_at": "2025-01-15T14:57:55.082076+00:00"} {"global_step": 3334, "acc_step": 0, "speed/wps": 12894.651112725565, "speed/FLOPS": 202528214832813.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11127064377069473, "optim/lr": 0.0019998, "optim/total_tokens": 1747976192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 3.2154288291931152, "created_at": "2025-01-15T14:58:05.252866+00:00"} {"global_step": 3335, "acc_step": 0, "speed/wps": 12891.183574336344, "speed/FLOPS": 202473752377513.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11669651418924332, "optim/lr": 0.0020004, "optim/total_tokens": 1748500480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 3.2986533641815186, "created_at": "2025-01-15T14:58:15.427317+00:00"} {"global_step": 3336, "acc_step": 0, "speed/wps": 12895.063356515955, "speed/FLOPS": 202534689688024.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11537224799394608, "optim/lr": 0.002001, "optim/total_tokens": 1749024768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 3.4437005519866943, "created_at": "2025-01-15T14:58:25.599548+00:00"} {"global_step": 3337, "acc_step": 0, "speed/wps": 12894.666413493016, "speed/FLOPS": 202528455152391.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1268075555562973, "optim/lr": 0.0020016, "optim/total_tokens": 1749549056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 3.2946152687072754, "created_at": "2025-01-15T14:58:35.769218+00:00"} {"global_step": 3338, "acc_step": 0, "speed/wps": 12889.442293566613, "speed/FLOPS": 202446403170254.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09650451689958572, "optim/lr": 0.0020022, "optim/total_tokens": 1750073344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 3.469700336456299, "created_at": "2025-01-15T14:58:45.942605+00:00"} {"global_step": 3339, "acc_step": 0, "speed/wps": 12898.689774903072, "speed/FLOPS": 202591647572014.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12594209611415863, "optim/lr": 0.0020028, "optim/total_tokens": 1750597632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 3.36488938331604, "created_at": "2025-01-15T14:58:56.112941+00:00"} {"global_step": 3340, "acc_step": 0, "speed/wps": 12891.50598643908, "speed/FLOPS": 202478816302628.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14774811267852783, "optim/lr": 0.0020034, "optim/total_tokens": 1751121920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.4469330310821533, "created_at": "2025-01-15T14:59:06.289703+00:00"} {"global_step": 3341, "acc_step": 0, "speed/wps": 12897.679201720299, "speed/FLOPS": 202575775131506.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13392478227615356, "optim/lr": 0.002004, "optim/total_tokens": 1751646208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.280703067779541, "created_at": "2025-01-15T14:59:16.455706+00:00"} {"global_step": 3342, "acc_step": 0, "speed/wps": 12895.48468942875, "speed/FLOPS": 202541307300391.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1270742267370224, "optim/lr": 0.0020046, "optim/total_tokens": 1752170496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 3.3537538051605225, "created_at": "2025-01-15T14:59:26.625829+00:00"} {"global_step": 3343, "acc_step": 0, "speed/wps": 12894.602421296313, "speed/FLOPS": 202527450067008.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1616721749305725, "optim/lr": 0.0020052, "optim/total_tokens": 1752694784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.4661765098571777, "created_at": "2025-01-15T14:59:36.794141+00:00"} {"global_step": 3344, "acc_step": 0, "speed/wps": 12894.16710336837, "speed/FLOPS": 202520612800760.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15973825752735138, "optim/lr": 0.0020058, "optim/total_tokens": 1753219072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369775, "loss/out": 3.275855541229248, "created_at": "2025-01-15T14:59:46.962905+00:00"} {"global_step": 3345, "acc_step": 0, "speed/wps": 12893.923219281007, "speed/FLOPS": 202516782266036.44, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12285465002059937, "optim/lr": 0.0020063999999999998, "optim/total_tokens": 1753743360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.3213255405426025, "created_at": "2025-01-15T14:59:57.133061+00:00"} {"global_step": 3346, "acc_step": 0, "speed/wps": 12893.370266811982, "speed/FLOPS": 202508097387674.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13836166262626648, "optim/lr": 0.002007, "optim/total_tokens": 1754267648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 3.2489264011383057, "created_at": "2025-01-15T15:00:07.302419+00:00"} {"global_step": 3347, "acc_step": 0, "speed/wps": 12887.699046110132, "speed/FLOPS": 202419023073477.66, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12787455320358276, "optim/lr": 0.0020076, "optim/total_tokens": 1754791936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.411623001098633, "created_at": "2025-01-15T15:00:17.476312+00:00"} {"global_step": 3348, "acc_step": 0, "speed/wps": 12892.510437185809, "speed/FLOPS": 202494592581865.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11689532548189163, "optim/lr": 0.0020082, "optim/total_tokens": 1755316224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372023, "loss/out": 3.3585472106933594, "created_at": "2025-01-15T15:00:27.647384+00:00"} {"global_step": 3349, "acc_step": 0, "speed/wps": 12896.025594000408, "speed/FLOPS": 202549802949970.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1311679184436798, "optim/lr": 0.0020088, "optim/total_tokens": 1755840512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.3232626914978027, "created_at": "2025-01-15T15:00:37.818107+00:00"} {"global_step": 3350, "acc_step": 0, "speed/wps": 12894.09104409272, "speed/FLOPS": 202519418185319.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11208468675613403, "optim/lr": 0.0020093999999999997, "optim/total_tokens": 1756364800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.4476511478424072, "created_at": "2025-01-15T15:00:47.987180+00:00"} {"global_step": 3351, "acc_step": 0, "speed/wps": 12892.413914923376, "speed/FLOPS": 202493076567098.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1220933124423027, "optim/lr": 0.00201, "optim/total_tokens": 1756889088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.3500125408172607, "created_at": "2025-01-15T15:00:58.157462+00:00"} {"global_step": 3352, "acc_step": 0, "speed/wps": 12896.462285372407, "speed/FLOPS": 202556661788044.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13915742933750153, "optim/lr": 0.0020106, "optim/total_tokens": 1757413376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.3910393714904785, "created_at": "2025-01-15T15:01:08.324334+00:00"} {"global_step": 3353, "acc_step": 0, "speed/wps": 12884.40765056848, "speed/FLOPS": 202367327183644.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12873885035514832, "optim/lr": 0.0020112, "optim/total_tokens": 1757937664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 3.2770237922668457, "created_at": "2025-01-15T15:01:18.500791+00:00"} {"global_step": 3354, "acc_step": 0, "speed/wps": 12882.973835062852, "speed/FLOPS": 202344807140859.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12404341250658035, "optim/lr": 0.0020118, "optim/total_tokens": 1758461952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.3559460639953613, "created_at": "2025-01-15T15:01:28.681348+00:00"} {"global_step": 3355, "acc_step": 0, "speed/wps": 12887.161694356539, "speed/FLOPS": 202410583225788.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1141357570886612, "optim/lr": 0.0020123999999999997, "optim/total_tokens": 1758986240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.250934600830078, "created_at": "2025-01-15T15:01:38.855644+00:00"} {"global_step": 3356, "acc_step": 0, "speed/wps": 12888.074062376865, "speed/FLOPS": 202424913219273.03, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13070236146450043, "optim/lr": 0.002013, "optim/total_tokens": 1759510528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 3.2399916648864746, "created_at": "2025-01-15T15:01:49.030829+00:00"} {"global_step": 3357, "acc_step": 0, "speed/wps": 12889.919837653382, "speed/FLOPS": 202453903656351.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18872356414794922, "optim/lr": 0.0020136, "optim/total_tokens": 1760034816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.30511474609375, "created_at": "2025-01-15T15:01:59.207058+00:00"} {"global_step": 3358, "acc_step": 0, "speed/wps": 12896.062483969823, "speed/FLOPS": 202550382357632.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18169063329696655, "optim/lr": 0.0020142, "optim/total_tokens": 1760559104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.342103958129883, "created_at": "2025-01-15T15:02:09.375171+00:00"} {"global_step": 3359, "acc_step": 0, "speed/wps": 12890.298604188507, "speed/FLOPS": 202459852705265.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1925436556339264, "optim/lr": 0.0020147999999999998, "optim/total_tokens": 1761083392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 3.3717551231384277, "created_at": "2025-01-15T15:02:19.550532+00:00"} {"global_step": 3360, "acc_step": 0, "speed/wps": 12892.419375980176, "speed/FLOPS": 202493162340499.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18702885508537292, "optim/lr": 0.0020154, "optim/total_tokens": 1761607680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 3.327939748764038, "created_at": "2025-01-15T15:02:29.722546+00:00"} {"global_step": 3361, "acc_step": 0, "speed/wps": 12895.610620609817, "speed/FLOPS": 202543285222637.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1513177901506424, "optim/lr": 0.002016, "optim/total_tokens": 1762131968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 3.3307137489318848, "created_at": "2025-01-15T15:02:39.895528+00:00"} {"global_step": 3362, "acc_step": 0, "speed/wps": 12893.17964062986, "speed/FLOPS": 202505103341536.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15297254920005798, "optim/lr": 0.0020166, "optim/total_tokens": 1762656256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.2338857650756836, "created_at": "2025-01-15T15:02:50.065627+00:00"} {"global_step": 3363, "acc_step": 0, "speed/wps": 12898.192910947722, "speed/FLOPS": 202583843640833.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14089713990688324, "optim/lr": 0.0020172000000000002, "optim/total_tokens": 1763180544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.328500986099243, "created_at": "2025-01-15T15:03:00.231185+00:00"} {"global_step": 3364, "acc_step": 0, "speed/wps": 12892.02537921995, "speed/FLOPS": 202486974079972.0, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12288685142993927, "optim/lr": 0.0020178, "optim/total_tokens": 1763704832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 3.3635356426239014, "created_at": "2025-01-15T15:03:10.407139+00:00"} {"global_step": 3365, "acc_step": 0, "speed/wps": 12890.232268113876, "speed/FLOPS": 202458810806058.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11990170925855637, "optim/lr": 0.0020184, "optim/total_tokens": 1764229120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.5056538581848145, "created_at": "2025-01-15T15:03:20.581950+00:00"} {"global_step": 3366, "acc_step": 0, "speed/wps": 12894.18779348003, "speed/FLOPS": 202520937767395.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13686370849609375, "optim/lr": 0.002019, "optim/total_tokens": 1764753408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411556, "loss/out": 3.3370161056518555, "created_at": "2025-01-15T15:03:30.750658+00:00"} {"global_step": 3367, "acc_step": 0, "speed/wps": 12894.30692416774, "speed/FLOPS": 202522808878548.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12983635067939758, "optim/lr": 0.0020196000000000003, "optim/total_tokens": 1765277696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 3.3320889472961426, "created_at": "2025-01-15T15:03:40.920177+00:00"} {"global_step": 3368, "acc_step": 0, "speed/wps": 12894.728373052598, "speed/FLOPS": 202529428312416.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10571408271789551, "optim/lr": 0.0020202, "optim/total_tokens": 1765801984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.351062536239624, "created_at": "2025-01-15T15:03:51.092096+00:00"} {"global_step": 3369, "acc_step": 0, "speed/wps": 12895.58874432601, "speed/FLOPS": 202542941625539.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10527534037828445, "optim/lr": 0.0020208, "optim/total_tokens": 1766326272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.398883581161499, "created_at": "2025-01-15T15:04:01.260873+00:00"} {"global_step": 3370, "acc_step": 0, "speed/wps": 12898.283758205795, "speed/FLOPS": 202585270521858.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11226128041744232, "optim/lr": 0.0020214, "optim/total_tokens": 1766850560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.2754709720611572, "created_at": "2025-01-15T15:04:11.432238+00:00"} {"global_step": 3371, "acc_step": 0, "speed/wps": 12893.670575963808, "speed/FLOPS": 202512814155571.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11280663311481476, "optim/lr": 0.0020220000000000004, "optim/total_tokens": 1767374848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 3.1961493492126465, "created_at": "2025-01-15T15:04:21.604344+00:00"} {"global_step": 3372, "acc_step": 0, "speed/wps": 12888.488184962625, "speed/FLOPS": 202431417583546.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12945622205734253, "optim/lr": 0.0020226000000000003, "optim/total_tokens": 1767899136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.270662307739258, "created_at": "2025-01-15T15:04:31.777634+00:00"} {"global_step": 3373, "acc_step": 0, "speed/wps": 12889.638535928261, "speed/FLOPS": 202449485426209.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11426789313554764, "optim/lr": 0.0020232, "optim/total_tokens": 1768423424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.3742928504943848, "created_at": "2025-01-15T15:04:41.951433+00:00"} {"global_step": 3374, "acc_step": 0, "speed/wps": 12890.825585068167, "speed/FLOPS": 202468129664128.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11087773740291595, "optim/lr": 0.0020238, "optim/total_tokens": 1768947712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 3.2780601978302, "created_at": "2025-01-15T15:04:52.126012+00:00"} {"global_step": 3375, "acc_step": 0, "speed/wps": 12894.973419469681, "speed/FLOPS": 202533277103125.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11432282626628876, "optim/lr": 0.0020244, "optim/total_tokens": 1769472000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 3.292893409729004, "created_at": "2025-01-15T15:05:02.298213+00:00"} {"global_step": 3376, "acc_step": 0, "speed/wps": 12894.231001461303, "speed/FLOPS": 202521616408115.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12495261430740356, "optim/lr": 0.0020250000000000003, "optim/total_tokens": 1769996288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.360142707824707, "created_at": "2025-01-15T15:05:12.471006+00:00"} {"global_step": 3377, "acc_step": 0, "speed/wps": 12892.919531081006, "speed/FLOPS": 202501017963643.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1250566840171814, "optim/lr": 0.0020256000000000002, "optim/total_tokens": 1770520576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 3.271009683609009, "created_at": "2025-01-15T15:05:22.641741+00:00"} {"global_step": 3378, "acc_step": 0, "speed/wps": 12896.315362841136, "speed/FLOPS": 202554354167798.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0965256467461586, "optim/lr": 0.0020262, "optim/total_tokens": 1771044864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.287199020385742, "created_at": "2025-01-15T15:05:32.813229+00:00"} {"global_step": 3379, "acc_step": 0, "speed/wps": 12895.009261337107, "speed/FLOPS": 202533840048905.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10315828025341034, "optim/lr": 0.0020268, "optim/total_tokens": 1771569152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.2392992973327637, "created_at": "2025-01-15T15:05:42.981499+00:00"} {"global_step": 3380, "acc_step": 0, "speed/wps": 12891.472983916581, "speed/FLOPS": 202478297952661.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.14809222519397736, "optim/lr": 0.0020274, "optim/total_tokens": 1772093440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.308645725250244, "created_at": "2025-01-15T15:05:53.157776+00:00"} {"global_step": 3381, "acc_step": 0, "speed/wps": 12896.541271531698, "speed/FLOPS": 202557902374214.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13611453771591187, "optim/lr": 0.0020280000000000003, "optim/total_tokens": 1772617728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354159, "loss/out": 3.235335111618042, "created_at": "2025-01-15T15:06:03.327423+00:00"} {"global_step": 3382, "acc_step": 0, "speed/wps": 12893.203210804333, "speed/FLOPS": 202505473543514.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10353387892246246, "optim/lr": 0.0020286, "optim/total_tokens": 1773142016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.3422391414642334, "created_at": "2025-01-15T15:06:13.496981+00:00"} {"global_step": 3383, "acc_step": 0, "speed/wps": 12894.152705150107, "speed/FLOPS": 202520386656957.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14267180860042572, "optim/lr": 0.0020292, "optim/total_tokens": 1773666304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.3064169883728027, "created_at": "2025-01-15T15:06:23.668665+00:00"} {"global_step": 3384, "acc_step": 0, "speed/wps": 12893.917374209113, "speed/FLOPS": 202516690461150.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13604633510112762, "optim/lr": 0.0020298, "optim/total_tokens": 1774190592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 3.330078125, "created_at": "2025-01-15T15:06:33.838034+00:00"} {"global_step": 3385, "acc_step": 0, "speed/wps": 12896.582067894946, "speed/FLOPS": 202558543137159.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13937515020370483, "optim/lr": 0.0020304, "optim/total_tokens": 1774714880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.351797103881836, "created_at": "2025-01-15T15:06:44.005036+00:00"} {"global_step": 3386, "acc_step": 0, "speed/wps": 12892.399300950276, "speed/FLOPS": 202492847034569.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14788037538528442, "optim/lr": 0.0020310000000000003, "optim/total_tokens": 1775239168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.226595878601074, "created_at": "2025-01-15T15:06:54.175143+00:00"} {"global_step": 3387, "acc_step": 0, "speed/wps": 12893.981320623332, "speed/FLOPS": 202517694827457.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15260449051856995, "optim/lr": 0.0020316, "optim/total_tokens": 1775763456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415290, "loss/out": 3.280721664428711, "created_at": "2025-01-15T15:07:04.344470+00:00"} {"global_step": 3388, "acc_step": 0, "speed/wps": 12892.814872013658, "speed/FLOPS": 202499374149173.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13564662635326385, "optim/lr": 0.0020322, "optim/total_tokens": 1776287744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 3.362708806991577, "created_at": "2025-01-15T15:07:14.515510+00:00"} {"global_step": 3389, "acc_step": 0, "speed/wps": 12890.784749549812, "speed/FLOPS": 202467488286198.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13267453014850616, "optim/lr": 0.0020328, "optim/total_tokens": 1776812032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.323148727416992, "created_at": "2025-01-15T15:07:24.687755+00:00"} {"global_step": 3390, "acc_step": 0, "speed/wps": 12894.668079318859, "speed/FLOPS": 202528481316475.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15291449427604675, "optim/lr": 0.0020334, "optim/total_tokens": 1777336320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433452, "loss/out": 3.347693920135498, "created_at": "2025-01-15T15:07:34.857805+00:00"} {"global_step": 3391, "acc_step": 0, "speed/wps": 12891.810184885278, "speed/FLOPS": 202483594157238.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.13542647659778595, "optim/lr": 0.0020340000000000002, "optim/total_tokens": 1777860608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 3.3235790729522705, "created_at": "2025-01-15T15:07:45.028348+00:00"} {"global_step": 3392, "acc_step": 0, "speed/wps": 12894.294902550278, "speed/FLOPS": 202522620062526.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11099980026483536, "optim/lr": 0.0020346, "optim/total_tokens": 1778384896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.2558462619781494, "created_at": "2025-01-15T15:07:55.197119+00:00"} {"global_step": 3393, "acc_step": 0, "speed/wps": 12889.962026533525, "speed/FLOPS": 202454566290688.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11622757464647293, "optim/lr": 0.0020352, "optim/total_tokens": 1778909184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.384855270385742, "created_at": "2025-01-15T15:08:05.369427+00:00"} {"global_step": 3394, "acc_step": 0, "speed/wps": 12886.826374355025, "speed/FLOPS": 202405316564387.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12229756265878677, "optim/lr": 0.0020358, "optim/total_tokens": 1779433472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.37165904045105, "created_at": "2025-01-15T15:08:15.544061+00:00"} {"global_step": 3395, "acc_step": 0, "speed/wps": 12894.820013897053, "speed/FLOPS": 202530867657806.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11368881911039352, "optim/lr": 0.0020364, "optim/total_tokens": 1779957760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340384, "loss/out": 3.35815691947937, "created_at": "2025-01-15T15:08:25.713251+00:00"} {"global_step": 3396, "acc_step": 0, "speed/wps": 12892.410717892488, "speed/FLOPS": 202493026353335.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09956088662147522, "optim/lr": 0.002037, "optim/total_tokens": 1780482048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.3450162410736084, "created_at": "2025-01-15T15:08:35.885371+00:00"} {"global_step": 3397, "acc_step": 0, "speed/wps": 12888.674350039888, "speed/FLOPS": 202434341561902.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1104278489947319, "optim/lr": 0.0020376, "optim/total_tokens": 1781006336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.294846296310425, "created_at": "2025-01-15T15:08:46.061848+00:00"} {"global_step": 3398, "acc_step": 0, "speed/wps": 12895.513207083552, "speed/FLOPS": 202541755209347.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11667019873857498, "optim/lr": 0.0020382, "optim/total_tokens": 1781530624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388798, "loss/out": 3.3292434215545654, "created_at": "2025-01-15T15:08:56.229888+00:00"} {"global_step": 3399, "acc_step": 0, "speed/wps": 12894.45319842687, "speed/FLOPS": 202525106316789.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11135820299386978, "optim/lr": 0.0020388, "optim/total_tokens": 1782054912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 3.301353693008423, "created_at": "2025-01-15T15:09:06.398459+00:00"} {"global_step": 3400, "acc_step": 0, "speed/wps": 12893.991051090037, "speed/FLOPS": 202517847657807.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1477312296628952, "optim/lr": 0.0020394, "optim/total_tokens": 1782579200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.2367665767669678, "created_at": "2025-01-15T15:09:16.571879+00:00"} {"global_step": 3401, "acc_step": 0, "speed/wps": 12892.811347620112, "speed/FLOPS": 202499318793730.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16775742173194885, "optim/lr": 0.00204, "optim/total_tokens": 1783103488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.286925792694092, "created_at": "2025-01-15T15:09:26.742654+00:00"} {"global_step": 3402, "acc_step": 0, "speed/wps": 12894.906394908854, "speed/FLOPS": 202532224390296.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17108866572380066, "optim/lr": 0.0020406, "optim/total_tokens": 1783627776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 3.250917911529541, "created_at": "2025-01-15T15:09:36.912726+00:00"} {"global_step": 3403, "acc_step": 0, "speed/wps": 12886.182228990125, "speed/FLOPS": 202395199376282.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1369982659816742, "optim/lr": 0.0020412, "optim/total_tokens": 1784152064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.347996234893799, "created_at": "2025-01-15T15:09:47.088664+00:00"} {"global_step": 3404, "acc_step": 0, "speed/wps": 12880.227723404849, "speed/FLOPS": 202301675683717.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12148943543434143, "optim/lr": 0.0020418, "optim/total_tokens": 1784676352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451791, "loss/out": 3.40694522857666, "created_at": "2025-01-15T15:09:57.271944+00:00"} {"global_step": 3405, "acc_step": 0, "speed/wps": 12893.004718492548, "speed/FLOPS": 202502355949001.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15992514789104462, "optim/lr": 0.0020423999999999998, "optim/total_tokens": 1785200640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.2825660705566406, "created_at": "2025-01-15T15:10:07.441783+00:00"} {"global_step": 3406, "acc_step": 0, "speed/wps": 12894.92210389393, "speed/FLOPS": 202532471121493.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12601040303707123, "optim/lr": 0.002043, "optim/total_tokens": 1785724928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 3.2774829864501953, "created_at": "2025-01-15T15:10:17.610596+00:00"} {"global_step": 3407, "acc_step": 0, "speed/wps": 12895.436320182569, "speed/FLOPS": 202540547594914.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11284641921520233, "optim/lr": 0.0020436, "optim/total_tokens": 1786249216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 3.317946434020996, "created_at": "2025-01-15T15:10:27.780383+00:00"} {"global_step": 3408, "acc_step": 0, "speed/wps": 12895.34055520269, "speed/FLOPS": 202539043474310.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12174267321825027, "optim/lr": 0.0020442, "optim/total_tokens": 1786773504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.314605712890625, "created_at": "2025-01-15T15:10:37.949226+00:00"} {"global_step": 3409, "acc_step": 0, "speed/wps": 12885.251065799275, "speed/FLOPS": 202380574178818.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.10679508000612259, "optim/lr": 0.0020448, "optim/total_tokens": 1787297792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.3228161334991455, "created_at": "2025-01-15T15:10:48.124956+00:00"} {"global_step": 3410, "acc_step": 0, "speed/wps": 12891.954446866604, "speed/FLOPS": 202485859989891.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.1323627233505249, "optim/lr": 0.0020453999999999997, "optim/total_tokens": 1787822080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 3.2951314449310303, "created_at": "2025-01-15T15:10:58.297157+00:00"} {"global_step": 3411, "acc_step": 0, "speed/wps": 12892.969320059528, "speed/FLOPS": 202501799967968.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14047697186470032, "optim/lr": 0.002046, "optim/total_tokens": 1788346368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.3412551879882812, "created_at": "2025-01-15T15:11:08.470555+00:00"} {"global_step": 3412, "acc_step": 0, "speed/wps": 12893.67772609909, "speed/FLOPS": 202512926458271.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1263304352760315, "optim/lr": 0.0020466, "optim/total_tokens": 1788870656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 3.4228734970092773, "created_at": "2025-01-15T15:11:18.640804+00:00"} {"global_step": 3413, "acc_step": 0, "speed/wps": 12888.985158461606, "speed/FLOPS": 202439223235261.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11798304319381714, "optim/lr": 0.0020472, "optim/total_tokens": 1789394944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.3493638038635254, "created_at": "2025-01-15T15:11:28.813594+00:00"} {"global_step": 3414, "acc_step": 0, "speed/wps": 12891.703837435682, "speed/FLOPS": 202481923824405.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10484645515680313, "optim/lr": 0.0020478, "optim/total_tokens": 1789919232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366127, "loss/out": 3.419412136077881, "created_at": "2025-01-15T15:11:38.985142+00:00"} {"global_step": 3415, "acc_step": 0, "speed/wps": 12901.956352864558, "speed/FLOPS": 202642953667648.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11511854082345963, "optim/lr": 0.0020484, "optim/total_tokens": 1790443520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.3643834590911865, "created_at": "2025-01-15T15:11:49.152472+00:00"} {"global_step": 3416, "acc_step": 0, "speed/wps": 12893.340006924047, "speed/FLOPS": 202507622114552.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15503109991550446, "optim/lr": 0.002049, "optim/total_tokens": 1790967808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.392167091369629, "created_at": "2025-01-15T15:11:59.323352+00:00"} {"global_step": 3417, "acc_step": 0, "speed/wps": 12897.45590427052, "speed/FLOPS": 202572267938214.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1871793270111084, "optim/lr": 0.0020496, "optim/total_tokens": 1791492096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 3.3467273712158203, "created_at": "2025-01-15T15:12:09.491840+00:00"} {"global_step": 3418, "acc_step": 0, "speed/wps": 12891.665854098424, "speed/FLOPS": 202481327243899.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12283892184495926, "optim/lr": 0.0020502, "optim/total_tokens": 1792016384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457124, "loss/out": 3.3826394081115723, "created_at": "2025-01-15T15:12:19.663559+00:00"} {"global_step": 3419, "acc_step": 0, "speed/wps": 12891.967074667351, "speed/FLOPS": 202486058326854.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13233715295791626, "optim/lr": 0.0020508, "optim/total_tokens": 1792540672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.3341588973999023, "created_at": "2025-01-15T15:12:29.834519+00:00"} {"global_step": 3420, "acc_step": 0, "speed/wps": 12895.25826597107, "speed/FLOPS": 202537751008850.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11059506237506866, "optim/lr": 0.0020514, "optim/total_tokens": 1793064960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.275517225265503, "created_at": "2025-01-15T15:12:40.003891+00:00"} {"global_step": 3421, "acc_step": 0, "speed/wps": 12885.998592342965, "speed/FLOPS": 202392315110395.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12148351222276688, "optim/lr": 0.002052, "optim/total_tokens": 1793589248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.326514959335327, "created_at": "2025-01-15T15:12:50.181795+00:00"} {"global_step": 3422, "acc_step": 0, "speed/wps": 12895.461029382792, "speed/FLOPS": 202540935686857.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13727225363254547, "optim/lr": 0.0020526000000000003, "optim/total_tokens": 1794113536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 3.335613250732422, "created_at": "2025-01-15T15:13:00.349394+00:00"} {"global_step": 3423, "acc_step": 0, "speed/wps": 12896.873037675703, "speed/FLOPS": 202563113217400.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13386394083499908, "optim/lr": 0.0020532000000000003, "optim/total_tokens": 1794637824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.282586097717285, "created_at": "2025-01-15T15:13:10.518995+00:00"} {"global_step": 3424, "acc_step": 0, "speed/wps": 12899.115269907425, "speed/FLOPS": 202598330555749.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1270017772912979, "optim/lr": 0.0020538, "optim/total_tokens": 1795162112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.3315138816833496, "created_at": "2025-01-15T15:13:20.683957+00:00"} {"global_step": 3425, "acc_step": 0, "speed/wps": 12895.930845956176, "speed/FLOPS": 202548314801738.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11115377396345139, "optim/lr": 0.0020544, "optim/total_tokens": 1795686400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.206050157546997, "created_at": "2025-01-15T15:13:30.852179+00:00"} {"global_step": 3426, "acc_step": 0, "speed/wps": 12886.163493663955, "speed/FLOPS": 202394905112240.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11692028492689133, "optim/lr": 0.0020550000000000004, "optim/total_tokens": 1796210688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.2201955318450928, "created_at": "2025-01-15T15:13:41.027195+00:00"} {"global_step": 3427, "acc_step": 0, "speed/wps": 12892.121979513606, "speed/FLOPS": 202488491320326.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11723556369543076, "optim/lr": 0.0020556000000000003, "optim/total_tokens": 1796734976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 3.3028974533081055, "created_at": "2025-01-15T15:13:51.202705+00:00"} {"global_step": 3428, "acc_step": 0, "speed/wps": 12893.60221833033, "speed/FLOPS": 202511740505004.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1454576551914215, "optim/lr": 0.0020562, "optim/total_tokens": 1797259264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.349639415740967, "created_at": "2025-01-15T15:14:01.372126+00:00"} {"global_step": 3429, "acc_step": 0, "speed/wps": 12893.866853777232, "speed/FLOPS": 202515896968345.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13572700321674347, "optim/lr": 0.0020568, "optim/total_tokens": 1797783552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 3.301058053970337, "created_at": "2025-01-15T15:14:11.541107+00:00"} {"global_step": 3430, "acc_step": 0, "speed/wps": 12892.729101294299, "speed/FLOPS": 202498027002165.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13065747916698456, "optim/lr": 0.0020574, "optim/total_tokens": 1798307840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 3.353041172027588, "created_at": "2025-01-15T15:14:21.714893+00:00"} {"global_step": 3431, "acc_step": 0, "speed/wps": 12894.37365305806, "speed/FLOPS": 202523856947465.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13278992474079132, "optim/lr": 0.0020580000000000004, "optim/total_tokens": 1798832128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.297799587249756, "created_at": "2025-01-15T15:14:31.886591+00:00"} {"global_step": 3432, "acc_step": 0, "speed/wps": 12897.647459490374, "speed/FLOPS": 202575276576166.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1371508091688156, "optim/lr": 0.0020586000000000003, "optim/total_tokens": 1799356416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412041, "loss/out": 3.2620253562927246, "created_at": "2025-01-15T15:14:42.055021+00:00"} {"global_step": 3433, "acc_step": 0, "speed/wps": 12899.957095348964, "speed/FLOPS": 202611552581098.3, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12150400876998901, "optim/lr": 0.0020592, "optim/total_tokens": 1799880704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 3.344481945037842, "created_at": "2025-01-15T15:14:52.220673+00:00"} {"global_step": 3434, "acc_step": 0, "speed/wps": 12897.848694043303, "speed/FLOPS": 202578437241345.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11526547372341156, "optim/lr": 0.0020598, "optim/total_tokens": 1800404992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.3385214805603027, "created_at": "2025-01-15T15:15:02.386943+00:00"} {"global_step": 3435, "acc_step": 0, "speed/wps": 12895.409601221385, "speed/FLOPS": 202540127936912.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10315004736185074, "optim/lr": 0.0020604, "optim/total_tokens": 1800929280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.221264123916626, "created_at": "2025-01-15T15:15:12.556966+00:00"} {"global_step": 3436, "acc_step": 0, "speed/wps": 12883.657244457345, "speed/FLOPS": 202355541024503.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11635476350784302, "optim/lr": 0.0020610000000000003, "optim/total_tokens": 1801453568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 3.308311700820923, "created_at": "2025-01-15T15:15:22.738008+00:00"} {"global_step": 3437, "acc_step": 0, "speed/wps": 12883.509751405183, "speed/FLOPS": 202353224443442.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11359988898038864, "optim/lr": 0.0020616000000000002, "optim/total_tokens": 1801977856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.319063663482666, "created_at": "2025-01-15T15:15:32.915379+00:00"} {"global_step": 3438, "acc_step": 0, "speed/wps": 12891.866013983561, "speed/FLOPS": 202484471029943.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11749912053346634, "optim/lr": 0.0020622, "optim/total_tokens": 1802502144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.221792221069336, "created_at": "2025-01-15T15:15:43.088289+00:00"} {"global_step": 3439, "acc_step": 0, "speed/wps": 12892.524073506294, "speed/FLOPS": 202494806759016.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13525736331939697, "optim/lr": 0.0020628, "optim/total_tokens": 1803026432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.2819418907165527, "created_at": "2025-01-15T15:15:53.263126+00:00"} {"global_step": 3440, "acc_step": 0, "speed/wps": 12893.156420164894, "speed/FLOPS": 202504738632225.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1258905529975891, "optim/lr": 0.0020634, "optim/total_tokens": 1803550720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 3.3458824157714844, "created_at": "2025-01-15T15:16:03.432606+00:00"} {"global_step": 3441, "acc_step": 0, "speed/wps": 12893.902208705736, "speed/FLOPS": 202516452266080.28, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11550237983465195, "optim/lr": 0.002064, "optim/total_tokens": 1804075008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.234562635421753, "created_at": "2025-01-15T15:16:13.603873+00:00"} {"global_step": 3442, "acc_step": 0, "speed/wps": 12893.210867143143, "speed/FLOPS": 202505593796836.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10468433797359467, "optim/lr": 0.0020646, "optim/total_tokens": 1804599296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.2919793128967285, "created_at": "2025-01-15T15:16:23.777442+00:00"} {"global_step": 3443, "acc_step": 0, "speed/wps": 12891.855572266306, "speed/FLOPS": 202484307028425.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11447356641292572, "optim/lr": 0.0020652, "optim/total_tokens": 1805123584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459256, "loss/out": 3.2312049865722656, "created_at": "2025-01-15T15:16:33.948112+00:00"} {"global_step": 3444, "acc_step": 0, "speed/wps": 12894.064694554492, "speed/FLOPS": 202519004329613.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11285992711782455, "optim/lr": 0.0020658, "optim/total_tokens": 1805647872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 3.2483811378479004, "created_at": "2025-01-15T15:16:44.118136+00:00"} {"global_step": 3445, "acc_step": 0, "speed/wps": 12889.855735545583, "speed/FLOPS": 202452896844663.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11646317690610886, "optim/lr": 0.0020664, "optim/total_tokens": 1806172160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.2766873836517334, "created_at": "2025-01-15T15:16:54.290309+00:00"} {"global_step": 3446, "acc_step": 0, "speed/wps": 12893.187475296683, "speed/FLOPS": 202505226395744.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12249485403299332, "optim/lr": 0.002067, "optim/total_tokens": 1806696448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.322402238845825, "created_at": "2025-01-15T15:17:04.462665+00:00"} {"global_step": 3447, "acc_step": 0, "speed/wps": 12889.546039663748, "speed/FLOPS": 202448032645272.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1466313600540161, "optim/lr": 0.0020676, "optim/total_tokens": 1807220736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.2219536304473877, "created_at": "2025-01-15T15:17:14.636097+00:00"} {"global_step": 3448, "acc_step": 0, "speed/wps": 12894.033661185797, "speed/FLOPS": 202518516907914.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.12956055998802185, "optim/lr": 0.0020682, "optim/total_tokens": 1807745024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 3.3392505645751953, "created_at": "2025-01-15T15:17:24.806469+00:00"} {"global_step": 3449, "acc_step": 0, "speed/wps": 12892.165200813522, "speed/FLOPS": 202489170170234.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.13002265989780426, "optim/lr": 0.0020688, "optim/total_tokens": 1808269312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 3.3756284713745117, "created_at": "2025-01-15T15:17:34.986322+00:00"} {"global_step": 3450, "acc_step": 0, "speed/wps": 12896.857591032316, "speed/FLOPS": 202562870606640.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12213029712438583, "optim/lr": 0.0020694, "optim/total_tokens": 1808793600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.2181711196899414, "created_at": "2025-01-15T15:17:45.155904+00:00"} {"global_step": 3451, "acc_step": 0, "speed/wps": 12895.803497636514, "speed/FLOPS": 202546314621382.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10571327060461044, "optim/lr": 0.00207, "optim/total_tokens": 1809317888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.1186294555664062, "created_at": "2025-01-15T15:17:55.325717+00:00"} {"global_step": 3452, "acc_step": 0, "speed/wps": 12890.86828405385, "speed/FLOPS": 202468800310373.03, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12960843741893768, "optim/lr": 0.0020706, "optim/total_tokens": 1809842176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.2028510570526123, "created_at": "2025-01-15T15:18:05.498257+00:00"} {"global_step": 3453, "acc_step": 0, "speed/wps": 12894.980913683332, "speed/FLOPS": 202533394810049.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13968095183372498, "optim/lr": 0.0020712, "optim/total_tokens": 1810366464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.299248695373535, "created_at": "2025-01-15T15:18:15.667356+00:00"} {"global_step": 3454, "acc_step": 0, "speed/wps": 12891.182411249936, "speed/FLOPS": 202473734109643.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1053505390882492, "optim/lr": 0.0020718, "optim/total_tokens": 1810890752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 3.2501566410064697, "created_at": "2025-01-15T15:18:25.839802+00:00"} {"global_step": 3455, "acc_step": 0, "speed/wps": 12893.443301948235, "speed/FLOPS": 202509244504849.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12220712006092072, "optim/lr": 0.0020724, "optim/total_tokens": 1811415040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.3528246879577637, "created_at": "2025-01-15T15:18:36.011863+00:00"} {"global_step": 3456, "acc_step": 0, "speed/wps": 12893.885116548126, "speed/FLOPS": 202516183810257.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12316413223743439, "optim/lr": 0.0020729999999999998, "optim/total_tokens": 1811939328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 3.337718963623047, "created_at": "2025-01-15T15:18:46.186013+00:00"} {"global_step": 3457, "acc_step": 0, "speed/wps": 12895.580742497626, "speed/FLOPS": 202542815945829.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12153785675764084, "optim/lr": 0.0020736, "optim/total_tokens": 1812463616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.2453513145446777, "created_at": "2025-01-15T15:18:56.353990+00:00"} {"global_step": 3458, "acc_step": 0, "speed/wps": 12896.139159722097, "speed/FLOPS": 202551586655685.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13236334919929504, "optim/lr": 0.0020742, "optim/total_tokens": 1812987904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326440, "loss/out": 3.301497459411621, "created_at": "2025-01-15T15:19:06.522998+00:00"} {"global_step": 3459, "acc_step": 0, "speed/wps": 12896.990426283464, "speed/FLOPS": 202564956966794.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10275942087173462, "optim/lr": 0.0020748, "optim/total_tokens": 1813512192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.2751898765563965, "created_at": "2025-01-15T15:19:16.691300+00:00"} {"global_step": 3460, "acc_step": 0, "speed/wps": 12893.608634837981, "speed/FLOPS": 202511841285074.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12434747070074081, "optim/lr": 0.0020754, "optim/total_tokens": 1814036480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 3.181112766265869, "created_at": "2025-01-15T15:19:26.864748+00:00"} {"global_step": 3461, "acc_step": 0, "speed/wps": 12894.927695857617, "speed/FLOPS": 202532558950967.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13486500084400177, "optim/lr": 0.0020759999999999997, "optim/total_tokens": 1814560768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.2954440116882324, "created_at": "2025-01-15T15:19:37.033748+00:00"} {"global_step": 3462, "acc_step": 0, "speed/wps": 12895.006599419416, "speed/FLOPS": 202533798239830.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.12712976336479187, "optim/lr": 0.0020766, "optim/total_tokens": 1815085056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 3.2679271697998047, "created_at": "2025-01-15T15:19:47.205551+00:00"} {"global_step": 3463, "acc_step": 0, "speed/wps": 12890.064031357777, "speed/FLOPS": 202456168416622.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11961272358894348, "optim/lr": 0.0020772, "optim/total_tokens": 1815609344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 3.274529218673706, "created_at": "2025-01-15T15:19:57.378117+00:00"} {"global_step": 3464, "acc_step": 0, "speed/wps": 12893.346765717206, "speed/FLOPS": 202507728270686.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1204797700047493, "optim/lr": 0.0020778, "optim/total_tokens": 1816133632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.2418460845947266, "created_at": "2025-01-15T15:20:07.551087+00:00"} {"global_step": 3465, "acc_step": 0, "speed/wps": 12894.580658305158, "speed/FLOPS": 202527108249326.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11958855390548706, "optim/lr": 0.0020783999999999998, "optim/total_tokens": 1816657920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 3.3141441345214844, "created_at": "2025-01-15T15:20:17.719617+00:00"} {"global_step": 3466, "acc_step": 0, "speed/wps": 12890.568626106275, "speed/FLOPS": 202464093770533.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11135963350534439, "optim/lr": 0.0020789999999999997, "optim/total_tokens": 1817182208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 3.2938504219055176, "created_at": "2025-01-15T15:20:27.891189+00:00"} {"global_step": 3467, "acc_step": 0, "speed/wps": 12893.946305476711, "speed/FLOPS": 202517144866463.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1230311319231987, "optim/lr": 0.0020796, "optim/total_tokens": 1817706496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.379880666732788, "created_at": "2025-01-15T15:20:38.062153+00:00"} {"global_step": 3468, "acc_step": 0, "speed/wps": 12891.997629252883, "speed/FLOPS": 202486538228606.78, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11112404614686966, "optim/lr": 0.0020802, "optim/total_tokens": 1818230784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.345843553543091, "created_at": "2025-01-15T15:20:48.235716+00:00"} {"global_step": 3469, "acc_step": 0, "speed/wps": 12891.478160771681, "speed/FLOPS": 202478379262285.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12433361262083054, "optim/lr": 0.0020808, "optim/total_tokens": 1818755072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 3.339611530303955, "created_at": "2025-01-15T15:20:58.406568+00:00"} {"global_step": 3470, "acc_step": 0, "speed/wps": 12891.173431285284, "speed/FLOPS": 202473593066958.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13100853562355042, "optim/lr": 0.0020813999999999997, "optim/total_tokens": 1819279360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.3844411373138428, "created_at": "2025-01-15T15:21:08.577598+00:00"} {"global_step": 3471, "acc_step": 0, "speed/wps": 12891.974734261901, "speed/FLOPS": 202486178631312.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10141877830028534, "optim/lr": 0.002082, "optim/total_tokens": 1819803648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.243147850036621, "created_at": "2025-01-15T15:21:18.750516+00:00"} {"global_step": 3472, "acc_step": 0, "speed/wps": 12893.654901921665, "speed/FLOPS": 202512567973201.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09445732086896896, "optim/lr": 0.0020826, "optim/total_tokens": 1820327936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.3159332275390625, "created_at": "2025-01-15T15:21:28.921481+00:00"} {"global_step": 3473, "acc_step": 0, "speed/wps": 12892.708798532864, "speed/FLOPS": 202497708119397.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12126422673463821, "optim/lr": 0.0020832, "optim/total_tokens": 1820852224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370586, "loss/out": 3.3190367221832275, "created_at": "2025-01-15T15:21:39.096923+00:00"} {"global_step": 3474, "acc_step": 0, "speed/wps": 12891.95490507123, "speed/FLOPS": 202485867186624.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13316987454891205, "optim/lr": 0.0020838000000000002, "optim/total_tokens": 1821376512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 3.2911040782928467, "created_at": "2025-01-15T15:21:49.267799+00:00"} {"global_step": 3475, "acc_step": 0, "speed/wps": 12893.453251308783, "speed/FLOPS": 202509400773229.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12910045683383942, "optim/lr": 0.0020844, "optim/total_tokens": 1821900800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.258697271347046, "created_at": "2025-01-15T15:21:59.439541+00:00"} {"global_step": 3476, "acc_step": 0, "speed/wps": 12897.735640748382, "speed/FLOPS": 202576661583997.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11421992629766464, "optim/lr": 0.002085, "optim/total_tokens": 1822425088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 3.4077115058898926, "created_at": "2025-01-15T15:22:09.605930+00:00"} {"global_step": 3477, "acc_step": 0, "speed/wps": 12894.201158094133, "speed/FLOPS": 202521147677024.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10632500797510147, "optim/lr": 0.0020856, "optim/total_tokens": 1822949376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 3.2618560791015625, "created_at": "2025-01-15T15:22:19.778143+00:00"} {"global_step": 3478, "acc_step": 0, "speed/wps": 12894.214375066715, "speed/FLOPS": 202521355267740.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12954388558864594, "optim/lr": 0.0020862000000000003, "optim/total_tokens": 1823473664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.3508167266845703, "created_at": "2025-01-15T15:22:29.947426+00:00"} {"global_step": 3479, "acc_step": 0, "speed/wps": 12889.001311017808, "speed/FLOPS": 202439476933352.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1402948498725891, "optim/lr": 0.0020868, "optim/total_tokens": 1823997952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.486593246459961, "created_at": "2025-01-15T15:22:40.122548+00:00"} {"global_step": 3480, "acc_step": 0, "speed/wps": 12890.818885593113, "speed/FLOPS": 202468024439666.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11237172782421112, "optim/lr": 0.0020874, "optim/total_tokens": 1824522240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.3686978816986084, "created_at": "2025-01-15T15:22:50.296647+00:00"} {"global_step": 3481, "acc_step": 0, "speed/wps": 12892.814218866853, "speed/FLOPS": 202499363890605.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1325623095035553, "optim/lr": 0.002088, "optim/total_tokens": 1825046528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 3.230151414871216, "created_at": "2025-01-15T15:23:00.469613+00:00"} {"global_step": 3482, "acc_step": 0, "speed/wps": 12897.07065279195, "speed/FLOPS": 202566217034352.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09986173361539841, "optim/lr": 0.0020886000000000004, "optim/total_tokens": 1825570816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 3.249889850616455, "created_at": "2025-01-15T15:23:10.637155+00:00"} {"global_step": 3483, "acc_step": 0, "speed/wps": 12891.016122754232, "speed/FLOPS": 202471122320314.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10492971539497375, "optim/lr": 0.0020892000000000003, "optim/total_tokens": 1826095104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.28287410736084, "created_at": "2025-01-15T15:23:20.808373+00:00"} {"global_step": 3484, "acc_step": 0, "speed/wps": 12891.255160642417, "speed/FLOPS": 202474876738820.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11168841272592545, "optim/lr": 0.0020898, "optim/total_tokens": 1826619392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 3.229292869567871, "created_at": "2025-01-15T15:23:30.982644+00:00"} {"global_step": 3485, "acc_step": 0, "speed/wps": 12896.3231505045, "speed/FLOPS": 202554476483752.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11872754991054535, "optim/lr": 0.0020904, "optim/total_tokens": 1827143680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 3.2017812728881836, "created_at": "2025-01-15T15:23:41.149983+00:00"} {"global_step": 3486, "acc_step": 0, "speed/wps": 12893.903745572776, "speed/FLOPS": 202516476404689.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11127348989248276, "optim/lr": 0.002091, "optim/total_tokens": 1827667968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 3.2889199256896973, "created_at": "2025-01-15T15:23:51.318945+00:00"} {"global_step": 3487, "acc_step": 0, "speed/wps": 12899.423262178621, "speed/FLOPS": 202603167997594.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.11827220022678375, "optim/lr": 0.0020916000000000003, "optim/total_tokens": 1828192256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.3689918518066406, "created_at": "2025-01-15T15:24:01.485404+00:00"} {"global_step": 3488, "acc_step": 0, "speed/wps": 12896.581122464875, "speed/FLOPS": 202558528287880.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13233473896980286, "optim/lr": 0.0020922000000000002, "optim/total_tokens": 1828716544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 3.337761640548706, "created_at": "2025-01-15T15:24:11.652344+00:00"} {"global_step": 3489, "acc_step": 0, "speed/wps": 12895.424257467106, "speed/FLOPS": 202540358133391.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.17031089961528778, "optim/lr": 0.0020928, "optim/total_tokens": 1829240832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 3.339163303375244, "created_at": "2025-01-15T15:24:21.826393+00:00"} {"global_step": 3490, "acc_step": 0, "speed/wps": 12897.851478882605, "speed/FLOPS": 202578480981073.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1628693789243698, "optim/lr": 0.0020934, "optim/total_tokens": 1829765120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.266383171081543, "created_at": "2025-01-15T15:24:31.994147+00:00"} {"global_step": 3491, "acc_step": 0, "speed/wps": 12890.891516350375, "speed/FLOPS": 202469165205514.5, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1506766676902771, "optim/lr": 0.002094, "optim/total_tokens": 1830289408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 3.375917911529541, "created_at": "2025-01-15T15:24:42.165557+00:00"} {"global_step": 3492, "acc_step": 0, "speed/wps": 12889.997785060448, "speed/FLOPS": 202455127927490.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.15183891355991364, "optim/lr": 0.0020946000000000003, "optim/total_tokens": 1830813696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.3722457885742188, "created_at": "2025-01-15T15:24:52.338525+00:00"} {"global_step": 3493, "acc_step": 0, "speed/wps": 12893.817446507897, "speed/FLOPS": 202515120959287.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1215137243270874, "optim/lr": 0.0020952, "optim/total_tokens": 1831337984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 3.2252962589263916, "created_at": "2025-01-15T15:25:02.509285+00:00"} {"global_step": 3494, "acc_step": 0, "speed/wps": 12896.853484145982, "speed/FLOPS": 202562806102346.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11075720191001892, "optim/lr": 0.0020958, "optim/total_tokens": 1831862272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2666332721710205, "created_at": "2025-01-15T15:25:12.676946+00:00"} {"global_step": 3495, "acc_step": 0, "speed/wps": 12894.687406474415, "speed/FLOPS": 202528784876012.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13649486005306244, "optim/lr": 0.0020964, "optim/total_tokens": 1832386560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 3.463839530944824, "created_at": "2025-01-15T15:25:22.846606+00:00"} {"global_step": 3496, "acc_step": 0, "speed/wps": 12894.888918471635, "speed/FLOPS": 202531949898835.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13023804128170013, "optim/lr": 0.002097, "optim/total_tokens": 1832910848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.339430570602417, "created_at": "2025-01-15T15:25:33.018589+00:00"} {"global_step": 3497, "acc_step": 0, "speed/wps": 12895.579271400606, "speed/FLOPS": 202542792840228.84, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.1289234310388565, "optim/lr": 0.0020976000000000002, "optim/total_tokens": 1833435136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.296959400177002, "created_at": "2025-01-15T15:25:43.186673+00:00"} {"global_step": 3498, "acc_step": 0, "speed/wps": 12901.350587535355, "speed/FLOPS": 202633439290744.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11497494578361511, "optim/lr": 0.0020982, "optim/total_tokens": 1833959424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 3.35086727142334, "created_at": "2025-01-15T15:25:53.350933+00:00"} {"global_step": 3499, "acc_step": 0, "speed/wps": 12893.411997530618, "speed/FLOPS": 202508752825954.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1204376071691513, "optim/lr": 0.0020988, "optim/total_tokens": 1834483712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.2776994705200195, "created_at": "2025-01-15T15:26:03.521765+00:00"} {"global_step": 3500, "acc_step": 0, "speed/wps": 12896.098778573916, "speed/FLOPS": 202550952414263.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10156525671482086, "optim/lr": 0.0020994, "optim/total_tokens": 1835008000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.1969552040100098, "created_at": "2025-01-15T15:26:13.701261+00:00"} {"global_step": 3501, "acc_step": 0, "speed/wps": 12896.346290201316, "speed/FLOPS": 202554839924488.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13814890384674072, "optim/lr": 0.0021, "optim/total_tokens": 1835532288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.266737937927246, "created_at": "2025-01-15T15:26:23.869201+00:00"} {"global_step": 3502, "acc_step": 0, "speed/wps": 12900.19699940732, "speed/FLOPS": 202615320603997.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13295133411884308, "optim/lr": 0.0021006, "optim/total_tokens": 1836056576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.1724977493286133, "created_at": "2025-01-15T15:26:34.033667+00:00"} {"global_step": 3503, "acc_step": 0, "speed/wps": 12901.514700500697, "speed/FLOPS": 202636016910380.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13423343002796173, "optim/lr": 0.0021012, "optim/total_tokens": 1836580864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.272753953933716, "created_at": "2025-01-15T15:26:44.196732+00:00"} {"global_step": 3504, "acc_step": 0, "speed/wps": 12891.59111029897, "speed/FLOPS": 202480153289820.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15676550567150116, "optim/lr": 0.0021018, "optim/total_tokens": 1837105152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 3.1935606002807617, "created_at": "2025-01-15T15:26:54.369656+00:00"} {"global_step": 3505, "acc_step": 0, "speed/wps": 12893.594089030654, "speed/FLOPS": 202511612823182.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15498767793178558, "optim/lr": 0.0021024, "optim/total_tokens": 1837629440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 3.3621630668640137, "created_at": "2025-01-15T15:27:04.538916+00:00"} {"global_step": 3506, "acc_step": 0, "speed/wps": 12898.140314382936, "speed/FLOPS": 202583017539508.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13043688237667084, "optim/lr": 0.002103, "optim/total_tokens": 1838153728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.339223861694336, "created_at": "2025-01-15T15:27:14.705236+00:00"} {"global_step": 3507, "acc_step": 0, "speed/wps": 12893.059411048254, "speed/FLOPS": 202503214970744.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12893642485141754, "optim/lr": 0.0021036, "optim/total_tokens": 1838678016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.303720235824585, "created_at": "2025-01-15T15:27:24.878412+00:00"} {"global_step": 3508, "acc_step": 0, "speed/wps": 12890.073572920746, "speed/FLOPS": 202456318279980.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1406862735748291, "optim/lr": 0.0021042, "optim/total_tokens": 1839202304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.2798171043395996, "created_at": "2025-01-15T15:27:35.051260+00:00"} {"global_step": 3509, "acc_step": 0, "speed/wps": 12894.629797331969, "speed/FLOPS": 202527880045266.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14940661191940308, "optim/lr": 0.0021048, "optim/total_tokens": 1839726592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.2535088062286377, "created_at": "2025-01-15T15:27:45.221984+00:00"} {"global_step": 3510, "acc_step": 0, "speed/wps": 12899.029350316836, "speed/FLOPS": 202596981070513.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13268381357192993, "optim/lr": 0.0021054, "optim/total_tokens": 1840250880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.302671432495117, "created_at": "2025-01-15T15:27:55.386907+00:00"} {"global_step": 3511, "acc_step": 0, "speed/wps": 12897.905855593208, "speed/FLOPS": 202579335042035.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1320638656616211, "optim/lr": 0.002106, "optim/total_tokens": 1840775168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.2797844409942627, "created_at": "2025-01-15T15:28:05.556239+00:00"} {"global_step": 3512, "acc_step": 0, "speed/wps": 12895.875846620138, "speed/FLOPS": 202547450961588.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13092607259750366, "optim/lr": 0.0021066, "optim/total_tokens": 1841299456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 3.281792402267456, "created_at": "2025-01-15T15:28:15.723693+00:00"} {"global_step": 3513, "acc_step": 0, "speed/wps": 12893.353720596, "speed/FLOPS": 202507837506615.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1338859498500824, "optim/lr": 0.0021072, "optim/total_tokens": 1841823744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.248721122741699, "created_at": "2025-01-15T15:28:25.893251+00:00"} {"global_step": 3514, "acc_step": 0, "speed/wps": 12900.26869176627, "speed/FLOPS": 202616446631010.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13745377957820892, "optim/lr": 0.0021078, "optim/total_tokens": 1842348032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.351586103439331, "created_at": "2025-01-15T15:28:36.057955+00:00"} {"global_step": 3515, "acc_step": 0, "speed/wps": 12895.779045903124, "speed/FLOPS": 202545930573309.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12089318782091141, "optim/lr": 0.0021084, "optim/total_tokens": 1842872320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330478, "loss/out": 3.2512803077697754, "created_at": "2025-01-15T15:28:46.225767+00:00"} {"global_step": 3516, "acc_step": 0, "speed/wps": 12897.132091790012, "speed/FLOPS": 202567182018243.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11449160426855087, "optim/lr": 0.0021089999999999998, "optim/total_tokens": 1843396608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 3.2646889686584473, "created_at": "2025-01-15T15:28:56.394231+00:00"} {"global_step": 3517, "acc_step": 0, "speed/wps": 12902.638402937557, "speed/FLOPS": 202653666201279.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11583919078111649, "optim/lr": 0.0021096, "optim/total_tokens": 1843920896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482329, "loss/out": 3.130976438522339, "created_at": "2025-01-15T15:29:06.558618+00:00"} {"global_step": 3518, "acc_step": 0, "speed/wps": 12895.752949425321, "speed/FLOPS": 202545520692265.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10213331878185272, "optim/lr": 0.0021102, "optim/total_tokens": 1844445184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.3502397537231445, "created_at": "2025-01-15T15:29:16.726933+00:00"} {"global_step": 3519, "acc_step": 0, "speed/wps": 12888.781105491937, "speed/FLOPS": 202436018302974.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10215147584676743, "optim/lr": 0.0021108, "optim/total_tokens": 1844969472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 3.292851209640503, "created_at": "2025-01-15T15:29:26.899921+00:00"} {"global_step": 3520, "acc_step": 0, "speed/wps": 12893.117784539225, "speed/FLOPS": 202504131806633.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1151319295167923, "optim/lr": 0.0021114, "optim/total_tokens": 1845493760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.387944459915161, "created_at": "2025-01-15T15:29:37.073035+00:00"} {"global_step": 3521, "acc_step": 0, "speed/wps": 12889.741846025705, "speed/FLOPS": 202451108053252.97, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12010566890239716, "optim/lr": 0.0021119999999999997, "optim/total_tokens": 1846018048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.1590142250061035, "created_at": "2025-01-15T15:29:47.246030+00:00"} {"global_step": 3522, "acc_step": 0, "speed/wps": 12891.642952019653, "speed/FLOPS": 202480967535281.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10662613064050674, "optim/lr": 0.0021126, "optim/total_tokens": 1846542336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.255699872970581, "created_at": "2025-01-15T15:29:57.416845+00:00"} {"global_step": 3523, "acc_step": 0, "speed/wps": 12895.149061411104, "speed/FLOPS": 202536035801173.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12577533721923828, "optim/lr": 0.0021132, "optim/total_tokens": 1847066624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 3.343595027923584, "created_at": "2025-01-15T15:30:07.587738+00:00"} {"global_step": 3524, "acc_step": 0, "speed/wps": 12890.114951733125, "speed/FLOPS": 202456968191088.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09752833843231201, "optim/lr": 0.0021138, "optim/total_tokens": 1847590912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.3530867099761963, "created_at": "2025-01-15T15:30:17.760751+00:00"} {"global_step": 3525, "acc_step": 0, "speed/wps": 12901.014844509924, "speed/FLOPS": 202628165985170.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0990620106458664, "optim/lr": 0.0021144, "optim/total_tokens": 1848115200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 3.191317081451416, "created_at": "2025-01-15T15:30:27.928044+00:00"} {"global_step": 3526, "acc_step": 0, "speed/wps": 12894.68323953525, "speed/FLOPS": 202528719428506.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11605895310640335, "optim/lr": 0.002115, "optim/total_tokens": 1848639488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 3.2911601066589355, "created_at": "2025-01-15T15:30:38.096430+00:00"} {"global_step": 3527, "acc_step": 0, "speed/wps": 12898.69733963886, "speed/FLOPS": 202591766386585.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14409595727920532, "optim/lr": 0.0021156, "optim/total_tokens": 1849163776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.3184444904327393, "created_at": "2025-01-15T15:30:48.262501+00:00"} {"global_step": 3528, "acc_step": 0, "speed/wps": 12886.095022108722, "speed/FLOPS": 202393829672375.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12819233536720276, "optim/lr": 0.0021162, "optim/total_tokens": 1849688064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.270310878753662, "created_at": "2025-01-15T15:30:58.437627+00:00"} {"global_step": 3529, "acc_step": 0, "speed/wps": 12886.86971001582, "speed/FLOPS": 202405997210489.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12582240998744965, "optim/lr": 0.0021168, "optim/total_tokens": 1850212352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.2112414836883545, "created_at": "2025-01-15T15:31:08.616731+00:00"} {"global_step": 3530, "acc_step": 0, "speed/wps": 12898.842816234694, "speed/FLOPS": 202594051296428.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11037951707839966, "optim/lr": 0.0021174, "optim/total_tokens": 1850736640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.2809176445007324, "created_at": "2025-01-15T15:31:18.786744+00:00"} {"global_step": 3531, "acc_step": 0, "speed/wps": 12899.184982954639, "speed/FLOPS": 202599425494951.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1262677013874054, "optim/lr": 0.002118, "optim/total_tokens": 1851260928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.2988343238830566, "created_at": "2025-01-15T15:31:28.951676+00:00"} {"global_step": 3532, "acc_step": 0, "speed/wps": 12896.688362299803, "speed/FLOPS": 202560212636851.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13508878648281097, "optim/lr": 0.0021186, "optim/total_tokens": 1851785216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.283998489379883, "created_at": "2025-01-15T15:31:39.119774+00:00"} {"global_step": 3533, "acc_step": 0, "speed/wps": 12896.638143963912, "speed/FLOPS": 202559423888878.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1317158043384552, "optim/lr": 0.0021192000000000003, "optim/total_tokens": 1852309504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 3.2691617012023926, "created_at": "2025-01-15T15:31:49.288158+00:00"} {"global_step": 3534, "acc_step": 0, "speed/wps": 12898.944456856083, "speed/FLOPS": 202595647702056.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.129235178232193, "optim/lr": 0.0021198000000000002, "optim/total_tokens": 1852833792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 3.2981982231140137, "created_at": "2025-01-15T15:31:59.453309+00:00"} {"global_step": 3535, "acc_step": 0, "speed/wps": 12893.857671615782, "speed/FLOPS": 202515752749882.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1649812012910843, "optim/lr": 0.0021204, "optim/total_tokens": 1853358080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 3.324111223220825, "created_at": "2025-01-15T15:32:09.627175+00:00"} {"global_step": 3536, "acc_step": 0, "speed/wps": 12890.60377521954, "speed/FLOPS": 202464645835656.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1453620195388794, "optim/lr": 0.002121, "optim/total_tokens": 1853882368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 3.2948875427246094, "created_at": "2025-01-15T15:32:19.805352+00:00"} {"global_step": 3537, "acc_step": 0, "speed/wps": 12899.30189226524, "speed/FLOPS": 202601261716325.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15230613946914673, "optim/lr": 0.0021216000000000004, "optim/total_tokens": 1854406656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.3098690509796143, "created_at": "2025-01-15T15:32:29.972276+00:00"} {"global_step": 3538, "acc_step": 0, "speed/wps": 12899.766901005187, "speed/FLOPS": 202608565317574.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1282772272825241, "optim/lr": 0.0021222000000000003, "optim/total_tokens": 1854930944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.3052310943603516, "created_at": "2025-01-15T15:32:40.140522+00:00"} {"global_step": 3539, "acc_step": 0, "speed/wps": 12896.44854026995, "speed/FLOPS": 202556445902322.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1316079944372177, "optim/lr": 0.0021228, "optim/total_tokens": 1855455232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 3.301703929901123, "created_at": "2025-01-15T15:32:50.310722+00:00"} {"global_step": 3540, "acc_step": 0, "speed/wps": 12893.373569445557, "speed/FLOPS": 202508149260072.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13246728479862213, "optim/lr": 0.0021234, "optim/total_tokens": 1855979520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 3.3045263290405273, "created_at": "2025-01-15T15:33:00.481414+00:00"} {"global_step": 3541, "acc_step": 0, "speed/wps": 12898.995812726556, "speed/FLOPS": 202596454316572.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1120879054069519, "optim/lr": 0.002124, "optim/total_tokens": 1856503808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 3.334296941757202, "created_at": "2025-01-15T15:33:10.647096+00:00"} {"global_step": 3542, "acc_step": 0, "speed/wps": 12897.469805173918, "speed/FLOPS": 202572486271004.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14610201120376587, "optim/lr": 0.0021246000000000004, "optim/total_tokens": 1857028096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 3.3029115200042725, "created_at": "2025-01-15T15:33:20.814597+00:00"} {"global_step": 3543, "acc_step": 0, "speed/wps": 12898.238792477161, "speed/FLOPS": 202584564273301.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12184049934148788, "optim/lr": 0.0021252000000000003, "optim/total_tokens": 1857552384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.3192567825317383, "created_at": "2025-01-15T15:33:30.982562+00:00"} {"global_step": 3544, "acc_step": 0, "speed/wps": 12898.953690006143, "speed/FLOPS": 202595792721365.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13526104390621185, "optim/lr": 0.0021258, "optim/total_tokens": 1858076672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.3214478492736816, "created_at": "2025-01-15T15:33:41.152217+00:00"} {"global_step": 3545, "acc_step": 0, "speed/wps": 12895.545422082609, "speed/FLOPS": 202542261190175.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10389799624681473, "optim/lr": 0.0021264, "optim/total_tokens": 1858600960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.2115039825439453, "created_at": "2025-01-15T15:33:51.321167+00:00"} {"global_step": 3546, "acc_step": 0, "speed/wps": 12888.548203739265, "speed/FLOPS": 202432360260908.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1009814515709877, "optim/lr": 0.002127, "optim/total_tokens": 1859125248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333550, "loss/out": 3.2152562141418457, "created_at": "2025-01-15T15:34:01.496381+00:00"} {"global_step": 3547, "acc_step": 0, "speed/wps": 12899.422465302781, "speed/FLOPS": 202603155481564.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.12276457250118256, "optim/lr": 0.0021276000000000003, "optim/total_tokens": 1859649536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 3.20798397064209, "created_at": "2025-01-15T15:34:11.662746+00:00"} {"global_step": 3548, "acc_step": 0, "speed/wps": 12898.272681942371, "speed/FLOPS": 202585096553921.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10645097494125366, "optim/lr": 0.0021282000000000002, "optim/total_tokens": 1860173824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.2200772762298584, "created_at": "2025-01-15T15:34:21.831329+00:00"} {"global_step": 3549, "acc_step": 0, "speed/wps": 12901.600989093067, "speed/FLOPS": 202637372191296.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10263028740882874, "optim/lr": 0.0021288, "optim/total_tokens": 1860698112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.2346506118774414, "created_at": "2025-01-15T15:34:31.994177+00:00"} {"global_step": 3550, "acc_step": 0, "speed/wps": 12898.061893754622, "speed/FLOPS": 202581785835779.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09017814695835114, "optim/lr": 0.0021294, "optim/total_tokens": 1861222400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.2908570766448975, "created_at": "2025-01-15T15:34:42.163616+00:00"} {"global_step": 3551, "acc_step": 0, "speed/wps": 12898.466216672721, "speed/FLOPS": 202588136282805.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09591455012559891, "optim/lr": 0.00213, "optim/total_tokens": 1861746688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.3158481121063232, "created_at": "2025-01-15T15:34:52.332588+00:00"} {"global_step": 3552, "acc_step": 0, "speed/wps": 12893.056667255601, "speed/FLOPS": 202503171875710.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09493765980005264, "optim/lr": 0.0021306000000000003, "optim/total_tokens": 1862270976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.403390645980835, "created_at": "2025-01-15T15:35:02.502432+00:00"} {"global_step": 3553, "acc_step": 0, "speed/wps": 12895.366567301891, "speed/FLOPS": 202539452030072.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1040160208940506, "optim/lr": 0.0021312, "optim/total_tokens": 1862795264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 3.3407347202301025, "created_at": "2025-01-15T15:35:12.671571+00:00"} {"global_step": 3554, "acc_step": 0, "speed/wps": 12898.863024828044, "speed/FLOPS": 202594368700156.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10850640386343002, "optim/lr": 0.0021318, "optim/total_tokens": 1863319552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.2925565242767334, "created_at": "2025-01-15T15:35:22.841256+00:00"} {"global_step": 3555, "acc_step": 0, "speed/wps": 12895.597373597724, "speed/FLOPS": 202543077160109.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16101141273975372, "optim/lr": 0.0021324, "optim/total_tokens": 1863843840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.226494073867798, "created_at": "2025-01-15T15:35:33.010011+00:00"} {"global_step": 3556, "acc_step": 0, "speed/wps": 12895.958677850344, "speed/FLOPS": 202548751939881.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14795753359794617, "optim/lr": 0.002133, "optim/total_tokens": 1864368128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 3.395479917526245, "created_at": "2025-01-15T15:35:43.178604+00:00"} {"global_step": 3557, "acc_step": 0, "speed/wps": 12900.693992947192, "speed/FLOPS": 202623126570481.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14181554317474365, "optim/lr": 0.0021336000000000003, "optim/total_tokens": 1864892416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.3014273643493652, "created_at": "2025-01-15T15:35:53.345426+00:00"} {"global_step": 3558, "acc_step": 0, "speed/wps": 12891.933793445007, "speed/FLOPS": 202485535599524.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1274946928024292, "optim/lr": 0.0021342, "optim/total_tokens": 1865416704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.363607883453369, "created_at": "2025-01-15T15:36:03.518092+00:00"} {"global_step": 3559, "acc_step": 0, "speed/wps": 12888.136147437734, "speed/FLOPS": 202425888350467.78, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10407884418964386, "optim/lr": 0.0021348, "optim/total_tokens": 1865940992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.373505115509033, "created_at": "2025-01-15T15:36:13.692828+00:00"} {"global_step": 3560, "acc_step": 0, "speed/wps": 12878.819019130808, "speed/FLOPS": 202279550047331.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11130280047655106, "optim/lr": 0.0021354, "optim/total_tokens": 1866465280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 3.2655656337738037, "created_at": "2025-01-15T15:36:23.876939+00:00"} {"global_step": 3561, "acc_step": 0, "speed/wps": 12882.007363295515, "speed/FLOPS": 202329627373682.3, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.1364370435476303, "optim/lr": 0.002136, "optim/total_tokens": 1866989568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.219283103942871, "created_at": "2025-01-15T15:36:34.058279+00:00"} {"global_step": 3562, "acc_step": 0, "speed/wps": 12888.760686504102, "speed/FLOPS": 202435697594711.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1625041663646698, "optim/lr": 0.0021366000000000002, "optim/total_tokens": 1867513856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 3.2875514030456543, "created_at": "2025-01-15T15:36:44.231498+00:00"} {"global_step": 3563, "acc_step": 0, "speed/wps": 12896.054639900103, "speed/FLOPS": 202550259155738.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12905491888523102, "optim/lr": 0.0021372, "optim/total_tokens": 1868038144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.238893985748291, "created_at": "2025-01-15T15:36:54.399112+00:00"} {"global_step": 3564, "acc_step": 0, "speed/wps": 12896.14488391094, "speed/FLOPS": 202551676561936.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13458430767059326, "optim/lr": 0.0021378, "optim/total_tokens": 1868562432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 3.2164320945739746, "created_at": "2025-01-15T15:37:04.568609+00:00"} {"global_step": 3565, "acc_step": 0, "speed/wps": 12900.676781080183, "speed/FLOPS": 202622856234458.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.15462946891784668, "optim/lr": 0.0021384, "optim/total_tokens": 1869086720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.342376708984375, "created_at": "2025-01-15T15:37:14.735361+00:00"} {"global_step": 3566, "acc_step": 0, "speed/wps": 12898.586114445638, "speed/FLOPS": 202590019442086.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1231844574213028, "optim/lr": 0.002139, "optim/total_tokens": 1869611008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.164127826690674, "created_at": "2025-01-15T15:37:24.904643+00:00"} {"global_step": 3567, "acc_step": 0, "speed/wps": 12898.59673375263, "speed/FLOPS": 202590186232895.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11965157091617584, "optim/lr": 0.0021395999999999997, "optim/total_tokens": 1870135296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.230130195617676, "created_at": "2025-01-15T15:37:35.076356+00:00"} {"global_step": 3568, "acc_step": 0, "speed/wps": 12897.740361654518, "speed/FLOPS": 202576735732315.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12007589638233185, "optim/lr": 0.0021402, "optim/total_tokens": 1870659584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.291865825653076, "created_at": "2025-01-15T15:37:45.242530+00:00"} {"global_step": 3569, "acc_step": 0, "speed/wps": 12889.775042738225, "speed/FLOPS": 202451629453240.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.12457878142595291, "optim/lr": 0.0021408, "optim/total_tokens": 1871183872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 3.2086262702941895, "created_at": "2025-01-15T15:37:55.415410+00:00"} {"global_step": 3570, "acc_step": 0, "speed/wps": 12884.183901832057, "speed/FLOPS": 202363812902276.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13717171549797058, "optim/lr": 0.0021414, "optim/total_tokens": 1871708160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.2756032943725586, "created_at": "2025-01-15T15:38:05.594994+00:00"} {"global_step": 3571, "acc_step": 0, "speed/wps": 12898.357145728603, "speed/FLOPS": 202586423173749.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12674029171466827, "optim/lr": 0.002142, "optim/total_tokens": 1872232448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 3.253749132156372, "created_at": "2025-01-15T15:38:15.764668+00:00"} {"global_step": 3572, "acc_step": 0, "speed/wps": 12895.4159084601, "speed/FLOPS": 202540227000763.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11005184799432755, "optim/lr": 0.0021425999999999997, "optim/total_tokens": 1872756736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 3.2534334659576416, "created_at": "2025-01-15T15:38:25.934309+00:00"} {"global_step": 3573, "acc_step": 0, "speed/wps": 12899.63684382332, "speed/FLOPS": 202606522590816.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11809934675693512, "optim/lr": 0.0021432, "optim/total_tokens": 1873281024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 3.2076761722564697, "created_at": "2025-01-15T15:38:36.101778+00:00"} {"global_step": 3574, "acc_step": 0, "speed/wps": 12896.13195840669, "speed/FLOPS": 202551473549130.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1382599025964737, "optim/lr": 0.0021438, "optim/total_tokens": 1873805312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.2661075592041016, "created_at": "2025-01-15T15:38:46.269941+00:00"} {"global_step": 3575, "acc_step": 0, "speed/wps": 12900.846740079814, "speed/FLOPS": 202625525674096.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12562903761863708, "optim/lr": 0.0021444, "optim/total_tokens": 1874329600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.256991386413574, "created_at": "2025-01-15T15:38:56.438507+00:00"} {"global_step": 3576, "acc_step": 0, "speed/wps": 12901.926681893518, "speed/FLOPS": 202642487644276.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11961524188518524, "optim/lr": 0.0021449999999999998, "optim/total_tokens": 1874853888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.3342766761779785, "created_at": "2025-01-15T15:39:06.606951+00:00"} {"global_step": 3577, "acc_step": 0, "speed/wps": 12896.294054944126, "speed/FLOPS": 202554019497995.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12351803481578827, "optim/lr": 0.0021455999999999997, "optim/total_tokens": 1875378176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.2394542694091797, "created_at": "2025-01-15T15:39:16.778440+00:00"} {"global_step": 3578, "acc_step": 0, "speed/wps": 12900.252754543935, "speed/FLOPS": 202616196315033.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10693620890378952, "optim/lr": 0.0021462, "optim/total_tokens": 1875902464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 3.29947566986084, "created_at": "2025-01-15T15:39:26.952523+00:00"} {"global_step": 3579, "acc_step": 0, "speed/wps": 12894.457041514232, "speed/FLOPS": 202525166677757.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.125337153673172, "optim/lr": 0.0021468, "optim/total_tokens": 1876426752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.3035507202148438, "created_at": "2025-01-15T15:39:37.121471+00:00"} {"global_step": 3580, "acc_step": 0, "speed/wps": 12899.602578298727, "speed/FLOPS": 202605984403666.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13740435242652893, "optim/lr": 0.0021474, "optim/total_tokens": 1876951040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.2976958751678467, "created_at": "2025-01-15T15:39:47.288990+00:00"} {"global_step": 3581, "acc_step": 0, "speed/wps": 12898.170606263786, "speed/FLOPS": 202583493315122.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11041779816150665, "optim/lr": 0.002148, "optim/total_tokens": 1877475328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 3.3605122566223145, "created_at": "2025-01-15T15:39:57.454671+00:00"} {"global_step": 3582, "acc_step": 0, "speed/wps": 12896.311967706471, "speed/FLOPS": 202554300842542.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11628880351781845, "optim/lr": 0.0021486, "optim/total_tokens": 1877999616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.3955297470092773, "created_at": "2025-01-15T15:40:07.624730+00:00"} {"global_step": 3583, "acc_step": 0, "speed/wps": 12894.24916012271, "speed/FLOPS": 202521901614845.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12292581051588058, "optim/lr": 0.0021492, "optim/total_tokens": 1878523904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.324840784072876, "created_at": "2025-01-15T15:40:17.794950+00:00"} {"global_step": 3584, "acc_step": 0, "speed/wps": 12899.93225658762, "speed/FLOPS": 202611162454219.34, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10375389456748962, "optim/lr": 0.0021498, "optim/total_tokens": 1879048192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417976, "loss/out": 3.179621458053589, "created_at": "2025-01-15T15:40:27.963540+00:00"} {"global_step": 3585, "acc_step": 0, "speed/wps": 12890.910711750239, "speed/FLOPS": 202469466695646.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.11592920124530792, "optim/lr": 0.0021504000000000002, "optim/total_tokens": 1879572480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 3.3108584880828857, "created_at": "2025-01-15T15:40:38.136025+00:00"} {"global_step": 3586, "acc_step": 0, "speed/wps": 12894.953653124212, "speed/FLOPS": 202532966645508.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10938158631324768, "optim/lr": 0.002151, "optim/total_tokens": 1880096768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.3004820346832275, "created_at": "2025-01-15T15:40:48.304187+00:00"} {"global_step": 3587, "acc_step": 0, "speed/wps": 12896.310529531276, "speed/FLOPS": 202554278254025.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11900344491004944, "optim/lr": 0.0021516, "optim/total_tokens": 1880621056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.3704090118408203, "created_at": "2025-01-15T15:40:58.471768+00:00"} {"global_step": 3588, "acc_step": 0, "speed/wps": 12897.37459622806, "speed/FLOPS": 202570990883678.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12524113059043884, "optim/lr": 0.0021522, "optim/total_tokens": 1881145344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 3.193873882293701, "created_at": "2025-01-15T15:41:08.638194+00:00"} {"global_step": 3589, "acc_step": 0, "speed/wps": 12896.415716880541, "speed/FLOPS": 202555930365887.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10586266964673996, "optim/lr": 0.0021528000000000003, "optim/total_tokens": 1881669632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.280614137649536, "created_at": "2025-01-15T15:41:18.808829+00:00"} {"global_step": 3590, "acc_step": 0, "speed/wps": 12897.11980843474, "speed/FLOPS": 202566989091269.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10249508172273636, "optim/lr": 0.0021534, "optim/total_tokens": 1882193920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.2613463401794434, "created_at": "2025-01-15T15:41:28.990660+00:00"} {"global_step": 3591, "acc_step": 0, "speed/wps": 12895.180875233347, "speed/FLOPS": 202536535480968.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09467208385467529, "optim/lr": 0.002154, "optim/total_tokens": 1882718208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.254093647003174, "created_at": "2025-01-15T15:41:39.158740+00:00"} {"global_step": 3592, "acc_step": 0, "speed/wps": 12895.172842009833, "speed/FLOPS": 202536409308154.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10570163279771805, "optim/lr": 0.0021546, "optim/total_tokens": 1883242496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304572, "loss/out": 3.195383071899414, "created_at": "2025-01-15T15:41:49.330117+00:00"} {"global_step": 3593, "acc_step": 0, "speed/wps": 12892.060859860838, "speed/FLOPS": 202487531352192.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10923102498054504, "optim/lr": 0.0021552000000000003, "optim/total_tokens": 1883766784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 3.2516260147094727, "created_at": "2025-01-15T15:41:59.506359+00:00"} {"global_step": 3594, "acc_step": 0, "speed/wps": 12898.985790969902, "speed/FLOPS": 202596296911113.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11933427304029465, "optim/lr": 0.0021558000000000003, "optim/total_tokens": 1884291072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 3.297475814819336, "created_at": "2025-01-15T15:42:09.672112+00:00"} {"global_step": 3595, "acc_step": 0, "speed/wps": 12895.557967117533, "speed/FLOPS": 202542458227188.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14362531900405884, "optim/lr": 0.0021564, "optim/total_tokens": 1884815360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 3.2734878063201904, "created_at": "2025-01-15T15:42:19.872371+00:00"} {"global_step": 3596, "acc_step": 0, "speed/wps": 12894.313962778464, "speed/FLOPS": 202522919429601.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13155902922153473, "optim/lr": 0.002157, "optim/total_tokens": 1885339648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413950, "loss/out": 3.2200191020965576, "created_at": "2025-01-15T15:42:30.041870+00:00"} {"global_step": 3597, "acc_step": 0, "speed/wps": 12901.548289036906, "speed/FLOPS": 202636544464496.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.14144214987754822, "optim/lr": 0.0021576, "optim/total_tokens": 1885863936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.238311290740967, "created_at": "2025-01-15T15:42:40.204944+00:00"} {"global_step": 3598, "acc_step": 0, "speed/wps": 12875.158504142279, "speed/FLOPS": 202222056629362.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11604753136634827, "optim/lr": 0.0021582000000000003, "optim/total_tokens": 1886388224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.27439546585083, "created_at": "2025-01-15T15:42:50.390029+00:00"} {"global_step": 3599, "acc_step": 0, "speed/wps": 12889.20430030463, "speed/FLOPS": 202442665159038.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12822262942790985, "optim/lr": 0.0021588, "optim/total_tokens": 1886912512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.205765724182129, "created_at": "2025-01-15T15:43:00.565697+00:00"} {"global_step": 3600, "acc_step": 0, "speed/wps": 12882.562235345495, "speed/FLOPS": 202338342401701.3, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13540394604206085, "optim/lr": 0.0021594, "optim/total_tokens": 1887436800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.358978509902954, "created_at": "2025-01-15T15:43:10.743990+00:00"} {"global_step": 3601, "acc_step": 0, "speed/wps": 12883.932668529254, "speed/FLOPS": 202359866938024.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11661653965711594, "optim/lr": 0.00216, "optim/total_tokens": 1887961088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.307333469390869, "created_at": "2025-01-15T15:43:20.921213+00:00"} {"global_step": 3602, "acc_step": 0, "speed/wps": 12900.832999820637, "speed/FLOPS": 202625309864444.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11817491799592972, "optim/lr": 0.0021606, "optim/total_tokens": 1888485376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.309831380844116, "created_at": "2025-01-15T15:43:31.092832+00:00"} {"global_step": 3603, "acc_step": 0, "speed/wps": 12902.33675801639, "speed/FLOPS": 202648928453289.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10840911418199539, "optim/lr": 0.0021612000000000003, "optim/total_tokens": 1889009664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 3.2226033210754395, "created_at": "2025-01-15T15:43:41.256365+00:00"} {"global_step": 3604, "acc_step": 0, "speed/wps": 12899.44366767229, "speed/FLOPS": 202603488493912.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11180209368467331, "optim/lr": 0.0021618, "optim/total_tokens": 1889533952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.369173765182495, "created_at": "2025-01-15T15:43:51.444274+00:00"} {"global_step": 3605, "acc_step": 0, "speed/wps": 12887.68498923696, "speed/FLOPS": 202418802290968.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.14114612340927124, "optim/lr": 0.0021624, "optim/total_tokens": 1890058240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.2838785648345947, "created_at": "2025-01-15T15:44:01.618619+00:00"} {"global_step": 3606, "acc_step": 0, "speed/wps": 12884.071886479493, "speed/FLOPS": 202362053547238.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13253159821033478, "optim/lr": 0.002163, "optim/total_tokens": 1890582528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 3.3111424446105957, "created_at": "2025-01-15T15:44:11.795496+00:00"} {"global_step": 3607, "acc_step": 0, "speed/wps": 12900.53824584829, "speed/FLOPS": 202620680348274.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11523418128490448, "optim/lr": 0.0021636, "optim/total_tokens": 1891106816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.358337879180908, "created_at": "2025-01-15T15:44:21.959491+00:00"} {"global_step": 3608, "acc_step": 0, "speed/wps": 12896.820202508354, "speed/FLOPS": 202562283368494.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.123930424451828, "optim/lr": 0.0021642000000000002, "optim/total_tokens": 1891631104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.291843891143799, "created_at": "2025-01-15T15:44:32.130138+00:00"} {"global_step": 3609, "acc_step": 0, "speed/wps": 12896.32503538043, "speed/FLOPS": 202554506088319.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13090503215789795, "optim/lr": 0.0021648, "optim/total_tokens": 1892155392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.246433734893799, "created_at": "2025-01-15T15:44:42.298402+00:00"} {"global_step": 3610, "acc_step": 0, "speed/wps": 12896.448717534433, "speed/FLOPS": 202556448686504.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10432194918394089, "optim/lr": 0.0021654, "optim/total_tokens": 1892679680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.3263728618621826, "created_at": "2025-01-15T15:44:52.465331+00:00"} {"global_step": 3611, "acc_step": 0, "speed/wps": 12904.2193574887, "speed/FLOPS": 202678497264964.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13259463012218475, "optim/lr": 0.002166, "optim/total_tokens": 1893203968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 3.3112144470214844, "created_at": "2025-01-15T15:45:02.628880+00:00"} {"global_step": 3612, "acc_step": 0, "speed/wps": 12895.504886319532, "speed/FLOPS": 202541624520314.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1430336982011795, "optim/lr": 0.0021666, "optim/total_tokens": 1893728256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 3.2500576972961426, "created_at": "2025-01-15T15:45:12.798415+00:00"} {"global_step": 3613, "acc_step": 0, "speed/wps": 12896.15658164461, "speed/FLOPS": 202551860290918.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15516303479671478, "optim/lr": 0.0021672, "optim/total_tokens": 1894252544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.362584114074707, "created_at": "2025-01-15T15:45:22.967371+00:00"} {"global_step": 3614, "acc_step": 0, "speed/wps": 12888.152081881522, "speed/FLOPS": 202426138622803.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14976787567138672, "optim/lr": 0.0021678, "optim/total_tokens": 1894776832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 3.2974886894226074, "created_at": "2025-01-15T15:45:33.145043+00:00"} {"global_step": 3615, "acc_step": 0, "speed/wps": 12897.419123340926, "speed/FLOPS": 202571690243173.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11437930166721344, "optim/lr": 0.0021684, "optim/total_tokens": 1895301120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 3.3231658935546875, "created_at": "2025-01-15T15:45:43.311199+00:00"} {"global_step": 3616, "acc_step": 0, "speed/wps": 12901.405651319687, "speed/FLOPS": 202634304143143.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10457158833742142, "optim/lr": 0.002169, "optim/total_tokens": 1895825408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.3639817237854004, "created_at": "2025-01-15T15:45:53.479200+00:00"} {"global_step": 3617, "acc_step": 0, "speed/wps": 12897.749541072786, "speed/FLOPS": 202576879907693.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10229001194238663, "optim/lr": 0.0021696, "optim/total_tokens": 1896349696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 3.3190054893493652, "created_at": "2025-01-15T15:46:03.646791+00:00"} {"global_step": 3618, "acc_step": 0, "speed/wps": 12899.953829543585, "speed/FLOPS": 202611501287137.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11680756509304047, "optim/lr": 0.0021702, "optim/total_tokens": 1896873984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 3.235508918762207, "created_at": "2025-01-15T15:46:13.811459+00:00"} {"global_step": 3619, "acc_step": 0, "speed/wps": 12898.06322829926, "speed/FLOPS": 202581806796636.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12020125240087509, "optim/lr": 0.0021708, "optim/total_tokens": 1897398272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 3.3345489501953125, "created_at": "2025-01-15T15:46:23.977999+00:00"} {"global_step": 3620, "acc_step": 0, "speed/wps": 12899.964728968562, "speed/FLOPS": 202611672477583.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10454510897397995, "optim/lr": 0.0021714, "optim/total_tokens": 1897922560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.3344337940216064, "created_at": "2025-01-15T15:46:34.145587+00:00"} {"global_step": 3621, "acc_step": 0, "speed/wps": 12898.40380389819, "speed/FLOPS": 202587156004417.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1207275465130806, "optim/lr": 0.002172, "optim/total_tokens": 1898446848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.2125842571258545, "created_at": "2025-01-15T15:46:44.311402+00:00"} {"global_step": 3622, "acc_step": 0, "speed/wps": 12894.385399559626, "speed/FLOPS": 202524041442414.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10221324115991592, "optim/lr": 0.0021726, "optim/total_tokens": 1898971136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 3.2385740280151367, "created_at": "2025-01-15T15:46:54.480534+00:00"} {"global_step": 3623, "acc_step": 0, "speed/wps": 12896.904330537662, "speed/FLOPS": 202563604714798.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11996611207723618, "optim/lr": 0.0021732, "optim/total_tokens": 1899495424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 3.298464775085449, "created_at": "2025-01-15T15:47:04.649303+00:00"} {"global_step": 3624, "acc_step": 0, "speed/wps": 12896.914794617156, "speed/FLOPS": 202563769067545.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10850197076797485, "optim/lr": 0.0021738, "optim/total_tokens": 1900019712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.204561948776245, "created_at": "2025-01-15T15:47:14.818543+00:00"} {"global_step": 3625, "acc_step": 0, "speed/wps": 12895.543412185245, "speed/FLOPS": 202542229621975.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12462559342384338, "optim/lr": 0.0021744, "optim/total_tokens": 1900544000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 3.199253559112549, "created_at": "2025-01-15T15:47:24.986598+00:00"} {"global_step": 3626, "acc_step": 0, "speed/wps": 12897.911235015343, "speed/FLOPS": 202579419533252.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1603369563817978, "optim/lr": 0.002175, "optim/total_tokens": 1901068288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 3.387815475463867, "created_at": "2025-01-15T15:47:35.153650+00:00"} {"global_step": 3627, "acc_step": 0, "speed/wps": 12897.0808027615, "speed/FLOPS": 202566376453572.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1600569486618042, "optim/lr": 0.0021755999999999998, "optim/total_tokens": 1901592576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 3.3086495399475098, "created_at": "2025-01-15T15:47:45.321360+00:00"} {"global_step": 3628, "acc_step": 0, "speed/wps": 12897.726613827568, "speed/FLOPS": 202576519803802.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1591109037399292, "optim/lr": 0.0021762, "optim/total_tokens": 1902116864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 3.289083957672119, "created_at": "2025-01-15T15:47:55.487673+00:00"} {"global_step": 3629, "acc_step": 0, "speed/wps": 12895.409566955774, "speed/FLOPS": 202540127398724.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16895028948783875, "optim/lr": 0.0021768, "optim/total_tokens": 1902641152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.3232827186584473, "created_at": "2025-01-15T15:48:05.657325+00:00"} {"global_step": 3630, "acc_step": 0, "speed/wps": 12898.751261247473, "speed/FLOPS": 202592613299544.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13059328496456146, "optim/lr": 0.0021774, "optim/total_tokens": 1903165440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.2884445190429688, "created_at": "2025-01-15T15:48:15.822682+00:00"} {"global_step": 3631, "acc_step": 0, "speed/wps": 12894.486538733205, "speed/FLOPS": 202525629972115.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15437176823616028, "optim/lr": 0.002178, "optim/total_tokens": 1903689728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.199702262878418, "created_at": "2025-01-15T15:48:25.991547+00:00"} {"global_step": 3632, "acc_step": 0, "speed/wps": 12893.687782134588, "speed/FLOPS": 202513084402127.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11995184421539307, "optim/lr": 0.0021785999999999997, "optim/total_tokens": 1904214016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 3.1927762031555176, "created_at": "2025-01-15T15:48:36.161507+00:00"} {"global_step": 3633, "acc_step": 0, "speed/wps": 12892.723387223854, "speed/FLOPS": 202497937254836.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.11101362109184265, "optim/lr": 0.0021792, "optim/total_tokens": 1904738304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.181753158569336, "created_at": "2025-01-15T15:48:46.333375+00:00"} {"global_step": 3634, "acc_step": 0, "speed/wps": 12899.134745173751, "speed/FLOPS": 202598636441569.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10184922814369202, "optim/lr": 0.0021798, "optim/total_tokens": 1905262592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.25840425491333, "created_at": "2025-01-15T15:48:56.498279+00:00"} {"global_step": 3635, "acc_step": 0, "speed/wps": 12904.174732532492, "speed/FLOPS": 202677796368706.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.130111575126648, "optim/lr": 0.0021804, "optim/total_tokens": 1905786880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 3.3775806427001953, "created_at": "2025-01-15T15:49:06.661282+00:00"} {"global_step": 3636, "acc_step": 0, "speed/wps": 12899.479303842214, "speed/FLOPS": 202604048208930.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16528622806072235, "optim/lr": 0.0021809999999999998, "optim/total_tokens": 1906311168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 3.3216826915740967, "created_at": "2025-01-15T15:49:16.827075+00:00"} {"global_step": 3637, "acc_step": 0, "speed/wps": 12898.524409796953, "speed/FLOPS": 202589050285787.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17039811611175537, "optim/lr": 0.0021816, "optim/total_tokens": 1906835456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.3474326133728027, "created_at": "2025-01-15T15:49:26.992457+00:00"} {"global_step": 3638, "acc_step": 0, "speed/wps": 12896.676234606182, "speed/FLOPS": 202560022154757.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2121305912733078, "optim/lr": 0.0021822, "optim/total_tokens": 1907359744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 3.216132640838623, "created_at": "2025-01-15T15:49:37.159239+00:00"} {"global_step": 3639, "acc_step": 0, "speed/wps": 12893.693828993511, "speed/FLOPS": 202513179376356.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2329498827457428, "optim/lr": 0.0021828, "optim/total_tokens": 1907884032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.4122402667999268, "created_at": "2025-01-15T15:49:47.340942+00:00"} {"global_step": 3640, "acc_step": 0, "speed/wps": 12899.416550217093, "speed/FLOPS": 202603062577015.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2238215208053589, "optim/lr": 0.0021834, "optim/total_tokens": 1908408320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.5024471282958984, "created_at": "2025-01-15T15:49:57.505484+00:00"} {"global_step": 3641, "acc_step": 0, "speed/wps": 12893.018597991257, "speed/FLOPS": 202502573945601.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17482584714889526, "optim/lr": 0.002184, "optim/total_tokens": 1908932608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421131, "loss/out": 3.408431053161621, "created_at": "2025-01-15T15:50:07.676151+00:00"} {"global_step": 3642, "acc_step": 0, "speed/wps": 12896.075285256855, "speed/FLOPS": 202550583419435.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.14451183378696442, "optim/lr": 0.0021846, "optim/total_tokens": 1909456896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.303957939147949, "created_at": "2025-01-15T15:50:17.845503+00:00"} {"global_step": 3643, "acc_step": 0, "speed/wps": 12899.001636394589, "speed/FLOPS": 202596545785281.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12174704670906067, "optim/lr": 0.0021852, "optim/total_tokens": 1909981184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 3.298616409301758, "created_at": "2025-01-15T15:50:28.014347+00:00"} {"global_step": 3644, "acc_step": 0, "speed/wps": 12896.313236893395, "speed/FLOPS": 202554320776867.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12321989238262177, "optim/lr": 0.0021858000000000003, "optim/total_tokens": 1910505472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.3909592628479004, "created_at": "2025-01-15T15:50:38.184209+00:00"} {"global_step": 3645, "acc_step": 0, "speed/wps": 12896.632233795064, "speed/FLOPS": 202559331061555.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.104631207883358, "optim/lr": 0.0021864000000000002, "optim/total_tokens": 1911029760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 3.3219404220581055, "created_at": "2025-01-15T15:50:48.353532+00:00"} {"global_step": 3646, "acc_step": 0, "speed/wps": 12896.866443006287, "speed/FLOPS": 202563009639055.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12476619333028793, "optim/lr": 0.002187, "optim/total_tokens": 1911554048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.2016334533691406, "created_at": "2025-01-15T15:50:58.521010+00:00"} {"global_step": 3647, "acc_step": 0, "speed/wps": 12893.695925727516, "speed/FLOPS": 202513212308444.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11599185317754745, "optim/lr": 0.0021876, "optim/total_tokens": 1912078336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.391087532043457, "created_at": "2025-01-15T15:51:08.694342+00:00"} {"global_step": 3648, "acc_step": 0, "speed/wps": 12898.589993097317, "speed/FLOPS": 202590080361640.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10002385824918747, "optim/lr": 0.0021882000000000004, "optim/total_tokens": 1912602624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 3.232937812805176, "created_at": "2025-01-15T15:51:18.861659+00:00"} {"global_step": 3649, "acc_step": 0, "speed/wps": 12892.18948177366, "speed/FLOPS": 202489551536078.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11314238607883453, "optim/lr": 0.0021888000000000003, "optim/total_tokens": 1913126912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.2174508571624756, "created_at": "2025-01-15T15:51:29.034876+00:00"} {"global_step": 3650, "acc_step": 0, "speed/wps": 12890.143020624237, "speed/FLOPS": 202457409051594.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09902948141098022, "optim/lr": 0.0021894, "optim/total_tokens": 1913651200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.2544925212860107, "created_at": "2025-01-15T15:51:39.208457+00:00"} {"global_step": 3651, "acc_step": 0, "speed/wps": 12894.174524582697, "speed/FLOPS": 202520729361129.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10274939239025116, "optim/lr": 0.00219, "optim/total_tokens": 1914175488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.2704896926879883, "created_at": "2025-01-15T15:51:49.382297+00:00"} {"global_step": 3652, "acc_step": 0, "speed/wps": 12893.480740587069, "speed/FLOPS": 202509832530118.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11661890894174576, "optim/lr": 0.0021906, "optim/total_tokens": 1914699776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 3.204568862915039, "created_at": "2025-01-15T15:51:59.551666+00:00"} {"global_step": 3653, "acc_step": 0, "speed/wps": 12898.057125330784, "speed/FLOPS": 202581710941130.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12648449838161469, "optim/lr": 0.0021912000000000004, "optim/total_tokens": 1915224064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.320824146270752, "created_at": "2025-01-15T15:52:09.720832+00:00"} {"global_step": 3654, "acc_step": 0, "speed/wps": 12900.939327650027, "speed/FLOPS": 202626979889115.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12085583060979843, "optim/lr": 0.0021918000000000003, "optim/total_tokens": 1915748352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.4104106426239014, "created_at": "2025-01-15T15:52:19.884225+00:00"} {"global_step": 3655, "acc_step": 0, "speed/wps": 12895.835244970584, "speed/FLOPS": 202546813256889.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12603245675563812, "optim/lr": 0.0021924, "optim/total_tokens": 1916272640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 3.330683469772339, "created_at": "2025-01-15T15:52:30.054866+00:00"} {"global_step": 3656, "acc_step": 0, "speed/wps": 12894.811085574027, "speed/FLOPS": 202530727426224.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11306022852659225, "optim/lr": 0.002193, "optim/total_tokens": 1916796928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.3160228729248047, "created_at": "2025-01-15T15:52:40.223086+00:00"} {"global_step": 3657, "acc_step": 0, "speed/wps": 12896.41980812522, "speed/FLOPS": 202555994624507.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11537490785121918, "optim/lr": 0.0021936, "optim/total_tokens": 1917321216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.283377170562744, "created_at": "2025-01-15T15:52:50.391033+00:00"} {"global_step": 3658, "acc_step": 0, "speed/wps": 12897.182681542952, "speed/FLOPS": 202567976599830.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1287815123796463, "optim/lr": 0.0021942000000000003, "optim/total_tokens": 1917845504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.3016226291656494, "created_at": "2025-01-15T15:53:00.561590+00:00"} {"global_step": 3659, "acc_step": 0, "speed/wps": 12900.399474729256, "speed/FLOPS": 202618500757158.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09007031470537186, "optim/lr": 0.0021948000000000002, "optim/total_tokens": 1918369792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 3.3432376384735107, "created_at": "2025-01-15T15:53:10.729345+00:00"} {"global_step": 3660, "acc_step": 0, "speed/wps": 12898.34805173957, "speed/FLOPS": 202586280340155.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11457514762878418, "optim/lr": 0.0021954, "optim/total_tokens": 1918894080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.2420315742492676, "created_at": "2025-01-15T15:53:20.898768+00:00"} {"global_step": 3661, "acc_step": 0, "speed/wps": 12900.940829534891, "speed/FLOPS": 202627003478280.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11461497843265533, "optim/lr": 0.002196, "optim/total_tokens": 1919418368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427332, "loss/out": 3.295746088027954, "created_at": "2025-01-15T15:53:31.065612+00:00"} {"global_step": 3662, "acc_step": 0, "speed/wps": 12895.48377251764, "speed/FLOPS": 202541292899042.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13085143268108368, "optim/lr": 0.0021966, "optim/total_tokens": 1919942656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 3.2985501289367676, "created_at": "2025-01-15T15:53:41.233829+00:00"} {"global_step": 3663, "acc_step": 0, "speed/wps": 12895.217900973143, "speed/FLOPS": 202537117021090.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1163272112607956, "optim/lr": 0.0021972000000000003, "optim/total_tokens": 1920466944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 3.322859048843384, "created_at": "2025-01-15T15:53:51.402559+00:00"} {"global_step": 3664, "acc_step": 0, "speed/wps": 12898.744597272611, "speed/FLOPS": 202592508632662.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11472122371196747, "optim/lr": 0.0021978, "optim/total_tokens": 1920991232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.2594308853149414, "created_at": "2025-01-15T15:54:01.567773+00:00"} {"global_step": 3665, "acc_step": 0, "speed/wps": 12890.684288309523, "speed/FLOPS": 202465910404371.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10913515836000443, "optim/lr": 0.0021984, "optim/total_tokens": 1921515520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394163, "loss/out": 3.3032119274139404, "created_at": "2025-01-15T15:54:11.741995+00:00"} {"global_step": 3666, "acc_step": 0, "speed/wps": 12897.070210769885, "speed/FLOPS": 202566210091788.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1002785712480545, "optim/lr": 0.002199, "optim/total_tokens": 1922039808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 3.143315315246582, "created_at": "2025-01-15T15:54:21.908496+00:00"} {"global_step": 3667, "acc_step": 0, "speed/wps": 12897.430428617337, "speed/FLOPS": 202571867808074.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12966421246528625, "optim/lr": 0.0021996, "optim/total_tokens": 1922564096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 3.1993017196655273, "created_at": "2025-01-15T15:54:32.080554+00:00"} {"global_step": 3668, "acc_step": 0, "speed/wps": 12901.38546077721, "speed/FLOPS": 202633987022928.7, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12693388760089874, "optim/lr": 0.0022002000000000002, "optim/total_tokens": 1923088384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 3.2925500869750977, "created_at": "2025-01-15T15:54:42.244807+00:00"} {"global_step": 3669, "acc_step": 0, "speed/wps": 12898.133626215822, "speed/FLOPS": 202582912492653.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1051778495311737, "optim/lr": 0.0022008, "optim/total_tokens": 1923612672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.2502212524414062, "created_at": "2025-01-15T15:54:52.413722+00:00"} {"global_step": 3670, "acc_step": 0, "speed/wps": 12896.395489956252, "speed/FLOPS": 202555612674247.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1268378347158432, "optim/lr": 0.0022014, "optim/total_tokens": 1924136960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.3626108169555664, "created_at": "2025-01-15T15:55:02.583176+00:00"} {"global_step": 3671, "acc_step": 0, "speed/wps": 12894.85999360709, "speed/FLOPS": 202531495594090.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11709197610616684, "optim/lr": 0.002202, "optim/total_tokens": 1924661248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.2909045219421387, "created_at": "2025-01-15T15:55:12.751712+00:00"} {"global_step": 3672, "acc_step": 0, "speed/wps": 12892.443856310678, "speed/FLOPS": 202493546837729.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1155080646276474, "optim/lr": 0.0022026, "optim/total_tokens": 1925185536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 3.3406858444213867, "created_at": "2025-01-15T15:55:22.922587+00:00"} {"global_step": 3673, "acc_step": 0, "speed/wps": 12899.551409134387, "speed/FLOPS": 202605180721627.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1020459234714508, "optim/lr": 0.0022032, "optim/total_tokens": 1925709824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.1162679195404053, "created_at": "2025-01-15T15:55:33.088335+00:00"} {"global_step": 3674, "acc_step": 0, "speed/wps": 12900.274950548208, "speed/FLOPS": 202616544933781.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1025472953915596, "optim/lr": 0.0022038, "optim/total_tokens": 1926234112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.201037883758545, "created_at": "2025-01-15T15:55:43.252423+00:00"} {"global_step": 3675, "acc_step": 0, "speed/wps": 12893.330468793012, "speed/FLOPS": 202507472305097.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.090064138174057, "optim/lr": 0.0022044, "optim/total_tokens": 1926758400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.3561770915985107, "created_at": "2025-01-15T15:55:53.424613+00:00"} {"global_step": 3676, "acc_step": 0, "speed/wps": 12898.926319252034, "speed/FLOPS": 202595362826061.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10378065705299377, "optim/lr": 0.002205, "optim/total_tokens": 1927282688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 3.267836570739746, "created_at": "2025-01-15T15:56:03.594842+00:00"} {"global_step": 3677, "acc_step": 0, "speed/wps": 12898.041026902627, "speed/FLOPS": 202581458093194.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08172748237848282, "optim/lr": 0.0022056, "optim/total_tokens": 1927806976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.1986618041992188, "created_at": "2025-01-15T15:56:13.762697+00:00"} {"global_step": 3678, "acc_step": 0, "speed/wps": 12894.549373174174, "speed/FLOPS": 202526616873354.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09172480553388596, "optim/lr": 0.0022062, "optim/total_tokens": 1928331264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389084, "loss/out": 3.368516445159912, "created_at": "2025-01-15T15:56:23.934282+00:00"} {"global_step": 3679, "acc_step": 0, "speed/wps": 12894.978843705745, "speed/FLOPS": 202533362298206.94, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09257689863443375, "optim/lr": 0.0022068, "optim/total_tokens": 1928855552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 3.1898791790008545, "created_at": "2025-01-15T15:56:34.104474+00:00"} {"global_step": 3680, "acc_step": 0, "speed/wps": 12896.596723260547, "speed/FLOPS": 202558773319814.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.09766934812068939, "optim/lr": 0.0022074, "optim/total_tokens": 1929379840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.332962989807129, "created_at": "2025-01-15T15:56:44.278755+00:00"} {"global_step": 3681, "acc_step": 0, "speed/wps": 12900.73167355619, "speed/FLOPS": 202623718396224.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0849323645234108, "optim/lr": 0.002208, "optim/total_tokens": 1929904128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 3.1957149505615234, "created_at": "2025-01-15T15:56:54.444590+00:00"} {"global_step": 3682, "acc_step": 0, "speed/wps": 12893.556589490665, "speed/FLOPS": 202511023841377.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10555525869131088, "optim/lr": 0.0022086, "optim/total_tokens": 1930428416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 3.2177982330322266, "created_at": "2025-01-15T15:57:04.614410+00:00"} {"global_step": 3683, "acc_step": 0, "speed/wps": 12893.630522122821, "speed/FLOPS": 202512185054959.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1347050815820694, "optim/lr": 0.0022092, "optim/total_tokens": 1930952704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.3028693199157715, "created_at": "2025-01-15T15:57:14.783897+00:00"} {"global_step": 3684, "acc_step": 0, "speed/wps": 12896.670233403529, "speed/FLOPS": 202559927897623.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11525395512580872, "optim/lr": 0.0022098, "optim/total_tokens": 1931476992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.2528910636901855, "created_at": "2025-01-15T15:57:24.952693+00:00"} {"global_step": 3685, "acc_step": 0, "speed/wps": 12891.716430499497, "speed/FLOPS": 202482121615777.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1075601726770401, "optim/lr": 0.0022104, "optim/total_tokens": 1932001280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.2570910453796387, "created_at": "2025-01-15T15:57:35.123284+00:00"} {"global_step": 3686, "acc_step": 0, "speed/wps": 12895.13932447044, "speed/FLOPS": 202535882869140.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1157454401254654, "optim/lr": 0.002211, "optim/total_tokens": 1932525568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 3.172959327697754, "created_at": "2025-01-15T15:57:45.297287+00:00"} {"global_step": 3687, "acc_step": 0, "speed/wps": 12894.984908316617, "speed/FLOPS": 202533457551254.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16400401294231415, "optim/lr": 0.0022115999999999998, "optim/total_tokens": 1933049856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.272212028503418, "created_at": "2025-01-15T15:57:55.476832+00:00"} {"global_step": 3688, "acc_step": 0, "speed/wps": 12890.773849101859, "speed/FLOPS": 202467317079684.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17005957663059235, "optim/lr": 0.0022122, "optim/total_tokens": 1933574144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.343203544616699, "created_at": "2025-01-15T15:58:05.648356+00:00"} {"global_step": 3689, "acc_step": 0, "speed/wps": 12887.851890474258, "speed/FLOPS": 202421423704246.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.15412046015262604, "optim/lr": 0.0022128, "optim/total_tokens": 1934098432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 3.2876720428466797, "created_at": "2025-01-15T15:58:15.827751+00:00"} {"global_step": 3690, "acc_step": 0, "speed/wps": 12894.856409021771, "speed/FLOPS": 202531439293252.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14261280000209808, "optim/lr": 0.0022134, "optim/total_tokens": 1934622720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455791, "loss/out": 3.2545700073242188, "created_at": "2025-01-15T15:58:25.996350+00:00"} {"global_step": 3691, "acc_step": 0, "speed/wps": 12897.359629428782, "speed/FLOPS": 202570755809529.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.125204399228096, "optim/lr": 0.002214, "optim/total_tokens": 1935147008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 3.35591983795166, "created_at": "2025-01-15T15:58:36.162589+00:00"} {"global_step": 3692, "acc_step": 0, "speed/wps": 12893.308317940791, "speed/FLOPS": 202507124395525.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11898793280124664, "optim/lr": 0.0022146, "optim/total_tokens": 1935671296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.325322151184082, "created_at": "2025-01-15T15:58:46.333104+00:00"} {"global_step": 3693, "acc_step": 0, "speed/wps": 12894.443911467455, "speed/FLOPS": 202524960452330.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10753370076417923, "optim/lr": 0.0022152, "optim/total_tokens": 1936195584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 3.288189649581909, "created_at": "2025-01-15T15:58:56.504618+00:00"} {"global_step": 3694, "acc_step": 0, "speed/wps": 12895.437161465947, "speed/FLOPS": 202540560808425.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13839736580848694, "optim/lr": 0.0022158, "optim/total_tokens": 1936719872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.256679058074951, "created_at": "2025-01-15T15:59:06.674770+00:00"} {"global_step": 3695, "acc_step": 0, "speed/wps": 12888.029230357291, "speed/FLOPS": 202424209070799.97, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11033119261264801, "optim/lr": 0.0022164, "optim/total_tokens": 1937244160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.278921127319336, "created_at": "2025-01-15T15:59:16.848402+00:00"} {"global_step": 3696, "acc_step": 0, "speed/wps": 12894.87829699866, "speed/FLOPS": 202531783074006.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11028972268104553, "optim/lr": 0.0022170000000000002, "optim/total_tokens": 1937768448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402169, "loss/out": 3.228456974029541, "created_at": "2025-01-15T15:59:27.017042+00:00"} {"global_step": 3697, "acc_step": 0, "speed/wps": 12898.544974401142, "speed/FLOPS": 202589373281155.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13488554954528809, "optim/lr": 0.0022176, "optim/total_tokens": 1938292736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 3.2732417583465576, "created_at": "2025-01-15T15:59:37.184288+00:00"} {"global_step": 3698, "acc_step": 0, "speed/wps": 12890.581761249618, "speed/FLOPS": 202464300076008.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11473243683576584, "optim/lr": 0.0022182, "optim/total_tokens": 1938817024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.261112689971924, "created_at": "2025-01-15T15:59:47.358475+00:00"} {"global_step": 3699, "acc_step": 0, "speed/wps": 12895.620418524415, "speed/FLOPS": 202543439112350.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10197152942419052, "optim/lr": 0.0022188, "optim/total_tokens": 1939341312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2577950954437256, "created_at": "2025-01-15T15:59:57.528090+00:00"} {"global_step": 3700, "acc_step": 0, "speed/wps": 12893.973210927872, "speed/FLOPS": 202517567453546.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09989550709724426, "optim/lr": 0.0022194000000000003, "optim/total_tokens": 1939865600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412896, "loss/out": 3.34938907623291, "created_at": "2025-01-15T16:00:07.698382+00:00"} {"global_step": 3701, "acc_step": 0, "speed/wps": 12897.743068435007, "speed/FLOPS": 202576778246022.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12205279618501663, "optim/lr": 0.00222, "optim/total_tokens": 1940389888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.347726345062256, "created_at": "2025-01-15T16:00:17.866475+00:00"} {"global_step": 3702, "acc_step": 0, "speed/wps": 12895.385986350564, "speed/FLOPS": 202539757032916.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11620482057332993, "optim/lr": 0.0022206, "optim/total_tokens": 1940914176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2389864921569824, "created_at": "2025-01-15T16:00:28.034879+00:00"} {"global_step": 3703, "acc_step": 0, "speed/wps": 12900.193435503159, "speed/FLOPS": 202615264627985.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11425304412841797, "optim/lr": 0.0022212, "optim/total_tokens": 1941438464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367082, "loss/out": 3.1032662391662598, "created_at": "2025-01-15T16:00:38.201563+00:00"} {"global_step": 3704, "acc_step": 0, "speed/wps": 12898.070298198876, "speed/FLOPS": 202581917839125.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11067990958690643, "optim/lr": 0.0022218000000000003, "optim/total_tokens": 1941962752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.325468063354492, "created_at": "2025-01-15T16:00:48.371107+00:00"} {"global_step": 3705, "acc_step": 0, "speed/wps": 12896.980842567446, "speed/FLOPS": 202564806441364.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10948877036571503, "optim/lr": 0.0022224000000000002, "optim/total_tokens": 1942487040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 3.208214282989502, "created_at": "2025-01-15T16:00:58.539269+00:00"} {"global_step": 3706, "acc_step": 0, "speed/wps": 12897.223227854704, "speed/FLOPS": 202568613435373.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12457004934549332, "optim/lr": 0.002223, "optim/total_tokens": 1943011328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.249851942062378, "created_at": "2025-01-15T16:01:08.707195+00:00"} {"global_step": 3707, "acc_step": 0, "speed/wps": 12890.718861865407, "speed/FLOPS": 202466453429575.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12036880850791931, "optim/lr": 0.0022236, "optim/total_tokens": 1943535616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 3.2330079078674316, "created_at": "2025-01-15T16:01:18.880192+00:00"} {"global_step": 3708, "acc_step": 0, "speed/wps": 12894.581716858373, "speed/FLOPS": 202527124875359.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13279293477535248, "optim/lr": 0.0022242, "optim/total_tokens": 1944059904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 3.3406338691711426, "created_at": "2025-01-15T16:01:29.050652+00:00"} {"global_step": 3709, "acc_step": 0, "speed/wps": 12898.395255982783, "speed/FLOPS": 202587021747660.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16024023294448853, "optim/lr": 0.0022248000000000003, "optim/total_tokens": 1944584192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.2743372917175293, "created_at": "2025-01-15T16:01:39.219982+00:00"} {"global_step": 3710, "acc_step": 0, "speed/wps": 12890.004354998571, "speed/FLOPS": 202455231117396.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16327807307243347, "optim/lr": 0.0022254, "optim/total_tokens": 1945108480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 3.279658555984497, "created_at": "2025-01-15T16:01:49.393780+00:00"} {"global_step": 3711, "acc_step": 0, "speed/wps": 12888.18629840307, "speed/FLOPS": 202426676040293.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14042937755584717, "optim/lr": 0.002226, "optim/total_tokens": 1945632768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.3754522800445557, "created_at": "2025-01-15T16:01:59.568207+00:00"} {"global_step": 3712, "acc_step": 0, "speed/wps": 12894.902049424703, "speed/FLOPS": 202532156138496.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15635178983211517, "optim/lr": 0.0022266, "optim/total_tokens": 1946157056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 3.284240484237671, "created_at": "2025-01-15T16:02:09.738475+00:00"} {"global_step": 3713, "acc_step": 0, "speed/wps": 12897.054228265148, "speed/FLOPS": 202565959064589.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10185453295707703, "optim/lr": 0.0022272, "optim/total_tokens": 1946681344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.2536063194274902, "created_at": "2025-01-15T16:02:19.906955+00:00"} {"global_step": 3714, "acc_step": 0, "speed/wps": 12895.251722577783, "speed/FLOPS": 202537648235867.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14255282282829285, "optim/lr": 0.0022278000000000003, "optim/total_tokens": 1947205632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 3.214372158050537, "created_at": "2025-01-15T16:02:30.081422+00:00"} {"global_step": 3715, "acc_step": 0, "speed/wps": 12893.033246440225, "speed/FLOPS": 202502804019621.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12949946522712708, "optim/lr": 0.0022284, "optim/total_tokens": 1947729920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 3.245450496673584, "created_at": "2025-01-15T16:02:40.254699+00:00"} {"global_step": 3716, "acc_step": 0, "speed/wps": 12898.40091479372, "speed/FLOPS": 202587110627062.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13253682851791382, "optim/lr": 0.002229, "optim/total_tokens": 1948254208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.389155387878418, "created_at": "2025-01-15T16:02:50.420147+00:00"} {"global_step": 3717, "acc_step": 0, "speed/wps": 12895.621460705788, "speed/FLOPS": 202543455481240.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1280232071876526, "optim/lr": 0.0022296, "optim/total_tokens": 1948778496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 3.222018241882324, "created_at": "2025-01-15T16:03:00.590254+00:00"} {"global_step": 3718, "acc_step": 0, "speed/wps": 12891.529319045769, "speed/FLOPS": 202479182773278.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1291464865207672, "optim/lr": 0.0022302, "optim/total_tokens": 1949302784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.3104279041290283, "created_at": "2025-01-15T16:03:10.761250+00:00"} {"global_step": 3719, "acc_step": 0, "speed/wps": 12894.181625660307, "speed/FLOPS": 202520840893310.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11838099360466003, "optim/lr": 0.0022308000000000002, "optim/total_tokens": 1949827072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.3295257091522217, "created_at": "2025-01-15T16:03:20.934811+00:00"} {"global_step": 3720, "acc_step": 0, "speed/wps": 12899.361159362112, "speed/FLOPS": 202602192587522.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.09552806615829468, "optim/lr": 0.0022314, "optim/total_tokens": 1950351360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.2789688110351562, "created_at": "2025-01-15T16:03:31.104848+00:00"} {"global_step": 3721, "acc_step": 0, "speed/wps": 12895.193826006433, "speed/FLOPS": 202536738890655.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11599688231945038, "optim/lr": 0.002232, "optim/total_tokens": 1950875648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.2866721153259277, "created_at": "2025-01-15T16:03:41.272830+00:00"} {"global_step": 3722, "acc_step": 0, "speed/wps": 12896.585103908195, "speed/FLOPS": 202558590821919.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11790834367275238, "optim/lr": 0.0022326, "optim/total_tokens": 1951399936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 3.359633445739746, "created_at": "2025-01-15T16:03:51.441341+00:00"} {"global_step": 3723, "acc_step": 0, "speed/wps": 12892.44877175683, "speed/FLOPS": 202493624041566.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10440763831138611, "optim/lr": 0.0022332, "optim/total_tokens": 1951924224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.2682089805603027, "created_at": "2025-01-15T16:04:01.642889+00:00"} {"global_step": 3724, "acc_step": 0, "speed/wps": 12891.061740699444, "speed/FLOPS": 202471838812829.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.11466966569423676, "optim/lr": 0.0022338, "optim/total_tokens": 1952448512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 3.221409320831299, "created_at": "2025-01-15T16:04:11.814134+00:00"} {"global_step": 3725, "acc_step": 0, "speed/wps": 12892.551685194898, "speed/FLOPS": 202495240438528.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12517806887626648, "optim/lr": 0.0022344, "optim/total_tokens": 1952972800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.253836154937744, "created_at": "2025-01-15T16:04:21.989775+00:00"} {"global_step": 3726, "acc_step": 0, "speed/wps": 12893.22181069165, "speed/FLOPS": 202505765680304.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14602720737457275, "optim/lr": 0.002235, "optim/total_tokens": 1953497088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 3.289034366607666, "created_at": "2025-01-15T16:04:32.159870+00:00"} {"global_step": 3727, "acc_step": 0, "speed/wps": 12895.77905653789, "speed/FLOPS": 202545930740343.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14348942041397095, "optim/lr": 0.0022356, "optim/total_tokens": 1954021376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 3.3105385303497314, "created_at": "2025-01-15T16:04:42.328303+00:00"} {"global_step": 3728, "acc_step": 0, "speed/wps": 12892.432917625843, "speed/FLOPS": 202493375030653.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14910250902175903, "optim/lr": 0.0022362, "optim/total_tokens": 1954545664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.3401479721069336, "created_at": "2025-01-15T16:04:52.498819+00:00"} {"global_step": 3729, "acc_step": 0, "speed/wps": 12892.094741669647, "speed/FLOPS": 202488063512558.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1396753191947937, "optim/lr": 0.0022368, "optim/total_tokens": 1955069952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 3.3673534393310547, "created_at": "2025-01-15T16:05:02.669479+00:00"} {"global_step": 3730, "acc_step": 0, "speed/wps": 12892.87398856956, "speed/FLOPS": 202500302655919.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1292152851819992, "optim/lr": 0.0022374, "optim/total_tokens": 1955594240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.118816375732422, "created_at": "2025-01-15T16:05:12.841335+00:00"} {"global_step": 3731, "acc_step": 0, "speed/wps": 12892.181692015978, "speed/FLOPS": 202489429187229.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11143338680267334, "optim/lr": 0.002238, "optim/total_tokens": 1956118528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.250505208969116, "created_at": "2025-01-15T16:05:23.011520+00:00"} {"global_step": 3732, "acc_step": 0, "speed/wps": 12893.393594355532, "speed/FLOPS": 202508463778800.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11946337670087814, "optim/lr": 0.0022386, "optim/total_tokens": 1956642816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.25651478767395, "created_at": "2025-01-15T16:05:33.182228+00:00"} {"global_step": 3733, "acc_step": 0, "speed/wps": 12883.322324979463, "speed/FLOPS": 202350280653875.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1238318681716919, "optim/lr": 0.0022392, "optim/total_tokens": 1957167104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 3.2251620292663574, "created_at": "2025-01-15T16:05:43.362245+00:00"} {"global_step": 3734, "acc_step": 0, "speed/wps": 12893.352766189635, "speed/FLOPS": 202507822516351.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11083023250102997, "optim/lr": 0.0022398, "optim/total_tokens": 1957691392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.248485565185547, "created_at": "2025-01-15T16:05:53.531849+00:00"} {"global_step": 3735, "acc_step": 0, "speed/wps": 12894.133778889964, "speed/FLOPS": 202520089394034.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14943768084049225, "optim/lr": 0.0022404, "optim/total_tokens": 1958215680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.350135087966919, "created_at": "2025-01-15T16:06:03.701361+00:00"} {"global_step": 3736, "acc_step": 0, "speed/wps": 12895.139652933918, "speed/FLOPS": 202535888028110.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1286008208990097, "optim/lr": 0.002241, "optim/total_tokens": 1958739968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430484, "loss/out": 3.347320556640625, "created_at": "2025-01-15T16:06:13.872426+00:00"} {"global_step": 3737, "acc_step": 0, "speed/wps": 12896.365116599844, "speed/FLOPS": 202555135618947.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09868157655000687, "optim/lr": 0.0022416, "optim/total_tokens": 1959264256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 3.1796822547912598, "created_at": "2025-01-15T16:06:24.049905+00:00"} {"global_step": 3738, "acc_step": 0, "speed/wps": 12892.073157200832, "speed/FLOPS": 202487724498815.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12105243653059006, "optim/lr": 0.0022421999999999997, "optim/total_tokens": 1959788544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.2285141944885254, "created_at": "2025-01-15T16:06:34.223788+00:00"} {"global_step": 3739, "acc_step": 0, "speed/wps": 12887.832995688503, "speed/FLOPS": 202421126935671.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1092018336057663, "optim/lr": 0.0022428, "optim/total_tokens": 1960312832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.340985059738159, "created_at": "2025-01-15T16:06:44.401407+00:00"} {"global_step": 3740, "acc_step": 0, "speed/wps": 12895.38158973576, "speed/FLOPS": 202539687978039.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11059543490409851, "optim/lr": 0.0022434, "optim/total_tokens": 1960837120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.211146831512451, "created_at": "2025-01-15T16:06:54.575129+00:00"} {"global_step": 3741, "acc_step": 0, "speed/wps": 12894.953155717256, "speed/FLOPS": 202532958833048.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10752817243337631, "optim/lr": 0.002244, "optim/total_tokens": 1961361408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 3.1730823516845703, "created_at": "2025-01-15T16:07:04.744913+00:00"} {"global_step": 3742, "acc_step": 0, "speed/wps": 12892.680785787597, "speed/FLOPS": 202497268140739.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10407787561416626, "optim/lr": 0.0022446, "optim/total_tokens": 1961885696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.353231191635132, "created_at": "2025-01-15T16:07:14.919477+00:00"} {"global_step": 3743, "acc_step": 0, "speed/wps": 12889.98121326245, "speed/FLOPS": 202454867644631.1, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12072288244962692, "optim/lr": 0.0022451999999999997, "optim/total_tokens": 1962409984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 3.217921733856201, "created_at": "2025-01-15T16:07:25.095239+00:00"} {"global_step": 3744, "acc_step": 0, "speed/wps": 12895.55721680158, "speed/FLOPS": 202542446442444.94, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11661528795957565, "optim/lr": 0.0022458, "optim/total_tokens": 1962934272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.1824424266815186, "created_at": "2025-01-15T16:07:35.263060+00:00"} {"global_step": 3745, "acc_step": 0, "speed/wps": 12892.979104511865, "speed/FLOPS": 202501953646237.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11987069249153137, "optim/lr": 0.0022464, "optim/total_tokens": 1963458560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371444, "loss/out": 3.255138874053955, "created_at": "2025-01-15T16:07:45.432715+00:00"} {"global_step": 3746, "acc_step": 0, "speed/wps": 12895.219055335554, "speed/FLOPS": 202537135151938.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10620944201946259, "optim/lr": 0.002247, "optim/total_tokens": 1963982848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 3.2701284885406494, "created_at": "2025-01-15T16:07:55.600741+00:00"} {"global_step": 3747, "acc_step": 0, "speed/wps": 12894.889936905225, "speed/FLOPS": 202531965894734.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12296103686094284, "optim/lr": 0.0022475999999999998, "optim/total_tokens": 1964507136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.3379855155944824, "created_at": "2025-01-15T16:08:05.769795+00:00"} {"global_step": 3748, "acc_step": 0, "speed/wps": 12892.037381253522, "speed/FLOPS": 202487162588401.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1446569710969925, "optim/lr": 0.0022482, "optim/total_tokens": 1965031424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 278279, "loss/out": 3.2436981201171875, "created_at": "2025-01-15T16:08:15.942061+00:00"} {"global_step": 3749, "acc_step": 0, "speed/wps": 12895.833172355402, "speed/FLOPS": 202546780703620.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.13942034542560577, "optim/lr": 0.0022488, "optim/total_tokens": 1965555712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.2292041778564453, "created_at": "2025-01-15T16:08:26.114807+00:00"} {"global_step": 3750, "acc_step": 0, "speed/wps": 12895.339452807162, "speed/FLOPS": 202539026159673.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11621500551700592, "optim/lr": 0.0022494, "optim/total_tokens": 1966080000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457124, "loss/out": 3.1901984214782715, "created_at": "2025-01-15T16:08:36.283108+00:00"} {"global_step": 3751, "acc_step": 0, "speed/wps": 12891.489087208744, "speed/FLOPS": 202478550876993.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10080941021442413, "optim/lr": 0.0022500000000000003, "optim/total_tokens": 1966604288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.2720980644226074, "created_at": "2025-01-15T16:08:46.457213+00:00"} {"global_step": 3752, "acc_step": 0, "speed/wps": 12892.111689726647, "speed/FLOPS": 202488329705082.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11384604871273041, "optim/lr": 0.0022506, "optim/total_tokens": 1967128576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 3.247130870819092, "created_at": "2025-01-15T16:08:56.629357+00:00"} {"global_step": 3753, "acc_step": 0, "speed/wps": 12891.815519072798, "speed/FLOPS": 202483677937982.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1201319471001625, "optim/lr": 0.0022512, "optim/total_tokens": 1967652864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.332892417907715, "created_at": "2025-01-15T16:09:06.800709+00:00"} {"global_step": 3754, "acc_step": 0, "speed/wps": 12892.928318593988, "speed/FLOPS": 202501155983610.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11218151450157166, "optim/lr": 0.0022518, "optim/total_tokens": 1968177152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.1612484455108643, "created_at": "2025-01-15T16:09:16.972511+00:00"} {"global_step": 3755, "acc_step": 0, "speed/wps": 12895.906173950381, "speed/FLOPS": 202547927293984.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1309199333190918, "optim/lr": 0.0022524000000000003, "optim/total_tokens": 1968701440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 3.2441067695617676, "created_at": "2025-01-15T16:09:27.148923+00:00"} {"global_step": 3756, "acc_step": 0, "speed/wps": 12893.124276158587, "speed/FLOPS": 202504233766435.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11896970868110657, "optim/lr": 0.0022530000000000002, "optim/total_tokens": 1969225728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 3.460012197494507, "created_at": "2025-01-15T16:09:37.321317+00:00"} {"global_step": 3757, "acc_step": 0, "speed/wps": 12896.20234469669, "speed/FLOPS": 202552579062536.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12557952105998993, "optim/lr": 0.0022536, "optim/total_tokens": 1969750016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.225668430328369, "created_at": "2025-01-15T16:09:47.494394+00:00"} {"global_step": 3758, "acc_step": 0, "speed/wps": 12890.335459038968, "speed/FLOPS": 202460431561336.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12329055368900299, "optim/lr": 0.0022542, "optim/total_tokens": 1970274304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.28249454498291, "created_at": "2025-01-15T16:09:57.667158+00:00"} {"global_step": 3759, "acc_step": 0, "speed/wps": 12893.978902480005, "speed/FLOPS": 202517656847193.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10635960847139359, "optim/lr": 0.0022548000000000004, "optim/total_tokens": 1970798592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.258382558822632, "created_at": "2025-01-15T16:10:07.837145+00:00"} {"global_step": 3760, "acc_step": 0, "speed/wps": 12892.637611816384, "speed/FLOPS": 202496590034194.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10558196902275085, "optim/lr": 0.0022554000000000003, "optim/total_tokens": 1971322880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 3.2206668853759766, "created_at": "2025-01-15T16:10:18.008068+00:00"} {"global_step": 3761, "acc_step": 0, "speed/wps": 12890.852688849118, "speed/FLOPS": 202468555366252.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11093905568122864, "optim/lr": 0.002256, "optim/total_tokens": 1971847168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.1208415031433105, "created_at": "2025-01-15T16:10:28.179399+00:00"} {"global_step": 3762, "acc_step": 0, "speed/wps": 12895.442993736877, "speed/FLOPS": 202540652412254.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.09637980908155441, "optim/lr": 0.0022566, "optim/total_tokens": 1972371456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.3100368976593018, "created_at": "2025-01-15T16:10:38.350811+00:00"} {"global_step": 3763, "acc_step": 0, "speed/wps": 12894.976936774925, "speed/FLOPS": 202533332347238.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10961141437292099, "optim/lr": 0.0022572, "optim/total_tokens": 1972895744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 3.1771187782287598, "created_at": "2025-01-15T16:10:48.519931+00:00"} {"global_step": 3764, "acc_step": 0, "speed/wps": 12895.686745078785, "speed/FLOPS": 202544480862028.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11333981156349182, "optim/lr": 0.0022578000000000003, "optim/total_tokens": 1973420032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.2252635955810547, "created_at": "2025-01-15T16:10:58.687528+00:00"} {"global_step": 3765, "acc_step": 0, "speed/wps": 12893.79119495252, "speed/FLOPS": 202514708642537.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11179382354021072, "optim/lr": 0.0022584000000000003, "optim/total_tokens": 1973944320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.338122844696045, "created_at": "2025-01-15T16:11:08.859938+00:00"} {"global_step": 3766, "acc_step": 0, "speed/wps": 12891.329588808452, "speed/FLOPS": 202476045735443.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1054639220237732, "optim/lr": 0.002259, "optim/total_tokens": 1974468608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 3.278252124786377, "created_at": "2025-01-15T16:11:19.036515+00:00"} {"global_step": 3767, "acc_step": 0, "speed/wps": 12896.862996762373, "speed/FLOPS": 202562955511058.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1015431135892868, "optim/lr": 0.0022596, "optim/total_tokens": 1974992896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.1815407276153564, "created_at": "2025-01-15T16:11:29.203752+00:00"} {"global_step": 3768, "acc_step": 0, "speed/wps": 12895.199811656721, "speed/FLOPS": 202536832903518.78, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12161571532487869, "optim/lr": 0.0022602, "optim/total_tokens": 1975517184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.2864651679992676, "created_at": "2025-01-15T16:11:39.372941+00:00"} {"global_step": 3769, "acc_step": 0, "speed/wps": 12893.353478450807, "speed/FLOPS": 202507833703392.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12768959999084473, "optim/lr": 0.0022608000000000003, "optim/total_tokens": 1976041472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 3.2191219329833984, "created_at": "2025-01-15T16:11:49.546204+00:00"} {"global_step": 3770, "acc_step": 0, "speed/wps": 12897.284193176058, "speed/FLOPS": 202569570979520.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13654202222824097, "optim/lr": 0.0022614, "optim/total_tokens": 1976565760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 3.318084239959717, "created_at": "2025-01-15T16:11:59.725072+00:00"} {"global_step": 3771, "acc_step": 0, "speed/wps": 12893.568711317728, "speed/FLOPS": 202511214231328.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12957927584648132, "optim/lr": 0.002262, "optim/total_tokens": 1977090048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.197035312652588, "created_at": "2025-01-15T16:12:09.894256+00:00"} {"global_step": 3772, "acc_step": 0, "speed/wps": 12891.611218238351, "speed/FLOPS": 202480469112639.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12458530813455582, "optim/lr": 0.0022626, "optim/total_tokens": 1977614336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.2833924293518066, "created_at": "2025-01-15T16:12:20.068425+00:00"} {"global_step": 3773, "acc_step": 0, "speed/wps": 12892.956540292626, "speed/FLOPS": 202501599244167.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12790478765964508, "optim/lr": 0.0022632, "optim/total_tokens": 1978138624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2259087562561035, "created_at": "2025-01-15T16:12:30.240671+00:00"} {"global_step": 3774, "acc_step": 0, "speed/wps": 12896.757180479464, "speed/FLOPS": 202561293520929.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10405456274747849, "optim/lr": 0.0022638000000000003, "optim/total_tokens": 1978662912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444353, "loss/out": 3.3304176330566406, "created_at": "2025-01-15T16:12:40.408174+00:00"} {"global_step": 3775, "acc_step": 0, "speed/wps": 12890.47678475127, "speed/FLOPS": 202462651275849.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12095799297094345, "optim/lr": 0.0022644, "optim/total_tokens": 1979187200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.3543896675109863, "created_at": "2025-01-15T16:12:50.579946+00:00"} {"global_step": 3776, "acc_step": 0, "speed/wps": 12897.513225449133, "speed/FLOPS": 202573168246092.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12283844500780106, "optim/lr": 0.002265, "optim/total_tokens": 1979711488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.4319217205047607, "created_at": "2025-01-15T16:13:00.747463+00:00"} {"global_step": 3777, "acc_step": 0, "speed/wps": 12897.53720624638, "speed/FLOPS": 202573544897466.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12847228348255157, "optim/lr": 0.0022656, "optim/total_tokens": 1980235776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 3.2572124004364014, "created_at": "2025-01-15T16:13:10.913516+00:00"} {"global_step": 3778, "acc_step": 0, "speed/wps": 12891.213408460077, "speed/FLOPS": 202474220963423.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11740673333406448, "optim/lr": 0.0022662, "optim/total_tokens": 1980760064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.309838056564331, "created_at": "2025-01-15T16:13:21.084645+00:00"} {"global_step": 3779, "acc_step": 0, "speed/wps": 12888.133404559896, "speed/FLOPS": 202425845269802.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.099556103348732, "optim/lr": 0.0022668000000000002, "optim/total_tokens": 1981284352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 3.275303840637207, "created_at": "2025-01-15T16:13:31.266619+00:00"} {"global_step": 3780, "acc_step": 0, "speed/wps": 12896.982011427843, "speed/FLOPS": 202564824799923.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0951492041349411, "optim/lr": 0.0022674, "optim/total_tokens": 1981808640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.3430981636047363, "created_at": "2025-01-15T16:13:41.434207+00:00"} {"global_step": 3781, "acc_step": 0, "speed/wps": 12893.62881758391, "speed/FLOPS": 202512158282833.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09351859241724014, "optim/lr": 0.002268, "optim/total_tokens": 1982332928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.2178702354431152, "created_at": "2025-01-15T16:13:51.603533+00:00"} {"global_step": 3782, "acc_step": 0, "speed/wps": 12893.71031111306, "speed/FLOPS": 202513438250692.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09951908141374588, "optim/lr": 0.0022686, "optim/total_tokens": 1982857216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 3.336019992828369, "created_at": "2025-01-15T16:14:01.776749+00:00"} {"global_step": 3783, "acc_step": 0, "speed/wps": 12897.17959088107, "speed/FLOPS": 202567928056738.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10573922097682953, "optim/lr": 0.0022692, "optim/total_tokens": 1983381504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.3654580116271973, "created_at": "2025-01-15T16:14:11.943685+00:00"} {"global_step": 3784, "acc_step": 0, "speed/wps": 12894.51993591962, "speed/FLOPS": 202526154520820.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10358447581529617, "optim/lr": 0.0022698, "optim/total_tokens": 1983905792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361859, "loss/out": 3.3770017623901367, "created_at": "2025-01-15T16:14:22.113233+00:00"} {"global_step": 3785, "acc_step": 0, "speed/wps": 12888.8573535554, "speed/FLOPS": 202437215883586.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11034303903579712, "optim/lr": 0.0022704, "optim/total_tokens": 1984430080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 3.206477642059326, "created_at": "2025-01-15T16:14:32.288723+00:00"} {"global_step": 3786, "acc_step": 0, "speed/wps": 12889.539389919, "speed/FLOPS": 202447928201893.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10381721705198288, "optim/lr": 0.002271, "optim/total_tokens": 1984954368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.256941318511963, "created_at": "2025-01-15T16:14:42.461605+00:00"} {"global_step": 3787, "acc_step": 0, "speed/wps": 12892.491259461187, "speed/FLOPS": 202494291369347.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11556588858366013, "optim/lr": 0.0022716, "optim/total_tokens": 1985478656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.2031781673431396, "created_at": "2025-01-15T16:14:52.636870+00:00"} {"global_step": 3788, "acc_step": 0, "speed/wps": 12893.391992643932, "speed/FLOPS": 202508438621719.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1309482902288437, "optim/lr": 0.0022722, "optim/total_tokens": 1986002944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 3.161033868789673, "created_at": "2025-01-15T16:15:02.806142+00:00"} {"global_step": 3789, "acc_step": 0, "speed/wps": 12891.816916094524, "speed/FLOPS": 202483699880128.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13633130490779877, "optim/lr": 0.0022728, "optim/total_tokens": 1986527232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.269164562225342, "created_at": "2025-01-15T16:15:12.977452+00:00"} {"global_step": 3790, "acc_step": 0, "speed/wps": 12894.104601063651, "speed/FLOPS": 202519631116177.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14796686172485352, "optim/lr": 0.0022734, "optim/total_tokens": 1987051520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.2484869956970215, "created_at": "2025-01-15T16:15:23.147498+00:00"} {"global_step": 3791, "acc_step": 0, "speed/wps": 12897.80177630118, "speed/FLOPS": 202577700333734.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16066807508468628, "optim/lr": 0.002274, "optim/total_tokens": 1987575808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.303575038909912, "created_at": "2025-01-15T16:15:33.313434+00:00"} {"global_step": 3792, "acc_step": 0, "speed/wps": 12893.062916679139, "speed/FLOPS": 202503270031494.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1259692907333374, "optim/lr": 0.0022746, "optim/total_tokens": 1988100096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316449, "loss/out": 3.295736789703369, "created_at": "2025-01-15T16:15:43.483109+00:00"} {"global_step": 3793, "acc_step": 0, "speed/wps": 12893.7177058138, "speed/FLOPS": 202513554394628.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10120700299739838, "optim/lr": 0.0022752, "optim/total_tokens": 1988624384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.0679335594177246, "created_at": "2025-01-15T16:15:53.655330+00:00"} {"global_step": 3794, "acc_step": 0, "speed/wps": 12893.291965598106, "speed/FLOPS": 202506867559513.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.12017938494682312, "optim/lr": 0.0022758, "optim/total_tokens": 1989148672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 3.196024179458618, "created_at": "2025-01-15T16:16:03.827625+00:00"} {"global_step": 3795, "acc_step": 0, "speed/wps": 12897.379846376258, "speed/FLOPS": 202571073344470.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10473216325044632, "optim/lr": 0.0022764, "optim/total_tokens": 1989672960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 3.1821300983428955, "created_at": "2025-01-15T16:16:13.995257+00:00"} {"global_step": 3796, "acc_step": 0, "speed/wps": 12900.88023407978, "speed/FLOPS": 202626051743391.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11789174377918243, "optim/lr": 0.002277, "optim/total_tokens": 1990197248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 3.178976058959961, "created_at": "2025-01-15T16:16:24.163800+00:00"} {"global_step": 3797, "acc_step": 0, "speed/wps": 12884.91808065523, "speed/FLOPS": 202375344189561.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11956316232681274, "optim/lr": 0.0022776, "optim/total_tokens": 1990721536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.2815937995910645, "created_at": "2025-01-15T16:16:34.341681+00:00"} {"global_step": 3798, "acc_step": 0, "speed/wps": 12885.460230941339, "speed/FLOPS": 202383859404797.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1121697798371315, "optim/lr": 0.0022781999999999998, "optim/total_tokens": 1991245824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.310105323791504, "created_at": "2025-01-15T16:16:44.518941+00:00"} {"global_step": 3799, "acc_step": 0, "speed/wps": 12895.785069909247, "speed/FLOPS": 202546025188603.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1182873323559761, "optim/lr": 0.0022788, "optim/total_tokens": 1991770112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 3.1456832885742188, "created_at": "2025-01-15T16:16:54.687045+00:00"} {"global_step": 3800, "acc_step": 0, "speed/wps": 12896.863164583536, "speed/FLOPS": 202562958146920.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11581022292375565, "optim/lr": 0.0022794, "optim/total_tokens": 1992294400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.2452268600463867, "created_at": "2025-01-15T16:17:04.856929+00:00"} {"global_step": 3801, "acc_step": 0, "speed/wps": 12895.493421359375, "speed/FLOPS": 202541444447361.0, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11817804723978043, "optim/lr": 0.00228, "optim/total_tokens": 1992818688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.1756250858306885, "created_at": "2025-01-15T16:17:15.025921+00:00"} {"global_step": 3802, "acc_step": 0, "speed/wps": 12894.77772114168, "speed/FLOPS": 202530203391966.9, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10426487028598785, "optim/lr": 0.0022806, "optim/total_tokens": 1993342976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 3.2364115715026855, "created_at": "2025-01-15T16:17:25.194731+00:00"} {"global_step": 3803, "acc_step": 0, "speed/wps": 12892.863931441194, "speed/FLOPS": 202500144694898.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.09235063195228577, "optim/lr": 0.0022812, "optim/total_tokens": 1993867264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.2979183197021484, "created_at": "2025-01-15T16:17:35.365011+00:00"} {"global_step": 3804, "acc_step": 0, "speed/wps": 12890.425656082376, "speed/FLOPS": 202461848229846.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12690556049346924, "optim/lr": 0.0022818, "optim/total_tokens": 1994391552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 3.337023973464966, "created_at": "2025-01-15T16:17:45.540296+00:00"} {"global_step": 3805, "acc_step": 0, "speed/wps": 12895.187972693306, "speed/FLOPS": 202536646956329.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15001937747001648, "optim/lr": 0.0022824, "optim/total_tokens": 1994915840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.28916597366333, "created_at": "2025-01-15T16:17:55.708292+00:00"} {"global_step": 3806, "acc_step": 0, "speed/wps": 12893.228413446037, "speed/FLOPS": 202505869385634.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1266874521970749, "optim/lr": 0.002283, "optim/total_tokens": 1995440128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370389, "loss/out": 3.320556640625, "created_at": "2025-01-15T16:18:05.879929+00:00"} {"global_step": 3807, "acc_step": 0, "speed/wps": 12891.568376169156, "speed/FLOPS": 202479796219071.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1098770871758461, "optim/lr": 0.0022836, "optim/total_tokens": 1995964416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.3429975509643555, "created_at": "2025-01-15T16:18:16.052923+00:00"} {"global_step": 3808, "acc_step": 0, "speed/wps": 12897.250825399791, "speed/FLOPS": 202569046892741.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1476297676563263, "optim/lr": 0.0022842, "optim/total_tokens": 1996488704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.3424453735351562, "created_at": "2025-01-15T16:18:26.222122+00:00"} {"global_step": 3809, "acc_step": 0, "speed/wps": 12893.872099888724, "speed/FLOPS": 202515979365735.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13266944885253906, "optim/lr": 0.0022848, "optim/total_tokens": 1997012992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.3033254146575928, "created_at": "2025-01-15T16:18:36.392335+00:00"} {"global_step": 3810, "acc_step": 0, "speed/wps": 12893.362913851679, "speed/FLOPS": 202507981899328.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1159408763051033, "optim/lr": 0.0022854, "optim/total_tokens": 1997537280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294366, "loss/out": 3.3975906372070312, "created_at": "2025-01-15T16:18:46.565048+00:00"} {"global_step": 3811, "acc_step": 0, "speed/wps": 12897.002575841572, "speed/FLOPS": 202565147792299.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11764052510261536, "optim/lr": 0.0022860000000000003, "optim/total_tokens": 1998061568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.175070285797119, "created_at": "2025-01-15T16:18:56.732717+00:00"} {"global_step": 3812, "acc_step": 0, "speed/wps": 12895.762081112673, "speed/FLOPS": 202545664117963.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13096919655799866, "optim/lr": 0.0022866, "optim/total_tokens": 1998585856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.4133293628692627, "created_at": "2025-01-15T16:19:06.901697+00:00"} {"global_step": 3813, "acc_step": 0, "speed/wps": 12894.800390978682, "speed/FLOPS": 202530559452908.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10783526301383972, "optim/lr": 0.0022872, "optim/total_tokens": 1999110144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 3.393855333328247, "created_at": "2025-01-15T16:19:17.072065+00:00"} {"global_step": 3814, "acc_step": 0, "speed/wps": 12894.643909527638, "speed/FLOPS": 202528101696691.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11810184270143509, "optim/lr": 0.0022878, "optim/total_tokens": 1999634432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.229942798614502, "created_at": "2025-01-15T16:19:27.243186+00:00"} {"global_step": 3815, "acc_step": 0, "speed/wps": 12897.204949716946, "speed/FLOPS": 202568326352103.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13754183053970337, "optim/lr": 0.0022884000000000003, "optim/total_tokens": 2000158720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.25761079788208, "created_at": "2025-01-15T16:19:37.416510+00:00"} {"global_step": 3816, "acc_step": 0, "speed/wps": 12895.811935768901, "speed/FLOPS": 202546447153847.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12018275260925293, "optim/lr": 0.0022890000000000002, "optim/total_tokens": 2000683008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.156965732574463, "created_at": "2025-01-15T16:19:47.587681+00:00"} {"global_step": 3817, "acc_step": 0, "speed/wps": 12895.672209943354, "speed/FLOPS": 202544252567753.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10683124512434006, "optim/lr": 0.0022896, "optim/total_tokens": 2001207296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 3.337714672088623, "created_at": "2025-01-15T16:19:57.757764+00:00"} {"global_step": 3818, "acc_step": 0, "speed/wps": 12894.915847941993, "speed/FLOPS": 202532372863171.97, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12473238259553909, "optim/lr": 0.0022902, "optim/total_tokens": 2001731584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 3.2443156242370605, "created_at": "2025-01-15T16:20:07.927920+00:00"} {"global_step": 3819, "acc_step": 0, "speed/wps": 12896.77272379178, "speed/FLOPS": 202561537650007.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1175495907664299, "optim/lr": 0.0022908, "optim/total_tokens": 2002255872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 3.2691640853881836, "created_at": "2025-01-15T16:20:18.095325+00:00"} {"global_step": 3820, "acc_step": 0, "speed/wps": 12899.949941794806, "speed/FLOPS": 202611440224700.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09947525709867477, "optim/lr": 0.0022914000000000003, "optim/total_tokens": 2002780160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 3.2838134765625, "created_at": "2025-01-15T16:20:28.259910+00:00"} {"global_step": 3821, "acc_step": 0, "speed/wps": 12901.391012223956, "speed/FLOPS": 202634074216028.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09064259380102158, "optim/lr": 0.002292, "optim/total_tokens": 2003304448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.2846603393554688, "created_at": "2025-01-15T16:20:38.424514+00:00"} {"global_step": 3822, "acc_step": 0, "speed/wps": 12899.252319195686, "speed/FLOPS": 202600483103147.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09819374233484268, "optim/lr": 0.0022926, "optim/total_tokens": 2003828736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2242116928100586, "created_at": "2025-01-15T16:20:48.589595+00:00"} {"global_step": 3823, "acc_step": 0, "speed/wps": 12894.683297425796, "speed/FLOPS": 202528720337757.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10399765521287918, "optim/lr": 0.0022932, "optim/total_tokens": 2004353024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400071, "loss/out": 3.1544063091278076, "created_at": "2025-01-15T16:20:58.762364+00:00"} {"global_step": 3824, "acc_step": 0, "speed/wps": 12898.46182268582, "speed/FLOPS": 202588067269203.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11590327322483063, "optim/lr": 0.0022938, "optim/total_tokens": 2004877312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 3.377732753753662, "created_at": "2025-01-15T16:21:08.931591+00:00"} {"global_step": 3825, "acc_step": 0, "speed/wps": 12895.645618800796, "speed/FLOPS": 202543834917319.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1031484380364418, "optim/lr": 0.0022944000000000003, "optim/total_tokens": 2005401600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.2506356239318848, "created_at": "2025-01-15T16:21:19.103026+00:00"} {"global_step": 3826, "acc_step": 0, "speed/wps": 12896.253964712601, "speed/FLOPS": 202553389825819.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09207774698734283, "optim/lr": 0.002295, "optim/total_tokens": 2005925888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.220386028289795, "created_at": "2025-01-15T16:21:29.270306+00:00"} {"global_step": 3827, "acc_step": 0, "speed/wps": 12896.675668521644, "speed/FLOPS": 202560013263622.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09104743599891663, "optim/lr": 0.0022956, "optim/total_tokens": 2006450176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449409, "loss/out": 3.310576915740967, "created_at": "2025-01-15T16:21:39.437236+00:00"} {"global_step": 3828, "acc_step": 0, "speed/wps": 12897.003378327783, "speed/FLOPS": 202565160396448.16, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10545136034488678, "optim/lr": 0.0022962, "optim/total_tokens": 2006974464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 3.2563259601593018, "created_at": "2025-01-15T16:21:49.605659+00:00"} {"global_step": 3829, "acc_step": 0, "speed/wps": 12893.967042132126, "speed/FLOPS": 202517470564132.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13065341114997864, "optim/lr": 0.0022968, "optim/total_tokens": 2007498752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.132681369781494, "created_at": "2025-01-15T16:21:59.775633+00:00"} {"global_step": 3830, "acc_step": 0, "speed/wps": 12898.544495631444, "speed/FLOPS": 202589365761419.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11275030672550201, "optim/lr": 0.0022974000000000002, "optim/total_tokens": 2008023040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366944, "loss/out": 3.229505777359009, "created_at": "2025-01-15T16:22:09.950276+00:00"} {"global_step": 3831, "acc_step": 0, "speed/wps": 12902.455065833203, "speed/FLOPS": 202650786640124.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10892900824546814, "optim/lr": 0.002298, "optim/total_tokens": 2008547328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 3.2425265312194824, "created_at": "2025-01-15T16:22:20.113523+00:00"} {"global_step": 3832, "acc_step": 0, "speed/wps": 12899.361483311513, "speed/FLOPS": 202602197675593.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1236526221036911, "optim/lr": 0.0022986, "optim/total_tokens": 2009071616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456028, "loss/out": 3.0245609283447266, "created_at": "2025-01-15T16:22:30.282184+00:00"} {"global_step": 3833, "acc_step": 0, "speed/wps": 12893.259455931673, "speed/FLOPS": 202506356950528.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15783889591693878, "optim/lr": 0.0022992, "optim/total_tokens": 2009595904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 3.3197035789489746, "created_at": "2025-01-15T16:22:40.451973+00:00"} {"global_step": 3834, "acc_step": 0, "speed/wps": 12899.049284032779, "speed/FLOPS": 202597294156914.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1015099436044693, "optim/lr": 0.0022998, "optim/total_tokens": 2010120192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.286639928817749, "created_at": "2025-01-15T16:22:50.617175+00:00"} {"global_step": 3835, "acc_step": 0, "speed/wps": 12890.418772840298, "speed/FLOPS": 202461740119071.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.1856359988451004, "optim/lr": 0.0023004, "optim/total_tokens": 2010644480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 3.300279140472412, "created_at": "2025-01-15T16:23:00.790928+00:00"} {"global_step": 3836, "acc_step": 0, "speed/wps": 12888.239592788274, "speed/FLOPS": 202427513101846.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13070084154605865, "optim/lr": 0.002301, "optim/total_tokens": 2011168768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.2630558013916016, "created_at": "2025-01-15T16:23:10.965389+00:00"} {"global_step": 3837, "acc_step": 0, "speed/wps": 12894.96282266881, "speed/FLOPS": 202533110665806.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10819584876298904, "optim/lr": 0.0023016, "optim/total_tokens": 2011693056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370300, "loss/out": 3.2867279052734375, "created_at": "2025-01-15T16:23:21.136261+00:00"} {"global_step": 3838, "acc_step": 0, "speed/wps": 12900.001128309927, "speed/FLOPS": 202612244179257.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1296626776456833, "optim/lr": 0.0023022, "optim/total_tokens": 2012217344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 3.1858348846435547, "created_at": "2025-01-15T16:23:31.301761+00:00"} {"global_step": 3839, "acc_step": 0, "speed/wps": 12896.500770869743, "speed/FLOPS": 202557266255664.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1297387331724167, "optim/lr": 0.0023028, "optim/total_tokens": 2012741632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.305541753768921, "created_at": "2025-01-15T16:23:41.469035+00:00"} {"global_step": 3840, "acc_step": 0, "speed/wps": 12896.729686683006, "speed/FLOPS": 202560861693075.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14958465099334717, "optim/lr": 0.0023034, "optim/total_tokens": 2013265920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 3.2337136268615723, "created_at": "2025-01-15T16:23:51.638479+00:00"} {"global_step": 3841, "acc_step": 0, "speed/wps": 12891.403031833534, "speed/FLOPS": 202477199259073.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.13042466342449188, "optim/lr": 0.002304, "optim/total_tokens": 2013790208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 3.146231174468994, "created_at": "2025-01-15T16:24:01.809391+00:00"} {"global_step": 3842, "acc_step": 0, "speed/wps": 12896.775252892676, "speed/FLOPS": 202561577373012.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1867581009864807, "optim/lr": 0.0023046, "optim/total_tokens": 2014314496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.315854787826538, "created_at": "2025-01-15T16:24:11.976922+00:00"} {"global_step": 3843, "acc_step": 0, "speed/wps": 12892.202687536583, "speed/FLOPS": 202489758950731.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1721564680337906, "optim/lr": 0.0023052, "optim/total_tokens": 2014838784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.335155725479126, "created_at": "2025-01-15T16:24:22.148664+00:00"} {"global_step": 3844, "acc_step": 0, "speed/wps": 12894.96101853241, "speed/FLOPS": 202533082329365.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14342817664146423, "optim/lr": 0.0023057999999999998, "optim/total_tokens": 2015363072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 3.1719329357147217, "created_at": "2025-01-15T16:24:32.316746+00:00"} {"global_step": 3845, "acc_step": 0, "speed/wps": 12893.917213552402, "speed/FLOPS": 202516687937815.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11374730616807938, "optim/lr": 0.0023064, "optim/total_tokens": 2015887360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.3576483726501465, "created_at": "2025-01-15T16:24:42.486020+00:00"} {"global_step": 3846, "acc_step": 0, "speed/wps": 12894.4586080485, "speed/FLOPS": 202525191282331.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12134069204330444, "optim/lr": 0.002307, "optim/total_tokens": 2016411648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.373093605041504, "created_at": "2025-01-15T16:24:52.658882+00:00"} {"global_step": 3847, "acc_step": 0, "speed/wps": 12895.28531041293, "speed/FLOPS": 202538175778972.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11531724780797958, "optim/lr": 0.0023076, "optim/total_tokens": 2016935936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 3.204716444015503, "created_at": "2025-01-15T16:25:02.828829+00:00"} {"global_step": 3848, "acc_step": 0, "speed/wps": 12898.487677151092, "speed/FLOPS": 202588473349108.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10279341787099838, "optim/lr": 0.0023082, "optim/total_tokens": 2017460224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.226780652999878, "created_at": "2025-01-15T16:25:12.995958+00:00"} {"global_step": 3849, "acc_step": 0, "speed/wps": 12899.087159563276, "speed/FLOPS": 202597889044167.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10387909412384033, "optim/lr": 0.0023087999999999997, "optim/total_tokens": 2017984512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.2507801055908203, "created_at": "2025-01-15T16:25:23.163685+00:00"} {"global_step": 3850, "acc_step": 0, "speed/wps": 12897.70306257146, "speed/FLOPS": 202576149898962.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09312066435813904, "optim/lr": 0.0023094, "optim/total_tokens": 2018508800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 3.228438377380371, "created_at": "2025-01-15T16:25:33.332724+00:00"} {"global_step": 3851, "acc_step": 0, "speed/wps": 12900.78464957251, "speed/FLOPS": 202624550457357.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.104599229991436, "optim/lr": 0.00231, "optim/total_tokens": 2019033088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.1814494132995605, "created_at": "2025-01-15T16:25:43.497460+00:00"} {"global_step": 3852, "acc_step": 0, "speed/wps": 12894.405383901321, "speed/FLOPS": 202524355323962.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12343605607748032, "optim/lr": 0.0023106, "optim/total_tokens": 2019557376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 3.305135726928711, "created_at": "2025-01-15T16:25:53.666269+00:00"} {"global_step": 3853, "acc_step": 0, "speed/wps": 12897.2114088137, "speed/FLOPS": 202568427801093.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1183677390217781, "optim/lr": 0.0023112, "optim/total_tokens": 2020081664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.2256035804748535, "created_at": "2025-01-15T16:26:03.832707+00:00"} {"global_step": 3854, "acc_step": 0, "speed/wps": 12893.208158721109, "speed/FLOPS": 202505551257347.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12951181828975677, "optim/lr": 0.0023117999999999997, "optim/total_tokens": 2020605952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.268603801727295, "created_at": "2025-01-15T16:26:14.005718+00:00"} {"global_step": 3855, "acc_step": 0, "speed/wps": 12897.203995922238, "speed/FLOPS": 202568311371446.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10890361666679382, "optim/lr": 0.0023124, "optim/total_tokens": 2021130240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.3440017700195312, "created_at": "2025-01-15T16:26:24.173517+00:00"} {"global_step": 3856, "acc_step": 0, "speed/wps": 12895.932304135056, "speed/FLOPS": 202548337704441.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13977651298046112, "optim/lr": 0.002313, "optim/total_tokens": 2021654528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.1898133754730225, "created_at": "2025-01-15T16:26:34.344573+00:00"} {"global_step": 3857, "acc_step": 0, "speed/wps": 12899.001743977626, "speed/FLOPS": 202596547475021.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12919847667217255, "optim/lr": 0.0023136, "optim/total_tokens": 2022178816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.2055630683898926, "created_at": "2025-01-15T16:26:44.509601+00:00"} {"global_step": 3858, "acc_step": 0, "speed/wps": 12894.79808949274, "speed/FLOPS": 202530523304909.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08963153511285782, "optim/lr": 0.0023141999999999998, "optim/total_tokens": 2022703104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.1945343017578125, "created_at": "2025-01-15T16:26:54.679830+00:00"} {"global_step": 3859, "acc_step": 0, "speed/wps": 12898.093041081844, "speed/FLOPS": 202582275047354.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10524740070104599, "optim/lr": 0.0023148, "optim/total_tokens": 2023227392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.339308738708496, "created_at": "2025-01-15T16:27:04.849323+00:00"} {"global_step": 3860, "acc_step": 0, "speed/wps": 12893.536377440118, "speed/FLOPS": 202510706383348.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09613019973039627, "optim/lr": 0.0023154, "optim/total_tokens": 2023751680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.1691782474517822, "created_at": "2025-01-15T16:27:15.018963+00:00"} {"global_step": 3861, "acc_step": 0, "speed/wps": 12898.709757220951, "speed/FLOPS": 202591961421775.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11297647655010223, "optim/lr": 0.002316, "optim/total_tokens": 2024275968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.3545241355895996, "created_at": "2025-01-15T16:27:25.184205+00:00"} {"global_step": 3862, "acc_step": 0, "speed/wps": 12895.86222447825, "speed/FLOPS": 202547237007131.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10346385836601257, "optim/lr": 0.0023166000000000003, "optim/total_tokens": 2024800256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.244649887084961, "created_at": "2025-01-15T16:27:35.351570+00:00"} {"global_step": 3863, "acc_step": 0, "speed/wps": 12891.737053702152, "speed/FLOPS": 202482445531514.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12297119200229645, "optim/lr": 0.0023172, "optim/total_tokens": 2025324544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.22719144821167, "created_at": "2025-01-15T16:27:45.522644+00:00"} {"global_step": 3864, "acc_step": 0, "speed/wps": 12899.503248205267, "speed/FLOPS": 202604424288055.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11107650399208069, "optim/lr": 0.0023178, "optim/total_tokens": 2025848832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 3.2975306510925293, "created_at": "2025-01-15T16:27:55.692515+00:00"} {"global_step": 3865, "acc_step": 0, "speed/wps": 12896.425815003611, "speed/FLOPS": 202556088970786.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08845594525337219, "optim/lr": 0.0023184, "optim/total_tokens": 2026373120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.2407939434051514, "created_at": "2025-01-15T16:28:05.866463+00:00"} {"global_step": 3866, "acc_step": 0, "speed/wps": 12893.664151125737, "speed/FLOPS": 202512713244660.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10438954830169678, "optim/lr": 0.0023190000000000003, "optim/total_tokens": 2026897408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.262582778930664, "created_at": "2025-01-15T16:28:16.036079+00:00"} {"global_step": 3867, "acc_step": 0, "speed/wps": 12895.827127034652, "speed/FLOPS": 202546685753550.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10341402888298035, "optim/lr": 0.0023196000000000002, "optim/total_tokens": 2027421696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 3.197819232940674, "created_at": "2025-01-15T16:28:26.205973+00:00"} {"global_step": 3868, "acc_step": 0, "speed/wps": 12898.08121571834, "speed/FLOPS": 202582089313770.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12098883092403412, "optim/lr": 0.0023202, "optim/total_tokens": 2027945984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.252598285675049, "created_at": "2025-01-15T16:28:36.379786+00:00"} {"global_step": 3869, "acc_step": 0, "speed/wps": 12890.452927016642, "speed/FLOPS": 202462276557343.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11217356473207474, "optim/lr": 0.0023208, "optim/total_tokens": 2028470272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.2737176418304443, "created_at": "2025-01-15T16:28:46.551955+00:00"} {"global_step": 3870, "acc_step": 0, "speed/wps": 12894.859070877455, "speed/FLOPS": 202531481101353.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11469075083732605, "optim/lr": 0.0023214000000000004, "optim/total_tokens": 2028994560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.1814706325531006, "created_at": "2025-01-15T16:28:56.724257+00:00"} {"global_step": 3871, "acc_step": 0, "speed/wps": 12890.934015044859, "speed/FLOPS": 202469832705910.7, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10094385594129562, "optim/lr": 0.0023220000000000003, "optim/total_tokens": 2029518848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391306, "loss/out": 3.258361339569092, "created_at": "2025-01-15T16:29:06.895643+00:00"} {"global_step": 3872, "acc_step": 0, "speed/wps": 12895.347783991749, "speed/FLOPS": 202539157012375.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10597240924835205, "optim/lr": 0.0023226, "optim/total_tokens": 2030043136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.237178325653076, "created_at": "2025-01-15T16:29:17.063757+00:00"} {"global_step": 3873, "acc_step": 0, "speed/wps": 12895.310338054187, "speed/FLOPS": 202538568872470.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09921421855688095, "optim/lr": 0.0023232, "optim/total_tokens": 2030567424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.277895927429199, "created_at": "2025-01-15T16:29:27.234177+00:00"} {"global_step": 3874, "acc_step": 0, "speed/wps": 12893.499120386476, "speed/FLOPS": 202510121210124.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10799410939216614, "optim/lr": 0.0023238, "optim/total_tokens": 2031091712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 3.35249924659729, "created_at": "2025-01-15T16:29:37.404650+00:00"} {"global_step": 3875, "acc_step": 0, "speed/wps": 12892.397388877209, "speed/FLOPS": 202492817002834.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09887747466564178, "optim/lr": 0.0023244000000000003, "optim/total_tokens": 2031616000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.245544910430908, "created_at": "2025-01-15T16:29:47.577277+00:00"} {"global_step": 3876, "acc_step": 0, "speed/wps": 12894.873383244169, "speed/FLOPS": 202531705896740.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09193040430545807, "optim/lr": 0.0023250000000000002, "optim/total_tokens": 2032140288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451547, "loss/out": 3.2216179370880127, "created_at": "2025-01-15T16:29:57.747901+00:00"} {"global_step": 3877, "acc_step": 0, "speed/wps": 12894.220483850024, "speed/FLOPS": 202521451214577.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09988734126091003, "optim/lr": 0.0023256, "optim/total_tokens": 2032664576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.2974929809570312, "created_at": "2025-01-15T16:30:07.920664+00:00"} {"global_step": 3878, "acc_step": 0, "speed/wps": 12891.368301099395, "speed/FLOPS": 202476653765168.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10245731472969055, "optim/lr": 0.0023262, "optim/total_tokens": 2033188864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393938, "loss/out": 3.347249746322632, "created_at": "2025-01-15T16:30:18.095106+00:00"} {"global_step": 3879, "acc_step": 0, "speed/wps": 12896.033506572647, "speed/FLOPS": 202549927227790.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10353753715753555, "optim/lr": 0.0023268, "optim/total_tokens": 2033713152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.222329616546631, "created_at": "2025-01-15T16:30:28.262827+00:00"} {"global_step": 3880, "acc_step": 0, "speed/wps": 12893.318426553613, "speed/FLOPS": 202507283165180.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09720878303050995, "optim/lr": 0.0023274000000000003, "optim/total_tokens": 2034237440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.2486424446105957, "created_at": "2025-01-15T16:30:38.439764+00:00"} {"global_step": 3881, "acc_step": 0, "speed/wps": 12893.259225601809, "speed/FLOPS": 202506353332881.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10649455338716507, "optim/lr": 0.002328, "optim/total_tokens": 2034761728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 3.225220203399658, "created_at": "2025-01-15T16:30:48.611353+00:00"} {"global_step": 3882, "acc_step": 0, "speed/wps": 12896.019229444419, "speed/FLOPS": 202549702985872.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.10215030610561371, "optim/lr": 0.0023286, "optim/total_tokens": 2035286016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 3.2515461444854736, "created_at": "2025-01-15T16:30:58.779016+00:00"} {"global_step": 3883, "acc_step": 0, "speed/wps": 12892.414132231777, "speed/FLOPS": 202493079980225.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11319807916879654, "optim/lr": 0.0023292, "optim/total_tokens": 2035810304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.3171401023864746, "created_at": "2025-01-15T16:31:08.949725+00:00"} {"global_step": 3884, "acc_step": 0, "speed/wps": 12891.771380191502, "speed/FLOPS": 202482984676199.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19935242831707, "optim/lr": 0.0023298, "optim/total_tokens": 2036334592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.2803127765655518, "created_at": "2025-01-15T16:31:19.121115+00:00"} {"global_step": 3885, "acc_step": 0, "speed/wps": 12889.163885846527, "speed/FLOPS": 202442030394438.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0986751914024353, "optim/lr": 0.0023304000000000003, "optim/total_tokens": 2036858880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 3.2133326530456543, "created_at": "2025-01-15T16:31:29.294107+00:00"} {"global_step": 3886, "acc_step": 0, "speed/wps": 12893.252860232556, "speed/FLOPS": 202506253356010.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0898531898856163, "optim/lr": 0.002331, "optim/total_tokens": 2037383168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.2499990463256836, "created_at": "2025-01-15T16:31:39.469658+00:00"} {"global_step": 3887, "acc_step": 0, "speed/wps": 12892.383374898842, "speed/FLOPS": 202492596894046.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09668083488941193, "optim/lr": 0.0023316, "optim/total_tokens": 2037907456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.2197928428649902, "created_at": "2025-01-15T16:31:49.641654+00:00"} {"global_step": 3888, "acc_step": 0, "speed/wps": 12893.942083495227, "speed/FLOPS": 202517078554442.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09989462047815323, "optim/lr": 0.0023322, "optim/total_tokens": 2038431744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.227992534637451, "created_at": "2025-01-15T16:31:59.811526+00:00"} {"global_step": 3889, "acc_step": 0, "speed/wps": 12896.905950849741, "speed/FLOPS": 202563630164026.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12196087837219238, "optim/lr": 0.0023328, "optim/total_tokens": 2038956032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.2090706825256348, "created_at": "2025-01-15T16:32:09.979893+00:00"} {"global_step": 3890, "acc_step": 0, "speed/wps": 12896.412123157936, "speed/FLOPS": 202555873921535.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.133778378367424, "optim/lr": 0.0023334000000000002, "optim/total_tokens": 2039480320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.179995536804199, "created_at": "2025-01-15T16:32:20.149009+00:00"} {"global_step": 3891, "acc_step": 0, "speed/wps": 12895.4056051434, "speed/FLOPS": 202540065173016.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12402788549661636, "optim/lr": 0.002334, "optim/total_tokens": 2040004608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.2385611534118652, "created_at": "2025-01-15T16:32:30.317802+00:00"} {"global_step": 3892, "acc_step": 0, "speed/wps": 12895.662422584626, "speed/FLOPS": 202544098843835.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12887263298034668, "optim/lr": 0.0023346, "optim/total_tokens": 2040528896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 3.2475736141204834, "created_at": "2025-01-15T16:32:40.485482+00:00"} {"global_step": 3893, "acc_step": 0, "speed/wps": 12893.140648155588, "speed/FLOPS": 202504490911146.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.6474717259407043, "optim/lr": 0.0023352, "optim/total_tokens": 2041053184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.1942543983459473, "created_at": "2025-01-15T16:32:50.662632+00:00"} {"global_step": 3894, "acc_step": 0, "speed/wps": 12897.6336078472, "speed/FLOPS": 202575059017076.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1351601481437683, "optim/lr": 0.0023358, "optim/total_tokens": 2041577472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 3.25822114944458, "created_at": "2025-01-15T16:33:00.829588+00:00"} {"global_step": 3895, "acc_step": 0, "speed/wps": 12898.273129957353, "speed/FLOPS": 202585103590612.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11363589018583298, "optim/lr": 0.0023364, "optim/total_tokens": 2042101760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.2318806648254395, "created_at": "2025-01-15T16:33:10.996740+00:00"} {"global_step": 3896, "acc_step": 0, "speed/wps": 12891.734551375775, "speed/FLOPS": 202482406229039.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11078819632530212, "optim/lr": 0.002337, "optim/total_tokens": 2042626048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296188, "loss/out": 3.2477316856384277, "created_at": "2025-01-15T16:33:21.168848+00:00"} {"global_step": 3897, "acc_step": 0, "speed/wps": 12877.51586985389, "speed/FLOPS": 202259082297222.8, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.14616060256958008, "optim/lr": 0.0023376, "optim/total_tokens": 2043150336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 3.3249831199645996, "created_at": "2025-01-15T16:33:31.352759+00:00"} {"global_step": 3898, "acc_step": 0, "speed/wps": 12890.673980745423, "speed/FLOPS": 202465748509913.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1342170089483261, "optim/lr": 0.0023382, "optim/total_tokens": 2043674624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.1965553760528564, "created_at": "2025-01-15T16:33:41.526409+00:00"} {"global_step": 3899, "acc_step": 0, "speed/wps": 12891.193193128034, "speed/FLOPS": 202473903453854.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11018946766853333, "optim/lr": 0.0023388, "optim/total_tokens": 2044198912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.3124234676361084, "created_at": "2025-01-15T16:33:51.700684+00:00"} {"global_step": 3900, "acc_step": 0, "speed/wps": 12893.676844882584, "speed/FLOPS": 202512912617555.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09552710503339767, "optim/lr": 0.0023394, "optim/total_tokens": 2044723200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 3.187440872192383, "created_at": "2025-01-15T16:34:01.871246+00:00"} {"global_step": 3901, "acc_step": 0, "speed/wps": 12896.990172183028, "speed/FLOPS": 202564952975797.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11963863670825958, "optim/lr": 0.00234, "optim/total_tokens": 2045247488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.3043031692504883, "created_at": "2025-01-15T16:34:12.043483+00:00"} {"global_step": 3902, "acc_step": 0, "speed/wps": 12892.027899373526, "speed/FLOPS": 202487013662447.0, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11408138275146484, "optim/lr": 0.0023406, "optim/total_tokens": 2045771776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.3293771743774414, "created_at": "2025-01-15T16:34:22.214068+00:00"} {"global_step": 3903, "acc_step": 0, "speed/wps": 12896.882339952932, "speed/FLOPS": 202563259322446.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1601574569940567, "optim/lr": 0.0023412, "optim/total_tokens": 2046296064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.301514148712158, "created_at": "2025-01-15T16:34:32.382016+00:00"} {"global_step": 3904, "acc_step": 0, "speed/wps": 12892.900666315376, "speed/FLOPS": 202500721666576.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10097195208072662, "optim/lr": 0.0023418, "optim/total_tokens": 2046820352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 3.27085018157959, "created_at": "2025-01-15T16:34:42.555065+00:00"} {"global_step": 3905, "acc_step": 0, "speed/wps": 12896.988301295012, "speed/FLOPS": 202564923590930.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09019408375024796, "optim/lr": 0.0023424, "optim/total_tokens": 2047344640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.1728599071502686, "created_at": "2025-01-15T16:34:52.721817+00:00"} {"global_step": 3906, "acc_step": 0, "speed/wps": 12895.881433502753, "speed/FLOPS": 202547538711257.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10042732954025269, "optim/lr": 0.002343, "optim/total_tokens": 2047868928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2541897296905518, "created_at": "2025-01-15T16:35:02.897993+00:00"} {"global_step": 3907, "acc_step": 0, "speed/wps": 12896.31117239673, "speed/FLOPS": 202554288351110.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09776009619235992, "optim/lr": 0.0023436, "optim/total_tokens": 2048393216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 3.319004774093628, "created_at": "2025-01-15T16:35:13.068062+00:00"} {"global_step": 3908, "acc_step": 0, "speed/wps": 12891.561504696188, "speed/FLOPS": 202479688293145.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10154882073402405, "optim/lr": 0.0023442, "optim/total_tokens": 2048917504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.2031211853027344, "created_at": "2025-01-15T16:35:23.245685+00:00"} {"global_step": 3909, "acc_step": 0, "speed/wps": 12897.846199981961, "speed/FLOPS": 202578398068685.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11146879941225052, "optim/lr": 0.0023447999999999997, "optim/total_tokens": 2049441792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 3.2339468002319336, "created_at": "2025-01-15T16:35:33.415371+00:00"} {"global_step": 3910, "acc_step": 0, "speed/wps": 12896.428728040297, "speed/FLOPS": 202556134724030.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10127520561218262, "optim/lr": 0.0023454, "optim/total_tokens": 2049966080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.303770065307617, "created_at": "2025-01-15T16:35:43.582773+00:00"} {"global_step": 3911, "acc_step": 0, "speed/wps": 12895.66464993708, "speed/FLOPS": 202544133827466.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08928894251585007, "optim/lr": 0.002346, "optim/total_tokens": 2050490368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 3.142796039581299, "created_at": "2025-01-15T16:35:53.752766+00:00"} {"global_step": 3912, "acc_step": 0, "speed/wps": 12894.814983222059, "speed/FLOPS": 202530788644142.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13263802230358124, "optim/lr": 0.0023466, "optim/total_tokens": 2051014656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.1858325004577637, "created_at": "2025-01-15T16:36:03.924044+00:00"} {"global_step": 3913, "acc_step": 0, "speed/wps": 12898.523590571438, "speed/FLOPS": 202589037418725.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12188467383384705, "optim/lr": 0.0023472, "optim/total_tokens": 2051538944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.2683193683624268, "created_at": "2025-01-15T16:36:14.096903+00:00"} {"global_step": 3914, "acc_step": 0, "speed/wps": 12897.438305091362, "speed/FLOPS": 202571991518922.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3244222402572632, "optim/lr": 0.0023478, "optim/total_tokens": 2052063232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 3.254045248031616, "created_at": "2025-01-15T16:36:24.266653+00:00"} {"global_step": 3915, "acc_step": 0, "speed/wps": 12894.96865805516, "speed/FLOPS": 202533202318567.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09472169727087021, "optim/lr": 0.0023484, "optim/total_tokens": 2052587520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.2956509590148926, "created_at": "2025-01-15T16:36:34.440931+00:00"} {"global_step": 3916, "acc_step": 0, "speed/wps": 12894.726776916588, "speed/FLOPS": 202529403242907.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09944884479045868, "optim/lr": 0.002349, "optim/total_tokens": 2053111808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.3507163524627686, "created_at": "2025-01-15T16:36:44.609345+00:00"} {"global_step": 3917, "acc_step": 0, "speed/wps": 12887.58230099705, "speed/FLOPS": 202417189431052.1, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11198177933692932, "optim/lr": 0.0023496, "optim/total_tokens": 2053636096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.1961474418640137, "created_at": "2025-01-15T16:36:54.784963+00:00"} {"global_step": 3918, "acc_step": 0, "speed/wps": 12888.636928685335, "speed/FLOPS": 202433753808106.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12620650231838226, "optim/lr": 0.0023502, "optim/total_tokens": 2054160384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384815, "loss/out": 3.2099344730377197, "created_at": "2025-01-15T16:37:04.960507+00:00"} {"global_step": 3919, "acc_step": 0, "speed/wps": 12895.710206257902, "speed/FLOPS": 202544849352086.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11061368882656097, "optim/lr": 0.0023508, "optim/total_tokens": 2054684672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.1512250900268555, "created_at": "2025-01-15T16:37:15.129798+00:00"} {"global_step": 3920, "acc_step": 0, "speed/wps": 12897.14177382771, "speed/FLOPS": 202567334087949.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11532676964998245, "optim/lr": 0.0023514, "optim/total_tokens": 2055208960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.25416898727417, "created_at": "2025-01-15T16:37:25.296158+00:00"} {"global_step": 3921, "acc_step": 0, "speed/wps": 12893.668968277123, "speed/FLOPS": 202512788904642.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1319119781255722, "optim/lr": 0.002352, "optim/total_tokens": 2055733248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303224, "loss/out": 3.2589850425720215, "created_at": "2025-01-15T16:37:35.473976+00:00"} {"global_step": 3922, "acc_step": 0, "speed/wps": 12898.150367923963, "speed/FLOPS": 202583175444185.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11974421888589859, "optim/lr": 0.0023526000000000003, "optim/total_tokens": 2056257536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.276674509048462, "created_at": "2025-01-15T16:37:45.641412+00:00"} {"global_step": 3923, "acc_step": 0, "speed/wps": 12898.78781924803, "speed/FLOPS": 202593187493179.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12360190600156784, "optim/lr": 0.0023532, "optim/total_tokens": 2056781824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 3.2456068992614746, "created_at": "2025-01-15T16:37:55.809382+00:00"} {"global_step": 3924, "acc_step": 0, "speed/wps": 12896.332810056314, "speed/FLOPS": 202554628200287.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16496612131595612, "optim/lr": 0.0023538, "optim/total_tokens": 2057306112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.20493221282959, "created_at": "2025-01-15T16:38:05.977109+00:00"} {"global_step": 3925, "acc_step": 0, "speed/wps": 12894.777883001292, "speed/FLOPS": 202530205934194.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14004480838775635, "optim/lr": 0.0023544000000000004, "optim/total_tokens": 2057830400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 3.353158712387085, "created_at": "2025-01-15T16:38:16.146857+00:00"} {"global_step": 3926, "acc_step": 0, "speed/wps": 12896.284515987147, "speed/FLOPS": 202553869675568.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1326690912246704, "optim/lr": 0.0023550000000000003, "optim/total_tokens": 2058354688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.34865403175354, "created_at": "2025-01-15T16:38:26.323039+00:00"} {"global_step": 3927, "acc_step": 0, "speed/wps": 12900.94713390738, "speed/FLOPS": 202627102497113.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12757843732833862, "optim/lr": 0.0023556000000000002, "optim/total_tokens": 2058878976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.29929256439209, "created_at": "2025-01-15T16:38:36.487950+00:00"} {"global_step": 3928, "acc_step": 0, "speed/wps": 12897.918107348896, "speed/FLOPS": 202579527472694.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11851244419813156, "optim/lr": 0.0023562, "optim/total_tokens": 2059403264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 3.2541065216064453, "created_at": "2025-01-15T16:38:46.655974+00:00"} {"global_step": 3929, "acc_step": 0, "speed/wps": 12897.817043176201, "speed/FLOPS": 202577940120985.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12774799764156342, "optim/lr": 0.0023568, "optim/total_tokens": 2059927552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.1831765174865723, "created_at": "2025-01-15T16:38:56.824513+00:00"} {"global_step": 3930, "acc_step": 0, "speed/wps": 12902.605488887782, "speed/FLOPS": 202653149240899.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09253563731908798, "optim/lr": 0.0023574000000000004, "optim/total_tokens": 2060451840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.3548965454101562, "created_at": "2025-01-15T16:39:06.987303+00:00"} {"global_step": 3931, "acc_step": 0, "speed/wps": 12898.83453018591, "speed/FLOPS": 202593921152646.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11379293352365494, "optim/lr": 0.0023580000000000003, "optim/total_tokens": 2060976128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 3.259044647216797, "created_at": "2025-01-15T16:39:17.153588+00:00"} {"global_step": 3932, "acc_step": 0, "speed/wps": 12897.91965817578, "speed/FLOPS": 202579551830561.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10100036859512329, "optim/lr": 0.0023586, "optim/total_tokens": 2061500416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.2146239280700684, "created_at": "2025-01-15T16:39:27.327493+00:00"} {"global_step": 3933, "acc_step": 0, "speed/wps": 12896.139287346401, "speed/FLOPS": 202551588660200.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09110330790281296, "optim/lr": 0.0023592, "optim/total_tokens": 2062024704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.1957483291625977, "created_at": "2025-01-15T16:39:37.499309+00:00"} {"global_step": 3934, "acc_step": 0, "speed/wps": 12897.351983526958, "speed/FLOPS": 202570635720134.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11539603024721146, "optim/lr": 0.0023598, "optim/total_tokens": 2062548992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.333603858947754, "created_at": "2025-01-15T16:39:47.666716+00:00"} {"global_step": 3935, "acc_step": 0, "speed/wps": 12896.672721101077, "speed/FLOPS": 202559966970332.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10308027267456055, "optim/lr": 0.0023604000000000003, "optim/total_tokens": 2063073280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.1450603008270264, "created_at": "2025-01-15T16:39:57.839277+00:00"} {"global_step": 3936, "acc_step": 0, "speed/wps": 12895.613319406804, "speed/FLOPS": 202543327610953.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10906707495450974, "optim/lr": 0.0023610000000000003, "optim/total_tokens": 2063597568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 3.39424467086792, "created_at": "2025-01-15T16:40:08.012022+00:00"} {"global_step": 3937, "acc_step": 0, "speed/wps": 12896.668473697862, "speed/FLOPS": 202559900259028.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09621506929397583, "optim/lr": 0.0023616, "optim/total_tokens": 2064121856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.1425676345825195, "created_at": "2025-01-15T16:40:18.180947+00:00"} {"global_step": 3938, "acc_step": 0, "speed/wps": 12898.045128635042, "speed/FLOPS": 202581522516538.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09622900187969208, "optim/lr": 0.0023622, "optim/total_tokens": 2064646144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1084768772125244, "created_at": "2025-01-15T16:40:28.349579+00:00"} {"global_step": 3939, "acc_step": 0, "speed/wps": 12887.9047501414, "speed/FLOPS": 202422253937955.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11329896003007889, "optim/lr": 0.0023628, "optim/total_tokens": 2065170432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 3.2036871910095215, "created_at": "2025-01-15T16:40:38.527013+00:00"} {"global_step": 3940, "acc_step": 0, "speed/wps": 12894.557272120828, "speed/FLOPS": 202526740937165.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.09862595796585083, "optim/lr": 0.0023634, "optim/total_tokens": 2065694720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.256762981414795, "created_at": "2025-01-15T16:40:48.695606+00:00"} {"global_step": 3941, "acc_step": 0, "speed/wps": 12896.727540507814, "speed/FLOPS": 202560827984444.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10064949095249176, "optim/lr": 0.002364, "optim/total_tokens": 2066219008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.2754945755004883, "created_at": "2025-01-15T16:40:58.862756+00:00"} {"global_step": 3942, "acc_step": 0, "speed/wps": 12897.14479946747, "speed/FLOPS": 202567381609779.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10169357061386108, "optim/lr": 0.0023646, "optim/total_tokens": 2066743296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.2754480838775635, "created_at": "2025-01-15T16:41:09.029256+00:00"} {"global_step": 3943, "acc_step": 0, "speed/wps": 12893.424049394478, "speed/FLOPS": 202508942117037.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10129299014806747, "optim/lr": 0.0023652, "optim/total_tokens": 2067267584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 3.217841625213623, "created_at": "2025-01-15T16:41:19.198804+00:00"} {"global_step": 3944, "acc_step": 0, "speed/wps": 12899.335781405707, "speed/FLOPS": 202601793991844.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10267525911331177, "optim/lr": 0.0023658, "optim/total_tokens": 2067791872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.2789788246154785, "created_at": "2025-01-15T16:41:29.367765+00:00"} {"global_step": 3945, "acc_step": 0, "speed/wps": 12894.90479163989, "speed/FLOPS": 202532199208754.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08204200118780136, "optim/lr": 0.0023664, "optim/total_tokens": 2068316160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.238525152206421, "created_at": "2025-01-15T16:41:39.541434+00:00"} {"global_step": 3946, "acc_step": 0, "speed/wps": 12898.007488541198, "speed/FLOPS": 202580931327141.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10909775644540787, "optim/lr": 0.002367, "optim/total_tokens": 2068840448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 3.342623472213745, "created_at": "2025-01-15T16:41:49.711868+00:00"} {"global_step": 3947, "acc_step": 0, "speed/wps": 12899.699356025883, "speed/FLOPS": 202607504430858.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10612603276968002, "optim/lr": 0.0023676, "optim/total_tokens": 2069364736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.1620099544525146, "created_at": "2025-01-15T16:41:59.881290+00:00"} {"global_step": 3948, "acc_step": 0, "speed/wps": 12898.705625511386, "speed/FLOPS": 202591896527599.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10997222363948822, "optim/lr": 0.0023682, "optim/total_tokens": 2069889024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.206082344055176, "created_at": "2025-01-15T16:42:10.046529+00:00"} {"global_step": 3949, "acc_step": 0, "speed/wps": 12892.070713792027, "speed/FLOPS": 202487686121722.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10872425138950348, "optim/lr": 0.0023688, "optim/total_tokens": 2070413312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.232164144515991, "created_at": "2025-01-15T16:42:20.217878+00:00"} {"global_step": 3950, "acc_step": 0, "speed/wps": 12899.383857140954, "speed/FLOPS": 202602549087329.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11711908876895905, "optim/lr": 0.0023694, "optim/total_tokens": 2070937600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 3.322129726409912, "created_at": "2025-01-15T16:42:30.390358+00:00"} {"global_step": 3951, "acc_step": 0, "speed/wps": 12897.026622106458, "speed/FLOPS": 202565525471932.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15257026255130768, "optim/lr": 0.00237, "optim/total_tokens": 2071461888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.2044239044189453, "created_at": "2025-01-15T16:42:40.558009+00:00"} {"global_step": 3952, "acc_step": 0, "speed/wps": 12896.543552367883, "speed/FLOPS": 202557938197880.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13961546123027802, "optim/lr": 0.0023706, "optim/total_tokens": 2071986176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.247830629348755, "created_at": "2025-01-15T16:42:50.724843+00:00"} {"global_step": 3953, "acc_step": 0, "speed/wps": 12896.487756019309, "speed/FLOPS": 202557061839554.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1135190799832344, "optim/lr": 0.0023712, "optim/total_tokens": 2072510464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.324676036834717, "created_at": "2025-01-15T16:43:00.894898+00:00"} {"global_step": 3954, "acc_step": 0, "speed/wps": 12899.159451997628, "speed/FLOPS": 202599024496188.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15418067574501038, "optim/lr": 0.0023718, "optim/total_tokens": 2073034752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.2447988986968994, "created_at": "2025-01-15T16:43:11.061636+00:00"} {"global_step": 3955, "acc_step": 0, "speed/wps": 12895.824699927407, "speed/FLOPS": 202546647632496.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13837459683418274, "optim/lr": 0.0023723999999999998, "optim/total_tokens": 2073559040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.329158306121826, "created_at": "2025-01-15T16:43:21.234924+00:00"} {"global_step": 3956, "acc_step": 0, "speed/wps": 12896.244990679885, "speed/FLOPS": 202553248876304.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12403975427150726, "optim/lr": 0.002373, "optim/total_tokens": 2074083328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.2149858474731445, "created_at": "2025-01-15T16:43:31.405817+00:00"} {"global_step": 3957, "acc_step": 0, "speed/wps": 12894.284129659569, "speed/FLOPS": 202522450859474.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11754049360752106, "optim/lr": 0.0023736, "optim/total_tokens": 2074607616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 3.285684585571289, "created_at": "2025-01-15T16:43:41.574595+00:00"} {"global_step": 3958, "acc_step": 0, "speed/wps": 12899.40321617015, "speed/FLOPS": 202602853147485.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11308864504098892, "optim/lr": 0.0023742, "optim/total_tokens": 2075131904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.258715867996216, "created_at": "2025-01-15T16:43:51.741673+00:00"} {"global_step": 3959, "acc_step": 0, "speed/wps": 12899.503262393144, "speed/FLOPS": 202604424510895.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12108640372753143, "optim/lr": 0.0023748, "optim/total_tokens": 2075656192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.3095970153808594, "created_at": "2025-01-15T16:44:01.908490+00:00"} {"global_step": 3960, "acc_step": 0, "speed/wps": 12898.81124442426, "speed/FLOPS": 202593555417761.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12338363379240036, "optim/lr": 0.0023753999999999997, "optim/total_tokens": 2076180480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.238785743713379, "created_at": "2025-01-15T16:44:12.079697+00:00"} {"global_step": 3961, "acc_step": 0, "speed/wps": 12892.28704704205, "speed/FLOPS": 202491083932687.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12982942163944244, "optim/lr": 0.002376, "optim/total_tokens": 2076704768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.252570152282715, "created_at": "2025-01-15T16:44:22.249941+00:00"} {"global_step": 3962, "acc_step": 0, "speed/wps": 12898.030873057529, "speed/FLOPS": 202581298613104.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13566093146800995, "optim/lr": 0.0023766, "optim/total_tokens": 2077229056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.2804765701293945, "created_at": "2025-01-15T16:44:32.419716+00:00"} {"global_step": 3963, "acc_step": 0, "speed/wps": 12898.647965091151, "speed/FLOPS": 202590990891465.88, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1049492210149765, "optim/lr": 0.0023772, "optim/total_tokens": 2077753344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444591, "loss/out": 3.2207040786743164, "created_at": "2025-01-15T16:44:42.586821+00:00"} {"global_step": 3964, "acc_step": 0, "speed/wps": 12895.867387134293, "speed/FLOPS": 202547318093738.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14626750349998474, "optim/lr": 0.0023778, "optim/total_tokens": 2078277632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 3.2517151832580566, "created_at": "2025-01-15T16:44:52.758489+00:00"} {"global_step": 3965, "acc_step": 0, "speed/wps": 12898.858326744124, "speed/FLOPS": 202594294910293.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11705274879932404, "optim/lr": 0.0023783999999999997, "optim/total_tokens": 2078801920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.237788200378418, "created_at": "2025-01-15T16:45:02.931591+00:00"} {"global_step": 3966, "acc_step": 0, "speed/wps": 12891.631093607002, "speed/FLOPS": 202480781282615.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13786107301712036, "optim/lr": 0.002379, "optim/total_tokens": 2079326208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.1319022178649902, "created_at": "2025-01-15T16:45:13.105698+00:00"} {"global_step": 3967, "acc_step": 0, "speed/wps": 12897.005808242453, "speed/FLOPS": 202565198561597.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.12787902355194092, "optim/lr": 0.0023796, "optim/total_tokens": 2079850496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 3.206948757171631, "created_at": "2025-01-15T16:45:23.272264+00:00"} {"global_step": 3968, "acc_step": 0, "speed/wps": 12899.31901652119, "speed/FLOPS": 202601530676295.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12373745441436768, "optim/lr": 0.0023802, "optim/total_tokens": 2080374784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.321744203567505, "created_at": "2025-01-15T16:45:33.438120+00:00"} {"global_step": 3969, "acc_step": 0, "speed/wps": 12896.2355061548, "speed/FLOPS": 202553099908804.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11421538889408112, "optim/lr": 0.0023807999999999998, "optim/total_tokens": 2080899072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.258431911468506, "created_at": "2025-01-15T16:45:43.605456+00:00"} {"global_step": 3970, "acc_step": 0, "speed/wps": 12900.832803514539, "speed/FLOPS": 202625306781187.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.14227266609668732, "optim/lr": 0.0023814, "optim/total_tokens": 2081423360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 3.272017478942871, "created_at": "2025-01-15T16:45:53.770173+00:00"} {"global_step": 3971, "acc_step": 0, "speed/wps": 12895.334578874214, "speed/FLOPS": 202538949607859.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13694462180137634, "optim/lr": 0.002382, "optim/total_tokens": 2081947648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 3.286412239074707, "created_at": "2025-01-15T16:46:03.938806+00:00"} {"global_step": 3972, "acc_step": 0, "speed/wps": 12896.743014021473, "speed/FLOPS": 202561071017239.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12194806337356567, "optim/lr": 0.0023826, "optim/total_tokens": 2082471936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.1421592235565186, "created_at": "2025-01-15T16:46:14.115120+00:00"} {"global_step": 3973, "acc_step": 0, "speed/wps": 12901.141361504257, "speed/FLOPS": 202630153108419.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10585901141166687, "optim/lr": 0.0023832000000000002, "optim/total_tokens": 2082996224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.35980224609375, "created_at": "2025-01-15T16:46:24.280852+00:00"} {"global_step": 3974, "acc_step": 0, "speed/wps": 12900.89865153748, "speed/FLOPS": 202626341014872.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12053053081035614, "optim/lr": 0.0023838, "optim/total_tokens": 2083520512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447832, "loss/out": 3.339122772216797, "created_at": "2025-01-15T16:46:34.445410+00:00"} {"global_step": 3975, "acc_step": 0, "speed/wps": 12900.572548271959, "speed/FLOPS": 202621219114974.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1028975173830986, "optim/lr": 0.0023844, "optim/total_tokens": 2084044800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 3.145857572555542, "created_at": "2025-01-15T16:46:44.612377+00:00"} {"global_step": 3976, "acc_step": 0, "speed/wps": 12893.64544837469, "speed/FLOPS": 202512419492255.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08621563017368317, "optim/lr": 0.002385, "optim/total_tokens": 2084569088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.3525936603546143, "created_at": "2025-01-15T16:46:54.782485+00:00"} {"global_step": 3977, "acc_step": 0, "speed/wps": 12896.914648067925, "speed/FLOPS": 202563766765788.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09014125913381577, "optim/lr": 0.0023856000000000003, "optim/total_tokens": 2085093376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.307040214538574, "created_at": "2025-01-15T16:47:04.951161+00:00"} {"global_step": 3978, "acc_step": 0, "speed/wps": 12894.704804390136, "speed/FLOPS": 202529058134185.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08342285454273224, "optim/lr": 0.0023862, "optim/total_tokens": 2085617664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.196530342102051, "created_at": "2025-01-15T16:47:15.121991+00:00"} {"global_step": 3979, "acc_step": 0, "speed/wps": 12898.137842668195, "speed/FLOPS": 202582978717831.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09650512039661407, "optim/lr": 0.0023868, "optim/total_tokens": 2086141952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.178476572036743, "created_at": "2025-01-15T16:47:25.289646+00:00"} {"global_step": 3980, "acc_step": 0, "speed/wps": 12899.382979873511, "speed/FLOPS": 202602535308639.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09743592888116837, "optim/lr": 0.0023874, "optim/total_tokens": 2086666240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 3.139674663543701, "created_at": "2025-01-15T16:47:35.462443+00:00"} {"global_step": 3981, "acc_step": 0, "speed/wps": 12900.254847499364, "speed/FLOPS": 202616229187774.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11186820268630981, "optim/lr": 0.0023880000000000004, "optim/total_tokens": 2087190528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.1811537742614746, "created_at": "2025-01-15T16:47:45.626492+00:00"} {"global_step": 3982, "acc_step": 0, "speed/wps": 12896.423312038454, "speed/FLOPS": 202556049658279.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11550980806350708, "optim/lr": 0.0023886000000000003, "optim/total_tokens": 2087714816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.357311725616455, "created_at": "2025-01-15T16:47:55.794143+00:00"} {"global_step": 3983, "acc_step": 0, "speed/wps": 12898.731214963193, "speed/FLOPS": 202592298445103.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1106535792350769, "optim/lr": 0.0023892, "optim/total_tokens": 2088239104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492154, "loss/out": 3.2422261238098145, "created_at": "2025-01-15T16:48:05.959381+00:00"} {"global_step": 3984, "acc_step": 0, "speed/wps": 12892.751818258606, "speed/FLOPS": 202498383803305.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09371019899845123, "optim/lr": 0.0023898, "optim/total_tokens": 2088763392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 3.3213353157043457, "created_at": "2025-01-15T16:48:16.129282+00:00"} {"global_step": 3985, "acc_step": 0, "speed/wps": 12894.693187265188, "speed/FLOPS": 202528875671274.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13048648834228516, "optim/lr": 0.0023904, "optim/total_tokens": 2089287680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 3.196547031402588, "created_at": "2025-01-15T16:48:26.297756+00:00"} {"global_step": 3986, "acc_step": 0, "speed/wps": 12897.78936164926, "speed/FLOPS": 202577505344566.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11100485920906067, "optim/lr": 0.0023910000000000003, "optim/total_tokens": 2089811968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403804, "loss/out": 3.2483904361724854, "created_at": "2025-01-15T16:48:36.467378+00:00"} {"global_step": 3987, "acc_step": 0, "speed/wps": 12899.335451546134, "speed/FLOPS": 202601788810946.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10272610187530518, "optim/lr": 0.0023916000000000002, "optim/total_tokens": 2090336256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.1514134407043457, "created_at": "2025-01-15T16:48:46.632238+00:00"} {"global_step": 3988, "acc_step": 0, "speed/wps": 12894.102017487434, "speed/FLOPS": 202519590537562.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10891743004322052, "optim/lr": 0.0023922, "optim/total_tokens": 2090860544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.25815486907959, "created_at": "2025-01-15T16:48:56.804354+00:00"} {"global_step": 3989, "acc_step": 0, "speed/wps": 12893.503911427631, "speed/FLOPS": 202510196460009.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09701769053936005, "optim/lr": 0.0023928, "optim/total_tokens": 2091384832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.1673741340637207, "created_at": "2025-01-15T16:49:06.975921+00:00"} {"global_step": 3990, "acc_step": 0, "speed/wps": 12891.445396922629, "speed/FLOPS": 202477864661013.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10210201889276505, "optim/lr": 0.0023934, "optim/total_tokens": 2091909120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.180685043334961, "created_at": "2025-01-15T16:49:17.148138+00:00"} {"global_step": 3991, "acc_step": 0, "speed/wps": 12892.85135503311, "speed/FLOPS": 202499947165127.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10510620474815369, "optim/lr": 0.0023940000000000003, "optim/total_tokens": 2092433408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 3.187384605407715, "created_at": "2025-01-15T16:49:27.317841+00:00"} {"global_step": 3992, "acc_step": 0, "speed/wps": 12895.34553548065, "speed/FLOPS": 202539121696419.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10893633961677551, "optim/lr": 0.0023946, "optim/total_tokens": 2092957696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.2585606575012207, "created_at": "2025-01-15T16:49:37.486511+00:00"} {"global_step": 3993, "acc_step": 0, "speed/wps": 12895.608082512805, "speed/FLOPS": 202543245358336.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11369934678077698, "optim/lr": 0.0023952, "optim/total_tokens": 2093481984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463510, "loss/out": 3.3221113681793213, "created_at": "2025-01-15T16:49:47.657644+00:00"} {"global_step": 3994, "acc_step": 0, "speed/wps": 12898.029462868199, "speed/FLOPS": 202581276464143.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11687058210372925, "optim/lr": 0.0023958, "optim/total_tokens": 2094006272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 3.2723546028137207, "created_at": "2025-01-15T16:49:57.827476+00:00"} {"global_step": 3995, "acc_step": 0, "speed/wps": 12897.143768858768, "speed/FLOPS": 202567365422654.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09214037656784058, "optim/lr": 0.0023964, "optim/total_tokens": 2094530560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.2868919372558594, "created_at": "2025-01-15T16:50:07.995613+00:00"} {"global_step": 3996, "acc_step": 0, "speed/wps": 12891.983994046004, "speed/FLOPS": 202486324068946.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10996134579181671, "optim/lr": 0.0023970000000000003, "optim/total_tokens": 2095054848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.241746187210083, "created_at": "2025-01-15T16:50:18.167343+00:00"} {"global_step": 3997, "acc_step": 0, "speed/wps": 12892.192600749924, "speed/FLOPS": 202489600523886.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10870283097028732, "optim/lr": 0.0023976, "optim/total_tokens": 2095579136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.2003273963928223, "created_at": "2025-01-15T16:50:28.340265+00:00"} {"global_step": 3998, "acc_step": 0, "speed/wps": 12893.392853741048, "speed/FLOPS": 202508452146432.78, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11583214998245239, "optim/lr": 0.0023982, "optim/total_tokens": 2096103424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297240, "loss/out": 3.1375417709350586, "created_at": "2025-01-15T16:50:38.514054+00:00"} {"global_step": 3999, "acc_step": 0, "speed/wps": 12892.25909636025, "speed/FLOPS": 202490644928821.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10979178547859192, "optim/lr": 0.0023988, "optim/total_tokens": 2096627712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451263, "loss/out": 3.1822879314422607, "created_at": "2025-01-15T16:50:48.690395+00:00"} {"global_step": 4000, "acc_step": 0, "speed/wps": 12896.10830543912, "speed/FLOPS": 202551102046772.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1542440801858902, "optim/lr": 0.0023994, "optim/total_tokens": 2097152000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.3193745613098145, "created_at": "2025-01-15T16:50:58.857637+00:00"} {"global_step": 4001, "acc_step": 0, "speed/wps": 12592.820230953474, "speed/FLOPS": 197787546075487.75, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.2478, "optim/grad_norm": 0.17623457312583923, "optim/lr": 0.0024000000000000002, "optim/total_tokens": 2097676288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.2065162658691406, "created_at": "2025-01-15T16:51:09.269606+00:00"} {"global_step": 4002, "acc_step": 0, "speed/wps": 12916.75876078133, "speed/FLOPS": 202875445824621.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12411568313837051, "optim/lr": 0.0024006, "optim/total_tokens": 2098200576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 3.224438190460205, "created_at": "2025-01-15T16:51:19.421564+00:00"} {"global_step": 4003, "acc_step": 0, "speed/wps": 12909.182273655806, "speed/FLOPS": 202756446683130.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1050049364566803, "optim/lr": 0.0024012, "optim/total_tokens": 2098724864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.270781993865967, "created_at": "2025-01-15T16:51:29.578863+00:00"} {"global_step": 4004, "acc_step": 0, "speed/wps": 12914.814451858529, "speed/FLOPS": 202844907781228.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09434840828180313, "optim/lr": 0.0024018, "optim/total_tokens": 2099249152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460071, "loss/out": 3.329537868499756, "created_at": "2025-01-15T16:51:39.731388+00:00"} {"global_step": 4005, "acc_step": 0, "speed/wps": 12914.062796318627, "speed/FLOPS": 202833101998091.22, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10110431909561157, "optim/lr": 0.0024024, "optim/total_tokens": 2099773440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.152137279510498, "created_at": "2025-01-15T16:51:49.884429+00:00"} {"global_step": 4006, "acc_step": 0, "speed/wps": 12910.568843914365, "speed/FLOPS": 202778224674388.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09676653146743774, "optim/lr": 0.002403, "optim/total_tokens": 2100297728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 3.2207913398742676, "created_at": "2025-01-15T16:52:00.045442+00:00"} {"global_step": 4007, "acc_step": 0, "speed/wps": 12915.440653564741, "speed/FLOPS": 202854743139515.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17011937499046326, "optim/lr": 0.0024036, "optim/total_tokens": 2100822016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.22694730758667, "created_at": "2025-01-15T16:52:10.198644+00:00"} {"global_step": 4008, "acc_step": 0, "speed/wps": 12915.962117863908, "speed/FLOPS": 202862933452901.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11027535051107407, "optim/lr": 0.0024042, "optim/total_tokens": 2101346304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 3.213449239730835, "created_at": "2025-01-15T16:52:20.355296+00:00"} {"global_step": 4009, "acc_step": 0, "speed/wps": 12914.68542432454, "speed/FLOPS": 202842881226504.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09329778701066971, "optim/lr": 0.0024048, "optim/total_tokens": 2101870592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 3.2203354835510254, "created_at": "2025-01-15T16:52:30.511585+00:00"} {"global_step": 4010, "acc_step": 0, "speed/wps": 12913.156328023868, "speed/FLOPS": 202818864667904.06, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12802253663539886, "optim/lr": 0.0024054, "optim/total_tokens": 2102394880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.190842390060425, "created_at": "2025-01-15T16:52:40.665671+00:00"} {"global_step": 4011, "acc_step": 0, "speed/wps": 12910.036145351158, "speed/FLOPS": 202769857911448.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1506221890449524, "optim/lr": 0.002406, "optim/total_tokens": 2102919168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 3.207503080368042, "created_at": "2025-01-15T16:52:50.823003+00:00"} {"global_step": 4012, "acc_step": 0, "speed/wps": 12905.955553524116, "speed/FLOPS": 202705766609484.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12067226320505142, "optim/lr": 0.0024066, "optim/total_tokens": 2103443456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.349186658859253, "created_at": "2025-01-15T16:53:00.982558+00:00"} {"global_step": 4013, "acc_step": 0, "speed/wps": 12914.087258187103, "speed/FLOPS": 202833486205349.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14493204653263092, "optim/lr": 0.0024072, "optim/total_tokens": 2103967744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.225778579711914, "created_at": "2025-01-15T16:53:11.136994+00:00"} {"global_step": 4014, "acc_step": 0, "speed/wps": 12910.890839474312, "speed/FLOPS": 202783282057125.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14730720221996307, "optim/lr": 0.0024078, "optim/total_tokens": 2104492032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1377475261688232, "created_at": "2025-01-15T16:53:21.294124+00:00"} {"global_step": 4015, "acc_step": 0, "speed/wps": 12912.580185289138, "speed/FLOPS": 202809815554552.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.15124885737895966, "optim/lr": 0.0024083999999999998, "optim/total_tokens": 2105016320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 3.171666145324707, "created_at": "2025-01-15T16:53:31.449845+00:00"} {"global_step": 4016, "acc_step": 0, "speed/wps": 12915.445116019137, "speed/FLOPS": 202854813228494.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13378462195396423, "optim/lr": 0.002409, "optim/total_tokens": 2105540608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.230220317840576, "created_at": "2025-01-15T16:53:41.605505+00:00"} {"global_step": 4017, "acc_step": 0, "speed/wps": 12913.289855877223, "speed/FLOPS": 202820961906327.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11730577796697617, "optim/lr": 0.0024096, "optim/total_tokens": 2106064896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.1747117042541504, "created_at": "2025-01-15T16:53:51.762191+00:00"} {"global_step": 4018, "acc_step": 0, "speed/wps": 12913.0700127534, "speed/FLOPS": 202817508967971.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11193226277828217, "optim/lr": 0.0024102, "optim/total_tokens": 2106589184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.1738834381103516, "created_at": "2025-01-15T16:54:01.921362+00:00"} {"global_step": 4019, "acc_step": 0, "speed/wps": 12912.690135488045, "speed/FLOPS": 202811542473513.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09981101751327515, "optim/lr": 0.0024108, "optim/total_tokens": 2107113472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412572, "loss/out": 3.30485200881958, "created_at": "2025-01-15T16:54:12.080917+00:00"} {"global_step": 4020, "acc_step": 0, "speed/wps": 12915.793723241135, "speed/FLOPS": 202860288583952.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10026223212480545, "optim/lr": 0.0024113999999999997, "optim/total_tokens": 2107637760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.284395217895508, "created_at": "2025-01-15T16:54:22.232672+00:00"} {"global_step": 4021, "acc_step": 0, "speed/wps": 12915.934116586937, "speed/FLOPS": 202862493654368.56, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08737720549106598, "optim/lr": 0.002412, "optim/total_tokens": 2108162048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.268820285797119, "created_at": "2025-01-15T16:54:32.387939+00:00"} {"global_step": 4022, "acc_step": 0, "speed/wps": 12909.672460317384, "speed/FLOPS": 202764145738237.84, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.09439373761415482, "optim/lr": 0.0024126, "optim/total_tokens": 2108686336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 3.294471502304077, "created_at": "2025-01-15T16:54:42.545340+00:00"} {"global_step": 4023, "acc_step": 0, "speed/wps": 12915.61756824244, "speed/FLOPS": 202857521827635.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09747786074876785, "optim/lr": 0.0024132, "optim/total_tokens": 2109210624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 3.1755175590515137, "created_at": "2025-01-15T16:54:52.700132+00:00"} {"global_step": 4024, "acc_step": 0, "speed/wps": 12914.256768458583, "speed/FLOPS": 202836148597094.88, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09880739450454712, "optim/lr": 0.0024138, "optim/total_tokens": 2109734912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336346, "loss/out": 3.2188706398010254, "created_at": "2025-01-15T16:55:02.857082+00:00"} {"global_step": 4025, "acc_step": 0, "speed/wps": 12913.082948536956, "speed/FLOPS": 202817712142228.12, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10661695152521133, "optim/lr": 0.0024144, "optim/total_tokens": 2110259200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.3025286197662354, "created_at": "2025-01-15T16:55:13.013850+00:00"} {"global_step": 4026, "acc_step": 0, "speed/wps": 12911.097919383456, "speed/FLOPS": 202786534531656.5, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11294553428888321, "optim/lr": 0.002415, "optim/total_tokens": 2110783488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.269324541091919, "created_at": "2025-01-15T16:55:23.172914+00:00"} {"global_step": 4027, "acc_step": 0, "speed/wps": 12912.708361579573, "speed/FLOPS": 202811828739326.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09679000824689865, "optim/lr": 0.0024156, "optim/total_tokens": 2111307776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 3.307931900024414, "created_at": "2025-01-15T16:55:33.330479+00:00"} {"global_step": 4028, "acc_step": 0, "speed/wps": 12901.62400344283, "speed/FLOPS": 202637733663284.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09819449484348297, "optim/lr": 0.0024162, "optim/total_tokens": 2111832064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.1610677242279053, "created_at": "2025-01-15T16:55:43.495530+00:00"} {"global_step": 4029, "acc_step": 0, "speed/wps": 12904.426106302075, "speed/FLOPS": 202681744539181.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0969427302479744, "optim/lr": 0.0024168, "optim/total_tokens": 2112356352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 3.169158458709717, "created_at": "2025-01-15T16:55:53.656863+00:00"} {"global_step": 4030, "acc_step": 0, "speed/wps": 12898.9498749757, "speed/FLOPS": 202595732801070.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.09118214249610901, "optim/lr": 0.0024174, "optim/total_tokens": 2112880640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 3.202937364578247, "created_at": "2025-01-15T16:56:03.830188+00:00"} {"global_step": 4031, "acc_step": 0, "speed/wps": 12907.277359958016, "speed/FLOPS": 202726527395881.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08944366127252579, "optim/lr": 0.002418, "optim/total_tokens": 2113404928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.2041730880737305, "created_at": "2025-01-15T16:56:13.994900+00:00"} {"global_step": 4032, "acc_step": 0, "speed/wps": 12901.333170498992, "speed/FLOPS": 202633165732254.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10284002125263214, "optim/lr": 0.0024186000000000004, "optim/total_tokens": 2113929216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.248138904571533, "created_at": "2025-01-15T16:56:24.158338+00:00"} {"global_step": 4033, "acc_step": 0, "speed/wps": 12910.918348674852, "speed/FLOPS": 202783714126921.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12203428894281387, "optim/lr": 0.0024192000000000003, "optim/total_tokens": 2114453504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 3.2245969772338867, "created_at": "2025-01-15T16:56:34.313885+00:00"} {"global_step": 4034, "acc_step": 0, "speed/wps": 12916.649229696879, "speed/FLOPS": 202873725488428.56, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12322892248630524, "optim/lr": 0.0024198, "optim/total_tokens": 2114977792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.2715718746185303, "created_at": "2025-01-15T16:56:44.469877+00:00"} {"global_step": 4035, "acc_step": 0, "speed/wps": 12910.66069346831, "speed/FLOPS": 202779667297847.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12164011597633362, "optim/lr": 0.0024204, "optim/total_tokens": 2115502080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.184572219848633, "created_at": "2025-01-15T16:56:54.626001+00:00"} {"global_step": 4036, "acc_step": 0, "speed/wps": 12914.708830246627, "speed/FLOPS": 202843248848674.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11539521813392639, "optim/lr": 0.0024210000000000004, "optim/total_tokens": 2116026368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 3.1307668685913086, "created_at": "2025-01-15T16:57:04.781602+00:00"} {"global_step": 4037, "acc_step": 0, "speed/wps": 12913.15482329348, "speed/FLOPS": 202818841034045.6, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11156414449214935, "optim/lr": 0.0024216000000000003, "optim/total_tokens": 2116550656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.292388916015625, "created_at": "2025-01-15T16:57:14.937753+00:00"} {"global_step": 4038, "acc_step": 0, "speed/wps": 12917.190763637955, "speed/FLOPS": 202882231023118.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13568058609962463, "optim/lr": 0.0024222000000000002, "optim/total_tokens": 2117074944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393108, "loss/out": 3.226015090942383, "created_at": "2025-01-15T16:57:25.088533+00:00"} {"global_step": 4039, "acc_step": 0, "speed/wps": 12914.043213144685, "speed/FLOPS": 202832794417434.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.10204195231199265, "optim/lr": 0.0024228, "optim/total_tokens": 2117599232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 3.225858688354492, "created_at": "2025-01-15T16:57:35.247658+00:00"} {"global_step": 4040, "acc_step": 0, "speed/wps": 12910.980437274218, "speed/FLOPS": 202784689313693.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09931834787130356, "optim/lr": 0.0024234, "optim/total_tokens": 2118123520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402750, "loss/out": 3.2583322525024414, "created_at": "2025-01-15T16:57:45.403802+00:00"} {"global_step": 4041, "acc_step": 0, "speed/wps": 12907.494954380403, "speed/FLOPS": 202729945015290.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09758846461772919, "optim/lr": 0.0024240000000000004, "optim/total_tokens": 2118647808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.1645283699035645, "created_at": "2025-01-15T16:57:55.565870+00:00"} {"global_step": 4042, "acc_step": 0, "speed/wps": 12914.18245807749, "speed/FLOPS": 202834981450446.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09369485080242157, "optim/lr": 0.0024246000000000003, "optim/total_tokens": 2119172096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.2488088607788086, "created_at": "2025-01-15T16:58:05.720124+00:00"} {"global_step": 4043, "acc_step": 0, "speed/wps": 12914.977763860814, "speed/FLOPS": 202847472820638.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09217721968889236, "optim/lr": 0.0024252, "optim/total_tokens": 2119696384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 3.2059004306793213, "created_at": "2025-01-15T16:58:15.872478+00:00"} {"global_step": 4044, "acc_step": 0, "speed/wps": 12915.44130426507, "speed/FLOPS": 202854753359658.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10008205473423004, "optim/lr": 0.0024258, "optim/total_tokens": 2120220672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.2479376792907715, "created_at": "2025-01-15T16:58:26.026853+00:00"} {"global_step": 4045, "acc_step": 0, "speed/wps": 12905.394555737043, "speed/FLOPS": 202696955368346.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11196501553058624, "optim/lr": 0.0024264, "optim/total_tokens": 2120744960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.2717716693878174, "created_at": "2025-01-15T16:58:36.187936+00:00"} {"global_step": 4046, "acc_step": 0, "speed/wps": 12902.94029973702, "speed/FLOPS": 202658407905365.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1277473270893097, "optim/lr": 0.0024270000000000003, "optim/total_tokens": 2121269248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 3.3004322052001953, "created_at": "2025-01-15T16:58:46.351416+00:00"} {"global_step": 4047, "acc_step": 0, "speed/wps": 12906.816494539353, "speed/FLOPS": 202719288871185.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1541534662246704, "optim/lr": 0.0024276000000000002, "optim/total_tokens": 2121793536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294022, "loss/out": 3.254317283630371, "created_at": "2025-01-15T16:58:56.512726+00:00"} {"global_step": 4048, "acc_step": 0, "speed/wps": 12907.937313436709, "speed/FLOPS": 202736892872135.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12940973043441772, "optim/lr": 0.0024282, "optim/total_tokens": 2122317824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 3.289285182952881, "created_at": "2025-01-15T16:59:06.674022+00:00"} {"global_step": 4049, "acc_step": 0, "speed/wps": 12913.036848783364, "speed/FLOPS": 202816988082250.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13908936083316803, "optim/lr": 0.0024288, "optim/total_tokens": 2122842112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476020, "loss/out": 3.276916980743408, "created_at": "2025-01-15T16:59:16.829442+00:00"} {"global_step": 4050, "acc_step": 0, "speed/wps": 12907.636451808265, "speed/FLOPS": 202732167426833.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08617009967565536, "optim/lr": 0.0024294, "optim/total_tokens": 2123366400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 3.2555413246154785, "created_at": "2025-01-15T16:59:26.989725+00:00"} {"global_step": 4051, "acc_step": 0, "speed/wps": 12912.656898830874, "speed/FLOPS": 202811020446140.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12298420816659927, "optim/lr": 0.0024300000000000003, "optim/total_tokens": 2123890688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.205977439880371, "created_at": "2025-01-15T16:59:37.144151+00:00"} {"global_step": 4052, "acc_step": 0, "speed/wps": 12907.862160642937, "speed/FLOPS": 202735712494238.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10036413371562958, "optim/lr": 0.0024306, "optim/total_tokens": 2124414976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.2683677673339844, "created_at": "2025-01-15T16:59:47.305141+00:00"} {"global_step": 4053, "acc_step": 0, "speed/wps": 12913.126195568453, "speed/FLOPS": 202818391396285.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09563029557466507, "optim/lr": 0.0024312, "optim/total_tokens": 2124939264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.1619434356689453, "created_at": "2025-01-15T16:59:57.462089+00:00"} {"global_step": 4054, "acc_step": 0, "speed/wps": 12912.056190021722, "speed/FLOPS": 202801585488832.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11282987892627716, "optim/lr": 0.0024318, "optim/total_tokens": 2125463552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 3.3127870559692383, "created_at": "2025-01-15T17:00:07.618493+00:00"} {"global_step": 4055, "acc_step": 0, "speed/wps": 12914.5329110435, "speed/FLOPS": 202840485795850.7, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11565115302801132, "optim/lr": 0.0024324, "optim/total_tokens": 2125987840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 3.208160161972046, "created_at": "2025-01-15T17:00:17.776209+00:00"} {"global_step": 4056, "acc_step": 0, "speed/wps": 12905.655132204889, "speed/FLOPS": 202701048079843.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11661536246538162, "optim/lr": 0.0024330000000000003, "optim/total_tokens": 2126512128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 3.251037836074829, "created_at": "2025-01-15T17:00:27.936751+00:00"} {"global_step": 4057, "acc_step": 0, "speed/wps": 12913.063269995384, "speed/FLOPS": 202817403063691.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1032920852303505, "optim/lr": 0.0024336, "optim/total_tokens": 2127036416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.291311740875244, "created_at": "2025-01-15T17:00:38.090805+00:00"} {"global_step": 4058, "acc_step": 0, "speed/wps": 12915.07193273836, "speed/FLOPS": 202848951872261.88, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09471629559993744, "optim/lr": 0.0024342, "optim/total_tokens": 2127560704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423026, "loss/out": 3.2325143814086914, "created_at": "2025-01-15T17:00:48.243455+00:00"} {"global_step": 4059, "acc_step": 0, "speed/wps": 12915.336507984963, "speed/FLOPS": 202853107390074.28, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10478656738996506, "optim/lr": 0.0024348, "optim/total_tokens": 2128084992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.337399482727051, "created_at": "2025-01-15T17:00:58.400581+00:00"} {"global_step": 4060, "acc_step": 0, "speed/wps": 12915.539738556925, "speed/FLOPS": 202856299405475.0, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08969558775424957, "optim/lr": 0.0024354, "optim/total_tokens": 2128609280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377259, "loss/out": 3.1947038173675537, "created_at": "2025-01-15T17:01:08.557204+00:00"} {"global_step": 4061, "acc_step": 0, "speed/wps": 12915.591327405908, "speed/FLOPS": 202857109679238.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.095552459359169, "optim/lr": 0.0024360000000000002, "optim/total_tokens": 2129133568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.224443197250366, "created_at": "2025-01-15T17:01:18.709568+00:00"} {"global_step": 4062, "acc_step": 0, "speed/wps": 12909.855026792755, "speed/FLOPS": 202767013195606.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13609381020069122, "optim/lr": 0.0024366, "optim/total_tokens": 2129657856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.1681485176086426, "created_at": "2025-01-15T17:01:28.866399+00:00"} {"global_step": 4063, "acc_step": 0, "speed/wps": 12912.112913686087, "speed/FLOPS": 202802476411927.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09808430075645447, "optim/lr": 0.0024372, "optim/total_tokens": 2130182144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.2927310466766357, "created_at": "2025-01-15T17:01:39.024049+00:00"} {"global_step": 4064, "acc_step": 0, "speed/wps": 12913.872319707662, "speed/FLOPS": 202830110301173.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10497178137302399, "optim/lr": 0.0024378, "optim/total_tokens": 2130706432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.225412368774414, "created_at": "2025-01-15T17:01:49.178292+00:00"} {"global_step": 4065, "acc_step": 0, "speed/wps": 12913.832051384854, "speed/FLOPS": 202829477831829.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11245416104793549, "optim/lr": 0.0024384, "optim/total_tokens": 2131230720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.182965040206909, "created_at": "2025-01-15T17:01:59.340732+00:00"} {"global_step": 4066, "acc_step": 0, "speed/wps": 12909.286528411481, "speed/FLOPS": 202758084147329.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10899641364812851, "optim/lr": 0.0024389999999999998, "optim/total_tokens": 2131755008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.1768429279327393, "created_at": "2025-01-15T17:02:09.500731+00:00"} {"global_step": 4067, "acc_step": 0, "speed/wps": 12901.748729897205, "speed/FLOPS": 202639692663642.28, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1179666817188263, "optim/lr": 0.0024396, "optim/total_tokens": 2132279296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346379, "loss/out": 3.246396780014038, "created_at": "2025-01-15T17:02:19.665041+00:00"} {"global_step": 4068, "acc_step": 0, "speed/wps": 12904.037204750772, "speed/FLOPS": 202675636305911.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10442616045475006, "optim/lr": 0.0024402, "optim/total_tokens": 2132803584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.160672903060913, "created_at": "2025-01-15T17:02:29.829649+00:00"} {"global_step": 4069, "acc_step": 0, "speed/wps": 12907.864185036839, "speed/FLOPS": 202735744290126.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12339618802070618, "optim/lr": 0.0024408, "optim/total_tokens": 2133327872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.286759853363037, "created_at": "2025-01-15T17:02:39.987736+00:00"} {"global_step": 4070, "acc_step": 0, "speed/wps": 12899.640418061086, "speed/FLOPS": 202606578729132.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10131219774484634, "optim/lr": 0.0024414, "optim/total_tokens": 2133852160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.281158447265625, "created_at": "2025-01-15T17:02:50.154016+00:00"} {"global_step": 4071, "acc_step": 0, "speed/wps": 12912.389718055121, "speed/FLOPS": 202806824004912.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11724280565977097, "optim/lr": 0.0024419999999999997, "optim/total_tokens": 2134376448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.407938003540039, "created_at": "2025-01-15T17:03:00.308587+00:00"} {"global_step": 4072, "acc_step": 0, "speed/wps": 12915.320941228014, "speed/FLOPS": 202852862892765.4, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12515290081501007, "optim/lr": 0.0024426, "optim/total_tokens": 2134900736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 3.344761848449707, "created_at": "2025-01-15T17:03:10.461894+00:00"} {"global_step": 4073, "acc_step": 0, "speed/wps": 12911.530339664758, "speed/FLOPS": 202793326286382.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11119221150875092, "optim/lr": 0.0024432, "optim/total_tokens": 2135425024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 3.2471232414245605, "created_at": "2025-01-15T17:03:20.617091+00:00"} {"global_step": 4074, "acc_step": 0, "speed/wps": 12906.767816536967, "speed/FLOPS": 202718524316267.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11412449181079865, "optim/lr": 0.0024438, "optim/total_tokens": 2135949312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 3.263737201690674, "created_at": "2025-01-15T17:03:30.777378+00:00"} {"global_step": 4075, "acc_step": 0, "speed/wps": 12907.42659568935, "speed/FLOPS": 202728871348113.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1278681606054306, "optim/lr": 0.0024444, "optim/total_tokens": 2136473600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.2546474933624268, "created_at": "2025-01-15T17:03:40.936192+00:00"} {"global_step": 4076, "acc_step": 0, "speed/wps": 12897.626125915613, "speed/FLOPS": 202574941503059.88, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11397723108530045, "optim/lr": 0.0024449999999999997, "optim/total_tokens": 2136997888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.2624738216400146, "created_at": "2025-01-15T17:03:51.105240+00:00"} {"global_step": 4077, "acc_step": 0, "speed/wps": 12899.454579240648, "speed/FLOPS": 202603659875087.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09789326041936874, "optim/lr": 0.0024456, "optim/total_tokens": 2137522176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453149, "loss/out": 3.2495386600494385, "created_at": "2025-01-15T17:04:01.271093+00:00"} {"global_step": 4078, "acc_step": 0, "speed/wps": 12906.046423769258, "speed/FLOPS": 202707193851553.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09370578080415726, "optim/lr": 0.0024462, "optim/total_tokens": 2138046464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.1754226684570312, "created_at": "2025-01-15T17:04:11.433783+00:00"} {"global_step": 4079, "acc_step": 0, "speed/wps": 12906.363722212636, "speed/FLOPS": 202712177459620.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08771216124296188, "optim/lr": 0.0024468, "optim/total_tokens": 2138570752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383767, "loss/out": 3.1745541095733643, "created_at": "2025-01-15T17:04:21.595915+00:00"} {"global_step": 4080, "acc_step": 0, "speed/wps": 12913.92357766562, "speed/FLOPS": 202830915377839.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09012281894683838, "optim/lr": 0.0024473999999999997, "optim/total_tokens": 2139095040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.174579620361328, "created_at": "2025-01-15T17:04:31.749736+00:00"} {"global_step": 4081, "acc_step": 0, "speed/wps": 12908.15531369793, "speed/FLOPS": 202740316865801.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09640634804964066, "optim/lr": 0.002448, "optim/total_tokens": 2139619328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489248, "loss/out": 3.2919845581054688, "created_at": "2025-01-15T17:04:41.910584+00:00"} {"global_step": 4082, "acc_step": 0, "speed/wps": 12913.288173388002, "speed/FLOPS": 202820935480522.3, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09168749302625656, "optim/lr": 0.0024486, "optim/total_tokens": 2140143616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.2056593894958496, "created_at": "2025-01-15T17:04:52.064726+00:00"} {"global_step": 4083, "acc_step": 0, "speed/wps": 12915.022006090594, "speed/FLOPS": 202848167705652.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07959986478090286, "optim/lr": 0.0024492, "optim/total_tokens": 2140667904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.3234663009643555, "created_at": "2025-01-15T17:05:02.219992+00:00"} {"global_step": 4084, "acc_step": 0, "speed/wps": 12913.877050064159, "speed/FLOPS": 202830184597922.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10428828001022339, "optim/lr": 0.0024498000000000002, "optim/total_tokens": 2141192192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.2804813385009766, "created_at": "2025-01-15T17:05:12.376071+00:00"} {"global_step": 4085, "acc_step": 0, "speed/wps": 12908.240875417028, "speed/FLOPS": 202741660730173.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10189925134181976, "optim/lr": 0.0024504, "optim/total_tokens": 2141716480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.208674907684326, "created_at": "2025-01-15T17:05:22.537797+00:00"} {"global_step": 4086, "acc_step": 0, "speed/wps": 12912.093632583434, "speed/FLOPS": 202802173575715.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11772670596837997, "optim/lr": 0.002451, "optim/total_tokens": 2142240768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 3.1763105392456055, "created_at": "2025-01-15T17:05:32.693325+00:00"} {"global_step": 4087, "acc_step": 0, "speed/wps": 12910.521092184914, "speed/FLOPS": 202777474667859.78, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12072282284498215, "optim/lr": 0.0024516, "optim/total_tokens": 2142765056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.3174092769622803, "created_at": "2025-01-15T17:05:42.855011+00:00"} {"global_step": 4088, "acc_step": 0, "speed/wps": 12908.484894151508, "speed/FLOPS": 202745493379715.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09588383138179779, "optim/lr": 0.0024522000000000003, "optim/total_tokens": 2143289344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416384, "loss/out": 3.3040871620178223, "created_at": "2025-01-15T17:05:53.012801+00:00"} {"global_step": 4089, "acc_step": 0, "speed/wps": 12911.349701541858, "speed/FLOPS": 202790489116439.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10582684725522995, "optim/lr": 0.0024528, "optim/total_tokens": 2143813632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.2302465438842773, "created_at": "2025-01-15T17:06:03.168071+00:00"} {"global_step": 4090, "acc_step": 0, "speed/wps": 12909.242377569086, "speed/FLOPS": 202757390697680.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12030710279941559, "optim/lr": 0.0024534, "optim/total_tokens": 2144337920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.2126893997192383, "created_at": "2025-01-15T17:06:13.325035+00:00"} {"global_step": 4091, "acc_step": 0, "speed/wps": 12906.54572403751, "speed/FLOPS": 202715036048376.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1135135293006897, "optim/lr": 0.002454, "optim/total_tokens": 2144862208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 3.197136878967285, "created_at": "2025-01-15T17:06:23.487349+00:00"} {"global_step": 4092, "acc_step": 0, "speed/wps": 12909.408857164794, "speed/FLOPS": 202760005488497.7, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10716848075389862, "optim/lr": 0.0024546000000000004, "optim/total_tokens": 2145386496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.238135814666748, "created_at": "2025-01-15T17:06:33.649465+00:00"} {"global_step": 4093, "acc_step": 0, "speed/wps": 12913.466599072519, "speed/FLOPS": 202823737901080.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10585802793502808, "optim/lr": 0.0024552000000000003, "optim/total_tokens": 2145910784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.146131753921509, "created_at": "2025-01-15T17:06:43.805600+00:00"} {"global_step": 4094, "acc_step": 0, "speed/wps": 12908.510264313298, "speed/FLOPS": 202745891852969.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11864792555570602, "optim/lr": 0.0024558, "optim/total_tokens": 2146435072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.23048734664917, "created_at": "2025-01-15T17:06:53.963428+00:00"} {"global_step": 4095, "acc_step": 0, "speed/wps": 12911.124394253247, "speed/FLOPS": 202786950355866.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14807289838790894, "optim/lr": 0.0024564, "optim/total_tokens": 2146959360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 3.1835427284240723, "created_at": "2025-01-15T17:07:04.121984+00:00"} {"global_step": 4096, "acc_step": 0, "speed/wps": 12907.052538048281, "speed/FLOPS": 202722996258851.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18227633833885193, "optim/lr": 0.002457, "optim/total_tokens": 2147483648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.235063076019287, "created_at": "2025-01-15T17:07:14.283696+00:00"} {"global_step": 4097, "acc_step": 0, "speed/wps": 12908.717304568388, "speed/FLOPS": 202749143704678.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19996760785579681, "optim/lr": 0.0024576000000000003, "optim/total_tokens": 2148007936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2631335258483887, "created_at": "2025-01-15T17:07:24.441214+00:00"} {"global_step": 4098, "acc_step": 0, "speed/wps": 12912.758143572124, "speed/FLOPS": 202812610633926.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.28455525636672974, "optim/lr": 0.0024582000000000002, "optim/total_tokens": 2148532224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0877275466918945, "created_at": "2025-01-15T17:07:34.596515+00:00"} {"global_step": 4099, "acc_step": 0, "speed/wps": 12906.524816706029, "speed/FLOPS": 202714707670005.9, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16814549267292023, "optim/lr": 0.0024588, "optim/total_tokens": 2149056512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.211027145385742, "created_at": "2025-01-15T17:07:44.756350+00:00"} {"global_step": 4100, "acc_step": 0, "speed/wps": 12910.389965353168, "speed/FLOPS": 202775415140780.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14271104335784912, "optim/lr": 0.0024594, "optim/total_tokens": 2149580800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 3.266848087310791, "created_at": "2025-01-15T17:07:54.917005+00:00"} {"global_step": 4101, "acc_step": 0, "speed/wps": 12910.11456125299, "speed/FLOPS": 202771089540941.88, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12563352286815643, "optim/lr": 0.00246, "optim/total_tokens": 2150105088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.3220670223236084, "created_at": "2025-01-15T17:08:05.074356+00:00"} {"global_step": 4102, "acc_step": 0, "speed/wps": 12910.57994249063, "speed/FLOPS": 202778398992779.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1543203592300415, "optim/lr": 0.0024606000000000003, "optim/total_tokens": 2150629376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353101, "loss/out": 3.0839571952819824, "created_at": "2025-01-15T17:08:15.231251+00:00"} {"global_step": 4103, "acc_step": 0, "speed/wps": 12912.84367627952, "speed/FLOPS": 202813954042630.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.150105819106102, "optim/lr": 0.0024612, "optim/total_tokens": 2151153664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.14547061920166, "created_at": "2025-01-15T17:08:25.385360+00:00"} {"global_step": 4104, "acc_step": 0, "speed/wps": 12908.300864010214, "speed/FLOPS": 202742602933462.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.13328292965888977, "optim/lr": 0.0024618, "optim/total_tokens": 2151677952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.321504592895508, "created_at": "2025-01-15T17:08:35.544372+00:00"} {"global_step": 4105, "acc_step": 0, "speed/wps": 12913.475111285881, "speed/FLOPS": 202823871597088.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13262732326984406, "optim/lr": 0.0024624, "optim/total_tokens": 2152202240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.0898513793945312, "created_at": "2025-01-15T17:08:45.705809+00:00"} {"global_step": 4106, "acc_step": 0, "speed/wps": 12916.126907417793, "speed/FLOPS": 202865521699289.5, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13037452101707458, "optim/lr": 0.002463, "optim/total_tokens": 2152726528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 3.127577543258667, "created_at": "2025-01-15T17:08:55.859353+00:00"} {"global_step": 4107, "acc_step": 0, "speed/wps": 12913.623945310961, "speed/FLOPS": 202826209239971.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09440334886312485, "optim/lr": 0.0024636000000000003, "optim/total_tokens": 2153250816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.4412307739257812, "created_at": "2025-01-15T17:09:06.017956+00:00"} {"global_step": 4108, "acc_step": 0, "speed/wps": 12906.475460332811, "speed/FLOPS": 202713932460341.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10731048136949539, "optim/lr": 0.0024642, "optim/total_tokens": 2153775104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.2869091033935547, "created_at": "2025-01-15T17:09:16.179847+00:00"} {"global_step": 4109, "acc_step": 0, "speed/wps": 12903.325379071315, "speed/FLOPS": 202664456105464.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1052456870675087, "optim/lr": 0.0024648, "optim/total_tokens": 2154299392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 3.47571063041687, "created_at": "2025-01-15T17:09:26.341687+00:00"} {"global_step": 4110, "acc_step": 0, "speed/wps": 12910.198284883894, "speed/FLOPS": 202772404535613.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11050836741924286, "optim/lr": 0.0024654, "optim/total_tokens": 2154823680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.3065996170043945, "created_at": "2025-01-15T17:09:36.498861+00:00"} {"global_step": 4111, "acc_step": 0, "speed/wps": 12907.781883873804, "speed/FLOPS": 202734451637267.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12579570710659027, "optim/lr": 0.002466, "optim/total_tokens": 2155347968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 3.2876222133636475, "created_at": "2025-01-15T17:09:46.656957+00:00"} {"global_step": 4112, "acc_step": 0, "speed/wps": 12910.216361827133, "speed/FLOPS": 202772688458847.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.13513334095478058, "optim/lr": 0.0024666, "optim/total_tokens": 2155872256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.243248462677002, "created_at": "2025-01-15T17:09:56.817558+00:00"} {"global_step": 4113, "acc_step": 0, "speed/wps": 12911.40590492343, "speed/FLOPS": 202791371867778.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11749555915594101, "optim/lr": 0.0024672, "optim/total_tokens": 2156396544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306453, "loss/out": 3.3135175704956055, "created_at": "2025-01-15T17:10:06.975645+00:00"} {"global_step": 4114, "acc_step": 0, "speed/wps": 12912.482098203112, "speed/FLOPS": 202808274962080.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11347200721502304, "optim/lr": 0.0024678, "optim/total_tokens": 2156920832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.1181411743164062, "created_at": "2025-01-15T17:10:17.135601+00:00"} {"global_step": 4115, "acc_step": 0, "speed/wps": 12913.664868781876, "speed/FLOPS": 202826851999316.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10021571815013885, "optim/lr": 0.0024684, "optim/total_tokens": 2157445120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.3044328689575195, "created_at": "2025-01-15T17:10:27.291323+00:00"} {"global_step": 4116, "acc_step": 0, "speed/wps": 12911.850287651654, "speed/FLOPS": 202798351509170.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09813597798347473, "optim/lr": 0.002469, "optim/total_tokens": 2157969408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.2051210403442383, "created_at": "2025-01-15T17:10:37.449187+00:00"} {"global_step": 4117, "acc_step": 0, "speed/wps": 12916.977555851588, "speed/FLOPS": 202878882301855.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09644049406051636, "optim/lr": 0.0024696, "optim/total_tokens": 2158493696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.1772329807281494, "created_at": "2025-01-15T17:10:47.600786+00:00"} {"global_step": 4118, "acc_step": 0, "speed/wps": 12915.122534144104, "speed/FLOPS": 202849746636872.72, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0882466658949852, "optim/lr": 0.0024702, "optim/total_tokens": 2159017984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.336676836013794, "created_at": "2025-01-15T17:10:57.754840+00:00"} {"global_step": 4119, "acc_step": 0, "speed/wps": 12916.329022501492, "speed/FLOPS": 202868696194414.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08742551505565643, "optim/lr": 0.0024708, "optim/total_tokens": 2159542272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.1665964126586914, "created_at": "2025-01-15T17:11:07.906204+00:00"} {"global_step": 4120, "acc_step": 0, "speed/wps": 12909.409909865008, "speed/FLOPS": 202760022022601.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0860830619931221, "optim/lr": 0.0024714, "optim/total_tokens": 2160066560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.2411446571350098, "created_at": "2025-01-15T17:11:18.065714+00:00"} {"global_step": 4121, "acc_step": 0, "speed/wps": 12909.950402326956, "speed/FLOPS": 202768511199430.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09352463483810425, "optim/lr": 0.002472, "optim/total_tokens": 2160590848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.154822826385498, "created_at": "2025-01-15T17:11:28.226099+00:00"} {"global_step": 4122, "acc_step": 0, "speed/wps": 12910.170077658364, "speed/FLOPS": 202771961502375.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10371613502502441, "optim/lr": 0.0024726, "optim/total_tokens": 2161115136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 3.308730363845825, "created_at": "2025-01-15T17:11:38.382204+00:00"} {"global_step": 4123, "acc_step": 0, "speed/wps": 12910.860407276492, "speed/FLOPS": 202782804077640.8, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.09637832641601562, "optim/lr": 0.0024732, "optim/total_tokens": 2161639424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 3.34456205368042, "created_at": "2025-01-15T17:11:48.540146+00:00"} {"global_step": 4124, "acc_step": 0, "speed/wps": 12912.626992620199, "speed/FLOPS": 202810550728006.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09365524351596832, "optim/lr": 0.0024738, "optim/total_tokens": 2162163712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 3.2632312774658203, "created_at": "2025-01-15T17:11:58.694531+00:00"} {"global_step": 4125, "acc_step": 0, "speed/wps": 12911.814683786144, "speed/FLOPS": 202797792301538.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10364893823862076, "optim/lr": 0.0024744, "optim/total_tokens": 2162688000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.2355146408081055, "created_at": "2025-01-15T17:12:08.853207+00:00"} {"global_step": 4126, "acc_step": 0, "speed/wps": 12912.473840825052, "speed/FLOPS": 202808145268610.97, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09471164643764496, "optim/lr": 0.0024749999999999998, "optim/total_tokens": 2163212288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 3.246859073638916, "created_at": "2025-01-15T17:12:19.009914+00:00"} {"global_step": 4127, "acc_step": 0, "speed/wps": 12913.102115297426, "speed/FLOPS": 202818013182539.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08572898805141449, "optim/lr": 0.0024756, "optim/total_tokens": 2163736576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.128505229949951, "created_at": "2025-01-15T17:12:29.163827+00:00"} {"global_step": 4128, "acc_step": 0, "speed/wps": 12909.289225830995, "speed/FLOPS": 202758126514009.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10367508977651596, "optim/lr": 0.0024762, "optim/total_tokens": 2164260864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 3.2421157360076904, "created_at": "2025-01-15T17:12:39.320671+00:00"} {"global_step": 4129, "acc_step": 0, "speed/wps": 12914.453673536264, "speed/FLOPS": 202839241261915.78, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09164878726005554, "optim/lr": 0.0024768, "optim/total_tokens": 2164785152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.2204055786132812, "created_at": "2025-01-15T17:12:49.474673+00:00"} {"global_step": 4130, "acc_step": 0, "speed/wps": 12912.165555559472, "speed/FLOPS": 202803303224886.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09330590814352036, "optim/lr": 0.0024774, "optim/total_tokens": 2165309440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.28501558303833, "created_at": "2025-01-15T17:12:59.629403+00:00"} {"global_step": 4131, "acc_step": 0, "speed/wps": 12912.860267817703, "speed/FLOPS": 202814214635536.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09915561974048615, "optim/lr": 0.0024779999999999997, "optim/total_tokens": 2165833728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458694, "loss/out": 3.2113518714904785, "created_at": "2025-01-15T17:13:09.784534+00:00"} {"global_step": 4132, "acc_step": 0, "speed/wps": 12912.505101631801, "speed/FLOPS": 202808636262537.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09739930927753448, "optim/lr": 0.0024786, "optim/total_tokens": 2166358016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.2187938690185547, "created_at": "2025-01-15T17:13:19.940779+00:00"} {"global_step": 4133, "acc_step": 0, "speed/wps": 12909.941614095484, "speed/FLOPS": 202768373168178.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09149972349405289, "optim/lr": 0.0024792, "optim/total_tokens": 2166882304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.1395835876464844, "created_at": "2025-01-15T17:13:30.097055+00:00"} {"global_step": 4134, "acc_step": 0, "speed/wps": 12914.538922956332, "speed/FLOPS": 202840580221202.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09547550976276398, "optim/lr": 0.0024798, "optim/total_tokens": 2167406592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.3424699306488037, "created_at": "2025-01-15T17:13:40.249754+00:00"} {"global_step": 4135, "acc_step": 0, "speed/wps": 12909.018018846962, "speed/FLOPS": 202753866835648.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09153826534748077, "optim/lr": 0.0024804, "optim/total_tokens": 2167930880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 3.2319531440734863, "created_at": "2025-01-15T17:13:50.407498+00:00"} {"global_step": 4136, "acc_step": 0, "speed/wps": 12910.551449353501, "speed/FLOPS": 202777951468907.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08449145406484604, "optim/lr": 0.002481, "optim/total_tokens": 2168455168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 3.2651755809783936, "created_at": "2025-01-15T17:14:00.564211+00:00"} {"global_step": 4137, "acc_step": 0, "speed/wps": 12909.8985281312, "speed/FLOPS": 202767696443903.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09686164557933807, "optim/lr": 0.0024816, "optim/total_tokens": 2168979456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350651, "loss/out": 3.1421172618865967, "created_at": "2025-01-15T17:14:10.721325+00:00"} {"global_step": 4138, "acc_step": 0, "speed/wps": 12911.427795816331, "speed/FLOPS": 202791715694332.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09908924996852875, "optim/lr": 0.0024822, "optim/total_tokens": 2169503744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.195819139480591, "created_at": "2025-01-15T17:14:20.877957+00:00"} {"global_step": 4139, "acc_step": 0, "speed/wps": 12907.70361468932, "speed/FLOPS": 202733222312171.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11611481010913849, "optim/lr": 0.0024828, "optim/total_tokens": 2170028032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.1562275886535645, "created_at": "2025-01-15T17:14:31.036128+00:00"} {"global_step": 4140, "acc_step": 0, "speed/wps": 12910.175045711858, "speed/FLOPS": 202772039532482.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.132134810090065, "optim/lr": 0.0024834, "optim/total_tokens": 2170552320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.194026470184326, "created_at": "2025-01-15T17:14:41.192277+00:00"} {"global_step": 4141, "acc_step": 0, "speed/wps": 12915.951006476127, "speed/FLOPS": 202862758933287.4, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11872248351573944, "optim/lr": 0.002484, "optim/total_tokens": 2171076608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 3.114915370941162, "created_at": "2025-01-15T17:14:51.343861+00:00"} {"global_step": 4142, "acc_step": 0, "speed/wps": 12911.631731392696, "speed/FLOPS": 202794918782796.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08923487365245819, "optim/lr": 0.0024846, "optim/total_tokens": 2171600896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.2432870864868164, "created_at": "2025-01-15T17:15:01.502099+00:00"} {"global_step": 4143, "acc_step": 0, "speed/wps": 12911.65050057444, "speed/FLOPS": 202795213578587.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08105753362178802, "optim/lr": 0.0024852000000000003, "optim/total_tokens": 2172125184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.176947593688965, "created_at": "2025-01-15T17:15:11.658268+00:00"} {"global_step": 4144, "acc_step": 0, "speed/wps": 12912.969444377473, "speed/FLOPS": 202815929403432.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08471379429101944, "optim/lr": 0.0024858000000000002, "optim/total_tokens": 2172649472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.2889199256896973, "created_at": "2025-01-15T17:15:21.813308+00:00"} {"global_step": 4145, "acc_step": 0, "speed/wps": 12913.490250588404, "speed/FLOPS": 202824109380637.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11257380247116089, "optim/lr": 0.0024864, "optim/total_tokens": 2173173760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 3.189167022705078, "created_at": "2025-01-15T17:15:31.967538+00:00"} {"global_step": 4146, "acc_step": 0, "speed/wps": 12912.032329342208, "speed/FLOPS": 202801210724072.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10161919891834259, "optim/lr": 0.002487, "optim/total_tokens": 2173698048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.211897611618042, "created_at": "2025-01-15T17:15:42.124992+00:00"} {"global_step": 4147, "acc_step": 0, "speed/wps": 12913.535870141424, "speed/FLOPS": 202824825898405.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08924254775047302, "optim/lr": 0.0024876000000000004, "optim/total_tokens": 2174222336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.2287166118621826, "created_at": "2025-01-15T17:15:52.278498+00:00"} {"global_step": 4148, "acc_step": 0, "speed/wps": 12917.424351962527, "speed/FLOPS": 202885899848740.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11038342863321304, "optim/lr": 0.0024882000000000003, "optim/total_tokens": 2174746624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 3.318575620651245, "created_at": "2025-01-15T17:16:02.432583+00:00"} {"global_step": 4149, "acc_step": 0, "speed/wps": 12911.111725341736, "speed/FLOPS": 202786751373202.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12467411905527115, "optim/lr": 0.0024888, "optim/total_tokens": 2175270912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.2384209632873535, "created_at": "2025-01-15T17:16:12.592814+00:00"} {"global_step": 4150, "acc_step": 0, "speed/wps": 12908.36819482282, "speed/FLOPS": 202743660456397.1, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11035237461328506, "optim/lr": 0.0024894, "optim/total_tokens": 2175795200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.2355599403381348, "created_at": "2025-01-15T17:16:22.750674+00:00"} {"global_step": 4151, "acc_step": 0, "speed/wps": 12914.356872202921, "speed/FLOPS": 202837720863956.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09122846275568008, "optim/lr": 0.00249, "optim/total_tokens": 2176319488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.238861083984375, "created_at": "2025-01-15T17:16:32.904982+00:00"} {"global_step": 4152, "acc_step": 0, "speed/wps": 12913.097018215558, "speed/FLOPS": 202817933125864.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10151176899671555, "optim/lr": 0.0024906000000000004, "optim/total_tokens": 2176843776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 3.2457175254821777, "created_at": "2025-01-15T17:16:43.058920+00:00"} {"global_step": 4153, "acc_step": 0, "speed/wps": 12910.573975715532, "speed/FLOPS": 202778305276377.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.10342386364936829, "optim/lr": 0.0024912000000000003, "optim/total_tokens": 2177368064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.184262275695801, "created_at": "2025-01-15T17:16:53.217746+00:00"} {"global_step": 4154, "acc_step": 0, "speed/wps": 12910.578727343052, "speed/FLOPS": 202778379907217.34, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09419671446084976, "optim/lr": 0.0024918, "optim/total_tokens": 2177892352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.2394614219665527, "created_at": "2025-01-15T17:17:03.373877+00:00"} {"global_step": 4155, "acc_step": 0, "speed/wps": 12908.823630974302, "speed/FLOPS": 202750813706990.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10261976718902588, "optim/lr": 0.0024924, "optim/total_tokens": 2178416640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 3.22784686088562, "created_at": "2025-01-15T17:17:13.531454+00:00"} {"global_step": 4156, "acc_step": 0, "speed/wps": 12914.789872277755, "speed/FLOPS": 202844521725136.4, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09666290879249573, "optim/lr": 0.002493, "optim/total_tokens": 2178940928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.2357494831085205, "created_at": "2025-01-15T17:17:23.689369+00:00"} {"global_step": 4157, "acc_step": 0, "speed/wps": 12911.390384359684, "speed/FLOPS": 202791128095997.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10123130679130554, "optim/lr": 0.0024936000000000003, "optim/total_tokens": 2179465216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333415, "loss/out": 3.266104221343994, "created_at": "2025-01-15T17:17:33.846183+00:00"} {"global_step": 4158, "acc_step": 0, "speed/wps": 12909.90241358923, "speed/FLOPS": 202767757470361.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10836222767829895, "optim/lr": 0.0024942000000000002, "optim/total_tokens": 2179989504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.28292179107666, "created_at": "2025-01-15T17:17:44.003733+00:00"} {"global_step": 4159, "acc_step": 0, "speed/wps": 12912.659109547822, "speed/FLOPS": 202811055168487.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.11531080305576324, "optim/lr": 0.0024948, "optim/total_tokens": 2180513792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.246385335922241, "created_at": "2025-01-15T17:17:54.158157+00:00"} {"global_step": 4160, "acc_step": 0, "speed/wps": 12912.131171318064, "speed/FLOPS": 202802763173126.47, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.125758558511734, "optim/lr": 0.0024954, "optim/total_tokens": 2181038080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.310471534729004, "created_at": "2025-01-15T17:18:04.312703+00:00"} {"global_step": 4161, "acc_step": 0, "speed/wps": 12911.714737904844, "speed/FLOPS": 202796222514132.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1316344290971756, "optim/lr": 0.002496, "optim/total_tokens": 2181562368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374420, "loss/out": 3.3159687519073486, "created_at": "2025-01-15T17:18:14.469524+00:00"} {"global_step": 4162, "acc_step": 0, "speed/wps": 12908.80736594353, "speed/FLOPS": 202750558242332.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12298706918954849, "optim/lr": 0.0024966000000000003, "optim/total_tokens": 2182086656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.255728244781494, "created_at": "2025-01-15T17:18:24.633370+00:00"} {"global_step": 4163, "acc_step": 0, "speed/wps": 12912.41258965399, "speed/FLOPS": 202807183234801.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11401065438985825, "optim/lr": 0.0024972, "optim/total_tokens": 2182610944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.127537250518799, "created_at": "2025-01-15T17:18:34.788921+00:00"} {"global_step": 4164, "acc_step": 0, "speed/wps": 12916.696986462843, "speed/FLOPS": 202874475574062.72, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08895464986562729, "optim/lr": 0.0024978, "optim/total_tokens": 2183135232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.0946383476257324, "created_at": "2025-01-15T17:18:44.939989+00:00"} {"global_step": 4165, "acc_step": 0, "speed/wps": 12913.809484028343, "speed/FLOPS": 202829123380484.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10158251971006393, "optim/lr": 0.0024984, "optim/total_tokens": 2183659520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.125037431716919, "created_at": "2025-01-15T17:18:55.098278+00:00"} {"global_step": 4166, "acc_step": 0, "speed/wps": 12914.241249335328, "speed/FLOPS": 202835904847938.66, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09677522629499435, "optim/lr": 0.002499, "optim/total_tokens": 2184183808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.1524100303649902, "created_at": "2025-01-15T17:19:05.252384+00:00"} {"global_step": 4167, "acc_step": 0, "speed/wps": 12911.371868232622, "speed/FLOPS": 202790837274777.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09952349960803986, "optim/lr": 0.0024996000000000003, "optim/total_tokens": 2184708096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.2582311630249023, "created_at": "2025-01-15T17:19:15.407814+00:00"} {"global_step": 4168, "acc_step": 0, "speed/wps": 12917.956613441997, "speed/FLOPS": 202894259746679.2, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10119839012622833, "optim/lr": 0.0025002, "optim/total_tokens": 2185232384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.2994771003723145, "created_at": "2025-01-15T17:19:25.561885+00:00"} {"global_step": 4169, "acc_step": 0, "speed/wps": 12912.889262829658, "speed/FLOPS": 202814670042042.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11047573387622833, "optim/lr": 0.0025008, "optim/total_tokens": 2185756672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.198848247528076, "created_at": "2025-01-15T17:19:35.715964+00:00"} {"global_step": 4170, "acc_step": 0, "speed/wps": 12917.091139968788, "speed/FLOPS": 202880666296497.25, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.108640156686306, "optim/lr": 0.0025014, "optim/total_tokens": 2186280960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.228079319000244, "created_at": "2025-01-15T17:19:45.866935+00:00"} {"global_step": 4171, "acc_step": 0, "speed/wps": 12911.597312088985, "speed/FLOPS": 202794378180334.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10212361067533493, "optim/lr": 0.002502, "optim/total_tokens": 2186805248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.2109975814819336, "created_at": "2025-01-15T17:19:56.022270+00:00"} {"global_step": 4172, "acc_step": 0, "speed/wps": 12908.046593820687, "speed/FLOPS": 202738609270732.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12325803190469742, "optim/lr": 0.0025026000000000002, "optim/total_tokens": 2187329536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.1895675659179688, "created_at": "2025-01-15T17:20:06.180110+00:00"} {"global_step": 4173, "acc_step": 0, "speed/wps": 12912.996134262761, "speed/FLOPS": 202816348604756.6, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1329531967639923, "optim/lr": 0.0025032, "optim/total_tokens": 2187853824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.2365031242370605, "created_at": "2025-01-15T17:20:16.334166+00:00"} {"global_step": 4174, "acc_step": 0, "speed/wps": 12916.850826465019, "speed/FLOPS": 202876891842695.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12826481461524963, "optim/lr": 0.0025038, "optim/total_tokens": 2188378112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 3.1038427352905273, "created_at": "2025-01-15T17:20:26.488533+00:00"} {"global_step": 4175, "acc_step": 0, "speed/wps": 12912.934828375506, "speed/FLOPS": 202815385711554.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09641702473163605, "optim/lr": 0.0025044, "optim/total_tokens": 2188902400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.127891778945923, "created_at": "2025-01-15T17:20:36.642566+00:00"} {"global_step": 4176, "acc_step": 0, "speed/wps": 12912.235087511484, "speed/FLOPS": 202804395319739.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10023383051156998, "optim/lr": 0.002505, "optim/total_tokens": 2189426688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439566, "loss/out": 3.1787338256835938, "created_at": "2025-01-15T17:20:46.799200+00:00"} {"global_step": 4177, "acc_step": 0, "speed/wps": 12908.10589394572, "speed/FLOPS": 202739540660683.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.11565064638853073, "optim/lr": 0.0025056, "optim/total_tokens": 2189950976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.1490488052368164, "created_at": "2025-01-15T17:20:56.961846+00:00"} {"global_step": 4178, "acc_step": 0, "speed/wps": 12915.08285535512, "speed/FLOPS": 202849123426967.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08264829963445663, "optim/lr": 0.0025062, "optim/total_tokens": 2190475264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 3.318843126296997, "created_at": "2025-01-15T17:21:07.114328+00:00"} {"global_step": 4179, "acc_step": 0, "speed/wps": 12908.852978464818, "speed/FLOPS": 202751274649657.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1108052060008049, "optim/lr": 0.0025068, "optim/total_tokens": 2190999552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.241018772125244, "created_at": "2025-01-15T17:21:17.275191+00:00"} {"global_step": 4180, "acc_step": 0, "speed/wps": 12914.031531524128, "speed/FLOPS": 202832610941531.28, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10057098418474197, "optim/lr": 0.0025074, "optim/total_tokens": 2191523840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.167863368988037, "created_at": "2025-01-15T17:21:27.428283+00:00"} {"global_step": 4181, "acc_step": 0, "speed/wps": 12907.935996973754, "speed/FLOPS": 202736872195275.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1002352386713028, "optim/lr": 0.002508, "optim/total_tokens": 2192048128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.1521153450012207, "created_at": "2025-01-15T17:21:37.592090+00:00"} {"global_step": 4182, "acc_step": 0, "speed/wps": 12910.190028057952, "speed/FLOPS": 202772274850816.06, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09684669226408005, "optim/lr": 0.0025086, "optim/total_tokens": 2192572416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.1258349418640137, "created_at": "2025-01-15T17:21:47.751525+00:00"} {"global_step": 4183, "acc_step": 0, "speed/wps": 12913.294312106074, "speed/FLOPS": 202821031897525.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11625435948371887, "optim/lr": 0.0025092, "optim/total_tokens": 2193096704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306740, "loss/out": 3.1946492195129395, "created_at": "2025-01-15T17:21:57.906415+00:00"} {"global_step": 4184, "acc_step": 0, "speed/wps": 12911.719810204266, "speed/FLOPS": 202796302181564.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11571210622787476, "optim/lr": 0.0025098, "optim/total_tokens": 2193620992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.292025566101074, "created_at": "2025-01-15T17:22:08.069784+00:00"} {"global_step": 4185, "acc_step": 0, "speed/wps": 12906.85290888436, "speed/FLOPS": 202719860808512.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09837421774864197, "optim/lr": 0.0025104, "optim/total_tokens": 2194145280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 3.2834081649780273, "created_at": "2025-01-15T17:22:18.229420+00:00"} {"global_step": 4186, "acc_step": 0, "speed/wps": 12913.805328420283, "speed/FLOPS": 202829058110949.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10381852835416794, "optim/lr": 0.0025109999999999998, "optim/total_tokens": 2194669568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.212381362915039, "created_at": "2025-01-15T17:22:28.386021+00:00"} {"global_step": 4187, "acc_step": 0, "speed/wps": 12910.30857586886, "speed/FLOPS": 202774136807088.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10702329128980637, "optim/lr": 0.0025116, "optim/total_tokens": 2195193856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.143523693084717, "created_at": "2025-01-15T17:22:38.543589+00:00"} {"global_step": 4188, "acc_step": 0, "speed/wps": 12916.180296811357, "speed/FLOPS": 202866360253080.16, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0959821417927742, "optim/lr": 0.0025122, "optim/total_tokens": 2195718144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.131288528442383, "created_at": "2025-01-15T17:22:48.699656+00:00"} {"global_step": 4189, "acc_step": 0, "speed/wps": 12908.67957137594, "speed/FLOPS": 202748551053040.06, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09432753920555115, "optim/lr": 0.0025128, "optim/total_tokens": 2196242432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.263942241668701, "created_at": "2025-01-15T17:22:58.860490+00:00"} {"global_step": 4190, "acc_step": 0, "speed/wps": 12908.522896152619, "speed/FLOPS": 202746090253363.5, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09189863502979279, "optim/lr": 0.0025134, "optim/total_tokens": 2196766720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343977, "loss/out": 3.2323389053344727, "created_at": "2025-01-15T17:23:09.018621+00:00"} {"global_step": 4191, "acc_step": 0, "speed/wps": 12910.631233533746, "speed/FLOPS": 202779204589093.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11200562119483948, "optim/lr": 0.0025139999999999997, "optim/total_tokens": 2197291008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.2062368392944336, "created_at": "2025-01-15T17:23:19.179039+00:00"} {"global_step": 4192, "acc_step": 0, "speed/wps": 12915.201187257986, "speed/FLOPS": 202850981992106.94, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11534286290407181, "optim/lr": 0.0025146, "optim/total_tokens": 2197815296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.2578012943267822, "created_at": "2025-01-15T17:23:29.334165+00:00"} {"global_step": 4193, "acc_step": 0, "speed/wps": 12911.852493354067, "speed/FLOPS": 202798386152758.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12053139507770538, "optim/lr": 0.0025152, "optim/total_tokens": 2198339584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 3.193833827972412, "created_at": "2025-01-15T17:23:39.490162+00:00"} {"global_step": 4194, "acc_step": 0, "speed/wps": 12914.521202465286, "speed/FLOPS": 202840301896540.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12285145372152328, "optim/lr": 0.0025158, "optim/total_tokens": 2198863872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 3.1436736583709717, "created_at": "2025-01-15T17:23:49.645775+00:00"} {"global_step": 4195, "acc_step": 0, "speed/wps": 12906.588185161385, "speed/FLOPS": 202715702958674.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11031120270490646, "optim/lr": 0.0025164000000000002, "optim/total_tokens": 2199388160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.2198190689086914, "created_at": "2025-01-15T17:23:59.805834+00:00"} {"global_step": 4196, "acc_step": 0, "speed/wps": 12913.431959077243, "speed/FLOPS": 202823193832354.06, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10433382540941238, "optim/lr": 0.002517, "optim/total_tokens": 2199912448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.3021750450134277, "created_at": "2025-01-15T17:24:09.965343+00:00"} {"global_step": 4197, "acc_step": 0, "speed/wps": 12908.264789569812, "speed/FLOPS": 202742036334804.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09620853513479233, "optim/lr": 0.0025176, "optim/total_tokens": 2200436736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 3.235896587371826, "created_at": "2025-01-15T17:24:20.126044+00:00"} {"global_step": 4198, "acc_step": 0, "speed/wps": 12913.73785792968, "speed/FLOPS": 202827998394180.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08088191598653793, "optim/lr": 0.0025182, "optim/total_tokens": 2200961024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.1998000144958496, "created_at": "2025-01-15T17:24:30.282607+00:00"} {"global_step": 4199, "acc_step": 0, "speed/wps": 12908.03217879545, "speed/FLOPS": 202738382862952.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08578027039766312, "optim/lr": 0.0025188000000000003, "optim/total_tokens": 2201485312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.143352508544922, "created_at": "2025-01-15T17:24:40.440485+00:00"} {"global_step": 4200, "acc_step": 0, "speed/wps": 12909.308459483142, "speed/FLOPS": 202758428604946.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08780092746019363, "optim/lr": 0.0025194, "optim/total_tokens": 2202009600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.2512762546539307, "created_at": "2025-01-15T17:24:50.598861+00:00"} {"global_step": 4201, "acc_step": 0, "speed/wps": 12910.923033044068, "speed/FLOPS": 202783787701377.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0852866917848587, "optim/lr": 0.00252, "optim/total_tokens": 2202533888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382950, "loss/out": 3.256283760070801, "created_at": "2025-01-15T17:25:00.756261+00:00"} {"global_step": 4202, "acc_step": 0, "speed/wps": 12914.40317810132, "speed/FLOPS": 202838448161722.0, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08790808171033859, "optim/lr": 0.0025206, "optim/total_tokens": 2203058176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.25823712348938, "created_at": "2025-01-15T17:25:10.909398+00:00"} {"global_step": 4203, "acc_step": 0, "speed/wps": 12911.564551317751, "speed/FLOPS": 202793863627404.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10136491805315018, "optim/lr": 0.0025212000000000004, "optim/total_tokens": 2203582464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2298331260681152, "created_at": "2025-01-15T17:25:21.067683+00:00"} {"global_step": 4204, "acc_step": 0, "speed/wps": 12912.666501119793, "speed/FLOPS": 202811171263282.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10992459952831268, "optim/lr": 0.0025218000000000003, "optim/total_tokens": 2204106752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.2979273796081543, "created_at": "2025-01-15T17:25:31.226053+00:00"} {"global_step": 4205, "acc_step": 0, "speed/wps": 12911.043020381983, "speed/FLOPS": 202785672267399.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11114129424095154, "optim/lr": 0.0025224, "optim/total_tokens": 2204631040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.175813674926758, "created_at": "2025-01-15T17:25:41.381903+00:00"} {"global_step": 4206, "acc_step": 0, "speed/wps": 12904.08387326103, "speed/FLOPS": 202676369298994.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.11553920060396194, "optim/lr": 0.002523, "optim/total_tokens": 2205155328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 3.2438805103302, "created_at": "2025-01-15T17:25:51.548611+00:00"} {"global_step": 4207, "acc_step": 0, "speed/wps": 12905.114686925195, "speed/FLOPS": 202692559644076.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10633714497089386, "optim/lr": 0.0025236, "optim/total_tokens": 2205679616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.2937979698181152, "created_at": "2025-01-15T17:26:01.712107+00:00"} {"global_step": 4208, "acc_step": 0, "speed/wps": 12910.153701465248, "speed/FLOPS": 202771704291759.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11265946924686432, "optim/lr": 0.0025242000000000003, "optim/total_tokens": 2206203904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.2542080879211426, "created_at": "2025-01-15T17:26:11.870820+00:00"} {"global_step": 4209, "acc_step": 0, "speed/wps": 12915.63146920849, "speed/FLOPS": 202857740161408.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11058841645717621, "optim/lr": 0.0025248000000000002, "optim/total_tokens": 2206728192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.130612373352051, "created_at": "2025-01-15T17:26:22.026583+00:00"} {"global_step": 4210, "acc_step": 0, "speed/wps": 12916.914662683212, "speed/FLOPS": 202877894478222.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1017366349697113, "optim/lr": 0.0025254, "optim/total_tokens": 2207252480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.2159423828125, "created_at": "2025-01-15T17:26:32.177675+00:00"} {"global_step": 4211, "acc_step": 0, "speed/wps": 12910.632972180167, "speed/FLOPS": 202779231896924.44, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09235679358243942, "optim/lr": 0.002526, "optim/total_tokens": 2207776768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.2500014305114746, "created_at": "2025-01-15T17:26:42.338587+00:00"} {"global_step": 4212, "acc_step": 0, "speed/wps": 12911.231481661049, "speed/FLOPS": 202788632310759.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09224384278059006, "optim/lr": 0.0025266, "optim/total_tokens": 2208301056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.194112539291382, "created_at": "2025-01-15T17:26:52.498293+00:00"} {"global_step": 4213, "acc_step": 0, "speed/wps": 12912.14824317469, "speed/FLOPS": 202803031310093.97, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10527756065130234, "optim/lr": 0.0025272000000000003, "optim/total_tokens": 2208825344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 3.1556482315063477, "created_at": "2025-01-15T17:27:02.653790+00:00"} {"global_step": 4214, "acc_step": 0, "speed/wps": 12911.3376244349, "speed/FLOPS": 202790299428878.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10123685747385025, "optim/lr": 0.0025278, "optim/total_tokens": 2209349632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 3.204774856567383, "created_at": "2025-01-15T17:27:12.812688+00:00"} {"global_step": 4215, "acc_step": 0, "speed/wps": 12918.660211918374, "speed/FLOPS": 202905310727596.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10315611958503723, "optim/lr": 0.0025284, "optim/total_tokens": 2209873920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 3.203806161880493, "created_at": "2025-01-15T17:27:22.965247+00:00"} {"global_step": 4216, "acc_step": 0, "speed/wps": 12913.340230044532, "speed/FLOPS": 202821753101846.47, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09472383558750153, "optim/lr": 0.002529, "optim/total_tokens": 2210398208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.240802049636841, "created_at": "2025-01-15T17:27:33.121395+00:00"} {"global_step": 4217, "acc_step": 0, "speed/wps": 12914.803752500105, "speed/FLOPS": 202844739733101.84, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10783713310956955, "optim/lr": 0.0025296, "optim/total_tokens": 2210922496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492439, "loss/out": 3.151160955429077, "created_at": "2025-01-15T17:27:43.274009+00:00"} {"global_step": 4218, "acc_step": 0, "speed/wps": 12913.617136804294, "speed/FLOPS": 202826102303018.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09667224436998367, "optim/lr": 0.0025302000000000002, "optim/total_tokens": 2211446784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.2528538703918457, "created_at": "2025-01-15T17:27:53.429810+00:00"} {"global_step": 4219, "acc_step": 0, "speed/wps": 12916.056723084757, "speed/FLOPS": 202864419357895.8, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10339080542325974, "optim/lr": 0.0025308, "optim/total_tokens": 2211971072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345847, "loss/out": 3.1669931411743164, "created_at": "2025-01-15T17:28:03.582672+00:00"} {"global_step": 4220, "acc_step": 0, "speed/wps": 12904.787449884136, "speed/FLOPS": 202687419936675.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10082890838384628, "optim/lr": 0.0025314, "optim/total_tokens": 2212495360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.1405205726623535, "created_at": "2025-01-15T17:28:13.745542+00:00"} {"global_step": 4221, "acc_step": 0, "speed/wps": 12916.093858134484, "speed/FLOPS": 202865002614880.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10308558493852615, "optim/lr": 0.002532, "optim/total_tokens": 2213019648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.30421781539917, "created_at": "2025-01-15T17:28:23.899057+00:00"} {"global_step": 4222, "acc_step": 0, "speed/wps": 12912.184868862449, "speed/FLOPS": 202803606566849.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12542036175727844, "optim/lr": 0.0025326, "optim/total_tokens": 2213543936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 3.2099828720092773, "created_at": "2025-01-15T17:28:34.054718+00:00"} {"global_step": 4223, "acc_step": 0, "speed/wps": 12910.969278788076, "speed/FLOPS": 202784514054335.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1408247947692871, "optim/lr": 0.0025332, "optim/total_tokens": 2214068224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.134385585784912, "created_at": "2025-01-15T17:28:44.216031+00:00"} {"global_step": 4224, "acc_step": 0, "speed/wps": 12905.52529117456, "speed/FLOPS": 202699008748042.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1236754059791565, "optim/lr": 0.0025338, "optim/total_tokens": 2214592512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 3.1883578300476074, "created_at": "2025-01-15T17:28:54.379689+00:00"} {"global_step": 4225, "acc_step": 0, "speed/wps": 12911.382210108455, "speed/FLOPS": 202790999708149.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1344471126794815, "optim/lr": 0.0025344, "optim/total_tokens": 2215116800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.245121717453003, "created_at": "2025-01-15T17:29:04.536270+00:00"} {"global_step": 4226, "acc_step": 0, "speed/wps": 12914.494896164262, "speed/FLOPS": 202839888719934.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10574516654014587, "optim/lr": 0.002535, "optim/total_tokens": 2215641088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 3.188783884048462, "created_at": "2025-01-15T17:29:14.689308+00:00"} {"global_step": 4227, "acc_step": 0, "speed/wps": 12911.284806901598, "speed/FLOPS": 202789469856939.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11375901848077774, "optim/lr": 0.0025356, "optim/total_tokens": 2216165376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.1323957443237305, "created_at": "2025-01-15T17:29:24.848834+00:00"} {"global_step": 4228, "acc_step": 0, "speed/wps": 12911.594413517463, "speed/FLOPS": 202794332654285.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11894501745700836, "optim/lr": 0.0025362, "optim/total_tokens": 2216689664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.2854068279266357, "created_at": "2025-01-15T17:29:35.007767+00:00"} {"global_step": 4229, "acc_step": 0, "speed/wps": 12909.602462160827, "speed/FLOPS": 202763046321001.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1070915162563324, "optim/lr": 0.0025368, "optim/total_tokens": 2217213952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.20613956451416, "created_at": "2025-01-15T17:29:45.165860+00:00"} {"global_step": 4230, "acc_step": 0, "speed/wps": 12916.108662191591, "speed/FLOPS": 202865235132940.44, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09307083487510681, "optim/lr": 0.0025374, "optim/total_tokens": 2217738240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.315096616744995, "created_at": "2025-01-15T17:29:55.318762+00:00"} {"global_step": 4231, "acc_step": 0, "speed/wps": 12910.175293226423, "speed/FLOPS": 202772043420039.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10114166140556335, "optim/lr": 0.002538, "optim/total_tokens": 2218262528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.318462610244751, "created_at": "2025-01-15T17:30:05.474937+00:00"} {"global_step": 4232, "acc_step": 0, "speed/wps": 12910.525349901849, "speed/FLOPS": 202777541541154.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1139509305357933, "optim/lr": 0.0025386, "optim/total_tokens": 2218786816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.2777581214904785, "created_at": "2025-01-15T17:30:15.632969+00:00"} {"global_step": 4233, "acc_step": 0, "speed/wps": 12909.31958786386, "speed/FLOPS": 202758603391457.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09935461729764938, "optim/lr": 0.0025392, "optim/total_tokens": 2219311104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 3.1843149662017822, "created_at": "2025-01-15T17:30:25.798637+00:00"} {"global_step": 4234, "acc_step": 0, "speed/wps": 12914.164359360751, "speed/FLOPS": 202834697185230.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0949668362736702, "optim/lr": 0.0025398, "optim/total_tokens": 2219835392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.248249053955078, "created_at": "2025-01-15T17:30:35.953450+00:00"} {"global_step": 4235, "acc_step": 0, "speed/wps": 12912.340117789565, "speed/FLOPS": 202806044964583.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09226953238248825, "optim/lr": 0.0025404, "optim/total_tokens": 2220359680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.193831443786621, "created_at": "2025-01-15T17:30:46.115403+00:00"} {"global_step": 4236, "acc_step": 0, "speed/wps": 12909.764299485867, "speed/FLOPS": 202765588198579.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09739904850721359, "optim/lr": 0.002541, "optim/total_tokens": 2220883968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 3.218461513519287, "created_at": "2025-01-15T17:30:56.274348+00:00"} {"global_step": 4237, "acc_step": 0, "speed/wps": 12911.194337847162, "speed/FLOPS": 202788048916121.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09183082729578018, "optim/lr": 0.0025415999999999998, "optim/total_tokens": 2221408256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0212936401367188, "created_at": "2025-01-15T17:31:06.432548+00:00"} {"global_step": 4238, "acc_step": 0, "speed/wps": 12910.690523062236, "speed/FLOPS": 202780135812610.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08882059156894684, "optim/lr": 0.0025422, "optim/total_tokens": 2221932544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1329221725463867, "created_at": "2025-01-15T17:31:16.591177+00:00"} {"global_step": 4239, "acc_step": 0, "speed/wps": 12918.41852616977, "speed/FLOPS": 202901514720803.47, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10287729650735855, "optim/lr": 0.0025428, "optim/total_tokens": 2222456832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 3.2607600688934326, "created_at": "2025-01-15T17:31:26.740947+00:00"} {"global_step": 4240, "acc_step": 0, "speed/wps": 12913.464802788394, "speed/FLOPS": 202823709687969.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11048692464828491, "optim/lr": 0.0025434, "optim/total_tokens": 2222981120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.3364176750183105, "created_at": "2025-01-15T17:31:36.896694+00:00"} {"global_step": 4241, "acc_step": 0, "speed/wps": 12912.035357234097, "speed/FLOPS": 202801258281275.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11160776019096375, "optim/lr": 0.002544, "optim/total_tokens": 2223505408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.1798038482666016, "created_at": "2025-01-15T17:31:47.053584+00:00"} {"global_step": 4242, "acc_step": 0, "speed/wps": 12912.20582059781, "speed/FLOPS": 202803935642643.75, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1206590011715889, "optim/lr": 0.0025445999999999997, "optim/total_tokens": 2224029696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 3.212136745452881, "created_at": "2025-01-15T17:31:57.213395+00:00"} {"global_step": 4243, "acc_step": 0, "speed/wps": 12908.944757427837, "speed/FLOPS": 202752716164388.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10016005486249924, "optim/lr": 0.0025452, "optim/total_tokens": 2224553984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.317211627960205, "created_at": "2025-01-15T17:32:07.377277+00:00"} {"global_step": 4244, "acc_step": 0, "speed/wps": 12912.711562761388, "speed/FLOPS": 202811879018285.28, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08910849690437317, "optim/lr": 0.0025458, "optim/total_tokens": 2225078272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340097, "loss/out": 3.299144744873047, "created_at": "2025-01-15T17:32:17.563081+00:00"} {"global_step": 4245, "acc_step": 0, "speed/wps": 12917.89357319834, "speed/FLOPS": 202893269613026.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10320999473333359, "optim/lr": 0.0025464, "optim/total_tokens": 2225602560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.2112650871276855, "created_at": "2025-01-15T17:32:27.716836+00:00"} {"global_step": 4246, "acc_step": 0, "speed/wps": 12913.567298422346, "speed/FLOPS": 202825319522744.66, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11719997227191925, "optim/lr": 0.002547, "optim/total_tokens": 2226126848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.2516136169433594, "created_at": "2025-01-15T17:32:37.873388+00:00"} {"global_step": 4247, "acc_step": 0, "speed/wps": 12917.26398826952, "speed/FLOPS": 202883381116578.72, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11757306754589081, "optim/lr": 0.0025476, "optim/total_tokens": 2226651136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.109126091003418, "created_at": "2025-01-15T17:32:48.023965+00:00"} {"global_step": 4248, "acc_step": 0, "speed/wps": 12914.573503710879, "speed/FLOPS": 202841123359472.03, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0889861136674881, "optim/lr": 0.0025482, "optim/total_tokens": 2227175424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.2092528343200684, "created_at": "2025-01-15T17:32:58.178555+00:00"} {"global_step": 4249, "acc_step": 0, "speed/wps": 12907.154057143774, "speed/FLOPS": 202724590755747.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10362423211336136, "optim/lr": 0.0025488, "optim/total_tokens": 2227699712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 3.2333240509033203, "created_at": "2025-01-15T17:33:08.340209+00:00"} {"global_step": 4250, "acc_step": 0, "speed/wps": 12913.232908600226, "speed/FLOPS": 202820067471088.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11882183700799942, "optim/lr": 0.0025494, "optim/total_tokens": 2228224000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.3064632415771484, "created_at": "2025-01-15T17:33:18.495142+00:00"} {"global_step": 4251, "acc_step": 0, "speed/wps": 12907.007705286556, "speed/FLOPS": 202722292098721.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1205228790640831, "optim/lr": 0.00255, "optim/total_tokens": 2228748288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 3.211101531982422, "created_at": "2025-01-15T17:33:28.655093+00:00"} {"global_step": 4252, "acc_step": 0, "speed/wps": 12909.8739862065, "speed/FLOPS": 202767310979251.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10406314581632614, "optim/lr": 0.0025506, "optim/total_tokens": 2229272576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 3.197160005569458, "created_at": "2025-01-15T17:33:38.813990+00:00"} {"global_step": 4253, "acc_step": 0, "speed/wps": 12910.515717635531, "speed/FLOPS": 202777390253176.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09455471485853195, "optim/lr": 0.0025512, "optim/total_tokens": 2229796864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.3122453689575195, "created_at": "2025-01-15T17:33:48.972090+00:00"} {"global_step": 4254, "acc_step": 0, "speed/wps": 12916.735582154144, "speed/FLOPS": 202875081772430.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09980427473783493, "optim/lr": 0.0025518000000000003, "optim/total_tokens": 2230321152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.3022561073303223, "created_at": "2025-01-15T17:33:59.126225+00:00"} {"global_step": 4255, "acc_step": 0, "speed/wps": 12915.075073462856, "speed/FLOPS": 202849001201655.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07896420359611511, "optim/lr": 0.0025524000000000002, "optim/total_tokens": 2230845440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.2441346645355225, "created_at": "2025-01-15T17:34:09.280824+00:00"} {"global_step": 4256, "acc_step": 0, "speed/wps": 12913.051444416224, "speed/FLOPS": 202817217326720.8, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09822283685207367, "optim/lr": 0.002553, "optim/total_tokens": 2231369728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 3.191068410873413, "created_at": "2025-01-15T17:34:19.437373+00:00"} {"global_step": 4257, "acc_step": 0, "speed/wps": 12914.944286691136, "speed/FLOPS": 202846947015686.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09647030383348465, "optim/lr": 0.0025536, "optim/total_tokens": 2231894016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.219606399536133, "created_at": "2025-01-15T17:34:29.591856+00:00"} {"global_step": 4258, "acc_step": 0, "speed/wps": 12914.2602500752, "speed/FLOPS": 202836203280668.1, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10040999948978424, "optim/lr": 0.0025542000000000004, "optim/total_tokens": 2232418304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.186894416809082, "created_at": "2025-01-15T17:34:39.744861+00:00"} {"global_step": 4259, "acc_step": 0, "speed/wps": 12912.202749978493, "speed/FLOPS": 202803887414347.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.10008790343999863, "optim/lr": 0.0025548000000000003, "optim/total_tokens": 2232942592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.1345434188842773, "created_at": "2025-01-15T17:34:49.902583+00:00"} {"global_step": 4260, "acc_step": 0, "speed/wps": 12912.515406227714, "speed/FLOPS": 202808798110376.62, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09429693967103958, "optim/lr": 0.0025554, "optim/total_tokens": 2233466880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 3.2930803298950195, "created_at": "2025-01-15T17:35:00.060122+00:00"} {"global_step": 4261, "acc_step": 0, "speed/wps": 12915.334228793216, "speed/FLOPS": 202853071592235.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09950227290391922, "optim/lr": 0.002556, "optim/total_tokens": 2233991168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.1639537811279297, "created_at": "2025-01-15T17:35:10.213746+00:00"} {"global_step": 4262, "acc_step": 0, "speed/wps": 12917.57757853157, "speed/FLOPS": 202888306482562.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11413774639368057, "optim/lr": 0.0025566, "optim/total_tokens": 2234515456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374042, "loss/out": 3.125410556793213, "created_at": "2025-01-15T17:35:20.368932+00:00"} {"global_step": 4263, "acc_step": 0, "speed/wps": 12915.098748648837, "speed/FLOPS": 202849373052984.84, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1009296104311943, "optim/lr": 0.0025572000000000004, "optim/total_tokens": 2235039744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.194878101348877, "created_at": "2025-01-15T17:35:30.521251+00:00"} {"global_step": 4264, "acc_step": 0, "speed/wps": 12911.121581174732, "speed/FLOPS": 202786906172603.2, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09895241260528564, "optim/lr": 0.0025578000000000003, "optim/total_tokens": 2235564032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 3.2531652450561523, "created_at": "2025-01-15T17:35:40.678015+00:00"} {"global_step": 4265, "acc_step": 0, "speed/wps": 12911.870008695463, "speed/FLOPS": 202798661255263.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09222351759672165, "optim/lr": 0.0025584, "optim/total_tokens": 2236088320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.22611141204834, "created_at": "2025-01-15T17:35:50.840553+00:00"} {"global_step": 4266, "acc_step": 0, "speed/wps": 12914.70759654502, "speed/FLOPS": 202843229471695.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09004402905702591, "optim/lr": 0.002559, "optim/total_tokens": 2236612608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390425, "loss/out": 3.265155553817749, "created_at": "2025-01-15T17:36:01.001263+00:00"} {"global_step": 4267, "acc_step": 0, "speed/wps": 12909.414156191802, "speed/FLOPS": 202760088716998.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1044159010052681, "optim/lr": 0.0025596, "optim/total_tokens": 2237136896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 3.222990036010742, "created_at": "2025-01-15T17:36:11.158589+00:00"} {"global_step": 4268, "acc_step": 0, "speed/wps": 12909.950627332366, "speed/FLOPS": 202768514733449.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09153930842876434, "optim/lr": 0.0025602000000000003, "optim/total_tokens": 2237661184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.244692802429199, "created_at": "2025-01-15T17:36:21.317427+00:00"} {"global_step": 4269, "acc_step": 0, "speed/wps": 12906.785961996207, "speed/FLOPS": 202718809315639.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08708938211202621, "optim/lr": 0.0025608000000000002, "optim/total_tokens": 2238185472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 3.2607827186584473, "created_at": "2025-01-15T17:36:31.480291+00:00"} {"global_step": 4270, "acc_step": 0, "speed/wps": 12909.597187846884, "speed/FLOPS": 202762963480653.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10313815623521805, "optim/lr": 0.0025614, "optim/total_tokens": 2238709760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.232320547103882, "created_at": "2025-01-15T17:36:41.644832+00:00"} {"global_step": 4271, "acc_step": 0, "speed/wps": 12910.044721729455, "speed/FLOPS": 202769992615255.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1161334291100502, "optim/lr": 0.002562, "optim/total_tokens": 2239234048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 3.3193178176879883, "created_at": "2025-01-15T17:36:51.801482+00:00"} {"global_step": 4272, "acc_step": 0, "speed/wps": 12913.382064066056, "speed/FLOPS": 202822410162640.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12415409088134766, "optim/lr": 0.0025626, "optim/total_tokens": 2239758336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.2401976585388184, "created_at": "2025-01-15T17:37:01.957214+00:00"} {"global_step": 4273, "acc_step": 0, "speed/wps": 12916.620307892172, "speed/FLOPS": 202873271231742.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09972856193780899, "optim/lr": 0.0025632000000000003, "optim/total_tokens": 2240282624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.2956156730651855, "created_at": "2025-01-15T17:37:12.112359+00:00"} {"global_step": 4274, "acc_step": 0, "speed/wps": 12908.45832710046, "speed/FLOPS": 202745076107672.16, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11014449596405029, "optim/lr": 0.0025638, "optim/total_tokens": 2240806912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 3.152704954147339, "created_at": "2025-01-15T17:37:22.271304+00:00"} {"global_step": 4275, "acc_step": 0, "speed/wps": 12912.32426327086, "speed/FLOPS": 202805795947581.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13371212780475616, "optim/lr": 0.0025644, "optim/total_tokens": 2241331200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.23637056350708, "created_at": "2025-01-15T17:37:32.430263+00:00"} {"global_step": 4276, "acc_step": 0, "speed/wps": 12916.204835454884, "speed/FLOPS": 202866745666196.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13987351953983307, "optim/lr": 0.002565, "optim/total_tokens": 2241855488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.1269402503967285, "created_at": "2025-01-15T17:37:42.585463+00:00"} {"global_step": 4277, "acc_step": 0, "speed/wps": 12912.367393860544, "speed/FLOPS": 202806473372759.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10583437234163284, "optim/lr": 0.0025656, "optim/total_tokens": 2242379776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.0890843868255615, "created_at": "2025-01-15T17:37:52.744367+00:00"} {"global_step": 4278, "acc_step": 0, "speed/wps": 12914.249859742265, "speed/FLOPS": 202836040086211.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09774771332740784, "optim/lr": 0.0025662000000000003, "optim/total_tokens": 2242904064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 3.286787986755371, "created_at": "2025-01-15T17:38:02.898367+00:00"} {"global_step": 4279, "acc_step": 0, "speed/wps": 12913.57058771289, "speed/FLOPS": 202825371185572.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08963150531053543, "optim/lr": 0.0025668, "optim/total_tokens": 2243428352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.2428107261657715, "created_at": "2025-01-15T17:38:13.056429+00:00"} {"global_step": 4280, "acc_step": 0, "speed/wps": 12912.50510755534, "speed/FLOPS": 202808636355575.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10034544765949249, "optim/lr": 0.0025674, "optim/total_tokens": 2243952640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 3.266714334487915, "created_at": "2025-01-15T17:38:23.210760+00:00"} {"global_step": 4281, "acc_step": 0, "speed/wps": 12918.203021362251, "speed/FLOPS": 202898129921666.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10629382729530334, "optim/lr": 0.002568, "optim/total_tokens": 2244476928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.2170138359069824, "created_at": "2025-01-15T17:38:33.360573+00:00"} {"global_step": 4282, "acc_step": 0, "speed/wps": 12907.24197425727, "speed/FLOPS": 202725971614825.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11573311686515808, "optim/lr": 0.0025686, "optim/total_tokens": 2245001216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.209172248840332, "created_at": "2025-01-15T17:38:43.521764+00:00"} {"global_step": 4283, "acc_step": 0, "speed/wps": 12909.786592180393, "speed/FLOPS": 202765938335980.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10089454799890518, "optim/lr": 0.0025692, "optim/total_tokens": 2245525504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.165961265563965, "created_at": "2025-01-15T17:38:53.678162+00:00"} {"global_step": 4284, "acc_step": 0, "speed/wps": 12914.883471916364, "speed/FLOPS": 202845991836080.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10773937404155731, "optim/lr": 0.0025698, "optim/total_tokens": 2246049792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 3.3313565254211426, "created_at": "2025-01-15T17:39:03.834233+00:00"} {"global_step": 4285, "acc_step": 0, "speed/wps": 12912.190948483732, "speed/FLOPS": 202803702055655.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12608787417411804, "optim/lr": 0.0025704, "optim/total_tokens": 2246574080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.2025551795959473, "created_at": "2025-01-15T17:39:13.989263+00:00"} {"global_step": 4286, "acc_step": 0, "speed/wps": 12911.455072771121, "speed/FLOPS": 202792144116389.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.10175631195306778, "optim/lr": 0.002571, "optim/total_tokens": 2247098368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.146191120147705, "created_at": "2025-01-15T17:39:24.149197+00:00"} {"global_step": 4287, "acc_step": 0, "speed/wps": 12914.240820355817, "speed/FLOPS": 202835898110225.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10614308714866638, "optim/lr": 0.0025716, "optim/total_tokens": 2247622656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.158421039581299, "created_at": "2025-01-15T17:39:34.304658+00:00"} {"global_step": 4288, "acc_step": 0, "speed/wps": 12919.417896189918, "speed/FLOPS": 202917211200248.2, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08905065059661865, "optim/lr": 0.0025722, "optim/total_tokens": 2248146944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.303986072540283, "created_at": "2025-01-15T17:39:44.457309+00:00"} {"global_step": 4289, "acc_step": 0, "speed/wps": 12911.092261270613, "speed/FLOPS": 202786445663219.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08400727808475494, "optim/lr": 0.0025728, "optim/total_tokens": 2248671232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318692, "loss/out": 3.219071865081787, "created_at": "2025-01-15T17:39:54.615421+00:00"} {"global_step": 4290, "acc_step": 0, "speed/wps": 12917.086402515162, "speed/FLOPS": 202880591888278.12, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09521792083978653, "optim/lr": 0.0025734, "optim/total_tokens": 2249195520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465054, "loss/out": 3.1424646377563477, "created_at": "2025-01-15T17:40:04.768199+00:00"} {"global_step": 4291, "acc_step": 0, "speed/wps": 12911.003540767295, "speed/FLOPS": 202785052185799.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09310078620910645, "optim/lr": 0.002574, "optim/total_tokens": 2249719808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.279763698577881, "created_at": "2025-01-15T17:40:14.926505+00:00"} {"global_step": 4292, "acc_step": 0, "speed/wps": 12913.623409730295, "speed/FLOPS": 202826200827941.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09234713762998581, "optim/lr": 0.0025746, "optim/total_tokens": 2250244096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.214188575744629, "created_at": "2025-01-15T17:40:25.079940+00:00"} {"global_step": 4293, "acc_step": 0, "speed/wps": 12910.989881909658, "speed/FLOPS": 202784837654671.47, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11034587770700455, "optim/lr": 0.0025752, "optim/total_tokens": 2250768384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.987250328063965, "created_at": "2025-01-15T17:40:35.236496+00:00"} {"global_step": 4294, "acc_step": 0, "speed/wps": 12910.745310102633, "speed/FLOPS": 202780996318365.25, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10873991996049881, "optim/lr": 0.0025758, "optim/total_tokens": 2251292672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.260100841522217, "created_at": "2025-01-15T17:40:45.396336+00:00"} {"global_step": 4295, "acc_step": 0, "speed/wps": 12912.567544323205, "speed/FLOPS": 202809617010811.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0963691994547844, "optim/lr": 0.0025764, "optim/total_tokens": 2251816960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.2817211151123047, "created_at": "2025-01-15T17:40:55.551677+00:00"} {"global_step": 4296, "acc_step": 0, "speed/wps": 12910.489381460771, "speed/FLOPS": 202776976607362.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09023672342300415, "optim/lr": 0.002577, "optim/total_tokens": 2252341248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.270909547805786, "created_at": "2025-01-15T17:41:05.711066+00:00"} {"global_step": 4297, "acc_step": 0, "speed/wps": 12912.713053171523, "speed/FLOPS": 202811902427224.5, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09210504591464996, "optim/lr": 0.0025775999999999998, "optim/total_tokens": 2252865536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.209348440170288, "created_at": "2025-01-15T17:41:15.865920+00:00"} {"global_step": 4298, "acc_step": 0, "speed/wps": 12916.219272295679, "speed/FLOPS": 202866972416619.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09196872264146805, "optim/lr": 0.0025782, "optim/total_tokens": 2253389824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.279524803161621, "created_at": "2025-01-15T17:41:26.019491+00:00"} {"global_step": 4299, "acc_step": 0, "speed/wps": 12915.74620310286, "speed/FLOPS": 202859542214880.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08594035357236862, "optim/lr": 0.0025788, "optim/total_tokens": 2253914112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.2025721073150635, "created_at": "2025-01-15T17:41:36.171827+00:00"} {"global_step": 4300, "acc_step": 0, "speed/wps": 12911.027343160593, "speed/FLOPS": 202785426035094.56, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09943851828575134, "optim/lr": 0.0025794, "optim/total_tokens": 2254438400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.244014024734497, "created_at": "2025-01-15T17:41:46.328339+00:00"} {"global_step": 4301, "acc_step": 0, "speed/wps": 12916.139330132488, "speed/FLOPS": 202865716815094.38, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11004163324832916, "optim/lr": 0.00258, "optim/total_tokens": 2254962688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.1083321571350098, "created_at": "2025-01-15T17:41:56.479839+00:00"} {"global_step": 4302, "acc_step": 0, "speed/wps": 12909.853108354242, "speed/FLOPS": 202766983063893.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11844472587108612, "optim/lr": 0.0025805999999999997, "optim/total_tokens": 2255486976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 3.2941973209381104, "created_at": "2025-01-15T17:42:06.637250+00:00"} {"global_step": 4303, "acc_step": 0, "speed/wps": 12914.95958587955, "speed/FLOPS": 202847187310463.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09426574409008026, "optim/lr": 0.0025812, "optim/total_tokens": 2256011264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.171891689300537, "created_at": "2025-01-15T17:42:16.792909+00:00"} {"global_step": 4304, "acc_step": 0, "speed/wps": 12914.808009486844, "speed/FLOPS": 202844806594928.28, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09814418107271194, "optim/lr": 0.0025818, "optim/total_tokens": 2256535552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.2312722206115723, "created_at": "2025-01-15T17:42:26.946417+00:00"} {"global_step": 4305, "acc_step": 0, "speed/wps": 12916.213813580014, "speed/FLOPS": 202866886679988.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09159579128026962, "optim/lr": 0.0025824, "optim/total_tokens": 2257059840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.2834877967834473, "created_at": "2025-01-15T17:42:37.099981+00:00"} {"global_step": 4306, "acc_step": 0, "speed/wps": 12910.674112386343, "speed/FLOPS": 202779878060395.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09162082523107529, "optim/lr": 0.0025830000000000002, "optim/total_tokens": 2257584128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 3.121523380279541, "created_at": "2025-01-15T17:42:47.256736+00:00"} {"global_step": 4307, "acc_step": 0, "speed/wps": 12916.634298741235, "speed/FLOPS": 202873490977252.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09346850216388702, "optim/lr": 0.0025836, "optim/total_tokens": 2258108416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0989255905151367, "created_at": "2025-01-15T17:42:57.415935+00:00"} {"global_step": 4308, "acc_step": 0, "speed/wps": 12907.461568078541, "speed/FLOPS": 202729420637542.5, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10381880402565002, "optim/lr": 0.0025842, "optim/total_tokens": 2258632704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468042, "loss/out": 3.098491668701172, "created_at": "2025-01-15T17:43:07.577280+00:00"} {"global_step": 4309, "acc_step": 0, "speed/wps": 12912.420740331409, "speed/FLOPS": 202807311252390.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10289914906024933, "optim/lr": 0.0025848, "optim/total_tokens": 2259156992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.2503576278686523, "created_at": "2025-01-15T17:43:17.731620+00:00"} {"global_step": 4310, "acc_step": 0, "speed/wps": 12913.277424424963, "speed/FLOPS": 202820766653287.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10367413610219955, "optim/lr": 0.0025854000000000003, "optim/total_tokens": 2259681280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 3.1455938816070557, "created_at": "2025-01-15T17:43:27.915743+00:00"} {"global_step": 4311, "acc_step": 0, "speed/wps": 12914.666027693796, "speed/FLOPS": 202842576575765.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1092304214835167, "optim/lr": 0.002586, "optim/total_tokens": 2260205568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.2439939975738525, "created_at": "2025-01-15T17:43:38.071207+00:00"} {"global_step": 4312, "acc_step": 0, "speed/wps": 12914.287928875583, "speed/FLOPS": 202836638014262.78, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09932848066091537, "optim/lr": 0.0025866, "optim/total_tokens": 2260729856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 3.225292205810547, "created_at": "2025-01-15T17:43:48.224514+00:00"} {"global_step": 4313, "acc_step": 0, "speed/wps": 12908.843335669386, "speed/FLOPS": 202751123196304.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09725075215101242, "optim/lr": 0.0025872000000000004, "optim/total_tokens": 2261254144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 3.3532323837280273, "created_at": "2025-01-15T17:43:58.384639+00:00"} {"global_step": 4314, "acc_step": 0, "speed/wps": 12917.310111096484, "speed/FLOPS": 202884105538955.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11713560670614243, "optim/lr": 0.0025878000000000003, "optim/total_tokens": 2261778432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 3.2628986835479736, "created_at": "2025-01-15T17:44:08.537950+00:00"} {"global_step": 4315, "acc_step": 0, "speed/wps": 12914.512429341092, "speed/FLOPS": 202840164102569.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12199398130178452, "optim/lr": 0.0025884000000000002, "optim/total_tokens": 2262302720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370300, "loss/out": 3.302403688430786, "created_at": "2025-01-15T17:44:18.691253+00:00"} {"global_step": 4316, "acc_step": 0, "speed/wps": 12917.932377610541, "speed/FLOPS": 202893879089642.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08844250440597534, "optim/lr": 0.002589, "optim/total_tokens": 2262827008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.1403799057006836, "created_at": "2025-01-15T17:44:28.842478+00:00"} {"global_step": 4317, "acc_step": 0, "speed/wps": 12912.289182783985, "speed/FLOPS": 202805244960329.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09657163172960281, "optim/lr": 0.0025896, "optim/total_tokens": 2263351296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333550, "loss/out": 3.1452064514160156, "created_at": "2025-01-15T17:44:38.997223+00:00"} {"global_step": 4318, "acc_step": 0, "speed/wps": 12914.879218435308, "speed/FLOPS": 202845925029315.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12034729868173599, "optim/lr": 0.0025902, "optim/total_tokens": 2263875584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.238079071044922, "created_at": "2025-01-15T17:44:49.150601+00:00"} {"global_step": 4319, "acc_step": 0, "speed/wps": 12912.399587686235, "speed/FLOPS": 202806979021031.34, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10370638966560364, "optim/lr": 0.0025908000000000003, "optim/total_tokens": 2264399872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 3.2590367794036865, "created_at": "2025-01-15T17:44:59.306970+00:00"} {"global_step": 4320, "acc_step": 0, "speed/wps": 12911.639744861219, "speed/FLOPS": 202795044645331.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10015159845352173, "optim/lr": 0.0025914, "optim/total_tokens": 2264924160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.2822468280792236, "created_at": "2025-01-15T17:45:09.465644+00:00"} {"global_step": 4321, "acc_step": 0, "speed/wps": 12909.322992219102, "speed/FLOPS": 202758656861534.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10099586099386215, "optim/lr": 0.002592, "optim/total_tokens": 2265448448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489004, "loss/out": 3.2700161933898926, "created_at": "2025-01-15T17:45:19.623604+00:00"} {"global_step": 4322, "acc_step": 0, "speed/wps": 12915.052726944989, "speed/FLOPS": 202848650218885.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10596777498722076, "optim/lr": 0.0025926, "optim/total_tokens": 2265972736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2292916774749756, "created_at": "2025-01-15T17:45:29.781904+00:00"} {"global_step": 4323, "acc_step": 0, "speed/wps": 12916.865846834358, "speed/FLOPS": 202877127758235.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09769254177808762, "optim/lr": 0.0025932, "optim/total_tokens": 2266497024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.367727041244507, "created_at": "2025-01-15T17:45:39.935137+00:00"} {"global_step": 4324, "acc_step": 0, "speed/wps": 12910.909257099696, "speed/FLOPS": 202783571331240.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08711642026901245, "optim/lr": 0.0025938000000000003, "optim/total_tokens": 2267021312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.1252241134643555, "created_at": "2025-01-15T17:45:50.091270+00:00"} {"global_step": 4325, "acc_step": 0, "speed/wps": 12914.78808747674, "speed/FLOPS": 202844493692383.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09477312117815018, "optim/lr": 0.0025944, "optim/total_tokens": 2267545600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.113562822341919, "created_at": "2025-01-15T17:46:00.243772+00:00"} {"global_step": 4326, "acc_step": 0, "speed/wps": 12913.763810404816, "speed/FLOPS": 202828406013465.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08430014550685883, "optim/lr": 0.002595, "optim/total_tokens": 2268069888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.1001906394958496, "created_at": "2025-01-15T17:46:10.398971+00:00"} {"global_step": 4327, "acc_step": 0, "speed/wps": 12905.968408781166, "speed/FLOPS": 202705968518961.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08222486823797226, "optim/lr": 0.0025956, "optim/total_tokens": 2268594176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.2576863765716553, "created_at": "2025-01-15T17:46:20.563739+00:00"} {"global_step": 4328, "acc_step": 0, "speed/wps": 12911.630161867506, "speed/FLOPS": 202794894131246.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0872180163860321, "optim/lr": 0.0025962, "optim/total_tokens": 2269118464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 3.2836811542510986, "created_at": "2025-01-15T17:46:30.718686+00:00"} {"global_step": 4329, "acc_step": 0, "speed/wps": 12910.429721900997, "speed/FLOPS": 202776039571994.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09786175191402435, "optim/lr": 0.0025968000000000002, "optim/total_tokens": 2269642752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.2291784286499023, "created_at": "2025-01-15T17:46:40.878248+00:00"} {"global_step": 4330, "acc_step": 0, "speed/wps": 12910.624987201623, "speed/FLOPS": 202779106481863.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0950835794210434, "optim/lr": 0.0025974, "optim/total_tokens": 2270167040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409401, "loss/out": 3.289111614227295, "created_at": "2025-01-15T17:46:51.038161+00:00"} {"global_step": 4331, "acc_step": 0, "speed/wps": 12913.201074363518, "speed/FLOPS": 202819567470656.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0895228385925293, "optim/lr": 0.002598, "optim/total_tokens": 2270691328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.233218193054199, "created_at": "2025-01-15T17:47:01.193922+00:00"} {"global_step": 4332, "acc_step": 0, "speed/wps": 12909.626645430946, "speed/FLOPS": 202763426152490.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10511861741542816, "optim/lr": 0.0025986, "optim/total_tokens": 2271215616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.1154661178588867, "created_at": "2025-01-15T17:47:11.353719+00:00"} {"global_step": 4333, "acc_step": 0, "speed/wps": 12912.439055699313, "speed/FLOPS": 202807598920411.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09832259267568588, "optim/lr": 0.0025992, "optim/total_tokens": 2271739904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.313596725463867, "created_at": "2025-01-15T17:47:21.511515+00:00"} {"global_step": 4334, "acc_step": 0, "speed/wps": 12904.391275781165, "speed/FLOPS": 202681197477988.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1051551103591919, "optim/lr": 0.0025998, "optim/total_tokens": 2272264192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.1397647857666016, "created_at": "2025-01-15T17:47:31.672497+00:00"} {"global_step": 4335, "acc_step": 0, "speed/wps": 12911.163799985958, "speed/FLOPS": 202787569277048.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09355858713388443, "optim/lr": 0.0026004, "optim/total_tokens": 2272788480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374860, "loss/out": 3.1103525161743164, "created_at": "2025-01-15T17:47:41.833372+00:00"} {"global_step": 4336, "acc_step": 0, "speed/wps": 12913.81857020892, "speed/FLOPS": 202829266091436.4, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07803837954998016, "optim/lr": 0.002601, "optim/total_tokens": 2273312768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.2258753776550293, "created_at": "2025-01-15T17:47:51.989016+00:00"} {"global_step": 4337, "acc_step": 0, "speed/wps": 12917.134935497306, "speed/FLOPS": 202881354165454.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08132630586624146, "optim/lr": 0.0026016, "optim/total_tokens": 2273837056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 3.2772791385650635, "created_at": "2025-01-15T17:48:02.139914+00:00"} {"global_step": 4338, "acc_step": 0, "speed/wps": 12914.894682204886, "speed/FLOPS": 202846167909065.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08748076856136322, "optim/lr": 0.0026022, "optim/total_tokens": 2274361344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.242884874343872, "created_at": "2025-01-15T17:48:12.295098+00:00"} {"global_step": 4339, "acc_step": 0, "speed/wps": 12912.827242357962, "speed/FLOPS": 202813695925309.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08978912979364395, "optim/lr": 0.0026028, "optim/total_tokens": 2274885632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414480, "loss/out": 3.2074873447418213, "created_at": "2025-01-15T17:48:22.452266+00:00"} {"global_step": 4340, "acc_step": 0, "speed/wps": 12915.857538343227, "speed/FLOPS": 202861290887821.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10933691263198853, "optim/lr": 0.0026034, "optim/total_tokens": 2275409920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.168962240219116, "created_at": "2025-01-15T17:48:32.604986+00:00"} {"global_step": 4341, "acc_step": 0, "speed/wps": 12912.628736543113, "speed/FLOPS": 202810578118712.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0952664390206337, "optim/lr": 0.002604, "optim/total_tokens": 2275934208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.169485569000244, "created_at": "2025-01-15T17:48:42.759127+00:00"} {"global_step": 4342, "acc_step": 0, "speed/wps": 12916.207887826366, "speed/FLOPS": 202866793607884.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10094095021486282, "optim/lr": 0.0026046, "optim/total_tokens": 2276458496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.263666868209839, "created_at": "2025-01-15T17:48:52.911506+00:00"} {"global_step": 4343, "acc_step": 0, "speed/wps": 12909.990991061302, "speed/FLOPS": 202769148701278.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1211739107966423, "optim/lr": 0.0026052, "optim/total_tokens": 2276982784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.2544095516204834, "created_at": "2025-01-15T17:49:03.068652+00:00"} {"global_step": 4344, "acc_step": 0, "speed/wps": 12912.142690758938, "speed/FLOPS": 202802944101774.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0955933928489685, "optim/lr": 0.0026058, "optim/total_tokens": 2277507072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.1256906986236572, "created_at": "2025-01-15T17:49:13.223576+00:00"} {"global_step": 4345, "acc_step": 0, "speed/wps": 12912.656215238489, "speed/FLOPS": 202811009709382.7, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10140667110681534, "optim/lr": 0.0026064, "optim/total_tokens": 2278031360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.2724080085754395, "created_at": "2025-01-15T17:49:23.380798+00:00"} {"global_step": 4346, "acc_step": 0, "speed/wps": 12913.53788328151, "speed/FLOPS": 202824857517537.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09904547780752182, "optim/lr": 0.002607, "optim/total_tokens": 2278555648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 3.1858177185058594, "created_at": "2025-01-15T17:49:33.534935+00:00"} {"global_step": 4347, "acc_step": 0, "speed/wps": 12913.678900626128, "speed/FLOPS": 202827072388712.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1032501757144928, "optim/lr": 0.0026076, "optim/total_tokens": 2279079936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.1459624767303467, "created_at": "2025-01-15T17:49:43.688642+00:00"} {"global_step": 4348, "acc_step": 0, "speed/wps": 12912.204104036136, "speed/FLOPS": 202803908681683.84, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10517928004264832, "optim/lr": 0.0026081999999999998, "optim/total_tokens": 2279604224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 3.221726655960083, "created_at": "2025-01-15T17:49:53.843791+00:00"} {"global_step": 4349, "acc_step": 0, "speed/wps": 12906.757953145987, "speed/FLOPS": 202718369398158.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08763986825942993, "optim/lr": 0.0026088, "optim/total_tokens": 2280128512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.180917263031006, "created_at": "2025-01-15T17:50:04.002656+00:00"} {"global_step": 4350, "acc_step": 0, "speed/wps": 12909.725843732585, "speed/FLOPS": 202764984198130.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11539175361394882, "optim/lr": 0.0026094, "optim/total_tokens": 2280652800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 3.2579712867736816, "created_at": "2025-01-15T17:50:14.159048+00:00"} {"global_step": 4351, "acc_step": 0, "speed/wps": 12912.02810972372, "speed/FLOPS": 202801144449165.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1263669729232788, "optim/lr": 0.00261, "optim/total_tokens": 2281177088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.273685932159424, "created_at": "2025-01-15T17:50:24.315060+00:00"} {"global_step": 4352, "acc_step": 0, "speed/wps": 12915.481091929449, "speed/FLOPS": 202855378279601.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11161600798368454, "optim/lr": 0.0026106, "optim/total_tokens": 2281701376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 3.2434561252593994, "created_at": "2025-01-15T17:50:34.471782+00:00"} {"global_step": 4353, "acc_step": 0, "speed/wps": 12908.71313210876, "speed/FLOPS": 202749078170466.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.11235103011131287, "optim/lr": 0.0026111999999999997, "optim/total_tokens": 2282225664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.2542219161987305, "created_at": "2025-01-15T17:50:44.630471+00:00"} {"global_step": 4354, "acc_step": 0, "speed/wps": 12909.88969615392, "speed/FLOPS": 202767557725563.94, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10145560652017593, "optim/lr": 0.0026118, "optim/total_tokens": 2282749952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494587, "loss/out": 3.225029945373535, "created_at": "2025-01-15T17:50:54.787191+00:00"} {"global_step": 4355, "acc_step": 0, "speed/wps": 12917.082235313457, "speed/FLOPS": 202880526436649.0, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11355546861886978, "optim/lr": 0.0026124, "optim/total_tokens": 2283274240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.1594347953796387, "created_at": "2025-01-15T17:51:04.937872+00:00"} {"global_step": 4356, "acc_step": 0, "speed/wps": 12904.207730312806, "speed/FLOPS": 202678314644190.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.11040104925632477, "optim/lr": 0.002613, "optim/total_tokens": 2283798528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.155517101287842, "created_at": "2025-01-15T17:51:15.098783+00:00"} {"global_step": 4357, "acc_step": 0, "speed/wps": 12913.800317284265, "speed/FLOPS": 202828979404172.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11680518090724945, "optim/lr": 0.0026135999999999998, "optim/total_tokens": 2284322816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.225414752960205, "created_at": "2025-01-15T17:51:25.252070+00:00"} {"global_step": 4358, "acc_step": 0, "speed/wps": 12908.995280046529, "speed/FLOPS": 202753509691540.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11510267108678818, "optim/lr": 0.0026142, "optim/total_tokens": 2284847104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 3.226052761077881, "created_at": "2025-01-15T17:51:35.409385+00:00"} {"global_step": 4359, "acc_step": 0, "speed/wps": 12906.776454858167, "speed/FLOPS": 202718659992972.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10281883180141449, "optim/lr": 0.0026148, "optim/total_tokens": 2285371392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.3472366333007812, "created_at": "2025-01-15T17:51:45.568627+00:00"} {"global_step": 4360, "acc_step": 0, "speed/wps": 12911.106206982015, "speed/FLOPS": 202786664699780.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11339914053678513, "optim/lr": 0.0026154, "optim/total_tokens": 2285895680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.22084379196167, "created_at": "2025-01-15T17:51:55.728628+00:00"} {"global_step": 4361, "acc_step": 0, "speed/wps": 12902.312804315776, "speed/FLOPS": 202648552227505.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09926252067089081, "optim/lr": 0.002616, "optim/total_tokens": 2286419968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 3.2569570541381836, "created_at": "2025-01-15T17:52:05.892138+00:00"} {"global_step": 4362, "acc_step": 0, "speed/wps": 12912.79889805863, "speed/FLOPS": 202813250739139.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09390643239021301, "optim/lr": 0.0026166, "optim/total_tokens": 2286944256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 3.2291088104248047, "created_at": "2025-01-15T17:52:16.046093+00:00"} {"global_step": 4363, "acc_step": 0, "speed/wps": 12914.22892033333, "speed/FLOPS": 202835711204020.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10117145627737045, "optim/lr": 0.0026172, "optim/total_tokens": 2287468544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349355, "loss/out": 3.1375210285186768, "created_at": "2025-01-15T17:52:26.201923+00:00"} {"global_step": 4364, "acc_step": 0, "speed/wps": 12913.218420377456, "speed/FLOPS": 202819839913641.0, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10281018912792206, "optim/lr": 0.0026178, "optim/total_tokens": 2287992832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.180898427963257, "created_at": "2025-01-15T17:52:36.356864+00:00"} {"global_step": 4365, "acc_step": 0, "speed/wps": 12915.288754167304, "speed/FLOPS": 202852357350747.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10471588373184204, "optim/lr": 0.0026184000000000003, "optim/total_tokens": 2288517120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 3.3295764923095703, "created_at": "2025-01-15T17:52:46.513107+00:00"} {"global_step": 4366, "acc_step": 0, "speed/wps": 12905.695724705049, "speed/FLOPS": 202701685640838.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0886310413479805, "optim/lr": 0.0026190000000000002, "optim/total_tokens": 2289041408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.0803699493408203, "created_at": "2025-01-15T17:52:56.674695+00:00"} {"global_step": 4367, "acc_step": 0, "speed/wps": 12914.507075169573, "speed/FLOPS": 202840080007948.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09744904190301895, "optim/lr": 0.0026196, "optim/total_tokens": 2289565696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.236137866973877, "created_at": "2025-01-15T17:53:06.830778+00:00"} {"global_step": 4368, "acc_step": 0, "speed/wps": 12914.419556781506, "speed/FLOPS": 202838705411400.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09339743107557297, "optim/lr": 0.0026202, "optim/total_tokens": 2290089984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.2377610206604004, "created_at": "2025-01-15T17:53:16.985126+00:00"} {"global_step": 4369, "acc_step": 0, "speed/wps": 12917.77190407827, "speed/FLOPS": 202891358632304.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10079219192266464, "optim/lr": 0.0026208000000000004, "optim/total_tokens": 2290614272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.239626884460449, "created_at": "2025-01-15T17:53:27.135470+00:00"} {"global_step": 4370, "acc_step": 0, "speed/wps": 12913.81794692591, "speed/FLOPS": 202829256301920.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09660787880420685, "optim/lr": 0.0026214000000000003, "optim/total_tokens": 2291138560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.262627601623535, "created_at": "2025-01-15T17:53:37.293015+00:00"} {"global_step": 4371, "acc_step": 0, "speed/wps": 12912.528855065148, "speed/FLOPS": 202809009342849.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09344285726547241, "optim/lr": 0.002622, "optim/total_tokens": 2291662848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.213498592376709, "created_at": "2025-01-15T17:53:47.447280+00:00"} {"global_step": 4372, "acc_step": 0, "speed/wps": 12914.797864779193, "speed/FLOPS": 202844647258354.34, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10026062279939651, "optim/lr": 0.0026226, "optim/total_tokens": 2292187136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.2135424613952637, "created_at": "2025-01-15T17:53:57.604464+00:00"} {"global_step": 4373, "acc_step": 0, "speed/wps": 12915.643893340748, "speed/FLOPS": 202857935299478.62, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10254080593585968, "optim/lr": 0.0026232, "optim/total_tokens": 2292711424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2566022872924805, "created_at": "2025-01-15T17:54:07.759238+00:00"} {"global_step": 4374, "acc_step": 0, "speed/wps": 12912.94314913694, "speed/FLOPS": 202815516400546.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09312369674444199, "optim/lr": 0.0026238000000000004, "optim/total_tokens": 2293235712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 3.1434288024902344, "created_at": "2025-01-15T17:54:17.914939+00:00"} {"global_step": 4375, "acc_step": 0, "speed/wps": 12914.539942126023, "speed/FLOPS": 202840596228663.4, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09099776297807693, "optim/lr": 0.0026244000000000003, "optim/total_tokens": 2293760000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.285400867462158, "created_at": "2025-01-15T17:54:28.072711+00:00"} {"global_step": 4376, "acc_step": 0, "speed/wps": 12907.470447636559, "speed/FLOPS": 202729560103202.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09662219136953354, "optim/lr": 0.002625, "optim/total_tokens": 2294284288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 3.1859042644500732, "created_at": "2025-01-15T17:54:38.233007+00:00"} {"global_step": 4377, "acc_step": 0, "speed/wps": 12909.846257638648, "speed/FLOPS": 202766875463990.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09182791411876678, "optim/lr": 0.0026256, "optim/total_tokens": 2294808576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.1697707176208496, "created_at": "2025-01-15T17:54:48.392143+00:00"} {"global_step": 4378, "acc_step": 0, "speed/wps": 12914.505760921422, "speed/FLOPS": 202840059365875.25, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0904577448964119, "optim/lr": 0.0026262, "optim/total_tokens": 2295332864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 3.331627130508423, "created_at": "2025-01-15T17:54:58.545527+00:00"} {"global_step": 4379, "acc_step": 0, "speed/wps": 12913.723209793028, "speed/FLOPS": 202827768325065.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09662085771560669, "optim/lr": 0.0026268000000000003, "optim/total_tokens": 2295857152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.157257556915283, "created_at": "2025-01-15T17:55:08.704451+00:00"} {"global_step": 4380, "acc_step": 0, "speed/wps": 12916.807284188439, "speed/FLOPS": 202876207951408.16, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0951610580086708, "optim/lr": 0.0026274000000000002, "optim/total_tokens": 2296381440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469907, "loss/out": 3.23392915725708, "created_at": "2025-01-15T17:55:18.856766+00:00"} {"global_step": 4381, "acc_step": 0, "speed/wps": 12912.55402669161, "speed/FLOPS": 202809404697832.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11150245368480682, "optim/lr": 0.002628, "optim/total_tokens": 2296905728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.160379409790039, "created_at": "2025-01-15T17:55:29.011155+00:00"} {"global_step": 4382, "acc_step": 0, "speed/wps": 12918.854057742705, "speed/FLOPS": 202908355342639.5, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11713210493326187, "optim/lr": 0.0026286, "optim/total_tokens": 2297430016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.271725654602051, "created_at": "2025-01-15T17:55:39.162159+00:00"} {"global_step": 4383, "acc_step": 0, "speed/wps": 12909.763054886704, "speed/FLOPS": 202765568650438.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10041684657335281, "optim/lr": 0.0026292, "optim/total_tokens": 2297954304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.2770824432373047, "created_at": "2025-01-15T17:55:49.319019+00:00"} {"global_step": 4384, "acc_step": 0, "speed/wps": 12912.617981569498, "speed/FLOPS": 202810409197072.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08712425827980042, "optim/lr": 0.0026298000000000003, "optim/total_tokens": 2298478592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 3.160759925842285, "created_at": "2025-01-15T17:55:59.473385+00:00"} {"global_step": 4385, "acc_step": 0, "speed/wps": 12915.493740971071, "speed/FLOPS": 202855576950181.88, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11047300696372986, "optim/lr": 0.0026304, "optim/total_tokens": 2299002880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.1585426330566406, "created_at": "2025-01-15T17:56:09.626441+00:00"} {"global_step": 4386, "acc_step": 0, "speed/wps": 12911.822506763983, "speed/FLOPS": 202797915172155.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12643346190452576, "optim/lr": 0.002631, "optim/total_tokens": 2299527168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.2303521633148193, "created_at": "2025-01-15T17:56:19.783908+00:00"} {"global_step": 4387, "acc_step": 0, "speed/wps": 12912.408854325175, "speed/FLOPS": 202807124566329.6, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10844862461090088, "optim/lr": 0.0026316, "optim/total_tokens": 2300051456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478609, "loss/out": 3.2244796752929688, "created_at": "2025-01-15T17:56:29.943052+00:00"} {"global_step": 4388, "acc_step": 0, "speed/wps": 12912.5127204916, "speed/FLOPS": 202808755927200.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0923217311501503, "optim/lr": 0.0026322, "optim/total_tokens": 2300575744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.2119078636169434, "created_at": "2025-01-15T17:56:40.098579+00:00"} {"global_step": 4389, "acc_step": 0, "speed/wps": 12908.50712559691, "speed/FLOPS": 202745842555115.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12225464731454849, "optim/lr": 0.0026328000000000002, "optim/total_tokens": 2301100032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371642, "loss/out": 3.1574349403381348, "created_at": "2025-01-15T17:56:50.261403+00:00"} {"global_step": 4390, "acc_step": 0, "speed/wps": 12918.567027113517, "speed/FLOPS": 202903847132186.7, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11150545626878738, "optim/lr": 0.0026334, "optim/total_tokens": 2301624320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.304443359375, "created_at": "2025-01-15T17:57:00.410859+00:00"} {"global_step": 4391, "acc_step": 0, "speed/wps": 12913.216254492841, "speed/FLOPS": 202819805895446.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08556057512760162, "optim/lr": 0.002634, "optim/total_tokens": 2302148608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328312, "loss/out": 3.1748907566070557, "created_at": "2025-01-15T17:57:10.565732+00:00"} {"global_step": 4392, "acc_step": 0, "speed/wps": 12917.518198462316, "speed/FLOPS": 202887373836977.97, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09636428952217102, "optim/lr": 0.0026346, "optim/total_tokens": 2302672896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.254490375518799, "created_at": "2025-01-15T17:57:20.721693+00:00"} {"global_step": 4393, "acc_step": 0, "speed/wps": 12914.91441966047, "speed/FLOPS": 202846477912927.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0862005278468132, "optim/lr": 0.0026352, "optim/total_tokens": 2303197184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376816, "loss/out": 3.3290884494781494, "created_at": "2025-01-15T17:57:30.874977+00:00"} {"global_step": 4394, "acc_step": 0, "speed/wps": 12910.457486136605, "speed/FLOPS": 202776475647467.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08739221841096878, "optim/lr": 0.0026358, "optim/total_tokens": 2303721472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.1373143196105957, "created_at": "2025-01-15T17:57:41.030915+00:00"} {"global_step": 4395, "acc_step": 0, "speed/wps": 12916.16672419944, "speed/FLOPS": 202866147076559.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08545487374067307, "optim/lr": 0.0026364, "optim/total_tokens": 2304245760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.2151055335998535, "created_at": "2025-01-15T17:57:51.182297+00:00"} {"global_step": 4396, "acc_step": 0, "speed/wps": 12917.19982730921, "speed/FLOPS": 202882373380530.66, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09707098454236984, "optim/lr": 0.002637, "optim/total_tokens": 2304770048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.26423716545105, "created_at": "2025-01-15T17:58:01.332957+00:00"} {"global_step": 4397, "acc_step": 0, "speed/wps": 12914.6758948795, "speed/FLOPS": 202842731553475.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08135147392749786, "optim/lr": 0.0026376, "optim/total_tokens": 2305294336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 3.0814316272735596, "created_at": "2025-01-15T17:58:11.486525+00:00"} {"global_step": 4398, "acc_step": 0, "speed/wps": 12914.54223525842, "speed/FLOPS": 202840632245458.8, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08229253441095352, "optim/lr": 0.0026382, "optim/total_tokens": 2305818624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.1771013736724854, "created_at": "2025-01-15T17:58:21.641898+00:00"} {"global_step": 4399, "acc_step": 0, "speed/wps": 12916.60619615646, "speed/FLOPS": 202873049587542.0, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0895681381225586, "optim/lr": 0.0026388, "optim/total_tokens": 2306342912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 3.2794151306152344, "created_at": "2025-01-15T17:58:31.795689+00:00"} {"global_step": 4400, "acc_step": 0, "speed/wps": 12907.865746543446, "speed/FLOPS": 202735768815733.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08412513881921768, "optim/lr": 0.0026394, "optim/total_tokens": 2306867200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.249613046646118, "created_at": "2025-01-15T17:58:41.954774+00:00"} {"global_step": 4401, "acc_step": 0, "speed/wps": 12911.453865750404, "speed/FLOPS": 202792125158471.0, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08518128842115402, "optim/lr": 0.00264, "optim/total_tokens": 2307391488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1004929542541504, "created_at": "2025-01-15T17:58:52.111343+00:00"} {"global_step": 4402, "acc_step": 0, "speed/wps": 12914.836114913767, "speed/FLOPS": 202845248029279.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.137277752161026, "optim/lr": 0.0026406, "optim/total_tokens": 2307915776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 3.149514675140381, "created_at": "2025-01-15T17:59:02.267607+00:00"} {"global_step": 4403, "acc_step": 0, "speed/wps": 12907.514116393964, "speed/FLOPS": 202730245981045.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11544207483530045, "optim/lr": 0.0026412, "optim/total_tokens": 2308440064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.325162172317505, "created_at": "2025-01-15T17:59:12.425804+00:00"} {"global_step": 4404, "acc_step": 0, "speed/wps": 12913.527035556446, "speed/FLOPS": 202824687139108.75, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11091291904449463, "optim/lr": 0.0026418, "optim/total_tokens": 2308964352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.283634662628174, "created_at": "2025-01-15T17:59:22.581585+00:00"} {"global_step": 4405, "acc_step": 0, "speed/wps": 12911.925092420235, "speed/FLOPS": 202799526420854.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.12121833860874176, "optim/lr": 0.0026424, "optim/total_tokens": 2309488640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.1430842876434326, "created_at": "2025-01-15T17:59:32.736488+00:00"} {"global_step": 4406, "acc_step": 0, "speed/wps": 12911.945211719814, "speed/FLOPS": 202799842422100.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1319020539522171, "optim/lr": 0.002643, "optim/total_tokens": 2310012928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 3.23673415184021, "created_at": "2025-01-15T17:59:42.893588+00:00"} {"global_step": 4407, "acc_step": 0, "speed/wps": 12909.244717369065, "speed/FLOPS": 202757427447454.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10711489617824554, "optim/lr": 0.0026436, "optim/total_tokens": 2310537216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 3.2005674839019775, "created_at": "2025-01-15T17:59:53.053096+00:00"} {"global_step": 4408, "acc_step": 0, "speed/wps": 12916.784265428667, "speed/FLOPS": 202875846410154.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1142890602350235, "optim/lr": 0.0026441999999999998, "optim/total_tokens": 2311061504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.1088881492614746, "created_at": "2025-01-15T18:00:03.205867+00:00"} {"global_step": 4409, "acc_step": 0, "speed/wps": 12911.962466698968, "speed/FLOPS": 202800113435259.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09168379008769989, "optim/lr": 0.0026448, "optim/total_tokens": 2311585792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.189784288406372, "created_at": "2025-01-15T18:00:13.364428+00:00"} {"global_step": 4410, "acc_step": 0, "speed/wps": 12916.952365790283, "speed/FLOPS": 202878486657328.12, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09646417945623398, "optim/lr": 0.0026454, "optim/total_tokens": 2312110080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 3.168347120285034, "created_at": "2025-01-15T18:00:23.515627+00:00"} {"global_step": 4411, "acc_step": 0, "speed/wps": 12908.394546023317, "speed/FLOPS": 202744074338211.25, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.10663746297359467, "optim/lr": 0.002646, "optim/total_tokens": 2312634368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.313544511795044, "created_at": "2025-01-15T18:00:33.676562+00:00"} {"global_step": 4412, "acc_step": 0, "speed/wps": 12911.745246388953, "speed/FLOPS": 202796701691798.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09152386337518692, "optim/lr": 0.0026466, "optim/total_tokens": 2313158656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 3.1868162155151367, "created_at": "2025-01-15T18:00:43.831726+00:00"} {"global_step": 4413, "acc_step": 0, "speed/wps": 12904.93537490828, "speed/FLOPS": 202689743302450.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10306640714406967, "optim/lr": 0.0026471999999999997, "optim/total_tokens": 2313682944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.159003734588623, "created_at": "2025-01-15T18:00:53.993687+00:00"} {"global_step": 4414, "acc_step": 0, "speed/wps": 12908.349983349275, "speed/FLOPS": 202743374420179.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10260522365570068, "optim/lr": 0.0026478, "optim/total_tokens": 2314207232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.18215012550354, "created_at": "2025-01-15T18:01:04.153730+00:00"} {"global_step": 4415, "acc_step": 0, "speed/wps": 12914.822142181487, "speed/FLOPS": 202845028568318.0, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09392253309488297, "optim/lr": 0.0026484, "optim/total_tokens": 2314731520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334895, "loss/out": 3.2642087936401367, "created_at": "2025-01-15T18:01:14.309829+00:00"} {"global_step": 4416, "acc_step": 0, "speed/wps": 12908.971820233955, "speed/FLOPS": 202753141222947.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10600544512271881, "optim/lr": 0.002649, "optim/total_tokens": 2315255808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.0917608737945557, "created_at": "2025-01-15T18:01:24.473676+00:00"} {"global_step": 4417, "acc_step": 0, "speed/wps": 12904.306963673487, "speed/FLOPS": 202679873240480.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08978024870157242, "optim/lr": 0.0026496000000000002, "optim/total_tokens": 2315780096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307605, "loss/out": 3.286677360534668, "created_at": "2025-01-15T18:01:34.636267+00:00"} {"global_step": 4418, "acc_step": 0, "speed/wps": 12915.670569391797, "speed/FLOPS": 202858354283517.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08214204758405685, "optim/lr": 0.0026502, "optim/total_tokens": 2316304384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470341, "loss/out": 3.282160997390747, "created_at": "2025-01-15T18:01:44.789969+00:00"} {"global_step": 4419, "acc_step": 0, "speed/wps": 12910.167032877052, "speed/FLOPS": 202771913679901.34, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08181201666593552, "optim/lr": 0.0026508, "optim/total_tokens": 2316828672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374573, "loss/out": 3.172248363494873, "created_at": "2025-01-15T18:01:54.948111+00:00"} {"global_step": 4420, "acc_step": 0, "speed/wps": 12910.66655254586, "speed/FLOPS": 202779759322711.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09284361451864243, "optim/lr": 0.0026514, "optim/total_tokens": 2317352960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.2231011390686035, "created_at": "2025-01-15T18:02:05.103911+00:00"} {"global_step": 4421, "acc_step": 0, "speed/wps": 12915.712533487174, "speed/FLOPS": 202859013387299.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10134025663137436, "optim/lr": 0.0026520000000000003, "optim/total_tokens": 2317877248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.1780846118927, "created_at": "2025-01-15T18:02:15.256084+00:00"} {"global_step": 4422, "acc_step": 0, "speed/wps": 12910.154331500256, "speed/FLOPS": 202771714187325.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08966536074876785, "optim/lr": 0.0026526, "optim/total_tokens": 2318401536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.2287192344665527, "created_at": "2025-01-15T18:02:25.412832+00:00"} {"global_step": 4423, "acc_step": 0, "speed/wps": 12914.172218366462, "speed/FLOPS": 202834820621714.12, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.097930908203125, "optim/lr": 0.0026532, "optim/total_tokens": 2318925824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.3416922092437744, "created_at": "2025-01-15T18:02:35.568175+00:00"} {"global_step": 4424, "acc_step": 0, "speed/wps": 12905.896643588861, "speed/FLOPS": 202704841348000.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09240254759788513, "optim/lr": 0.0026538000000000004, "optim/total_tokens": 2319450112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.1393542289733887, "created_at": "2025-01-15T18:02:45.727956+00:00"} {"global_step": 4425, "acc_step": 0, "speed/wps": 12910.605637112152, "speed/FLOPS": 202778802562118.97, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1013210192322731, "optim/lr": 0.0026544000000000003, "optim/total_tokens": 2319974400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.1281700134277344, "created_at": "2025-01-15T18:02:55.883714+00:00"} {"global_step": 4426, "acc_step": 0, "speed/wps": 12916.758189375956, "speed/FLOPS": 202875436849914.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10705827921628952, "optim/lr": 0.0026550000000000002, "optim/total_tokens": 2320498688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.143134117126465, "created_at": "2025-01-15T18:03:06.040822+00:00"} {"global_step": 4427, "acc_step": 0, "speed/wps": 12914.148928150284, "speed/FLOPS": 202834454816864.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09331587702035904, "optim/lr": 0.0026556, "optim/total_tokens": 2321022976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 3.245169162750244, "created_at": "2025-01-15T18:03:16.195857+00:00"} {"global_step": 4428, "acc_step": 0, "speed/wps": 12911.269485621793, "speed/FLOPS": 202789229215186.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0950663685798645, "optim/lr": 0.0026562, "optim/total_tokens": 2321547264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 3.236334800720215, "created_at": "2025-01-15T18:03:26.356006+00:00"} {"global_step": 4429, "acc_step": 0, "speed/wps": 12908.912679103356, "speed/FLOPS": 202752212330222.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08305898308753967, "optim/lr": 0.0026568000000000004, "optim/total_tokens": 2322071552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.155958652496338, "created_at": "2025-01-15T18:03:36.525857+00:00"} {"global_step": 4430, "acc_step": 0, "speed/wps": 12908.362529630673, "speed/FLOPS": 202743571476769.62, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09753044694662094, "optim/lr": 0.0026574000000000003, "optim/total_tokens": 2322595840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.1105194091796875, "created_at": "2025-01-15T18:03:46.685694+00:00"} {"global_step": 4431, "acc_step": 0, "speed/wps": 12909.202429404459, "speed/FLOPS": 202756763256860.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08982304483652115, "optim/lr": 0.002658, "optim/total_tokens": 2323120128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.2552926540374756, "created_at": "2025-01-15T18:03:56.846884+00:00"} {"global_step": 4432, "acc_step": 0, "speed/wps": 12913.766603305925, "speed/FLOPS": 202828449879814.88, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09662462770938873, "optim/lr": 0.0026586, "optim/total_tokens": 2323644416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.2892353534698486, "created_at": "2025-01-15T18:04:07.001341+00:00"} {"global_step": 4433, "acc_step": 0, "speed/wps": 12911.400387496737, "speed/FLOPS": 202791285209010.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0763920396566391, "optim/lr": 0.0026592, "optim/total_tokens": 2324168704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350412, "loss/out": 3.1912970542907715, "created_at": "2025-01-15T18:04:17.163951+00:00"} {"global_step": 4434, "acc_step": 0, "speed/wps": 12912.755910310945, "speed/FLOPS": 202812575557491.0, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08962351083755493, "optim/lr": 0.0026598000000000004, "optim/total_tokens": 2324692992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.1655311584472656, "created_at": "2025-01-15T18:04:27.319120+00:00"} {"global_step": 4435, "acc_step": 0, "speed/wps": 12912.488903143558, "speed/FLOPS": 202808381843020.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08017899841070175, "optim/lr": 0.0026604000000000003, "optim/total_tokens": 2325217280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.2823102474212646, "created_at": "2025-01-15T18:04:37.473749+00:00"} {"global_step": 4436, "acc_step": 0, "speed/wps": 12917.05965185374, "speed/FLOPS": 202880171732381.22, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09216445684432983, "optim/lr": 0.002661, "optim/total_tokens": 2325741568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.242309808731079, "created_at": "2025-01-15T18:04:47.628775+00:00"} {"global_step": 4437, "acc_step": 0, "speed/wps": 12913.7604451798, "speed/FLOPS": 202828353157981.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10630424320697784, "optim/lr": 0.0026616, "optim/total_tokens": 2326265856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.108671188354492, "created_at": "2025-01-15T18:04:57.782093+00:00"} {"global_step": 4438, "acc_step": 0, "speed/wps": 12913.439252004195, "speed/FLOPS": 202823308377793.38, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11337372660636902, "optim/lr": 0.0026622, "optim/total_tokens": 2326790144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.1461057662963867, "created_at": "2025-01-15T18:05:07.936794+00:00"} {"global_step": 4439, "acc_step": 0, "speed/wps": 12908.112948813288, "speed/FLOPS": 202739651467072.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09750780463218689, "optim/lr": 0.0026628, "optim/total_tokens": 2327314432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.093174457550049, "created_at": "2025-01-15T18:05:18.101102+00:00"} {"global_step": 4440, "acc_step": 0, "speed/wps": 12916.38750494387, "speed/FLOPS": 202869614741535.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09252440929412842, "optim/lr": 0.0026634000000000002, "optim/total_tokens": 2327838720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334228, "loss/out": 3.1972153186798096, "created_at": "2025-01-15T18:05:28.255008+00:00"} {"global_step": 4441, "acc_step": 0, "speed/wps": 12913.268667208713, "speed/FLOPS": 202820629109172.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08194470405578613, "optim/lr": 0.002664, "optim/total_tokens": 2328363008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.2069969177246094, "created_at": "2025-01-15T18:05:38.409912+00:00"} {"global_step": 4442, "acc_step": 0, "speed/wps": 12911.72698511009, "speed/FLOPS": 202796414873320.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08653297275304794, "optim/lr": 0.0026646, "optim/total_tokens": 2328887296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 3.3158607482910156, "created_at": "2025-01-15T18:05:48.567357+00:00"} {"global_step": 4443, "acc_step": 0, "speed/wps": 12914.300922796123, "speed/FLOPS": 202836842101640.38, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09767655283212662, "optim/lr": 0.0026652, "optim/total_tokens": 2329411584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.127628803253174, "created_at": "2025-01-15T18:05:58.720175+00:00"} {"global_step": 4444, "acc_step": 0, "speed/wps": 12917.080172460692, "speed/FLOPS": 202880494036711.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09958001971244812, "optim/lr": 0.0026658, "optim/total_tokens": 2329935872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.2490663528442383, "created_at": "2025-01-15T18:06:08.875180+00:00"} {"global_step": 4445, "acc_step": 0, "speed/wps": 12915.59437000523, "speed/FLOPS": 202857157467442.12, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10088200867176056, "optim/lr": 0.0026664, "optim/total_tokens": 2330460160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.254483699798584, "created_at": "2025-01-15T18:06:19.029392+00:00"} {"global_step": 4446, "acc_step": 0, "speed/wps": 12908.800156393245, "speed/FLOPS": 202750445006438.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12925417721271515, "optim/lr": 0.002667, "optim/total_tokens": 2330984448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.206908702850342, "created_at": "2025-01-15T18:06:29.191158+00:00"} {"global_step": 4447, "acc_step": 0, "speed/wps": 12910.854840570999, "speed/FLOPS": 202782716644881.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.12743087112903595, "optim/lr": 0.0026676, "optim/total_tokens": 2331508736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.1458258628845215, "created_at": "2025-01-15T18:06:39.346695+00:00"} {"global_step": 4448, "acc_step": 0, "speed/wps": 12912.520216150499, "speed/FLOPS": 202808873656823.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26923874020576477, "optim/lr": 0.0026682, "optim/total_tokens": 2332033024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416629, "loss/out": 3.21947979927063, "created_at": "2025-01-15T18:06:49.505325+00:00"} {"global_step": 4449, "acc_step": 0, "speed/wps": 12910.980193282063, "speed/FLOPS": 202784685481461.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11970469355583191, "optim/lr": 0.0026688, "optim/total_tokens": 2332557312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.2706496715545654, "created_at": "2025-01-15T18:06:59.663521+00:00"} {"global_step": 4450, "acc_step": 0, "speed/wps": 12911.805279404223, "speed/FLOPS": 202797644592797.2, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1062210351228714, "optim/lr": 0.0026694, "optim/total_tokens": 2333081600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 3.2543959617614746, "created_at": "2025-01-15T18:07:09.820857+00:00"} {"global_step": 4451, "acc_step": 0, "speed/wps": 12910.839564134501, "speed/FLOPS": 202782476707454.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09346433728933334, "optim/lr": 0.00267, "optim/total_tokens": 2333605888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.255636692047119, "created_at": "2025-01-15T18:07:19.979627+00:00"} {"global_step": 4452, "acc_step": 0, "speed/wps": 12909.491209864676, "speed/FLOPS": 202761298950813.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09861362725496292, "optim/lr": 0.0026706, "optim/total_tokens": 2334130176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.2121710777282715, "created_at": "2025-01-15T18:07:30.139808+00:00"} {"global_step": 4453, "acc_step": 0, "speed/wps": 12908.339501889326, "speed/FLOPS": 202743209794448.34, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11252022534608841, "optim/lr": 0.0026712, "optim/total_tokens": 2334654464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.1173481941223145, "created_at": "2025-01-15T18:07:40.298351+00:00"} {"global_step": 4454, "acc_step": 0, "speed/wps": 12907.41468454621, "speed/FLOPS": 202728684267242.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10326631367206573, "optim/lr": 0.0026717999999999998, "optim/total_tokens": 2335178752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 3.2413713932037354, "created_at": "2025-01-15T18:07:50.458791+00:00"} {"global_step": 4455, "acc_step": 0, "speed/wps": 12916.52519415706, "speed/FLOPS": 202871777339833.03, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08825031667947769, "optim/lr": 0.0026724, "optim/total_tokens": 2335703040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 3.227327585220337, "created_at": "2025-01-15T18:08:00.613087+00:00"} {"global_step": 4456, "acc_step": 0, "speed/wps": 12897.869509445962, "speed/FLOPS": 202578764175848.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10674956440925598, "optim/lr": 0.002673, "optim/total_tokens": 2336227328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.967829942703247, "created_at": "2025-01-15T18:08:10.778906+00:00"} {"global_step": 4457, "acc_step": 0, "speed/wps": 12912.843036503838, "speed/FLOPS": 202813943994074.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09015952795743942, "optim/lr": 0.0026736, "optim/total_tokens": 2336751616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348781, "loss/out": 3.296224594116211, "created_at": "2025-01-15T18:08:20.933729+00:00"} {"global_step": 4458, "acc_step": 0, "speed/wps": 12912.390948944554, "speed/FLOPS": 202806843337722.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10652472078800201, "optim/lr": 0.0026742, "optim/total_tokens": 2337275904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.2139086723327637, "created_at": "2025-01-15T18:08:31.090675+00:00"} {"global_step": 4459, "acc_step": 0, "speed/wps": 12911.127895501317, "speed/FLOPS": 202787005347779.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09908998012542725, "optim/lr": 0.0026747999999999997, "optim/total_tokens": 2337800192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.273801326751709, "created_at": "2025-01-15T18:08:41.247015+00:00"} {"global_step": 4460, "acc_step": 0, "speed/wps": 12913.322787729865, "speed/FLOPS": 202821479146326.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08814734220504761, "optim/lr": 0.0026754, "optim/total_tokens": 2338324480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.2799811363220215, "created_at": "2025-01-15T18:08:51.401105+00:00"} {"global_step": 4461, "acc_step": 0, "speed/wps": 12910.98372169378, "speed/FLOPS": 202784740900016.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10568133741617203, "optim/lr": 0.002676, "optim/total_tokens": 2338848768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.2140071392059326, "created_at": "2025-01-15T18:09:01.560092+00:00"} {"global_step": 4462, "acc_step": 0, "speed/wps": 12913.286870051585, "speed/FLOPS": 202820915009832.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11098198592662811, "optim/lr": 0.0026766, "optim/total_tokens": 2339373056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.1624746322631836, "created_at": "2025-01-15T18:09:11.714387+00:00"} {"global_step": 4463, "acc_step": 0, "speed/wps": 12908.067749964808, "speed/FLOPS": 202738941557047.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10681235045194626, "optim/lr": 0.0026772, "optim/total_tokens": 2339897344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0854074954986572, "created_at": "2025-01-15T18:09:21.872270+00:00"} {"global_step": 4464, "acc_step": 0, "speed/wps": 12914.475856811696, "speed/FLOPS": 202839589680739.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09399009495973587, "optim/lr": 0.0026777999999999997, "optim/total_tokens": 2340421632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.0590972900390625, "created_at": "2025-01-15T18:09:32.025376+00:00"} {"global_step": 4465, "acc_step": 0, "speed/wps": 12916.355483266514, "speed/FLOPS": 202869111797089.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.09606607258319855, "optim/lr": 0.0026784, "optim/total_tokens": 2340945920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1510496139526367, "created_at": "2025-01-15T18:09:42.180089+00:00"} {"global_step": 4466, "acc_step": 0, "speed/wps": 12913.638852718968, "speed/FLOPS": 202826443381299.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11400553584098816, "optim/lr": 0.002679, "optim/total_tokens": 2341470208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316975, "loss/out": 3.2790355682373047, "created_at": "2025-01-15T18:09:52.334074+00:00"} {"global_step": 4467, "acc_step": 0, "speed/wps": 12909.4440996246, "speed/FLOPS": 202760559019757.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10811719298362732, "optim/lr": 0.0026796, "optim/total_tokens": 2341994496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.3391542434692383, "created_at": "2025-01-15T18:10:02.491332+00:00"} {"global_step": 4468, "acc_step": 0, "speed/wps": 12910.247116669547, "speed/FLOPS": 202773171505909.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12649385631084442, "optim/lr": 0.0026801999999999998, "optim/total_tokens": 2342518784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438464, "loss/out": 3.2977659702301025, "created_at": "2025-01-15T18:10:12.651857+00:00"} {"global_step": 4469, "acc_step": 0, "speed/wps": 12910.356963186934, "speed/FLOPS": 202774896796409.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11587920784950256, "optim/lr": 0.0026808, "optim/total_tokens": 2343043072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.261935234069824, "created_at": "2025-01-15T18:10:22.810551+00:00"} {"global_step": 4470, "acc_step": 0, "speed/wps": 12911.101316385551, "speed/FLOPS": 202786587886242.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0872802659869194, "optim/lr": 0.0026814, "optim/total_tokens": 2343567360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 3.1184511184692383, "created_at": "2025-01-15T18:10:32.970194+00:00"} {"global_step": 4471, "acc_step": 0, "speed/wps": 12909.22094878252, "speed/FLOPS": 202757054129140.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09046265482902527, "optim/lr": 0.002682, "optim/total_tokens": 2344091648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.1376423835754395, "created_at": "2025-01-15T18:10:43.132030+00:00"} {"global_step": 4472, "acc_step": 0, "speed/wps": 12912.536619672737, "speed/FLOPS": 202809131296681.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0798172578215599, "optim/lr": 0.0026826, "optim/total_tokens": 2344615936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.2265312671661377, "created_at": "2025-01-15T18:10:53.295576+00:00"} {"global_step": 4473, "acc_step": 0, "speed/wps": 12903.369857372463, "speed/FLOPS": 202665154698303.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08603572845458984, "optim/lr": 0.0026832, "optim/total_tokens": 2345140224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 3.248371124267578, "created_at": "2025-01-15T18:11:03.458110+00:00"} {"global_step": 4474, "acc_step": 0, "speed/wps": 12916.592336954094, "speed/FLOPS": 202872831909724.3, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09809034317731857, "optim/lr": 0.0026838, "optim/total_tokens": 2345664512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.211991548538208, "created_at": "2025-01-15T18:11:13.609410+00:00"} {"global_step": 4475, "acc_step": 0, "speed/wps": 12907.09263389331, "speed/FLOPS": 202723626019194.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10810092836618423, "optim/lr": 0.0026844, "optim/total_tokens": 2346188800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.2955925464630127, "created_at": "2025-01-15T18:11:23.773426+00:00"} {"global_step": 4476, "acc_step": 0, "speed/wps": 12908.979093915272, "speed/FLOPS": 202753255466107.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09219344705343246, "optim/lr": 0.0026850000000000003, "optim/total_tokens": 2346713088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.191650152206421, "created_at": "2025-01-15T18:11:33.934703+00:00"} {"global_step": 4477, "acc_step": 0, "speed/wps": 12908.661511760112, "speed/FLOPS": 202748267401956.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11533107608556747, "optim/lr": 0.0026856000000000002, "optim/total_tokens": 2347237376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 3.2294044494628906, "created_at": "2025-01-15T18:11:44.092000+00:00"} {"global_step": 4478, "acc_step": 0, "speed/wps": 12915.538802197938, "speed/FLOPS": 202856284698670.44, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.11850415170192719, "optim/lr": 0.0026862, "optim/total_tokens": 2347761664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.176780939102173, "created_at": "2025-01-15T18:11:54.247135+00:00"} {"global_step": 4479, "acc_step": 0, "speed/wps": 12913.943025389965, "speed/FLOPS": 202831220831074.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09896131604909897, "optim/lr": 0.0026868, "optim/total_tokens": 2348285952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.2077770233154297, "created_at": "2025-01-15T18:12:04.405042+00:00"} {"global_step": 4480, "acc_step": 0, "speed/wps": 12904.655198246463, "speed/FLOPS": 202685342742973.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11577829718589783, "optim/lr": 0.0026874000000000004, "optim/total_tokens": 2348810240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 3.179255247116089, "created_at": "2025-01-15T18:12:14.566844+00:00"} {"global_step": 4481, "acc_step": 0, "speed/wps": 12910.56241880557, "speed/FLOPS": 202778123759225.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10847008228302002, "optim/lr": 0.0026880000000000003, "optim/total_tokens": 2349334528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.2160301208496094, "created_at": "2025-01-15T18:12:24.723749+00:00"} {"global_step": 4482, "acc_step": 0, "speed/wps": 12912.607544120496, "speed/FLOPS": 202810245262593.5, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11076150834560394, "optim/lr": 0.0026886, "optim/total_tokens": 2349858816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425394, "loss/out": 3.1569342613220215, "created_at": "2025-01-15T18:12:34.878269+00:00"} {"global_step": 4483, "acc_step": 0, "speed/wps": 12915.949566288049, "speed/FLOPS": 202862736313154.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11520830541849136, "optim/lr": 0.0026892, "optim/total_tokens": 2350383104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.2087602615356445, "created_at": "2025-01-15T18:12:45.030349+00:00"} {"global_step": 4484, "acc_step": 0, "speed/wps": 12910.837664353794, "speed/FLOPS": 202782446868787.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10541248321533203, "optim/lr": 0.0026898, "optim/total_tokens": 2350907392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302869, "loss/out": 3.137423038482666, "created_at": "2025-01-15T18:12:55.187649+00:00"} {"global_step": 4485, "acc_step": 0, "speed/wps": 12914.397643892024, "speed/FLOPS": 202838361239360.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10189070552587509, "optim/lr": 0.0026904000000000003, "optim/total_tokens": 2351431680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 3.2660398483276367, "created_at": "2025-01-15T18:13:05.343553+00:00"} {"global_step": 4486, "acc_step": 0, "speed/wps": 12911.871494172972, "speed/FLOPS": 202798684586729.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0987926796078682, "optim/lr": 0.0026910000000000002, "optim/total_tokens": 2351955968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 3.2893595695495605, "created_at": "2025-01-15T18:13:15.499208+00:00"} {"global_step": 4487, "acc_step": 0, "speed/wps": 12915.821530702804, "speed/FLOPS": 202860725338349.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10554146021604538, "optim/lr": 0.0026916, "optim/total_tokens": 2352480256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.3278536796569824, "created_at": "2025-01-15T18:13:25.650903+00:00"} {"global_step": 4488, "acc_step": 0, "speed/wps": 12914.722978130698, "speed/FLOPS": 202843471060634.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09301603585481644, "optim/lr": 0.0026922, "optim/total_tokens": 2353004544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.222935438156128, "created_at": "2025-01-15T18:13:35.806613+00:00"} {"global_step": 4489, "acc_step": 0, "speed/wps": 12914.372758969695, "speed/FLOPS": 202837970387459.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09973935782909393, "optim/lr": 0.0026928, "optim/total_tokens": 2353528832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.2293381690979004, "created_at": "2025-01-15T18:13:45.962694+00:00"} {"global_step": 4490, "acc_step": 0, "speed/wps": 12920.322812225702, "speed/FLOPS": 202931424150075.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08189396560192108, "optim/lr": 0.0026934000000000003, "optim/total_tokens": 2354053120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.198075771331787, "created_at": "2025-01-15T18:13:56.112443+00:00"} {"global_step": 4491, "acc_step": 0, "speed/wps": 12913.823291046441, "speed/FLOPS": 202829340238676.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08345998823642731, "optim/lr": 0.002694, "optim/total_tokens": 2354577408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.1899867057800293, "created_at": "2025-01-15T18:14:06.265869+00:00"} {"global_step": 4492, "acc_step": 0, "speed/wps": 12909.14351469037, "speed/FLOPS": 202755837920318.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07981410622596741, "optim/lr": 0.0026946, "optim/total_tokens": 2355101696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.1779043674468994, "created_at": "2025-01-15T18:14:16.423558+00:00"} {"global_step": 4493, "acc_step": 0, "speed/wps": 12910.7356076295, "speed/FLOPS": 202780843927692.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08848855644464493, "optim/lr": 0.0026952, "optim/total_tokens": 2355625984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.249211072921753, "created_at": "2025-01-15T18:14:26.588549+00:00"} {"global_step": 4494, "acc_step": 0, "speed/wps": 12906.191817639585, "speed/FLOPS": 202709477462078.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09498859196901321, "optim/lr": 0.0026958, "optim/total_tokens": 2356150272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.2235517501831055, "created_at": "2025-01-15T18:14:36.749107+00:00"} {"global_step": 4495, "acc_step": 0, "speed/wps": 12914.856578680656, "speed/FLOPS": 202845569440858.16, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1014992743730545, "optim/lr": 0.0026964000000000003, "optim/total_tokens": 2356674560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376816, "loss/out": 3.262159585952759, "created_at": "2025-01-15T18:14:46.903886+00:00"} {"global_step": 4496, "acc_step": 0, "speed/wps": 12918.348971480642, "speed/FLOPS": 202900422268832.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09169124066829681, "optim/lr": 0.002697, "optim/total_tokens": 2357198848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.21990966796875, "created_at": "2025-01-15T18:14:57.080555+00:00"} {"global_step": 4497, "acc_step": 0, "speed/wps": 12916.459314544527, "speed/FLOPS": 202870742609990.12, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10301089286804199, "optim/lr": 0.0026976, "optim/total_tokens": 2357723136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 3.1719226837158203, "created_at": "2025-01-15T18:15:07.232261+00:00"} {"global_step": 4498, "acc_step": 0, "speed/wps": 12913.65751279937, "speed/FLOPS": 202826736463503.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09795635938644409, "optim/lr": 0.0026982, "optim/total_tokens": 2358247424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.1302738189697266, "created_at": "2025-01-15T18:15:17.385665+00:00"} {"global_step": 4499, "acc_step": 0, "speed/wps": 12916.306502131116, "speed/FLOPS": 202868342481051.25, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09861356019973755, "optim/lr": 0.0026988, "optim/total_tokens": 2358771712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.230822801589966, "created_at": "2025-01-15T18:15:27.539436+00:00"} {"global_step": 4500, "acc_step": 0, "speed/wps": 12912.113150613242, "speed/FLOPS": 202802480133194.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10826944559812546, "optim/lr": 0.0026994000000000002, "optim/total_tokens": 2359296000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.145108699798584, "created_at": "2025-01-15T18:15:37.698467+00:00"} {"global_step": 4501, "acc_step": 0, "speed/wps": 12912.462829029568, "speed/FLOPS": 202807972313231.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09477584809064865, "optim/lr": 0.0027, "optim/total_tokens": 2359820288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.2016892433166504, "created_at": "2025-01-15T18:15:47.853577+00:00"} {"global_step": 4502, "acc_step": 0, "speed/wps": 12915.401149421674, "speed/FLOPS": 202854122672664.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09202265739440918, "optim/lr": 0.0027006, "optim/total_tokens": 2360344576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329658, "loss/out": 3.1969094276428223, "created_at": "2025-01-15T18:15:58.012186+00:00"} {"global_step": 4503, "acc_step": 0, "speed/wps": 12915.419165082507, "speed/FLOPS": 202854405633373.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08852076530456543, "optim/lr": 0.0027012, "optim/total_tokens": 2360868864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 3.1318540573120117, "created_at": "2025-01-15T18:16:08.168087+00:00"} {"global_step": 4504, "acc_step": 0, "speed/wps": 12909.301900639095, "speed/FLOPS": 202758325589287.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09125793725252151, "optim/lr": 0.0027018, "optim/total_tokens": 2361393152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.1935317516326904, "created_at": "2025-01-15T18:16:18.325747+00:00"} {"global_step": 4505, "acc_step": 0, "speed/wps": 12911.916848785153, "speed/FLOPS": 202799396943237.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09503740072250366, "optim/lr": 0.0027024, "optim/total_tokens": 2361917440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.255248546600342, "created_at": "2025-01-15T18:16:28.483772+00:00"} {"global_step": 4506, "acc_step": 0, "speed/wps": 12913.689548353996, "speed/FLOPS": 202827239625910.25, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11584920436143875, "optim/lr": 0.002703, "optim/total_tokens": 2362441728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331823, "loss/out": 3.230952024459839, "created_at": "2025-01-15T18:16:38.638437+00:00"} {"global_step": 4507, "acc_step": 0, "speed/wps": 12911.651940988497, "speed/FLOPS": 202795236202269.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0941118523478508, "optim/lr": 0.0027036, "optim/total_tokens": 2362966016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.2231850624084473, "created_at": "2025-01-15T18:16:48.793537+00:00"} {"global_step": 4508, "acc_step": 0, "speed/wps": 12909.369935796722, "speed/FLOPS": 202759394174928.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08512196689844131, "optim/lr": 0.0027042, "optim/total_tokens": 2363490304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406438, "loss/out": 3.2396483421325684, "created_at": "2025-01-15T18:16:58.953823+00:00"} {"global_step": 4509, "acc_step": 0, "speed/wps": 12914.33171013767, "speed/FLOPS": 202837325659145.62, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07855115085840225, "optim/lr": 0.0027048, "optim/total_tokens": 2364014592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.1906707286834717, "created_at": "2025-01-15T18:17:09.106721+00:00"} {"global_step": 4510, "acc_step": 0, "speed/wps": 12911.99438012122, "speed/FLOPS": 202800614679409.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08658163249492645, "optim/lr": 0.0027054, "optim/total_tokens": 2364538880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.141509532928467, "created_at": "2025-01-15T18:17:19.261437+00:00"} {"global_step": 4511, "acc_step": 0, "speed/wps": 12909.636966748434, "speed/FLOPS": 202763588262964.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09619983285665512, "optim/lr": 0.002706, "optim/total_tokens": 2365063168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454173, "loss/out": 3.0871071815490723, "created_at": "2025-01-15T18:17:29.419768+00:00"} {"global_step": 4512, "acc_step": 0, "speed/wps": 12902.906905053376, "speed/FLOPS": 202657883395969.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08803562074899673, "optim/lr": 0.0027066, "optim/total_tokens": 2365587456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1127429008483887, "created_at": "2025-01-15T18:17:39.583391+00:00"} {"global_step": 4513, "acc_step": 0, "speed/wps": 12910.441813890138, "speed/FLOPS": 202776229493300.53, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10328849405050278, "optim/lr": 0.0027072, "optim/total_tokens": 2366111744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.1986947059631348, "created_at": "2025-01-15T18:17:49.739424+00:00"} {"global_step": 4514, "acc_step": 0, "speed/wps": 12915.400191748478, "speed/FLOPS": 202854107631090.94, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11627457290887833, "optim/lr": 0.0027078, "optim/total_tokens": 2366636032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.1433818340301514, "created_at": "2025-01-15T18:17:59.891611+00:00"} {"global_step": 4515, "acc_step": 0, "speed/wps": 12915.112714868535, "speed/FLOPS": 202849592411656.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09656576067209244, "optim/lr": 0.0027084, "optim/total_tokens": 2367160320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.138216495513916, "created_at": "2025-01-15T18:18:10.044519+00:00"} {"global_step": 4516, "acc_step": 0, "speed/wps": 12905.569700619486, "speed/FLOPS": 202699706259399.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07515458762645721, "optim/lr": 0.002709, "optim/total_tokens": 2367684608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.2127575874328613, "created_at": "2025-01-15T18:18:20.205515+00:00"} {"global_step": 4517, "acc_step": 0, "speed/wps": 12911.729634982392, "speed/FLOPS": 202796456493206.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07997199892997742, "optim/lr": 0.0027096, "optim/total_tokens": 2368208896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.024122714996338, "created_at": "2025-01-15T18:18:30.366149+00:00"} {"global_step": 4518, "acc_step": 0, "speed/wps": 12907.557490680641, "speed/FLOPS": 202730927233819.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07797227799892426, "optim/lr": 0.0027102, "optim/total_tokens": 2368733184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1956253051757812, "created_at": "2025-01-15T18:18:40.528643+00:00"} {"global_step": 4519, "acc_step": 0, "speed/wps": 12910.028279536175, "speed/FLOPS": 202769734368015.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08165343105792999, "optim/lr": 0.0027107999999999998, "optim/total_tokens": 2369257472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304572, "loss/out": 3.177072048187256, "created_at": "2025-01-15T18:18:50.687435+00:00"} {"global_step": 4520, "acc_step": 0, "speed/wps": 12916.551041662597, "speed/FLOPS": 202872183310425.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09533443301916122, "optim/lr": 0.0027114, "optim/total_tokens": 2369781760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.213196277618408, "created_at": "2025-01-15T18:19:00.842725+00:00"} {"global_step": 4521, "acc_step": 0, "speed/wps": 12913.946824409979, "speed/FLOPS": 202831280499903.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08285126835107803, "optim/lr": 0.002712, "optim/total_tokens": 2370306048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456307, "loss/out": 3.3060898780822754, "created_at": "2025-01-15T18:19:11.000023+00:00"} {"global_step": 4522, "acc_step": 0, "speed/wps": 12914.17559447122, "speed/FLOPS": 202834873648078.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07368922978639603, "optim/lr": 0.0027126, "optim/total_tokens": 2370830336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.1792964935302734, "created_at": "2025-01-15T18:19:21.153039+00:00"} {"global_step": 4523, "acc_step": 0, "speed/wps": 12914.28929403591, "speed/FLOPS": 202836659455981.62, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09270680695772171, "optim/lr": 0.0027132, "optim/total_tokens": 2371354624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311884, "loss/out": 3.187417507171631, "created_at": "2025-01-15T18:19:31.307444+00:00"} {"global_step": 4524, "acc_step": 0, "speed/wps": 12910.01175685576, "speed/FLOPS": 202769474856615.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08614606410264969, "optim/lr": 0.0027137999999999997, "optim/total_tokens": 2371878912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.179258346557617, "created_at": "2025-01-15T18:19:41.467946+00:00"} {"global_step": 4525, "acc_step": 0, "speed/wps": 12912.823202304347, "speed/FLOPS": 202813632470716.28, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09151912480592728, "optim/lr": 0.0027144, "optim/total_tokens": 2372403200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.1175460815429688, "created_at": "2025-01-15T18:19:51.630850+00:00"} {"global_step": 4526, "acc_step": 0, "speed/wps": 12909.072535932732, "speed/FLOPS": 202754723101394.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09463992714881897, "optim/lr": 0.002715, "optim/total_tokens": 2372927488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.050731658935547, "created_at": "2025-01-15T18:20:01.788284+00:00"} {"global_step": 4527, "acc_step": 0, "speed/wps": 12910.360861948375, "speed/FLOPS": 202774958031815.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07723779231309891, "optim/lr": 0.0027156, "optim/total_tokens": 2373451776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.2226364612579346, "created_at": "2025-01-15T18:20:11.947187+00:00"} {"global_step": 4528, "acc_step": 0, "speed/wps": 12912.282658820093, "speed/FLOPS": 202805142492511.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10804211348295212, "optim/lr": 0.0027162, "optim/total_tokens": 2373976064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.274693727493286, "created_at": "2025-01-15T18:20:22.105604+00:00"} {"global_step": 4529, "acc_step": 0, "speed/wps": 12911.243794280212, "speed/FLOPS": 202788825697363.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0903371274471283, "optim/lr": 0.0027168, "optim/total_tokens": 2374500352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.149913787841797, "created_at": "2025-01-15T18:20:32.266200+00:00"} {"global_step": 4530, "acc_step": 0, "speed/wps": 12913.461493408724, "speed/FLOPS": 202823657709614.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08855478465557098, "optim/lr": 0.0027174, "optim/total_tokens": 2375024640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 3.169290781021118, "created_at": "2025-01-15T18:20:42.420826+00:00"} {"global_step": 4531, "acc_step": 0, "speed/wps": 12912.52453679109, "speed/FLOPS": 202808941518421.16, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09626423567533493, "optim/lr": 0.002718, "optim/total_tokens": 2375548928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.1673200130462646, "created_at": "2025-01-15T18:20:52.581154+00:00"} {"global_step": 4532, "acc_step": 0, "speed/wps": 12915.394605719415, "speed/FLOPS": 202854019894828.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10830505937337875, "optim/lr": 0.0027186000000000003, "optim/total_tokens": 2376073216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423810, "loss/out": 3.2016968727111816, "created_at": "2025-01-15T18:21:02.734304+00:00"} {"global_step": 4533, "acc_step": 0, "speed/wps": 12911.752202177675, "speed/FLOPS": 202796810942018.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14594444632530212, "optim/lr": 0.0027192, "optim/total_tokens": 2376597504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.1680941581726074, "created_at": "2025-01-15T18:21:12.889701+00:00"} {"global_step": 4534, "acc_step": 0, "speed/wps": 12908.783774203155, "speed/FLOPS": 202750187701631.16, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1285974532365799, "optim/lr": 0.0027198, "optim/total_tokens": 2377121792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.104766368865967, "created_at": "2025-01-15T18:21:23.051638+00:00"} {"global_step": 4535, "acc_step": 0, "speed/wps": 12907.655601219953, "speed/FLOPS": 202732468194657.84, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1079481765627861, "optim/lr": 0.0027204000000000004, "optim/total_tokens": 2377646080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332493, "loss/out": 3.052666187286377, "created_at": "2025-01-15T18:21:33.209696+00:00"} {"global_step": 4536, "acc_step": 0, "speed/wps": 12909.541651501037, "speed/FLOPS": 202762091206026.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10295959562063217, "optim/lr": 0.0027210000000000003, "optim/total_tokens": 2378170368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.218179225921631, "created_at": "2025-01-15T18:21:43.367332+00:00"} {"global_step": 4537, "acc_step": 0, "speed/wps": 12913.889115332815, "speed/FLOPS": 202830374099546.34, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10530757158994675, "optim/lr": 0.0027216000000000002, "optim/total_tokens": 2378694656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 3.209226131439209, "created_at": "2025-01-15T18:21:53.520546+00:00"} {"global_step": 4538, "acc_step": 0, "speed/wps": 12909.422876209068, "speed/FLOPS": 202760225676852.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08499817550182343, "optim/lr": 0.0027222, "optim/total_tokens": 2379218944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.1740670204162598, "created_at": "2025-01-15T18:22:03.678572+00:00"} {"global_step": 4539, "acc_step": 0, "speed/wps": 12911.696030133422, "speed/FLOPS": 202795928682875.03, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10621530562639236, "optim/lr": 0.0027228, "optim/total_tokens": 2379743232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 3.1500937938690186, "created_at": "2025-01-15T18:22:13.836633+00:00"} {"global_step": 4540, "acc_step": 0, "speed/wps": 12911.79476508691, "speed/FLOPS": 202797479450995.6, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09015563875436783, "optim/lr": 0.0027234000000000004, "optim/total_tokens": 2380267520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0938706398010254, "created_at": "2025-01-15T18:22:23.991697+00:00"} {"global_step": 4541, "acc_step": 0, "speed/wps": 12908.58045308803, "speed/FLOPS": 202746994264125.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11749010533094406, "optim/lr": 0.0027240000000000003, "optim/total_tokens": 2380791808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 3.114069938659668, "created_at": "2025-01-15T18:22:34.150766+00:00"} {"global_step": 4542, "acc_step": 0, "speed/wps": 12907.610125011397, "speed/FLOPS": 202731753928311.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11716574430465698, "optim/lr": 0.0027246, "optim/total_tokens": 2381316096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.1236183643341064, "created_at": "2025-01-15T18:22:44.309477+00:00"} {"global_step": 4543, "acc_step": 0, "speed/wps": 12906.367894337194, "speed/FLOPS": 202712242988569.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11305709928274155, "optim/lr": 0.0027252, "optim/total_tokens": 2381840384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372985, "loss/out": 3.1547341346740723, "created_at": "2025-01-15T18:22:54.487394+00:00"} {"global_step": 4544, "acc_step": 0, "speed/wps": 12910.486407589084, "speed/FLOPS": 202776929898620.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12017667293548584, "optim/lr": 0.0027258, "optim/total_tokens": 2382364672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 3.138545036315918, "created_at": "2025-01-15T18:23:04.644511+00:00"} {"global_step": 4545, "acc_step": 0, "speed/wps": 12907.929221692499, "speed/FLOPS": 202736765780173.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12356649339199066, "optim/lr": 0.0027264000000000004, "optim/total_tokens": 2382888960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.233524799346924, "created_at": "2025-01-15T18:23:14.806908+00:00"} {"global_step": 4546, "acc_step": 0, "speed/wps": 12909.080420712477, "speed/FLOPS": 202754846942695.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10806764662265778, "optim/lr": 0.0027270000000000003, "optim/total_tokens": 2383413248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441677, "loss/out": 3.101830005645752, "created_at": "2025-01-15T18:23:24.964955+00:00"} {"global_step": 4547, "acc_step": 0, "speed/wps": 12911.968662202527, "speed/FLOPS": 202800210744156.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1153516098856926, "optim/lr": 0.0027276, "optim/total_tokens": 2383937536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.129563331604004, "created_at": "2025-01-15T18:23:35.122075+00:00"} {"global_step": 4548, "acc_step": 0, "speed/wps": 12907.428196154704, "speed/FLOPS": 202728896485620.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08612166345119476, "optim/lr": 0.0027282, "optim/total_tokens": 2384461824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.1736040115356445, "created_at": "2025-01-15T18:23:45.281661+00:00"} {"global_step": 4549, "acc_step": 0, "speed/wps": 12914.2399860974, "speed/FLOPS": 202835885007051.12, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09289439022541046, "optim/lr": 0.0027288, "optim/total_tokens": 2384986112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.0054948329925537, "created_at": "2025-01-15T18:23:55.434565+00:00"} {"global_step": 4550, "acc_step": 0, "speed/wps": 12912.288101182106, "speed/FLOPS": 202805227972285.5, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09379088133573532, "optim/lr": 0.0027294000000000003, "optim/total_tokens": 2385510400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 3.21785831451416, "created_at": "2025-01-15T18:24:05.593596+00:00"} {"global_step": 4551, "acc_step": 0, "speed/wps": 12904.903241958329, "speed/FLOPS": 202689238610316.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09796956181526184, "optim/lr": 0.0027300000000000002, "optim/total_tokens": 2386034688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.1993837356567383, "created_at": "2025-01-15T18:24:15.754312+00:00"} {"global_step": 4552, "acc_step": 0, "speed/wps": 12911.205465926128, "speed/FLOPS": 202788223697893.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10417849570512772, "optim/lr": 0.0027306, "optim/total_tokens": 2386558976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.2495338916778564, "created_at": "2025-01-15T18:24:25.909643+00:00"} {"global_step": 4553, "acc_step": 0, "speed/wps": 12905.320394644992, "speed/FLOPS": 202695790566487.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10408496111631393, "optim/lr": 0.0027312, "optim/total_tokens": 2387083264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 3.219668388366699, "created_at": "2025-01-15T18:24:36.072343+00:00"} {"global_step": 4554, "acc_step": 0, "speed/wps": 12908.786361298204, "speed/FLOPS": 202750228335513.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0791170597076416, "optim/lr": 0.0027318, "optim/total_tokens": 2387607552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.1833653450012207, "created_at": "2025-01-15T18:24:46.230557+00:00"} {"global_step": 4555, "acc_step": 0, "speed/wps": 12900.191028036166, "speed/FLOPS": 202615226815408.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08785437047481537, "optim/lr": 0.0027324000000000003, "optim/total_tokens": 2388131840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.287013530731201, "created_at": "2025-01-15T18:24:56.397057+00:00"} {"global_step": 4556, "acc_step": 0, "speed/wps": 12912.079890864412, "speed/FLOPS": 202801957743135.16, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08993182331323624, "optim/lr": 0.002733, "optim/total_tokens": 2388656128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.415501832962036, "created_at": "2025-01-15T18:25:06.551782+00:00"} {"global_step": 4557, "acc_step": 0, "speed/wps": 12910.307976610715, "speed/FLOPS": 202774127394915.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07743604481220245, "optim/lr": 0.0027336, "optim/total_tokens": 2389180416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 3.0715138912200928, "created_at": "2025-01-15T18:25:16.707952+00:00"} {"global_step": 4558, "acc_step": 0, "speed/wps": 12914.120395557691, "speed/FLOPS": 202834006673290.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.102384552359581, "optim/lr": 0.0027342, "optim/total_tokens": 2389704704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.2253148555755615, "created_at": "2025-01-15T18:25:26.863468+00:00"} {"global_step": 4559, "acc_step": 0, "speed/wps": 12912.264625845788, "speed/FLOPS": 202804859259870.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10051383823156357, "optim/lr": 0.0027348, "optim/total_tokens": 2390228992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.2191720008850098, "created_at": "2025-01-15T18:25:37.019150+00:00"} {"global_step": 4560, "acc_step": 0, "speed/wps": 12910.844851307687, "speed/FLOPS": 202782559749774.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09175845980644226, "optim/lr": 0.0027354000000000002, "optim/total_tokens": 2390753280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 3.1744303703308105, "created_at": "2025-01-15T18:25:47.176354+00:00"} {"global_step": 4561, "acc_step": 0, "speed/wps": 12909.470825857246, "speed/FLOPS": 202760978791966.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10970593988895416, "optim/lr": 0.002736, "optim/total_tokens": 2391277568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 3.1931052207946777, "created_at": "2025-01-15T18:25:57.338890+00:00"} {"global_step": 4562, "acc_step": 0, "speed/wps": 12907.171584598638, "speed/FLOPS": 202724866048511.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08951879292726517, "optim/lr": 0.0027366, "optim/total_tokens": 2391801856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420921, "loss/out": 3.2527034282684326, "created_at": "2025-01-15T18:26:07.501327+00:00"} {"global_step": 4563, "acc_step": 0, "speed/wps": 12910.064027504237, "speed/FLOPS": 202770295838977.06, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10556419938802719, "optim/lr": 0.0027372, "optim/total_tokens": 2392326144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.1483376026153564, "created_at": "2025-01-15T18:26:17.665133+00:00"} {"global_step": 4564, "acc_step": 0, "speed/wps": 12911.745281926002, "speed/FLOPS": 202796702249956.53, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09736859798431396, "optim/lr": 0.0027378, "optim/total_tokens": 2392850432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 3.151564598083496, "created_at": "2025-01-15T18:26:27.823120+00:00"} {"global_step": 4565, "acc_step": 0, "speed/wps": 12909.149550004762, "speed/FLOPS": 202755932713224.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08154870569705963, "optim/lr": 0.0027383999999999998, "optim/total_tokens": 2393374720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.2231955528259277, "created_at": "2025-01-15T18:26:37.981430+00:00"} {"global_step": 4566, "acc_step": 0, "speed/wps": 12907.829590481675, "speed/FLOPS": 202735200935099.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0873328372836113, "optim/lr": 0.002739, "optim/total_tokens": 2393899008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507176, "loss/out": 3.2261252403259277, "created_at": "2025-01-15T18:26:48.143824+00:00"} {"global_step": 4567, "acc_step": 0, "speed/wps": 12909.556009505375, "speed/FLOPS": 202762316718215.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09878989309072495, "optim/lr": 0.0027396, "optim/total_tokens": 2394423296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307513, "loss/out": 3.213609218597412, "created_at": "2025-01-15T18:26:58.302209+00:00"} {"global_step": 4568, "acc_step": 0, "speed/wps": 12909.388317111496, "speed/FLOPS": 202759682878735.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08738775551319122, "optim/lr": 0.0027402, "optim/total_tokens": 2394947584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484762, "loss/out": 3.2110722064971924, "created_at": "2025-01-15T18:27:08.467497+00:00"} {"global_step": 4569, "acc_step": 0, "speed/wps": 12911.092382084358, "speed/FLOPS": 202786447560765.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1015789806842804, "optim/lr": 0.0027408, "optim/total_tokens": 2395471872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350240, "loss/out": 3.0623435974121094, "created_at": "2025-01-15T18:27:18.623684+00:00"} {"global_step": 4570, "acc_step": 0, "speed/wps": 12906.381568272416, "speed/FLOPS": 202712457756512.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08951391279697418, "optim/lr": 0.0027413999999999997, "optim/total_tokens": 2395996160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.17648983001709, "created_at": "2025-01-15T18:27:28.785798+00:00"} {"global_step": 4571, "acc_step": 0, "speed/wps": 12908.669879111403, "speed/FLOPS": 202748398822706.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09527205675840378, "optim/lr": 0.002742, "optim/total_tokens": 2396520448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.3177294731140137, "created_at": "2025-01-15T18:27:38.943943+00:00"} {"global_step": 4572, "acc_step": 0, "speed/wps": 12911.571965327392, "speed/FLOPS": 202793980074613.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.087556391954422, "optim/lr": 0.0027426, "optim/total_tokens": 2397044736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.319979190826416, "created_at": "2025-01-15T18:27:49.098976+00:00"} {"global_step": 4573, "acc_step": 0, "speed/wps": 12911.17115312776, "speed/FLOPS": 202787684768244.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0918789878487587, "optim/lr": 0.0027432, "optim/total_tokens": 2397569024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.150963068008423, "created_at": "2025-01-15T18:27:59.254739+00:00"} {"global_step": 4574, "acc_step": 0, "speed/wps": 12911.251185416415, "speed/FLOPS": 202788941785313.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09359989315271378, "optim/lr": 0.0027438, "optim/total_tokens": 2398093312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328312, "loss/out": 3.230865001678467, "created_at": "2025-01-15T18:28:09.411317+00:00"} {"global_step": 4575, "acc_step": 0, "speed/wps": 12913.343269213616, "speed/FLOPS": 202821800836173.16, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09259922802448273, "optim/lr": 0.0027443999999999997, "optim/total_tokens": 2398617600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.0825438499450684, "created_at": "2025-01-15T18:28:19.565972+00:00"} {"global_step": 4576, "acc_step": 0, "speed/wps": 12911.299236299406, "speed/FLOPS": 202789696490459.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08962831646203995, "optim/lr": 0.002745, "optim/total_tokens": 2399141888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.1996536254882812, "created_at": "2025-01-15T18:28:29.723032+00:00"} {"global_step": 4577, "acc_step": 0, "speed/wps": 12913.686576562612, "speed/FLOPS": 202827192949842.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11001405864953995, "optim/lr": 0.0027456, "optim/total_tokens": 2399666176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.154501438140869, "created_at": "2025-01-15T18:28:39.877929+00:00"} {"global_step": 4578, "acc_step": 0, "speed/wps": 12909.113743160275, "speed/FLOPS": 202755370317527.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08718196302652359, "optim/lr": 0.0027462, "optim/total_tokens": 2400190464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.2576780319213867, "created_at": "2025-01-15T18:28:50.041293+00:00"} {"global_step": 4579, "acc_step": 0, "speed/wps": 12908.119493426311, "speed/FLOPS": 202739754259213.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09515857696533203, "optim/lr": 0.0027467999999999998, "optim/total_tokens": 2400714752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.155313014984131, "created_at": "2025-01-15T18:29:00.205656+00:00"} {"global_step": 4580, "acc_step": 0, "speed/wps": 12910.064919251945, "speed/FLOPS": 202770309845100.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1222456693649292, "optim/lr": 0.0027474, "optim/total_tokens": 2401239040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437124, "loss/out": 3.0728979110717773, "created_at": "2025-01-15T18:29:10.361949+00:00"} {"global_step": 4581, "acc_step": 0, "speed/wps": 12907.078573744873, "speed/FLOPS": 202723405185242.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11166530847549438, "optim/lr": 0.002748, "optim/total_tokens": 2401763328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.144422769546509, "created_at": "2025-01-15T18:29:20.521831+00:00"} {"global_step": 4582, "acc_step": 0, "speed/wps": 12901.157302104899, "speed/FLOPS": 202630403477457.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.09432162344455719, "optim/lr": 0.0027486, "optim/total_tokens": 2402287616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.1446104049682617, "created_at": "2025-01-15T18:29:30.693667+00:00"} {"global_step": 4583, "acc_step": 0, "speed/wps": 12909.386443808837, "speed/FLOPS": 202759653455943.22, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0794769823551178, "optim/lr": 0.0027492, "optim/total_tokens": 2402811904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472193, "loss/out": 3.3117265701293945, "created_at": "2025-01-15T18:29:40.850618+00:00"} {"global_step": 4584, "acc_step": 0, "speed/wps": 12908.512310222048, "speed/FLOPS": 202745923986777.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08236146718263626, "optim/lr": 0.0027498, "optim/total_tokens": 2403336192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.203943967819214, "created_at": "2025-01-15T18:29:51.009178+00:00"} {"global_step": 4585, "acc_step": 0, "speed/wps": 12912.24715326619, "speed/FLOPS": 202804584828997.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07591883838176727, "optim/lr": 0.0027504, "optim/total_tokens": 2403860480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 3.3333089351654053, "created_at": "2025-01-15T18:30:01.165169+00:00"} {"global_step": 4586, "acc_step": 0, "speed/wps": 12912.146388025221, "speed/FLOPS": 202803002172422.3, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09887972474098206, "optim/lr": 0.002751, "optim/total_tokens": 2404384768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.126667022705078, "created_at": "2025-01-15T18:30:11.322190+00:00"} {"global_step": 4587, "acc_step": 0, "speed/wps": 12909.757987677178, "speed/FLOPS": 202765489062950.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0713915079832077, "optim/lr": 0.0027516000000000003, "optim/total_tokens": 2404909056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.319681406021118, "created_at": "2025-01-15T18:30:21.478756+00:00"} {"global_step": 4588, "acc_step": 0, "speed/wps": 12910.863375398447, "speed/FLOPS": 202782850696074.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08735374361276627, "optim/lr": 0.0027522000000000002, "optim/total_tokens": 2405433344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1270105838775635, "created_at": "2025-01-15T18:30:31.637872+00:00"} {"global_step": 4589, "acc_step": 0, "speed/wps": 12908.176818269027, "speed/FLOPS": 202740654624640.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10045240819454193, "optim/lr": 0.0027528, "optim/total_tokens": 2405957632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.115518093109131, "created_at": "2025-01-15T18:30:41.797204+00:00"} {"global_step": 4590, "acc_step": 0, "speed/wps": 12909.227929086184, "speed/FLOPS": 202757163764401.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09627722948789597, "optim/lr": 0.0027534, "optim/total_tokens": 2406481920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.131791114807129, "created_at": "2025-01-15T18:30:51.956341+00:00"} {"global_step": 4591, "acc_step": 0, "speed/wps": 12908.404595428921, "speed/FLOPS": 202744232177935.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0989360511302948, "optim/lr": 0.0027540000000000004, "optim/total_tokens": 2407006208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1597695350646973, "created_at": "2025-01-15T18:31:02.115648+00:00"} {"global_step": 4592, "acc_step": 0, "speed/wps": 12912.87480855171, "speed/FLOPS": 202814443017745.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09178240597248077, "optim/lr": 0.0027546000000000003, "optim/total_tokens": 2407530496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.2515041828155518, "created_at": "2025-01-15T18:31:12.269853+00:00"} {"global_step": 4593, "acc_step": 0, "speed/wps": 12914.05978051071, "speed/FLOPS": 202833054630683.72, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09206746518611908, "optim/lr": 0.0027552, "optim/total_tokens": 2408054784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.1606006622314453, "created_at": "2025-01-15T18:31:22.425606+00:00"} {"global_step": 4594, "acc_step": 0, "speed/wps": 12910.689203668882, "speed/FLOPS": 202780115089724.88, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10661879181861877, "optim/lr": 0.0027558, "optim/total_tokens": 2408579072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 3.2222373485565186, "created_at": "2025-01-15T18:31:32.585984+00:00"} {"global_step": 4595, "acc_step": 0, "speed/wps": 12905.960514783976, "speed/FLOPS": 202705844532887.97, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09440333396196365, "optim/lr": 0.0027564, "optim/total_tokens": 2409103360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.2216224670410156, "created_at": "2025-01-15T18:31:42.753565+00:00"} {"global_step": 4596, "acc_step": 0, "speed/wps": 12909.693386137222, "speed/FLOPS": 202764474406993.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1073349341750145, "optim/lr": 0.0027570000000000003, "optim/total_tokens": 2409627648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385060, "loss/out": 3.186965227127075, "created_at": "2025-01-15T18:31:52.910155+00:00"} {"global_step": 4597, "acc_step": 0, "speed/wps": 12909.028462334534, "speed/FLOPS": 202754030864971.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09694099426269531, "optim/lr": 0.0027576000000000002, "optim/total_tokens": 2410151936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.140286922454834, "created_at": "2025-01-15T18:32:03.068982+00:00"} {"global_step": 4598, "acc_step": 0, "speed/wps": 12906.515607038533, "speed/FLOPS": 202714563019522.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08891171962022781, "optim/lr": 0.0027582, "optim/total_tokens": 2410676224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.1353559494018555, "created_at": "2025-01-15T18:32:13.228162+00:00"} {"global_step": 4599, "acc_step": 0, "speed/wps": 12906.710546819419, "speed/FLOPS": 202717624816656.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09446734189987183, "optim/lr": 0.0027588, "optim/total_tokens": 2411200512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 3.185999870300293, "created_at": "2025-01-15T18:32:23.387453+00:00"} {"global_step": 4600, "acc_step": 0, "speed/wps": 12909.329617395586, "speed/FLOPS": 202758760919035.44, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0926584005355835, "optim/lr": 0.0027594, "optim/total_tokens": 2411724800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.156555652618408, "created_at": "2025-01-15T18:32:33.544967+00:00"} {"global_step": 4601, "acc_step": 0, "speed/wps": 12906.529585474893, "speed/FLOPS": 202714782570073.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09166566282510757, "optim/lr": 0.0027600000000000003, "optim/total_tokens": 2412249088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454718, "loss/out": 3.1391096115112305, "created_at": "2025-01-15T18:32:43.705274+00:00"} {"global_step": 4602, "acc_step": 0, "speed/wps": 12904.46281243414, "speed/FLOPS": 202682321059425.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0855645090341568, "optim/lr": 0.0027606, "optim/total_tokens": 2412773376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.267031669616699, "created_at": "2025-01-15T18:32:53.869383+00:00"} {"global_step": 4603, "acc_step": 0, "speed/wps": 12907.714489331605, "speed/FLOPS": 202733393113372.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10510455816984177, "optim/lr": 0.0027612, "optim/total_tokens": 2413297664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459779, "loss/out": 3.1381070613861084, "created_at": "2025-01-15T18:33:04.028692+00:00"} {"global_step": 4604, "acc_step": 0, "speed/wps": 12907.906246403267, "speed/FLOPS": 202736404921684.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10208134353160858, "optim/lr": 0.0027618, "optim/total_tokens": 2413821952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.349600076675415, "created_at": "2025-01-15T18:33:14.189347+00:00"} {"global_step": 4605, "acc_step": 0, "speed/wps": 12911.68611538007, "speed/FLOPS": 202795772958049.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10665473341941833, "optim/lr": 0.0027624, "optim/total_tokens": 2414346240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.1079864501953125, "created_at": "2025-01-15T18:33:24.347404+00:00"} {"global_step": 4606, "acc_step": 0, "speed/wps": 12905.584926717624, "speed/FLOPS": 202699945406193.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08222076296806335, "optim/lr": 0.0027630000000000003, "optim/total_tokens": 2414870528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 3.1182892322540283, "created_at": "2025-01-15T18:33:34.512420+00:00"} {"global_step": 4607, "acc_step": 0, "speed/wps": 12911.35360800989, "speed/FLOPS": 202790550472887.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08080506324768066, "optim/lr": 0.0027636, "optim/total_tokens": 2415394816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.1885271072387695, "created_at": "2025-01-15T18:33:44.670857+00:00"} {"global_step": 4608, "acc_step": 0, "speed/wps": 12911.499421166665, "speed/FLOPS": 202792840668883.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0785837173461914, "optim/lr": 0.0027642, "optim/total_tokens": 2415919104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.2973480224609375, "created_at": "2025-01-15T18:33:54.826269+00:00"} {"global_step": 4609, "acc_step": 0, "speed/wps": 12914.10761056074, "speed/FLOPS": 202833805867344.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08852609246969223, "optim/lr": 0.0027648, "optim/total_tokens": 2416443392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.302136182785034, "created_at": "2025-01-15T18:34:04.981976+00:00"} {"global_step": 4610, "acc_step": 0, "speed/wps": 12905.73832461918, "speed/FLOPS": 202702354731028.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09266360104084015, "optim/lr": 0.0027654, "optim/total_tokens": 2416967680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466970, "loss/out": 3.204545021057129, "created_at": "2025-01-15T18:34:15.143903+00:00"} {"global_step": 4611, "acc_step": 0, "speed/wps": 12909.774202998504, "speed/FLOPS": 202765743746853.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0933917835354805, "optim/lr": 0.0027660000000000002, "optim/total_tokens": 2417491968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.1587023735046387, "created_at": "2025-01-15T18:34:25.300767+00:00"} {"global_step": 4612, "acc_step": 0, "speed/wps": 12906.028402291879, "speed/FLOPS": 202706910799486.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10235471278429031, "optim/lr": 0.0027666, "optim/total_tokens": 2418016256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1792945861816406, "created_at": "2025-01-15T18:34:35.460653+00:00"} {"global_step": 4613, "acc_step": 0, "speed/wps": 12906.369791619127, "speed/FLOPS": 202712272787989.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08822516351938248, "optim/lr": 0.0027672, "optim/total_tokens": 2418540544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 3.1535189151763916, "created_at": "2025-01-15T18:34:45.620648+00:00"} {"global_step": 4614, "acc_step": 0, "speed/wps": 12906.115373812216, "speed/FLOPS": 202708276806725.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11855790764093399, "optim/lr": 0.0027678, "optim/total_tokens": 2419064832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.0313892364501953, "created_at": "2025-01-15T18:34:55.780058+00:00"} {"global_step": 4615, "acc_step": 0, "speed/wps": 12904.42147515835, "speed/FLOPS": 202681671800705.44, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11802096664905548, "optim/lr": 0.0027684, "optim/total_tokens": 2419589120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 3.172651767730713, "created_at": "2025-01-15T18:35:05.942993+00:00"} {"global_step": 4616, "acc_step": 0, "speed/wps": 12912.413852534439, "speed/FLOPS": 202807203070074.34, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10785190761089325, "optim/lr": 0.002769, "optim/total_tokens": 2420113408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 3.2709884643554688, "created_at": "2025-01-15T18:35:16.100536+00:00"} {"global_step": 4617, "acc_step": 0, "speed/wps": 12912.988375026513, "speed/FLOPS": 202816226735289.0, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1044272929430008, "optim/lr": 0.0027696, "optim/total_tokens": 2420637696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 3.128167152404785, "created_at": "2025-01-15T18:35:26.258471+00:00"} {"global_step": 4618, "acc_step": 0, "speed/wps": 12912.313363089712, "speed/FLOPS": 202805624745257.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09089358150959015, "optim/lr": 0.0027702, "optim/total_tokens": 2421161984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.100166082382202, "created_at": "2025-01-15T18:35:36.413014+00:00"} {"global_step": 4619, "acc_step": 0, "speed/wps": 12907.735085618104, "speed/FLOPS": 202733716606353.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10216377675533295, "optim/lr": 0.0027708, "optim/total_tokens": 2421686272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.2473134994506836, "created_at": "2025-01-15T18:35:46.576061+00:00"} {"global_step": 4620, "acc_step": 0, "speed/wps": 12909.552252145773, "speed/FLOPS": 202762257703719.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10234013944864273, "optim/lr": 0.0027714, "optim/total_tokens": 2422210560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 3.3042593002319336, "created_at": "2025-01-15T18:35:56.734127+00:00"} {"global_step": 4621, "acc_step": 0, "speed/wps": 12911.297544852976, "speed/FLOPS": 202789669923969.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09220441430807114, "optim/lr": 0.002772, "optim/total_tokens": 2422734848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481042, "loss/out": 3.143775224685669, "created_at": "2025-01-15T18:36:06.892481+00:00"} {"global_step": 4622, "acc_step": 0, "speed/wps": 12908.441955249667, "speed/FLOPS": 202744818965258.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11171307414770126, "optim/lr": 0.0027726, "optim/total_tokens": 2423259136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 3.1797614097595215, "created_at": "2025-01-15T18:36:17.051463+00:00"} {"global_step": 4623, "acc_step": 0, "speed/wps": 12908.494112549517, "speed/FLOPS": 202745638167323.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.10618127882480621, "optim/lr": 0.0027732, "optim/total_tokens": 2423783424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.1560356616973877, "created_at": "2025-01-15T18:36:27.213914+00:00"} {"global_step": 4624, "acc_step": 0, "speed/wps": 12908.483426028715, "speed/FLOPS": 202745470320829.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0871395692229271, "optim/lr": 0.0027738, "optim/total_tokens": 2424307712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 3.1477909088134766, "created_at": "2025-01-15T18:36:37.371640+00:00"} {"global_step": 4625, "acc_step": 0, "speed/wps": 12908.10351312204, "speed/FLOPS": 202739503266575.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08994953334331512, "optim/lr": 0.0027743999999999998, "optim/total_tokens": 2424832000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.195927143096924, "created_at": "2025-01-15T18:36:47.529656+00:00"} {"global_step": 4626, "acc_step": 0, "speed/wps": 12906.11088111301, "speed/FLOPS": 202708206242710.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08793512731790543, "optim/lr": 0.002775, "optim/total_tokens": 2425356288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.289811611175537, "created_at": "2025-01-15T18:36:57.692135+00:00"} {"global_step": 4627, "acc_step": 0, "speed/wps": 12906.322237830105, "speed/FLOPS": 202711525890386.97, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0800294578075409, "optim/lr": 0.0027756, "optim/total_tokens": 2425880576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 3.0966200828552246, "created_at": "2025-01-15T18:37:07.852371+00:00"} {"global_step": 4628, "acc_step": 0, "speed/wps": 12909.215028259458, "speed/FLOPS": 202756961139190.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08101291209459305, "optim/lr": 0.0027762, "optim/total_tokens": 2426404864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.164275646209717, "created_at": "2025-01-15T18:37:18.013634+00:00"} {"global_step": 4629, "acc_step": 0, "speed/wps": 12913.567854141704, "speed/FLOPS": 202825328251080.8, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08040764927864075, "optim/lr": 0.0027768, "optim/total_tokens": 2426929152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.17744779586792, "created_at": "2025-01-15T18:37:28.168091+00:00"} {"global_step": 4630, "acc_step": 0, "speed/wps": 12907.063247057937, "speed/FLOPS": 202723164458563.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06878969073295593, "optim/lr": 0.0027773999999999997, "optim/total_tokens": 2427453440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.1412410736083984, "created_at": "2025-01-15T18:37:38.332236+00:00"} {"global_step": 4631, "acc_step": 0, "speed/wps": 12910.254445081355, "speed/FLOPS": 202773286608686.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07837177813053131, "optim/lr": 0.002778, "optim/total_tokens": 2427977728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 3.121115207672119, "created_at": "2025-01-15T18:37:48.488343+00:00"} {"global_step": 4632, "acc_step": 0, "speed/wps": 12908.324142609115, "speed/FLOPS": 202742968555846.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08355434983968735, "optim/lr": 0.0027786, "optim/total_tokens": 2428502016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.268556833267212, "created_at": "2025-01-15T18:37:58.650142+00:00"} {"global_step": 4633, "acc_step": 0, "speed/wps": 12910.084975910255, "speed/FLOPS": 202770624862479.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09316128492355347, "optim/lr": 0.0027792, "optim/total_tokens": 2429026304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 3.2581303119659424, "created_at": "2025-01-15T18:38:08.806557+00:00"} {"global_step": 4634, "acc_step": 0, "speed/wps": 12909.670279043481, "speed/FLOPS": 202764111478333.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08900980651378632, "optim/lr": 0.0027798, "optim/total_tokens": 2429550592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.300264596939087, "created_at": "2025-01-15T18:38:18.964350+00:00"} {"global_step": 4635, "acc_step": 0, "speed/wps": 12914.94484371417, "speed/FLOPS": 202846955764498.34, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08678456395864487, "optim/lr": 0.0027804, "optim/total_tokens": 2430074880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.2652759552001953, "created_at": "2025-01-15T18:38:29.116875+00:00"} {"global_step": 4636, "acc_step": 0, "speed/wps": 12908.880123769102, "speed/FLOPS": 202751701003962.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08733013272285461, "optim/lr": 0.002781, "optim/total_tokens": 2430599168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.13907527923584, "created_at": "2025-01-15T18:38:39.274247+00:00"} {"global_step": 4637, "acc_step": 0, "speed/wps": 12902.908179087552, "speed/FLOPS": 202657903406426.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09606803953647614, "optim/lr": 0.0027816, "optim/total_tokens": 2431123456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.2275071144104004, "created_at": "2025-01-15T18:38:49.436075+00:00"} {"global_step": 4638, "acc_step": 0, "speed/wps": 12913.323452435103, "speed/FLOPS": 202821489586435.5, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09787047654390335, "optim/lr": 0.0027822, "optim/total_tokens": 2431647744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.2025177478790283, "created_at": "2025-01-15T18:38:59.592267+00:00"} {"global_step": 4639, "acc_step": 0, "speed/wps": 12908.178056642788, "speed/FLOPS": 202740674075002.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0853530615568161, "optim/lr": 0.0027828, "optim/total_tokens": 2432172032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.193474292755127, "created_at": "2025-01-15T18:39:09.750729+00:00"} {"global_step": 4640, "acc_step": 0, "speed/wps": 12908.941250259395, "speed/FLOPS": 202752661079489.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09541469812393188, "optim/lr": 0.0027834, "optim/total_tokens": 2432696320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 3.2454559803009033, "created_at": "2025-01-15T18:39:19.908995+00:00"} {"global_step": 4641, "acc_step": 0, "speed/wps": 12909.339225393724, "speed/FLOPS": 202758911825848.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09489277750253677, "optim/lr": 0.002784, "optim/total_tokens": 2433220608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.205550193786621, "created_at": "2025-01-15T18:39:30.067009+00:00"} {"global_step": 4642, "acc_step": 0, "speed/wps": 12906.963437456892, "speed/FLOPS": 202721596811626.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08835135400295258, "optim/lr": 0.0027846, "optim/total_tokens": 2433744896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 3.169461488723755, "created_at": "2025-01-15T18:39:40.226062+00:00"} {"global_step": 4643, "acc_step": 0, "speed/wps": 12906.269103886258, "speed/FLOPS": 202710691348785.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08438289165496826, "optim/lr": 0.0027852000000000003, "optim/total_tokens": 2434269184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.2158281803131104, "created_at": "2025-01-15T18:39:50.388359+00:00"} {"global_step": 4644, "acc_step": 0, "speed/wps": 12908.387468352841, "speed/FLOPS": 202743963173670.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08726722747087479, "optim/lr": 0.0027858, "optim/total_tokens": 2434793472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 3.2025434970855713, "created_at": "2025-01-15T18:40:00.546307+00:00"} {"global_step": 4645, "acc_step": 0, "speed/wps": 12909.525538601745, "speed/FLOPS": 202761838130802.12, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08375727385282516, "optim/lr": 0.0027864, "optim/total_tokens": 2435317760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.2283334732055664, "created_at": "2025-01-15T18:40:10.707573+00:00"} {"global_step": 4646, "acc_step": 0, "speed/wps": 12910.26480174733, "speed/FLOPS": 202773449274357.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09418247640132904, "optim/lr": 0.0027870000000000004, "optim/total_tokens": 2435842048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.301679849624634, "created_at": "2025-01-15T18:40:20.874103+00:00"} {"global_step": 4647, "acc_step": 0, "speed/wps": 12906.906543597428, "speed/FLOPS": 202720703215383.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10886406898498535, "optim/lr": 0.0027876000000000003, "optim/total_tokens": 2436366336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348536, "loss/out": 3.025839328765869, "created_at": "2025-01-15T18:40:31.035951+00:00"} {"global_step": 4648, "acc_step": 0, "speed/wps": 12914.469380410836, "speed/FLOPS": 202839487959964.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09486399590969086, "optim/lr": 0.0027882000000000002, "optim/total_tokens": 2436890624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.2900350093841553, "created_at": "2025-01-15T18:40:41.188955+00:00"} {"global_step": 4649, "acc_step": 0, "speed/wps": 12905.666599856168, "speed/FLOPS": 202701228195065.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0787031352519989, "optim/lr": 0.0027888, "optim/total_tokens": 2437414912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 3.3129749298095703, "created_at": "2025-01-15T18:40:51.356802+00:00"} {"global_step": 4650, "acc_step": 0, "speed/wps": 12908.857769082933, "speed/FLOPS": 202751349892897.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12036126852035522, "optim/lr": 0.0027894, "optim/total_tokens": 2437939200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.054982900619507, "created_at": "2025-01-15T18:41:01.514211+00:00"} {"global_step": 4651, "acc_step": 0, "speed/wps": 12911.573396247632, "speed/FLOPS": 202794002549182.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10370847582817078, "optim/lr": 0.0027900000000000004, "optim/total_tokens": 2438463488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437710, "loss/out": 3.0716919898986816, "created_at": "2025-01-15T18:41:11.669541+00:00"} {"global_step": 4652, "acc_step": 0, "speed/wps": 12904.571292095325, "speed/FLOPS": 202684024881570.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10302641242742538, "optim/lr": 0.0027906000000000003, "optim/total_tokens": 2438987776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.2525229454040527, "created_at": "2025-01-15T18:41:21.832752+00:00"} {"global_step": 4653, "acc_step": 0, "speed/wps": 12906.179292147137, "speed/FLOPS": 202709280732008.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1052447184920311, "optim/lr": 0.0027912, "optim/total_tokens": 2439512064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.054378032684326, "created_at": "2025-01-15T18:41:31.993043+00:00"} {"global_step": 4654, "acc_step": 0, "speed/wps": 12908.239139782674, "speed/FLOPS": 202741633469651.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.09047084301710129, "optim/lr": 0.0027918, "optim/total_tokens": 2440036352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330191, "loss/out": 3.2346534729003906, "created_at": "2025-01-15T18:41:42.155293+00:00"} {"global_step": 4655, "acc_step": 0, "speed/wps": 12902.627163051615, "speed/FLOPS": 202653489663425.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09829916059970856, "optim/lr": 0.0027924, "optim/total_tokens": 2440560640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.1081037521362305, "created_at": "2025-01-15T18:41:52.318619+00:00"} {"global_step": 4656, "acc_step": 0, "speed/wps": 12910.912177867494, "speed/FLOPS": 202783617205912.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08707460016012192, "optim/lr": 0.0027930000000000003, "optim/total_tokens": 2441084928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294022, "loss/out": 3.1337084770202637, "created_at": "2025-01-15T18:42:02.475326+00:00"} {"global_step": 4657, "acc_step": 0, "speed/wps": 12905.49623218313, "speed/FLOPS": 202698552336650.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0904741957783699, "optim/lr": 0.0027936000000000002, "optim/total_tokens": 2441609216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.1892123222351074, "created_at": "2025-01-15T18:42:12.638904+00:00"} {"global_step": 4658, "acc_step": 0, "speed/wps": 12906.939370695367, "speed/FLOPS": 202721218810065.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08974862843751907, "optim/lr": 0.0027942, "optim/total_tokens": 2442133504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.1045260429382324, "created_at": "2025-01-15T18:42:22.797782+00:00"} {"global_step": 4659, "acc_step": 0, "speed/wps": 12904.730710100603, "speed/FLOPS": 202686528760406.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1054990291595459, "optim/lr": 0.0027948, "optim/total_tokens": 2442657792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306165, "loss/out": 3.1946892738342285, "created_at": "2025-01-15T18:42:32.961648+00:00"} {"global_step": 4660, "acc_step": 0, "speed/wps": 12909.248961218964, "speed/FLOPS": 202757494102948.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10275976359844208, "optim/lr": 0.0027954, "optim/total_tokens": 2443182080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.2058651447296143, "created_at": "2025-01-15T18:42:43.118652+00:00"} {"global_step": 4661, "acc_step": 0, "speed/wps": 12906.234270517072, "speed/FLOPS": 202710144242856.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0922199934720993, "optim/lr": 0.0027960000000000003, "optim/total_tokens": 2443706368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466705, "loss/out": 3.2637505531311035, "created_at": "2025-01-15T18:42:53.278088+00:00"} {"global_step": 4662, "acc_step": 0, "speed/wps": 12911.13983600334, "speed/FLOPS": 202787192889771.25, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10211437940597534, "optim/lr": 0.0027966, "optim/total_tokens": 2444230656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482092, "loss/out": 3.2096643447875977, "created_at": "2025-01-15T18:43:03.439874+00:00"} {"global_step": 4663, "acc_step": 0, "speed/wps": 12909.442331694725, "speed/FLOPS": 202760531251989.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08873384445905685, "optim/lr": 0.0027972, "optim/total_tokens": 2444754944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 3.1050381660461426, "created_at": "2025-01-15T18:43:13.598277+00:00"} {"global_step": 4664, "acc_step": 0, "speed/wps": 12909.177957619819, "speed/FLOPS": 202756378893854.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09748756885528564, "optim/lr": 0.0027978, "optim/total_tokens": 2445279232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.167733907699585, "created_at": "2025-01-15T18:43:23.756543+00:00"} {"global_step": 4665, "acc_step": 0, "speed/wps": 12908.209477792676, "speed/FLOPS": 202741167587338.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08750297129154205, "optim/lr": 0.0027984, "optim/total_tokens": 2445803520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.09442400932312, "created_at": "2025-01-15T18:43:33.916735+00:00"} {"global_step": 4666, "acc_step": 0, "speed/wps": 12905.031842168464, "speed/FLOPS": 202691258453331.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07636310905218124, "optim/lr": 0.0027990000000000003, "optim/total_tokens": 2446327808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337297, "loss/out": 3.2472643852233887, "created_at": "2025-01-15T18:43:44.083543+00:00"} {"global_step": 4667, "acc_step": 0, "speed/wps": 12906.234865846427, "speed/FLOPS": 202710153593321.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07177422940731049, "optim/lr": 0.0027996, "optim/total_tokens": 2446852096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.073737859725952, "created_at": "2025-01-15T18:43:54.247407+00:00"} {"global_step": 4668, "acc_step": 0, "speed/wps": 12908.996724604473, "speed/FLOPS": 202753532380308.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08353052288293839, "optim/lr": 0.0028002, "optim/total_tokens": 2447376384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.1454200744628906, "created_at": "2025-01-15T18:44:04.409099+00:00"} {"global_step": 4669, "acc_step": 0, "speed/wps": 12913.069118220874, "speed/FLOPS": 202817494918109.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07665611058473587, "optim/lr": 0.0028008, "optim/total_tokens": 2447900672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.1836023330688477, "created_at": "2025-01-15T18:44:14.563024+00:00"} {"global_step": 4670, "acc_step": 0, "speed/wps": 12906.151484764978, "speed/FLOPS": 202708843978859.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08977165818214417, "optim/lr": 0.0028014, "optim/total_tokens": 2448424960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2231698036193848, "created_at": "2025-01-15T18:44:24.724641+00:00"} {"global_step": 4671, "acc_step": 0, "speed/wps": 12909.855354822113, "speed/FLOPS": 202767018347758.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11759345233440399, "optim/lr": 0.0028020000000000002, "optim/total_tokens": 2448949248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.115262746810913, "created_at": "2025-01-15T18:44:34.881203+00:00"} {"global_step": 4672, "acc_step": 0, "speed/wps": 12904.680722564754, "speed/FLOPS": 202685743637466.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1075398251414299, "optim/lr": 0.0028026, "optim/total_tokens": 2449473536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.251326560974121, "created_at": "2025-01-15T18:44:45.044537+00:00"} {"global_step": 4673, "acc_step": 0, "speed/wps": 12907.70310564381, "speed/FLOPS": 202733214316912.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10030384361743927, "optim/lr": 0.0028032, "optim/total_tokens": 2449997824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398779, "loss/out": 3.172023296356201, "created_at": "2025-01-15T18:44:55.202703+00:00"} {"global_step": 4674, "acc_step": 0, "speed/wps": 12909.81957011479, "speed/FLOPS": 202766456299756.16, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11782771348953247, "optim/lr": 0.0028038, "optim/total_tokens": 2450522112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.1629750728607178, "created_at": "2025-01-15T18:45:05.360134+00:00"} {"global_step": 4675, "acc_step": 0, "speed/wps": 12905.134676137752, "speed/FLOPS": 202692873602127.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08322983235120773, "optim/lr": 0.0028044, "optim/total_tokens": 2451046400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.144320249557495, "created_at": "2025-01-15T18:45:15.520486+00:00"} {"global_step": 4676, "acc_step": 0, "speed/wps": 12908.795316088752, "speed/FLOPS": 202750368982804.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11849653720855713, "optim/lr": 0.002805, "optim/total_tokens": 2451570688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.1813597679138184, "created_at": "2025-01-15T18:45:25.682366+00:00"} {"global_step": 4677, "acc_step": 0, "speed/wps": 12906.257187777948, "speed/FLOPS": 202710504189929.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1459103375673294, "optim/lr": 0.0028056, "optim/total_tokens": 2452094976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.2140133380889893, "created_at": "2025-01-15T18:45:35.841658+00:00"} {"global_step": 4678, "acc_step": 0, "speed/wps": 12908.556548421471, "speed/FLOPS": 202746618808488.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11788245290517807, "optim/lr": 0.0028062, "optim/total_tokens": 2452619264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.166715621948242, "created_at": "2025-01-15T18:45:46.028575+00:00"} {"global_step": 4679, "acc_step": 0, "speed/wps": 12906.203163155287, "speed/FLOPS": 202709655658993.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11333081871271133, "optim/lr": 0.0028068, "optim/total_tokens": 2453143552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.1577205657958984, "created_at": "2025-01-15T18:45:56.192927+00:00"} {"global_step": 4680, "acc_step": 0, "speed/wps": 12908.34998453322, "speed/FLOPS": 202743374438775.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08551207184791565, "optim/lr": 0.0028074, "optim/total_tokens": 2453667840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.201511859893799, "created_at": "2025-01-15T18:46:06.350531+00:00"} {"global_step": 4681, "acc_step": 0, "speed/wps": 12914.244284188431, "speed/FLOPS": 202835952514476.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09885554760694504, "optim/lr": 0.002808, "optim/total_tokens": 2454192128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 3.1695563793182373, "created_at": "2025-01-15T18:46:16.510294+00:00"} {"global_step": 4682, "acc_step": 0, "speed/wps": 12909.937498871948, "speed/FLOPS": 202768308532938.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08416024595499039, "optim/lr": 0.0028086, "optim/total_tokens": 2454716416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.1812222003936768, "created_at": "2025-01-15T18:46:26.666707+00:00"} {"global_step": 4683, "acc_step": 0, "speed/wps": 12910.485019545898, "speed/FLOPS": 202776908097495.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08230460435152054, "optim/lr": 0.0028092, "optim/total_tokens": 2455240704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.308868885040283, "created_at": "2025-01-15T18:46:36.823595+00:00"} {"global_step": 4684, "acc_step": 0, "speed/wps": 12911.01344855022, "speed/FLOPS": 202785207801145.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08052174746990204, "optim/lr": 0.0028098, "optim/total_tokens": 2455764992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.1184864044189453, "created_at": "2025-01-15T18:46:46.979286+00:00"} {"global_step": 4685, "acc_step": 0, "speed/wps": 12903.527544453, "speed/FLOPS": 202667631390588.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08144450932741165, "optim/lr": 0.0028104, "optim/total_tokens": 2456289280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 3.1540932655334473, "created_at": "2025-01-15T18:46:57.143989+00:00"} {"global_step": 4686, "acc_step": 0, "speed/wps": 12911.91958876501, "speed/FLOPS": 202799439978385.84, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08570687472820282, "optim/lr": 0.002811, "optim/total_tokens": 2456813568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.2065086364746094, "created_at": "2025-01-15T18:47:07.311940+00:00"} {"global_step": 4687, "acc_step": 0, "speed/wps": 12912.637981001042, "speed/FLOPS": 202810723315627.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07873473316431046, "optim/lr": 0.0028116, "optim/total_tokens": 2457337856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2104990482330322, "created_at": "2025-01-15T18:47:17.468032+00:00"} {"global_step": 4688, "acc_step": 0, "speed/wps": 12912.058838844527, "speed/FLOPS": 202801627092234.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11152494698762894, "optim/lr": 0.0028122, "optim/total_tokens": 2457862144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 3.075770616531372, "created_at": "2025-01-15T18:47:27.625792+00:00"} {"global_step": 4689, "acc_step": 0, "speed/wps": 12910.845082266293, "speed/FLOPS": 202782563377296.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12110642343759537, "optim/lr": 0.0028128, "optim/total_tokens": 2458386432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.1765658855438232, "created_at": "2025-01-15T18:47:37.785771+00:00"} {"global_step": 4690, "acc_step": 0, "speed/wps": 12908.357831727979, "speed/FLOPS": 202743497689752.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08266453444957733, "optim/lr": 0.0028133999999999998, "optim/total_tokens": 2458910720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 3.137894630432129, "created_at": "2025-01-15T18:47:47.944740+00:00"} {"global_step": 4691, "acc_step": 0, "speed/wps": 12914.768201081788, "speed/FLOPS": 202844181349224.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11187152564525604, "optim/lr": 0.002814, "optim/total_tokens": 2459435008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.2153358459472656, "created_at": "2025-01-15T18:47:58.097266+00:00"} {"global_step": 4692, "acc_step": 0, "speed/wps": 12908.507981610319, "speed/FLOPS": 202745855999981.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10499288141727448, "optim/lr": 0.0028146, "optim/total_tokens": 2459959296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367755, "loss/out": 3.1105165481567383, "created_at": "2025-01-15T18:48:08.256072+00:00"} {"global_step": 4693, "acc_step": 0, "speed/wps": 12905.101041707976, "speed/FLOPS": 202692345327189.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08964896947145462, "optim/lr": 0.0028152, "optim/total_tokens": 2460483584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.1330690383911133, "created_at": "2025-01-15T18:48:18.418701+00:00"} {"global_step": 4694, "acc_step": 0, "speed/wps": 12911.62378071506, "speed/FLOPS": 202794793906478.53, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.09933613240718842, "optim/lr": 0.0028158, "optim/total_tokens": 2461007872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.2687153816223145, "created_at": "2025-01-15T18:48:28.578656+00:00"} {"global_step": 4695, "acc_step": 0, "speed/wps": 12908.828689162134, "speed/FLOPS": 202750893152781.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1029297485947609, "optim/lr": 0.0028164, "optim/total_tokens": 2461532160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0851047039031982, "created_at": "2025-01-15T18:48:38.735865+00:00"} {"global_step": 4696, "acc_step": 0, "speed/wps": 12913.217696441106, "speed/FLOPS": 202819828543225.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09895233809947968, "optim/lr": 0.002817, "optim/total_tokens": 2462056448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.132129192352295, "created_at": "2025-01-15T18:48:48.889597+00:00"} {"global_step": 4697, "acc_step": 0, "speed/wps": 12911.365130801407, "speed/FLOPS": 202790731454162.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10645037889480591, "optim/lr": 0.0028176, "optim/total_tokens": 2462580736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 3.079432964324951, "created_at": "2025-01-15T18:48:59.046930+00:00"} {"global_step": 4698, "acc_step": 0, "speed/wps": 12910.213280311085, "speed/FLOPS": 202772640059403.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09705962240695953, "optim/lr": 0.0028182000000000003, "optim/total_tokens": 2463105024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.1962428092956543, "created_at": "2025-01-15T18:49:09.203194+00:00"} {"global_step": 4699, "acc_step": 0, "speed/wps": 12906.334789638679, "speed/FLOPS": 202711723033788.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.09287415444850922, "optim/lr": 0.0028188, "optim/total_tokens": 2463629312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 3.110779285430908, "created_at": "2025-01-15T18:49:19.363085+00:00"} {"global_step": 4700, "acc_step": 0, "speed/wps": 12913.702970014174, "speed/FLOPS": 202827450431526.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08499833196401596, "optim/lr": 0.0028194, "optim/total_tokens": 2464153600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.2529923915863037, "created_at": "2025-01-15T18:49:29.517583+00:00"} {"global_step": 4701, "acc_step": 0, "speed/wps": 12912.086850567588, "speed/FLOPS": 202802067054837.16, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10174889862537384, "optim/lr": 0.00282, "optim/total_tokens": 2464677888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382664, "loss/out": 3.1028144359588623, "created_at": "2025-01-15T18:49:39.673514+00:00"} {"global_step": 4702, "acc_step": 0, "speed/wps": 12907.870386078952, "speed/FLOPS": 202735841686013.75, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09951288998126984, "optim/lr": 0.0028206000000000004, "optim/total_tokens": 2465202176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.226414203643799, "created_at": "2025-01-15T18:49:49.839374+00:00"} {"global_step": 4703, "acc_step": 0, "speed/wps": 12911.215382310265, "speed/FLOPS": 202788379448333.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10790198296308517, "optim/lr": 0.0028212000000000003, "optim/total_tokens": 2465726464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435028, "loss/out": 3.069504737854004, "created_at": "2025-01-15T18:49:59.998442+00:00"} {"global_step": 4704, "acc_step": 0, "speed/wps": 12914.325271802722, "speed/FLOPS": 202837224536248.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0866566151380539, "optim/lr": 0.0028218, "optim/total_tokens": 2466250752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.1796562671661377, "created_at": "2025-01-15T18:50:10.157975+00:00"} {"global_step": 4705, "acc_step": 0, "speed/wps": 12911.82606525068, "speed/FLOPS": 202797971063078.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09165884554386139, "optim/lr": 0.0028224, "optim/total_tokens": 2466775040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.1611814498901367, "created_at": "2025-01-15T18:50:20.313257+00:00"} {"global_step": 4706, "acc_step": 0, "speed/wps": 12907.546349948347, "speed/FLOPS": 202730752253310.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0979924127459526, "optim/lr": 0.002823, "optim/total_tokens": 2467299328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.2086782455444336, "created_at": "2025-01-15T18:50:30.471399+00:00"} {"global_step": 4707, "acc_step": 0, "speed/wps": 12908.7756636373, "speed/FLOPS": 202750060314049.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1012219712138176, "optim/lr": 0.0028236000000000003, "optim/total_tokens": 2467823616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.10111665725708, "created_at": "2025-01-15T18:50:40.632748+00:00"} {"global_step": 4708, "acc_step": 0, "speed/wps": 12909.637209505641, "speed/FLOPS": 202763592075799.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11505912244319916, "optim/lr": 0.0028242000000000002, "optim/total_tokens": 2468347904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367755, "loss/out": 3.2261581420898438, "created_at": "2025-01-15T18:50:50.794432+00:00"} {"global_step": 4709, "acc_step": 0, "speed/wps": 12904.259143483852, "speed/FLOPS": 202679122158690.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09253934770822525, "optim/lr": 0.0028248, "optim/total_tokens": 2468872192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.1338095664978027, "created_at": "2025-01-15T18:51:00.958784+00:00"} {"global_step": 4710, "acc_step": 0, "speed/wps": 12912.217538009254, "speed/FLOPS": 202804119680691.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08374027162790298, "optim/lr": 0.0028254, "optim/total_tokens": 2469396480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414480, "loss/out": 3.1537647247314453, "created_at": "2025-01-15T18:51:11.113842+00:00"} {"global_step": 4711, "acc_step": 0, "speed/wps": 12910.62277362804, "speed/FLOPS": 202779071714649.0, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08930481225252151, "optim/lr": 0.002826, "optim/total_tokens": 2469920768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.1374130249023438, "created_at": "2025-01-15T18:51:21.270698+00:00"} {"global_step": 4712, "acc_step": 0, "speed/wps": 12904.064223320831, "speed/FLOPS": 202676060669681.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10259795188903809, "optim/lr": 0.0028266000000000003, "optim/total_tokens": 2470445056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.1747968196868896, "created_at": "2025-01-15T18:51:31.431817+00:00"} {"global_step": 4713, "acc_step": 0, "speed/wps": 12906.065173165594, "speed/FLOPS": 202707488336586.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08968621492385864, "optim/lr": 0.0028272, "optim/total_tokens": 2470969344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 3.185197591781616, "created_at": "2025-01-15T18:51:41.592139+00:00"} {"global_step": 4714, "acc_step": 0, "speed/wps": 12906.22758342174, "speed/FLOPS": 202710039212834.8, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10272184759378433, "optim/lr": 0.0028278, "optim/total_tokens": 2471493632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.217437267303467, "created_at": "2025-01-15T18:51:51.756748+00:00"} {"global_step": 4715, "acc_step": 0, "speed/wps": 12909.900811326004, "speed/FLOPS": 202767732304615.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.11163073778152466, "optim/lr": 0.0028284, "optim/total_tokens": 2472017920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 3.1138956546783447, "created_at": "2025-01-15T18:52:01.915485+00:00"} {"global_step": 4716, "acc_step": 0, "speed/wps": 12908.687472251755, "speed/FLOPS": 202748675147151.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08916657418012619, "optim/lr": 0.002829, "optim/total_tokens": 2472542208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.159672260284424, "created_at": "2025-01-15T18:52:12.074319+00:00"} {"global_step": 4717, "acc_step": 0, "speed/wps": 12904.634617703976, "speed/FLOPS": 202685019497273.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08373721688985825, "optim/lr": 0.0028296000000000003, "optim/total_tokens": 2473066496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406246, "loss/out": 3.15730619430542, "created_at": "2025-01-15T18:52:22.234858+00:00"} {"global_step": 4718, "acc_step": 0, "speed/wps": 12910.929623145985, "speed/FLOPS": 202783891207983.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08250737190246582, "optim/lr": 0.0028302, "optim/total_tokens": 2473590784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.2401866912841797, "created_at": "2025-01-15T18:52:32.391165+00:00"} {"global_step": 4719, "acc_step": 0, "speed/wps": 12912.80023920893, "speed/FLOPS": 202813271803748.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10077906399965286, "optim/lr": 0.0028308, "optim/total_tokens": 2474115072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.0683236122131348, "created_at": "2025-01-15T18:52:42.549707+00:00"} {"global_step": 4720, "acc_step": 0, "speed/wps": 12912.659329908718, "speed/FLOPS": 202811058629558.66, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09025973081588745, "optim/lr": 0.0028314, "optim/total_tokens": 2474639360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481564, "loss/out": 3.1374430656433105, "created_at": "2025-01-15T18:52:52.707507+00:00"} {"global_step": 4721, "acc_step": 0, "speed/wps": 12906.765368738543, "speed/FLOPS": 202718485870229.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0706796795129776, "optim/lr": 0.002832, "optim/total_tokens": 2475163648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.125441074371338, "created_at": "2025-01-15T18:53:02.875268+00:00"} {"global_step": 4722, "acc_step": 0, "speed/wps": 12910.950812430125, "speed/FLOPS": 202784224014808.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09660030901432037, "optim/lr": 0.0028326000000000002, "optim/total_tokens": 2475687936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.1790552139282227, "created_at": "2025-01-15T18:53:13.031863+00:00"} {"global_step": 4723, "acc_step": 0, "speed/wps": 12910.568652049273, "speed/FLOPS": 202778221660883.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09632380306720734, "optim/lr": 0.0028332, "optim/total_tokens": 2476212224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.125122308731079, "created_at": "2025-01-15T18:53:23.190833+00:00"} {"global_step": 4724, "acc_step": 0, "speed/wps": 12911.903839523558, "speed/FLOPS": 202799192614907.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08368729799985886, "optim/lr": 0.0028338, "optim/total_tokens": 2476736512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340756, "loss/out": 3.065819263458252, "created_at": "2025-01-15T18:53:33.345533+00:00"} {"global_step": 4725, "acc_step": 0, "speed/wps": 12908.983158797235, "speed/FLOPS": 202753319310664.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09001249819993973, "optim/lr": 0.0028344, "optim/total_tokens": 2477260800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.15822434425354, "created_at": "2025-01-15T18:53:43.502917+00:00"} {"global_step": 4726, "acc_step": 0, "speed/wps": 12911.865239533143, "speed/FLOPS": 202798586349015.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08326692879199982, "optim/lr": 0.002835, "optim/total_tokens": 2477785088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.0152692794799805, "created_at": "2025-01-15T18:53:53.661732+00:00"} {"global_step": 4727, "acc_step": 0, "speed/wps": 12910.71179683683, "speed/FLOPS": 202780469946473.75, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09025850147008896, "optim/lr": 0.0028356, "optim/total_tokens": 2478309376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.1560792922973633, "created_at": "2025-01-15T18:54:03.820039+00:00"} {"global_step": 4728, "acc_step": 0, "speed/wps": 12907.3074485508, "speed/FLOPS": 202726999978576.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08763464540243149, "optim/lr": 0.0028362, "optim/total_tokens": 2478833664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.1588058471679688, "created_at": "2025-01-15T18:54:13.981060+00:00"} {"global_step": 4729, "acc_step": 0, "speed/wps": 12913.38346339607, "speed/FLOPS": 202822432141041.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10842500627040863, "optim/lr": 0.0028368, "optim/total_tokens": 2479357952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.1205861568450928, "created_at": "2025-01-15T18:54:24.135099+00:00"} {"global_step": 4730, "acc_step": 0, "speed/wps": 12916.351034407973, "speed/FLOPS": 202869041921653.03, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11553262919187546, "optim/lr": 0.0028374, "optim/total_tokens": 2479882240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470861, "loss/out": 3.25569486618042, "created_at": "2025-01-15T18:54:34.288484+00:00"} {"global_step": 4731, "acc_step": 0, "speed/wps": 12912.168347769302, "speed/FLOPS": 202803347080378.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09210207313299179, "optim/lr": 0.002838, "optim/total_tokens": 2480406528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.1392178535461426, "created_at": "2025-01-15T18:54:44.444793+00:00"} {"global_step": 4732, "acc_step": 0, "speed/wps": 12911.893275292716, "speed/FLOPS": 202799026689145.0, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11117661744356155, "optim/lr": 0.0028386, "optim/total_tokens": 2480930816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.2339839935302734, "created_at": "2025-01-15T18:54:54.604154+00:00"} {"global_step": 4733, "acc_step": 0, "speed/wps": 12918.313627024156, "speed/FLOPS": 202899867135575.06, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10970350354909897, "optim/lr": 0.0028392, "optim/total_tokens": 2481455104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 3.2311058044433594, "created_at": "2025-01-15T18:55:04.758805+00:00"} {"global_step": 4734, "acc_step": 0, "speed/wps": 12909.432612233573, "speed/FLOPS": 202760378594495.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08913835883140564, "optim/lr": 0.0028398, "optim/total_tokens": 2481979392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478894, "loss/out": 3.1862354278564453, "created_at": "2025-01-15T18:55:14.918597+00:00"} {"global_step": 4735, "acc_step": 0, "speed/wps": 12910.094277123799, "speed/FLOPS": 202770770950819.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08338358998298645, "optim/lr": 0.0028404, "optim/total_tokens": 2482503680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 3.303393602371216, "created_at": "2025-01-15T18:55:25.075253+00:00"} {"global_step": 4736, "acc_step": 0, "speed/wps": 12913.355969756742, "speed/FLOPS": 202822000315655.34, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09929363429546356, "optim/lr": 0.0028409999999999998, "optim/total_tokens": 2483027968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.1223506927490234, "created_at": "2025-01-15T18:55:35.231374+00:00"} {"global_step": 4737, "acc_step": 0, "speed/wps": 12915.586659798202, "speed/FLOPS": 202857036368045.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07543063163757324, "optim/lr": 0.0028416, "optim/total_tokens": 2483552256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 3.1133053302764893, "created_at": "2025-01-15T18:55:45.383974+00:00"} {"global_step": 4738, "acc_step": 0, "speed/wps": 12911.042347620416, "speed/FLOPS": 202785661700754.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07811658084392548, "optim/lr": 0.0028422, "optim/total_tokens": 2484076544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.1823015213012695, "created_at": "2025-01-15T18:55:55.542281+00:00"} {"global_step": 4739, "acc_step": 0, "speed/wps": 12905.898528897287, "speed/FLOPS": 202704870959360.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08726603537797928, "optim/lr": 0.0028428, "optim/total_tokens": 2484600832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.2831318378448486, "created_at": "2025-01-15T18:56:05.703332+00:00"} {"global_step": 4740, "acc_step": 0, "speed/wps": 12911.00289998951, "speed/FLOPS": 202785042121504.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07792283594608307, "optim/lr": 0.0028434, "optim/total_tokens": 2485125120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 3.215245485305786, "created_at": "2025-01-15T18:56:15.869770+00:00"} {"global_step": 4741, "acc_step": 0, "speed/wps": 12916.892122487901, "speed/FLOPS": 202877540453481.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08545316755771637, "optim/lr": 0.0028439999999999997, "optim/total_tokens": 2485649408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.1140952110290527, "created_at": "2025-01-15T18:56:26.023153+00:00"} {"global_step": 4742, "acc_step": 0, "speed/wps": 12912.521928057378, "speed/FLOPS": 202808900544673.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07691381126642227, "optim/lr": 0.0028446, "optim/total_tokens": 2486173696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.204451322555542, "created_at": "2025-01-15T18:56:36.183126+00:00"} {"global_step": 4743, "acc_step": 0, "speed/wps": 12912.564672564571, "speed/FLOPS": 202809571905896.0, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07703078538179398, "optim/lr": 0.0028452, "optim/total_tokens": 2486697984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.1521716117858887, "created_at": "2025-01-15T18:56:46.369480+00:00"} {"global_step": 4744, "acc_step": 0, "speed/wps": 12912.64443187053, "speed/FLOPS": 202810824635396.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08685663342475891, "optim/lr": 0.0028458, "optim/total_tokens": 2487222272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.080200433731079, "created_at": "2025-01-15T18:56:56.527404+00:00"} {"global_step": 4745, "acc_step": 0, "speed/wps": 12908.970039408783, "speed/FLOPS": 202753113252640.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09570380300283432, "optim/lr": 0.0028464, "optim/total_tokens": 2487746560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 3.203246593475342, "created_at": "2025-01-15T18:57:06.688985+00:00"} {"global_step": 4746, "acc_step": 0, "speed/wps": 12912.923513673693, "speed/FLOPS": 202815207998614.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.10240770876407623, "optim/lr": 0.002847, "optim/total_tokens": 2488270848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.080148220062256, "created_at": "2025-01-15T18:57:16.845752+00:00"} {"global_step": 4747, "acc_step": 0, "speed/wps": 12909.549553432174, "speed/FLOPS": 202762215316713.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11112630367279053, "optim/lr": 0.0028476, "optim/total_tokens": 2488795136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358724, "loss/out": 3.210876941680908, "created_at": "2025-01-15T18:57:27.006164+00:00"} {"global_step": 4748, "acc_step": 0, "speed/wps": 12910.359666975375, "speed/FLOPS": 202774939263122.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08497793227434158, "optim/lr": 0.0028482, "optim/total_tokens": 2489319424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.147592782974243, "created_at": "2025-01-15T18:57:37.169221+00:00"} {"global_step": 4749, "acc_step": 0, "speed/wps": 12909.8206548558, "speed/FLOPS": 202766473337104.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08794019371271133, "optim/lr": 0.0028488, "optim/total_tokens": 2489843712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.2363085746765137, "created_at": "2025-01-15T18:57:47.330427+00:00"} {"global_step": 4750, "acc_step": 0, "speed/wps": 12914.219235139022, "speed/FLOPS": 202835559084735.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0976836234331131, "optim/lr": 0.0028494, "optim/total_tokens": 2490368000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 3.249885320663452, "created_at": "2025-01-15T18:57:57.485576+00:00"} {"global_step": 4751, "acc_step": 0, "speed/wps": 12915.331100980162, "speed/FLOPS": 202853022465633.88, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0905815064907074, "optim/lr": 0.00285, "optim/total_tokens": 2490892288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406724, "loss/out": 3.3339102268218994, "created_at": "2025-01-15T18:58:07.640064+00:00"} {"global_step": 4752, "acc_step": 0, "speed/wps": 12911.214990250355, "speed/FLOPS": 202788373290493.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08116637915372849, "optim/lr": 0.0028506, "optim/total_tokens": 2491416576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2328662872314453, "created_at": "2025-01-15T18:58:17.795418+00:00"} {"global_step": 4753, "acc_step": 0, "speed/wps": 12911.50234220149, "speed/FLOPS": 202792886547748.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09079926460981369, "optim/lr": 0.0028512, "optim/total_tokens": 2491940864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.194295883178711, "created_at": "2025-01-15T18:58:27.950589+00:00"} {"global_step": 4754, "acc_step": 0, "speed/wps": 12907.948354214812, "speed/FLOPS": 202737066282726.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08751460164785385, "optim/lr": 0.0028518000000000003, "optim/total_tokens": 2492465152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 3.196756601333618, "created_at": "2025-01-15T18:58:38.113267+00:00"} {"global_step": 4755, "acc_step": 0, "speed/wps": 12909.986471978527, "speed/FLOPS": 202769077722873.3, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07935088872909546, "optim/lr": 0.0028524, "optim/total_tokens": 2492989440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.2757246494293213, "created_at": "2025-01-15T18:58:48.272791+00:00"} {"global_step": 4756, "acc_step": 0, "speed/wps": 12899.140962658703, "speed/FLOPS": 202598734095713.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08503485471010208, "optim/lr": 0.002853, "optim/total_tokens": 2493513728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.237536907196045, "created_at": "2025-01-15T18:58:58.446212+00:00"} {"global_step": 4757, "acc_step": 0, "speed/wps": 12897.31355086328, "speed/FLOPS": 202570032082339.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10902132838964462, "optim/lr": 0.0028536000000000004, "optim/total_tokens": 2494038016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.208747148513794, "created_at": "2025-01-15T18:59:08.616536+00:00"} {"global_step": 4758, "acc_step": 0, "speed/wps": 12910.085931611105, "speed/FLOPS": 202770639873074.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09439469873905182, "optim/lr": 0.0028542000000000003, "optim/total_tokens": 2494562304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.167571544647217, "created_at": "2025-01-15T18:59:18.774213+00:00"} {"global_step": 4759, "acc_step": 0, "speed/wps": 12909.646459154123, "speed/FLOPS": 202763737354239.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10379675775766373, "optim/lr": 0.0028548000000000002, "optim/total_tokens": 2495086592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0512375831604004, "created_at": "2025-01-15T18:59:28.934015+00:00"} {"global_step": 4760, "acc_step": 0, "speed/wps": 12907.881466993786, "speed/FLOPS": 202736015727007.97, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09106001257896423, "optim/lr": 0.0028554, "optim/total_tokens": 2495610880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.028745651245117, "created_at": "2025-01-15T18:59:39.093578+00:00"} {"global_step": 4761, "acc_step": 0, "speed/wps": 12907.566726694942, "speed/FLOPS": 202731072298116.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10502076148986816, "optim/lr": 0.002856, "optim/total_tokens": 2496135168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 3.2149453163146973, "created_at": "2025-01-15T18:59:49.256884+00:00"} {"global_step": 4762, "acc_step": 0, "speed/wps": 12903.164035563517, "speed/FLOPS": 202661921983965.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09748003631830215, "optim/lr": 0.0028566000000000004, "optim/total_tokens": 2496659456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 2.9752511978149414, "created_at": "2025-01-15T18:59:59.423591+00:00"} {"global_step": 4763, "acc_step": 0, "speed/wps": 12896.213810917441, "speed/FLOPS": 202552759155289.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07989779114723206, "optim/lr": 0.0028572000000000003, "optim/total_tokens": 2497183744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.216352939605713, "created_at": "2025-01-15T19:00:09.596761+00:00"} {"global_step": 4764, "acc_step": 0, "speed/wps": 12902.33750793863, "speed/FLOPS": 202648940231848.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08057428896427155, "optim/lr": 0.0028578, "optim/total_tokens": 2497708032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.2792253494262695, "created_at": "2025-01-15T19:00:19.764106+00:00"} {"global_step": 4765, "acc_step": 0, "speed/wps": 12904.85794133812, "speed/FLOPS": 202688527101826.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1846129149198532, "optim/lr": 0.0028584, "optim/total_tokens": 2498232320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.1518802642822266, "created_at": "2025-01-15T19:00:29.926944+00:00"} {"global_step": 4766, "acc_step": 0, "speed/wps": 12906.251625041235, "speed/FLOPS": 202710416819505.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09728638827800751, "optim/lr": 0.002859, "optim/total_tokens": 2498756608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.11767315864563, "created_at": "2025-01-15T19:00:40.088908+00:00"} {"global_step": 4767, "acc_step": 0, "speed/wps": 12906.583958447816, "speed/FLOPS": 202715636572329.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09495460242033005, "optim/lr": 0.0028596000000000003, "optim/total_tokens": 2499280896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333550, "loss/out": 3.2484841346740723, "created_at": "2025-01-15T19:00:50.248582+00:00"} {"global_step": 4768, "acc_step": 0, "speed/wps": 12914.022332470186, "speed/FLOPS": 202832466457748.28, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10495270043611526, "optim/lr": 0.0028602000000000002, "optim/total_tokens": 2499805184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 3.2617347240448, "created_at": "2025-01-15T19:01:00.402700+00:00"} {"global_step": 4769, "acc_step": 0, "speed/wps": 12909.168738245071, "speed/FLOPS": 202756234090905.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08855446428060532, "optim/lr": 0.0028608, "optim/total_tokens": 2500329472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.094536304473877, "created_at": "2025-01-15T19:01:10.564530+00:00"} {"global_step": 4770, "acc_step": 0, "speed/wps": 12912.183806232426, "speed/FLOPS": 202803589876784.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09544140845537186, "optim/lr": 0.0028614, "optim/total_tokens": 2500853760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.243086338043213, "created_at": "2025-01-15T19:01:20.722485+00:00"} {"global_step": 4771, "acc_step": 0, "speed/wps": 12912.074501986204, "speed/FLOPS": 202801873103397.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.079891137778759, "optim/lr": 0.002862, "optim/total_tokens": 2501378048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 3.2071003913879395, "created_at": "2025-01-15T19:01:30.879737+00:00"} {"global_step": 4772, "acc_step": 0, "speed/wps": 12908.930908741399, "speed/FLOPS": 202752498651738.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08975670486688614, "optim/lr": 0.0028626000000000003, "optim/total_tokens": 2501902336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.08467435836792, "created_at": "2025-01-15T19:01:41.039298+00:00"} {"global_step": 4773, "acc_step": 0, "speed/wps": 12910.105500416503, "speed/FLOPS": 202770947228053.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09465876966714859, "optim/lr": 0.0028632, "optim/total_tokens": 2502426624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.2023744583129883, "created_at": "2025-01-15T19:01:51.198197+00:00"} {"global_step": 4774, "acc_step": 0, "speed/wps": 12912.562698823613, "speed/FLOPS": 202809540905582.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.12048666179180145, "optim/lr": 0.0028638, "optim/total_tokens": 2502950912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.2388486862182617, "created_at": "2025-01-15T19:02:01.352514+00:00"} {"global_step": 4775, "acc_step": 0, "speed/wps": 12917.938836142452, "speed/FLOPS": 202893980529761.25, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13178877532482147, "optim/lr": 0.0028644, "optim/total_tokens": 2503475200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 3.1889753341674805, "created_at": "2025-01-15T19:02:11.502594+00:00"} {"global_step": 4776, "acc_step": 0, "speed/wps": 12906.314841683896, "speed/FLOPS": 202711409723747.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09399515390396118, "optim/lr": 0.002865, "optim/total_tokens": 2503999488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.063309907913208, "created_at": "2025-01-15T19:02:21.663965+00:00"} {"global_step": 4777, "acc_step": 0, "speed/wps": 12911.539436852761, "speed/FLOPS": 202793469170220.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16253536939620972, "optim/lr": 0.0028656000000000003, "optim/total_tokens": 2504523776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 3.296741247177124, "created_at": "2025-01-15T19:02:31.819307+00:00"} {"global_step": 4778, "acc_step": 0, "speed/wps": 12915.86700315398, "speed/FLOPS": 202861439545680.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13779287040233612, "optim/lr": 0.0028662, "optim/total_tokens": 2505048064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.226618766784668, "created_at": "2025-01-15T19:02:41.974852+00:00"} {"global_step": 4779, "acc_step": 0, "speed/wps": 12910.991278352414, "speed/FLOPS": 202784859587723.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12146786600351334, "optim/lr": 0.0028668, "optim/total_tokens": 2505572352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.0903522968292236, "created_at": "2025-01-15T19:02:52.130413+00:00"} {"global_step": 4780, "acc_step": 0, "speed/wps": 12909.618649855362, "speed/FLOPS": 202763300570988.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10059814900159836, "optim/lr": 0.0028674, "optim/total_tokens": 2506096640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.1143288612365723, "created_at": "2025-01-15T19:03:02.287087+00:00"} {"global_step": 4781, "acc_step": 0, "speed/wps": 12907.903345937157, "speed/FLOPS": 202736359365878.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12875252962112427, "optim/lr": 0.002868, "optim/total_tokens": 2506620928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.116658926010132, "created_at": "2025-01-15T19:03:12.445522+00:00"} {"global_step": 4782, "acc_step": 0, "speed/wps": 12911.074636701373, "speed/FLOPS": 202786168845140.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08689059317111969, "optim/lr": 0.0028686000000000002, "optim/total_tokens": 2507145216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.103977918624878, "created_at": "2025-01-15T19:03:22.602351+00:00"} {"global_step": 4783, "acc_step": 0, "speed/wps": 12903.293857477738, "speed/FLOPS": 202663961015522.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08717405796051025, "optim/lr": 0.0028692, "optim/total_tokens": 2507669504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.2947397232055664, "created_at": "2025-01-15T19:03:32.766713+00:00"} {"global_step": 4784, "acc_step": 0, "speed/wps": 12889.74108222, "speed/FLOPS": 202451096056634.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09308541566133499, "optim/lr": 0.0028698, "optim/total_tokens": 2508193792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.2213258743286133, "created_at": "2025-01-15T19:03:42.946814+00:00"} {"global_step": 4785, "acc_step": 0, "speed/wps": 12899.63934685366, "speed/FLOPS": 202606561904347.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07659785449504852, "optim/lr": 0.0028704, "optim/total_tokens": 2508718080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 3.1779732704162598, "created_at": "2025-01-15T19:03:53.111803+00:00"} {"global_step": 4786, "acc_step": 0, "speed/wps": 12901.426349293195, "speed/FLOPS": 202634629233258.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09872152656316757, "optim/lr": 0.002871, "optim/total_tokens": 2509242368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 3.2260475158691406, "created_at": "2025-01-15T19:04:03.278568+00:00"} {"global_step": 4787, "acc_step": 0, "speed/wps": 12905.551142011846, "speed/FLOPS": 202699414770964.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10707535594701767, "optim/lr": 0.0028716, "optim/total_tokens": 2509766656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.1403181552886963, "created_at": "2025-01-15T19:04:13.439784+00:00"} {"global_step": 4788, "acc_step": 0, "speed/wps": 12912.57463961022, "speed/FLOPS": 202809728452044.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09329014271497726, "optim/lr": 0.0028722, "optim/total_tokens": 2510290944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.2383127212524414, "created_at": "2025-01-15T19:04:23.599341+00:00"} {"global_step": 4789, "acc_step": 0, "speed/wps": 12909.407632764534, "speed/FLOPS": 202759986257609.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08562514930963516, "optim/lr": 0.0028728, "optim/total_tokens": 2510815232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 3.1035537719726562, "created_at": "2025-01-15T19:04:33.759485+00:00"} {"global_step": 4790, "acc_step": 0, "speed/wps": 12907.275689688917, "speed/FLOPS": 202726501162009.78, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1004008948802948, "optim/lr": 0.0028734, "optim/total_tokens": 2511339520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.2384159564971924, "created_at": "2025-01-15T19:04:43.919023+00:00"} {"global_step": 4791, "acc_step": 0, "speed/wps": 12910.223345575901, "speed/FLOPS": 202772798148218.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10096469521522522, "optim/lr": 0.002874, "optim/total_tokens": 2511863808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.1153650283813477, "created_at": "2025-01-15T19:04:54.076225+00:00"} {"global_step": 4792, "acc_step": 0, "speed/wps": 12910.014995780963, "speed/FLOPS": 202769525728386.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08573374897241592, "optim/lr": 0.0028746, "optim/total_tokens": 2512388096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 3.0858423709869385, "created_at": "2025-01-15T19:05:04.236228+00:00"} {"global_step": 4793, "acc_step": 0, "speed/wps": 12905.153667840414, "speed/FLOPS": 202693171892915.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0798010602593422, "optim/lr": 0.0028752, "optim/total_tokens": 2512912384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.176149845123291, "created_at": "2025-01-15T19:05:14.399135+00:00"} {"global_step": 4794, "acc_step": 0, "speed/wps": 12908.568801475998, "speed/FLOPS": 202746811259547.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10147591680288315, "optim/lr": 0.0028758, "optim/total_tokens": 2513436672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.296520233154297, "created_at": "2025-01-15T19:05:24.561033+00:00"} {"global_step": 4795, "acc_step": 0, "speed/wps": 12908.864100140061, "speed/FLOPS": 202751449330849.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08625099807977676, "optim/lr": 0.0028764, "optim/total_tokens": 2513960960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.2873778343200684, "created_at": "2025-01-15T19:05:34.720685+00:00"} {"global_step": 4796, "acc_step": 0, "speed/wps": 12910.188743111088, "speed/FLOPS": 202772254668959.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08583247661590576, "optim/lr": 0.002877, "optim/total_tokens": 2514485248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.075594663619995, "created_at": "2025-01-15T19:05:44.877685+00:00"} {"global_step": 4797, "acc_step": 0, "speed/wps": 12910.70456028722, "speed/FLOPS": 202780356286518.1, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.08544253557920456, "optim/lr": 0.0028776, "optim/total_tokens": 2515009536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.1935858726501465, "created_at": "2025-01-15T19:05:55.053179+00:00"} {"global_step": 4798, "acc_step": 0, "speed/wps": 12913.504924264173, "speed/FLOPS": 202824339850879.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07485304027795792, "optim/lr": 0.0028782, "optim/total_tokens": 2515533824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.0783934593200684, "created_at": "2025-01-15T19:06:05.209711+00:00"} {"global_step": 4799, "acc_step": 0, "speed/wps": 12905.866126048988, "speed/FLOPS": 202704362028101.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08376239240169525, "optim/lr": 0.0028788, "optim/total_tokens": 2516058112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.2258946895599365, "created_at": "2025-01-15T19:06:15.371440+00:00"} {"global_step": 4800, "acc_step": 0, "speed/wps": 12905.77333993309, "speed/FLOPS": 202702904694649.6, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09661360085010529, "optim/lr": 0.0028794, "optim/total_tokens": 2516582400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.2190206050872803, "created_at": "2025-01-15T19:06:25.532205+00:00"} {"global_step": 4801, "acc_step": 0, "speed/wps": 12908.14915855448, "speed/FLOPS": 202740220190815.47, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09050291776657104, "optim/lr": 0.0028799999999999997, "optim/total_tokens": 2517106688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.162898302078247, "created_at": "2025-01-15T19:06:35.690308+00:00"} {"global_step": 4802, "acc_step": 0, "speed/wps": 12911.216377266035, "speed/FLOPS": 202788395075480.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07563016563653946, "optim/lr": 0.0028806, "optim/total_tokens": 2517630976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.2556018829345703, "created_at": "2025-01-15T19:06:45.846262+00:00"} {"global_step": 4803, "acc_step": 0, "speed/wps": 12906.890613643396, "speed/FLOPS": 202720453013564.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07719198614358902, "optim/lr": 0.0028812, "optim/total_tokens": 2518155264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.1744441986083984, "created_at": "2025-01-15T19:06:56.005605+00:00"} {"global_step": 4804, "acc_step": 0, "speed/wps": 12911.983931834204, "speed/FLOPS": 202800450574704.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07640402764081955, "optim/lr": 0.0028818, "optim/total_tokens": 2518679552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 3.1325597763061523, "created_at": "2025-01-15T19:07:06.160399+00:00"} {"global_step": 4805, "acc_step": 0, "speed/wps": 12912.428062918254, "speed/FLOPS": 202807426263679.28, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08549989759922028, "optim/lr": 0.0028824, "optim/total_tokens": 2519203840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.1296451091766357, "created_at": "2025-01-15T19:07:16.315776+00:00"} {"global_step": 4806, "acc_step": 0, "speed/wps": 12912.543223269144, "speed/FLOPS": 202809235015236.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08041902631521225, "optim/lr": 0.002883, "optim/total_tokens": 2519728128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401883, "loss/out": 3.1929850578308105, "created_at": "2025-01-15T19:07:26.470190+00:00"} {"global_step": 4807, "acc_step": 0, "speed/wps": 12907.705882901897, "speed/FLOPS": 202733257937567.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09478256106376648, "optim/lr": 0.0028836, "optim/total_tokens": 2520252416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.260465621948242, "created_at": "2025-01-15T19:07:36.629492+00:00"} {"global_step": 4808, "acc_step": 0, "speed/wps": 12907.314199504235, "speed/FLOPS": 202727106011577.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14868049323558807, "optim/lr": 0.0028842, "optim/total_tokens": 2520776704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340630, "loss/out": 3.166790723800659, "created_at": "2025-01-15T19:07:46.788588+00:00"} {"global_step": 4809, "acc_step": 0, "speed/wps": 12912.318268820522, "speed/FLOPS": 202805701796501.47, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1330975443124771, "optim/lr": 0.0028848000000000003, "optim/total_tokens": 2521300992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.1236324310302734, "created_at": "2025-01-15T19:07:56.973774+00:00"} {"global_step": 4810, "acc_step": 0, "speed/wps": 12912.251592190161, "speed/FLOPS": 202804654548398.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08706211298704147, "optim/lr": 0.0028854, "optim/total_tokens": 2521825280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.089416742324829, "created_at": "2025-01-15T19:08:07.131030+00:00"} {"global_step": 4811, "acc_step": 0, "speed/wps": 12908.010172739847, "speed/FLOPS": 202738037227610.1, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09291788190603256, "optim/lr": 0.002886, "optim/total_tokens": 2522349568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419023, "loss/out": 3.3093056678771973, "created_at": "2025-01-15T19:08:17.291164+00:00"} {"global_step": 4812, "acc_step": 0, "speed/wps": 12909.959795717366, "speed/FLOPS": 202768658735534.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09971194714307785, "optim/lr": 0.0028866, "optim/total_tokens": 2522873856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501532, "loss/out": 3.1959898471832275, "created_at": "2025-01-15T19:08:27.450965+00:00"} {"global_step": 4813, "acc_step": 0, "speed/wps": 12910.767349172938, "speed/FLOPS": 202781342472249.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0829550102353096, "optim/lr": 0.0028872000000000004, "optim/total_tokens": 2523398144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.2523627281188965, "created_at": "2025-01-15T19:08:37.607648+00:00"} {"global_step": 4814, "acc_step": 0, "speed/wps": 12911.889554479489, "speed/FLOPS": 202798968248660.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10093657672405243, "optim/lr": 0.0028878000000000003, "optim/total_tokens": 2523922432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.0437331199645996, "created_at": "2025-01-15T19:08:47.765158+00:00"} {"global_step": 4815, "acc_step": 0, "speed/wps": 12913.877455320311, "speed/FLOPS": 202830190963027.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0932437852025032, "optim/lr": 0.0028884, "optim/total_tokens": 2524446720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433146, "loss/out": 3.206315040588379, "created_at": "2025-01-15T19:08:57.920591+00:00"} {"global_step": 4816, "acc_step": 0, "speed/wps": 12911.07890307454, "speed/FLOPS": 202786235854393.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0882456824183464, "optim/lr": 0.002889, "optim/total_tokens": 2524971008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.145120620727539, "created_at": "2025-01-15T19:09:08.079530+00:00"} {"global_step": 4817, "acc_step": 0, "speed/wps": 12909.34583756146, "speed/FLOPS": 202759015679029.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08770608901977539, "optim/lr": 0.0028896, "optim/total_tokens": 2525495296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.1684112548828125, "created_at": "2025-01-15T19:09:18.238216+00:00"} {"global_step": 4818, "acc_step": 0, "speed/wps": 12909.226096089682, "speed/FLOPS": 202757134974673.03, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08896096795797348, "optim/lr": 0.0028902000000000003, "optim/total_tokens": 2526019584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 3.183152675628662, "created_at": "2025-01-15T19:09:28.399623+00:00"} {"global_step": 4819, "acc_step": 0, "speed/wps": 12907.851370989309, "speed/FLOPS": 202735543027901.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07528713345527649, "optim/lr": 0.0028908000000000002, "optim/total_tokens": 2526543872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.183229446411133, "created_at": "2025-01-15T19:09:38.558908+00:00"} {"global_step": 4820, "acc_step": 0, "speed/wps": 12913.038335714173, "speed/FLOPS": 202817011436541.88, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08576856553554535, "optim/lr": 0.0028914, "optim/total_tokens": 2527068160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.167393922805786, "created_at": "2025-01-15T19:09:48.717945+00:00"} {"global_step": 4821, "acc_step": 0, "speed/wps": 12913.906491634423, "speed/FLOPS": 202830647018240.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07331705093383789, "optim/lr": 0.002892, "optim/total_tokens": 2527592448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.1608760356903076, "created_at": "2025-01-15T19:09:58.877975+00:00"} {"global_step": 4822, "acc_step": 0, "speed/wps": 12909.45435433791, "speed/FLOPS": 202760720084121.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07210966944694519, "optim/lr": 0.0028926, "optim/total_tokens": 2528116736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425702, "loss/out": 3.009652614593506, "created_at": "2025-01-15T19:10:09.038815+00:00"} {"global_step": 4823, "acc_step": 0, "speed/wps": 12910.118427892745, "speed/FLOPS": 202771150271832.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07935557514429092, "optim/lr": 0.0028932000000000003, "optim/total_tokens": 2528641024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.036609649658203, "created_at": "2025-01-15T19:10:19.198428+00:00"} {"global_step": 4824, "acc_step": 0, "speed/wps": 12908.824217070302, "speed/FLOPS": 202750822912434.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09519316256046295, "optim/lr": 0.0028938, "optim/total_tokens": 2529165312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.2200331687927246, "created_at": "2025-01-15T19:10:29.358991+00:00"} {"global_step": 4825, "acc_step": 0, "speed/wps": 12910.460085745892, "speed/FLOPS": 202776516477903.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08376667648553848, "optim/lr": 0.0028944, "optim/total_tokens": 2529689600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304045, "loss/out": 3.0759921073913574, "created_at": "2025-01-15T19:10:39.516003+00:00"} {"global_step": 4826, "acc_step": 0, "speed/wps": 12909.372968360734, "speed/FLOPS": 202759441805513.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08893173933029175, "optim/lr": 0.002895, "optim/total_tokens": 2530213888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.2629568576812744, "created_at": "2025-01-15T19:10:49.679581+00:00"} {"global_step": 4827, "acc_step": 0, "speed/wps": 12911.519175489968, "speed/FLOPS": 202793150937675.84, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09768493473529816, "optim/lr": 0.0028956, "optim/total_tokens": 2530738176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.171480894088745, "created_at": "2025-01-15T19:10:59.834775+00:00"} {"global_step": 4828, "acc_step": 0, "speed/wps": 12908.180832921349, "speed/FLOPS": 202740717680272.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11058587580919266, "optim/lr": 0.0028962000000000002, "optim/total_tokens": 2531262464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.235581159591675, "created_at": "2025-01-15T19:11:09.996259+00:00"} {"global_step": 4829, "acc_step": 0, "speed/wps": 12912.31499319932, "speed/FLOPS": 202805650348369.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08994495868682861, "optim/lr": 0.0028968, "optim/total_tokens": 2531786752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 3.1979010105133057, "created_at": "2025-01-15T19:11:20.153511+00:00"} {"global_step": 4830, "acc_step": 0, "speed/wps": 12907.363759981276, "speed/FLOPS": 202727884426972.53, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08224870264530182, "optim/lr": 0.0028974, "optim/total_tokens": 2532311040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.047926187515259, "created_at": "2025-01-15T19:11:30.314315+00:00"} {"global_step": 4831, "acc_step": 0, "speed/wps": 12905.344176088278, "speed/FLOPS": 202696164086733.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11032901704311371, "optim/lr": 0.002898, "optim/total_tokens": 2532835328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.200247287750244, "created_at": "2025-01-15T19:11:40.478626+00:00"} {"global_step": 4832, "acc_step": 0, "speed/wps": 12915.497454379767, "speed/FLOPS": 202855635274368.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1174459457397461, "optim/lr": 0.0028986, "optim/total_tokens": 2533359616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 3.215787649154663, "created_at": "2025-01-15T19:11:50.631671+00:00"} {"global_step": 4833, "acc_step": 0, "speed/wps": 12914.603139256766, "speed/FLOPS": 202841588826443.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11434170603752136, "optim/lr": 0.0028992, "optim/total_tokens": 2533883904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.206139326095581, "created_at": "2025-01-15T19:12:00.788966+00:00"} {"global_step": 4834, "acc_step": 0, "speed/wps": 12908.120310319502, "speed/FLOPS": 202739767089643.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1052609384059906, "optim/lr": 0.0028998, "optim/total_tokens": 2534408192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374042, "loss/out": 3.2420902252197266, "created_at": "2025-01-15T19:12:10.949376+00:00"} {"global_step": 4835, "acc_step": 0, "speed/wps": 12906.635526673017, "speed/FLOPS": 202716446522168.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07638536393642426, "optim/lr": 0.0029004, "optim/total_tokens": 2534932480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 3.189605236053467, "created_at": "2025-01-15T19:12:21.110545+00:00"} {"global_step": 4836, "acc_step": 0, "speed/wps": 12908.605277805009, "speed/FLOPS": 202747384170418.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09269417822360992, "optim/lr": 0.002901, "optim/total_tokens": 2535456768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 3.0660018920898438, "created_at": "2025-01-15T19:12:31.268766+00:00"} {"global_step": 4837, "acc_step": 0, "speed/wps": 12905.5556213036, "speed/FLOPS": 202699485124396.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08857248723506927, "optim/lr": 0.0029016, "optim/total_tokens": 2535981056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.162625312805176, "created_at": "2025-01-15T19:12:41.431405+00:00"} {"global_step": 4838, "acc_step": 0, "speed/wps": 12907.922612159731, "speed/FLOPS": 202736661968378.53, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08981291949748993, "optim/lr": 0.0029022, "optim/total_tokens": 2536505344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491222, "loss/out": 3.221498966217041, "created_at": "2025-01-15T19:12:51.589483+00:00"} {"global_step": 4839, "acc_step": 0, "speed/wps": 12912.640043619951, "speed/FLOPS": 202810755711891.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09021730720996857, "optim/lr": 0.0029028, "optim/total_tokens": 2537029632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 3.043893575668335, "created_at": "2025-01-15T19:13:01.747637+00:00"} {"global_step": 4840, "acc_step": 0, "speed/wps": 12908.36373844137, "speed/FLOPS": 202743590462802.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0741846039891243, "optim/lr": 0.0029034, "optim/total_tokens": 2537553920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 3.1821348667144775, "created_at": "2025-01-15T19:13:11.908851+00:00"} {"global_step": 4841, "acc_step": 0, "speed/wps": 12905.416498438402, "speed/FLOPS": 202697300008623.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07565312832593918, "optim/lr": 0.002904, "optim/total_tokens": 2538078208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 3.169429063796997, "created_at": "2025-01-15T19:13:22.071028+00:00"} {"global_step": 4842, "acc_step": 0, "speed/wps": 12908.22979033334, "speed/FLOPS": 202741486623702.12, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07832114398479462, "optim/lr": 0.0029046, "optim/total_tokens": 2538602496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.213871717453003, "created_at": "2025-01-15T19:13:32.228612+00:00"} {"global_step": 4843, "acc_step": 0, "speed/wps": 12905.25380328326, "speed/FLOPS": 202694744657645.84, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07348768413066864, "optim/lr": 0.0029052, "optim/total_tokens": 2539126784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 3.2265536785125732, "created_at": "2025-01-15T19:13:42.392694+00:00"} {"global_step": 4844, "acc_step": 0, "speed/wps": 12907.39037461288, "speed/FLOPS": 202728302446334.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07707614451646805, "optim/lr": 0.0029058, "optim/total_tokens": 2539651072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.0650854110717773, "created_at": "2025-01-15T19:13:52.553809+00:00"} {"global_step": 4845, "acc_step": 0, "speed/wps": 12910.852456364348, "speed/FLOPS": 202782679197639.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08091085404157639, "optim/lr": 0.0029064, "optim/total_tokens": 2540175360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.032231330871582, "created_at": "2025-01-15T19:14:02.710827+00:00"} {"global_step": 4846, "acc_step": 0, "speed/wps": 12906.028096943843, "speed/FLOPS": 202706906003576.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09283710271120071, "optim/lr": 0.002907, "optim/total_tokens": 2540699648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 3.270449161529541, "created_at": "2025-01-15T19:14:12.870584+00:00"} {"global_step": 4847, "acc_step": 0, "speed/wps": 12909.309238634944, "speed/FLOPS": 202758440842596.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08223102986812592, "optim/lr": 0.0029075999999999998, "optim/total_tokens": 2541223936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.08359956741333, "created_at": "2025-01-15T19:14:23.031645+00:00"} {"global_step": 4848, "acc_step": 0, "speed/wps": 12910.072615762194, "speed/FLOPS": 202770430729369.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08378241956233978, "optim/lr": 0.0029082, "optim/total_tokens": 2541748224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.2173166275024414, "created_at": "2025-01-15T19:14:33.188001+00:00"} {"global_step": 4849, "acc_step": 0, "speed/wps": 12912.507931899148, "speed/FLOPS": 202808680715775.7, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08868489414453506, "optim/lr": 0.0029088, "optim/total_tokens": 2542272512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.159691095352173, "created_at": "2025-01-15T19:14:43.346127+00:00"} {"global_step": 4850, "acc_step": 0, "speed/wps": 12906.499321784097, "speed/FLOPS": 202714307237223.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09746349602937698, "optim/lr": 0.0029094, "optim/total_tokens": 2542796800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 3.0758562088012695, "created_at": "2025-01-15T19:14:53.505195+00:00"} {"global_step": 4851, "acc_step": 0, "speed/wps": 12911.288979839383, "speed/FLOPS": 202789535398661.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08117365837097168, "optim/lr": 0.00291, "optim/total_tokens": 2543321088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.0797576904296875, "created_at": "2025-01-15T19:15:03.660978+00:00"} {"global_step": 4852, "acc_step": 0, "speed/wps": 12907.705067244986, "speed/FLOPS": 202733245126554.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09796752780675888, "optim/lr": 0.0029105999999999997, "optim/total_tokens": 2543845376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.1731467247009277, "created_at": "2025-01-15T19:15:13.821028+00:00"} {"global_step": 4853, "acc_step": 0, "speed/wps": 12901.540166272045, "speed/FLOPS": 202636416885312.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08601519465446472, "optim/lr": 0.0029112, "optim/total_tokens": 2544369664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 3.168405055999756, "created_at": "2025-01-15T19:15:23.984960+00:00"} {"global_step": 4854, "acc_step": 0, "speed/wps": 12912.229694949761, "speed/FLOPS": 202804310622147.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09246490895748138, "optim/lr": 0.0029118, "optim/total_tokens": 2544893952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.128278970718384, "created_at": "2025-01-15T19:15:34.144009+00:00"} {"global_step": 4855, "acc_step": 0, "speed/wps": 12907.783425235193, "speed/FLOPS": 202734475846465.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08054956048727036, "optim/lr": 0.0029124, "optim/total_tokens": 2545418240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 3.2025556564331055, "created_at": "2025-01-15T19:15:44.304985+00:00"} {"global_step": 4856, "acc_step": 0, "speed/wps": 12907.931157317294, "speed/FLOPS": 202736796181820.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0737169161438942, "optim/lr": 0.002913, "optim/total_tokens": 2545942528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9747185707092285, "created_at": "2025-01-15T19:15:54.465157+00:00"} {"global_step": 4857, "acc_step": 0, "speed/wps": 12910.055722295434, "speed/FLOPS": 202770165394260.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07979083061218262, "optim/lr": 0.0029136, "optim/total_tokens": 2546466816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370300, "loss/out": 3.178896427154541, "created_at": "2025-01-15T19:16:04.627306+00:00"} {"global_step": 4858, "acc_step": 0, "speed/wps": 12899.634397546613, "speed/FLOPS": 202606484168679.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09032696485519409, "optim/lr": 0.0029142, "optim/total_tokens": 2546991104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.1093099117279053, "created_at": "2025-01-15T19:16:14.795116+00:00"} {"global_step": 4859, "acc_step": 0, "speed/wps": 12896.45939241057, "speed/FLOPS": 202556616350103.34, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0903191938996315, "optim/lr": 0.0029148, "optim/total_tokens": 2547515392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.116635322570801, "created_at": "2025-01-15T19:16:24.964302+00:00"} {"global_step": 4860, "acc_step": 0, "speed/wps": 12902.797438421447, "speed/FLOPS": 202656164072092.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09406150132417679, "optim/lr": 0.0029154, "optim/total_tokens": 2548039680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 3.125382661819458, "created_at": "2025-01-15T19:16:35.127884+00:00"} {"global_step": 4861, "acc_step": 0, "speed/wps": 12908.285397083999, "speed/FLOPS": 202742360004132.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09466783702373505, "optim/lr": 0.002916, "optim/total_tokens": 2548563968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.2289600372314453, "created_at": "2025-01-15T19:16:45.286850+00:00"} {"global_step": 4862, "acc_step": 0, "speed/wps": 12907.03436224688, "speed/FLOPS": 202722710782913.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08920924365520477, "optim/lr": 0.0029166, "optim/total_tokens": 2549088256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 3.216644763946533, "created_at": "2025-01-15T19:16:55.449286+00:00"} {"global_step": 4863, "acc_step": 0, "speed/wps": 12906.988865558094, "speed/FLOPS": 202721996194897.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08396587520837784, "optim/lr": 0.0029172, "optim/total_tokens": 2549612544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 3.0936694145202637, "created_at": "2025-01-15T19:17:05.610227+00:00"} {"global_step": 4864, "acc_step": 0, "speed/wps": 12901.153449109459, "speed/FLOPS": 202630342960869.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08322995901107788, "optim/lr": 0.0029178, "optim/total_tokens": 2550136832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.1359410285949707, "created_at": "2025-01-15T19:17:15.775182+00:00"} {"global_step": 4865, "acc_step": 0, "speed/wps": 12908.175331273696, "speed/FLOPS": 202740631269335.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08614843338727951, "optim/lr": 0.0029184000000000002, "optim/total_tokens": 2550661120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374420, "loss/out": 3.1004416942596436, "created_at": "2025-01-15T19:17:25.933621+00:00"} {"global_step": 4866, "acc_step": 0, "speed/wps": 12902.983742658214, "speed/FLOPS": 202659090236139.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10566508769989014, "optim/lr": 0.002919, "optim/total_tokens": 2551185408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.119455575942993, "created_at": "2025-01-15T19:17:36.099018+00:00"} {"global_step": 4867, "acc_step": 0, "speed/wps": 12911.257431169997, "speed/FLOPS": 202789039883456.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10802844911813736, "optim/lr": 0.0029196, "optim/total_tokens": 2551709696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2012720108032227, "created_at": "2025-01-15T19:17:46.260882+00:00"} {"global_step": 4868, "acc_step": 0, "speed/wps": 12912.302376446018, "speed/FLOPS": 202805452184921.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09563135355710983, "optim/lr": 0.0029202000000000004, "optim/total_tokens": 2552233984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.1480584144592285, "created_at": "2025-01-15T19:17:56.415640+00:00"} {"global_step": 4869, "acc_step": 0, "speed/wps": 12907.841754551095, "speed/FLOPS": 202735391988524.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1131327673792839, "optim/lr": 0.0029208000000000003, "optim/total_tokens": 2552758272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.1734671592712402, "created_at": "2025-01-15T19:18:06.575029+00:00"} {"global_step": 4870, "acc_step": 0, "speed/wps": 12907.564164948559, "speed/FLOPS": 202731032062369.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13078510761260986, "optim/lr": 0.0029214, "optim/total_tokens": 2553282560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 3.188908100128174, "created_at": "2025-01-15T19:18:16.735459+00:00"} {"global_step": 4871, "acc_step": 0, "speed/wps": 12914.833756494578, "speed/FLOPS": 202845210987064.84, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10850455611944199, "optim/lr": 0.002922, "optim/total_tokens": 2553806848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.110806703567505, "created_at": "2025-01-15T19:18:26.888291+00:00"} {"global_step": 4872, "acc_step": 0, "speed/wps": 12900.588425975839, "speed/FLOPS": 202621468496132.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08735395222902298, "optim/lr": 0.0029226, "optim/total_tokens": 2554331136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424102, "loss/out": 3.14901065826416, "created_at": "2025-01-15T19:18:37.056455+00:00"} {"global_step": 4873, "acc_step": 0, "speed/wps": 12909.27182524513, "speed/FLOPS": 202757853213897.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10407636314630508, "optim/lr": 0.0029232000000000004, "optim/total_tokens": 2554855424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.1271917819976807, "created_at": "2025-01-15T19:18:47.215040+00:00"} {"global_step": 4874, "acc_step": 0, "speed/wps": 12905.614375461417, "speed/FLOPS": 202700407939181.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08873969316482544, "optim/lr": 0.0029238000000000003, "optim/total_tokens": 2555379712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.157809019088745, "created_at": "2025-01-15T19:18:57.378573+00:00"} {"global_step": 4875, "acc_step": 0, "speed/wps": 12911.761493948004, "speed/FLOPS": 202796956882039.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07640281319618225, "optim/lr": 0.0029244, "optim/total_tokens": 2555904000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.2243123054504395, "created_at": "2025-01-15T19:19:07.536709+00:00"} {"global_step": 4876, "acc_step": 0, "speed/wps": 12907.143735061203, "speed/FLOPS": 202724428633256.88, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08410224318504333, "optim/lr": 0.002925, "optim/total_tokens": 2556428288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.149209976196289, "created_at": "2025-01-15T19:19:17.697514+00:00"} {"global_step": 4877, "acc_step": 0, "speed/wps": 12910.4965372341, "speed/FLOPS": 202777088998615.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07786087691783905, "optim/lr": 0.0029256, "optim/total_tokens": 2556952576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 3.1689984798431396, "created_at": "2025-01-15T19:19:27.854866+00:00"} {"global_step": 4878, "acc_step": 0, "speed/wps": 12913.521531720387, "speed/FLOPS": 202824600693800.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07842318713665009, "optim/lr": 0.0029262000000000003, "optim/total_tokens": 2557476864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.154670000076294, "created_at": "2025-01-15T19:19:38.009037+00:00"} {"global_step": 4879, "acc_step": 0, "speed/wps": 12908.10496339991, "speed/FLOPS": 202739526045182.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08475712686777115, "optim/lr": 0.0029268000000000002, "optim/total_tokens": 2558001152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.083293914794922, "created_at": "2025-01-15T19:19:48.167799+00:00"} {"global_step": 4880, "acc_step": 0, "speed/wps": 12912.78951830766, "speed/FLOPS": 202813103417261.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10073336213827133, "optim/lr": 0.0029274, "optim/total_tokens": 2558525440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9447693824768066, "created_at": "2025-01-15T19:19:58.325985+00:00"} {"global_step": 4881, "acc_step": 0, "speed/wps": 12906.316022888946, "speed/FLOPS": 202711428276195.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07607245445251465, "optim/lr": 0.002928, "optim/total_tokens": 2559049728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457122, "loss/out": 3.2646055221557617, "created_at": "2025-01-15T19:20:08.485133+00:00"} {"global_step": 4882, "acc_step": 0, "speed/wps": 12910.33202988155, "speed/FLOPS": 202774505184586.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11309613287448883, "optim/lr": 0.0029286, "optim/total_tokens": 2559574016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.1168103218078613, "created_at": "2025-01-15T19:20:18.643021+00:00"} {"global_step": 4883, "acc_step": 0, "speed/wps": 12904.93774036358, "speed/FLOPS": 202689780455176.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12556591629981995, "optim/lr": 0.0029292000000000003, "optim/total_tokens": 2560098304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.10870099067688, "created_at": "2025-01-15T19:20:28.804565+00:00"} {"global_step": 4884, "acc_step": 0, "speed/wps": 12911.603136480408, "speed/FLOPS": 202794469660405.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09542122483253479, "optim/lr": 0.0029298, "optim/total_tokens": 2560622592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 3.223087787628174, "created_at": "2025-01-15T19:20:38.963437+00:00"} {"global_step": 4885, "acc_step": 0, "speed/wps": 12908.285382876797, "speed/FLOPS": 202742359780989.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09155218303203583, "optim/lr": 0.0029304, "optim/total_tokens": 2561146880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 3.127607583999634, "created_at": "2025-01-15T19:20:49.121114+00:00"} {"global_step": 4886, "acc_step": 0, "speed/wps": 12916.555163474433, "speed/FLOPS": 202872248049144.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0916716456413269, "optim/lr": 0.002931, "optim/total_tokens": 2561671168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372261, "loss/out": 3.0973925590515137, "created_at": "2025-01-15T19:20:59.278122+00:00"} {"global_step": 4887, "acc_step": 0, "speed/wps": 12912.309550814636, "speed/FLOPS": 202805564868239.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08597945421934128, "optim/lr": 0.0029316, "optim/total_tokens": 2562195456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.089555263519287, "created_at": "2025-01-15T19:21:09.434890+00:00"} {"global_step": 4888, "acc_step": 0, "speed/wps": 12909.860641189234, "speed/FLOPS": 202767101377417.84, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10050223767757416, "optim/lr": 0.0029322000000000003, "optim/total_tokens": 2562719744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 3.1539812088012695, "created_at": "2025-01-15T19:21:19.591237+00:00"} {"global_step": 4889, "acc_step": 0, "speed/wps": 12907.481922065515, "speed/FLOPS": 202729740324877.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09608620405197144, "optim/lr": 0.0029328, "optim/total_tokens": 2563244032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.418687343597412, "created_at": "2025-01-15T19:21:29.749601+00:00"} {"global_step": 4890, "acc_step": 0, "speed/wps": 12904.93954847607, "speed/FLOPS": 202689808854067.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09767051786184311, "optim/lr": 0.0029334, "optim/total_tokens": 2563768320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.1112375259399414, "created_at": "2025-01-15T19:21:39.914706+00:00"} {"global_step": 4891, "acc_step": 0, "speed/wps": 12908.992277261541, "speed/FLOPS": 202753462528676.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0920976921916008, "optim/lr": 0.002934, "optim/total_tokens": 2564292608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 3.201104164123535, "created_at": "2025-01-15T19:21:50.074156+00:00"} {"global_step": 4892, "acc_step": 0, "speed/wps": 12908.314105168167, "speed/FLOPS": 202742810904043.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07875220477581024, "optim/lr": 0.0029346, "optim/total_tokens": 2564816896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.2111713886260986, "created_at": "2025-01-15T19:22:00.238141+00:00"} {"global_step": 4893, "acc_step": 0, "speed/wps": 12913.38754053123, "speed/FLOPS": 202822496178051.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08446201682090759, "optim/lr": 0.0029352000000000002, "optim/total_tokens": 2565341184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 3.1176629066467285, "created_at": "2025-01-15T19:22:10.392745+00:00"} {"global_step": 4894, "acc_step": 0, "speed/wps": 12909.864552677373, "speed/FLOPS": 202767162812714.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.084471695125103, "optim/lr": 0.0029358, "optim/total_tokens": 2565865472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.1139369010925293, "created_at": "2025-01-15T19:22:20.554616+00:00"} {"global_step": 4895, "acc_step": 0, "speed/wps": 12910.077430970841, "speed/FLOPS": 202770506358838.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07272972911596298, "optim/lr": 0.0029364, "optim/total_tokens": 2566389760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381371, "loss/out": 3.1605300903320312, "created_at": "2025-01-15T19:22:30.713531+00:00"} {"global_step": 4896, "acc_step": 0, "speed/wps": 12914.684489274528, "speed/FLOPS": 202842866540259.3, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07761737704277039, "optim/lr": 0.002937, "optim/total_tokens": 2566914048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.12358021736145, "created_at": "2025-01-15T19:22:40.866183+00:00"} {"global_step": 4897, "acc_step": 0, "speed/wps": 12910.795223824965, "speed/FLOPS": 202781780281963.12, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07880779355764389, "optim/lr": 0.0029376, "optim/total_tokens": 2567438336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.1844522953033447, "created_at": "2025-01-15T19:22:51.022668+00:00"} {"global_step": 4898, "acc_step": 0, "speed/wps": 12910.575605385991, "speed/FLOPS": 202778330872591.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07895821332931519, "optim/lr": 0.0029382, "optim/total_tokens": 2567962624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.1095595359802246, "created_at": "2025-01-15T19:23:01.181931+00:00"} {"global_step": 4899, "acc_step": 0, "speed/wps": 12910.850635934925, "speed/FLOPS": 202782650605294.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08954692631959915, "optim/lr": 0.0029388, "optim/total_tokens": 2568486912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.1793508529663086, "created_at": "2025-01-15T19:23:11.342663+00:00"} {"global_step": 4900, "acc_step": 0, "speed/wps": 12904.868912950407, "speed/FLOPS": 202688699426073.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10027062892913818, "optim/lr": 0.0029394, "optim/total_tokens": 2569011200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 3.097834587097168, "created_at": "2025-01-15T19:23:21.506354+00:00"} {"global_step": 4901, "acc_step": 0, "speed/wps": 12914.63832594087, "speed/FLOPS": 202842141481669.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09902042150497437, "optim/lr": 0.00294, "optim/total_tokens": 2569535488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.235307216644287, "created_at": "2025-01-15T19:23:31.660765+00:00"} {"global_step": 4902, "acc_step": 0, "speed/wps": 12911.087375422505, "speed/FLOPS": 202786368924260.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10409627109766006, "optim/lr": 0.0029406, "optim/total_tokens": 2570059776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 3.225964069366455, "created_at": "2025-01-15T19:23:41.817161+00:00"} {"global_step": 4903, "acc_step": 0, "speed/wps": 12909.890999988904, "speed/FLOPS": 202767578204084.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09204056859016418, "optim/lr": 0.0029412, "optim/total_tokens": 2570584064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.1419949531555176, "created_at": "2025-01-15T19:23:51.976345+00:00"} {"global_step": 4904, "acc_step": 0, "speed/wps": 12904.702246325473, "speed/FLOPS": 202686081697705.12, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08899521082639694, "optim/lr": 0.0029418, "optim/total_tokens": 2571108352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469093, "loss/out": 3.194150447845459, "created_at": "2025-01-15T19:24:02.137101+00:00"} {"global_step": 4905, "acc_step": 0, "speed/wps": 12901.826022077174, "speed/FLOPS": 202640906643539.28, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0832139179110527, "optim/lr": 0.0029424, "optim/total_tokens": 2571632640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.13474440574646, "created_at": "2025-01-15T19:24:12.299865+00:00"} {"global_step": 4906, "acc_step": 0, "speed/wps": 12906.452936434698, "speed/FLOPS": 202713578691570.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07446323335170746, "optim/lr": 0.002943, "optim/total_tokens": 2572156928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.2116541862487793, "created_at": "2025-01-15T19:24:22.459693+00:00"} {"global_step": 4907, "acc_step": 0, "speed/wps": 12916.441574557859, "speed/FLOPS": 202870463979122.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08737236261367798, "optim/lr": 0.0029435999999999998, "optim/total_tokens": 2572681216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.2017111778259277, "created_at": "2025-01-15T19:24:32.617211+00:00"} {"global_step": 4908, "acc_step": 0, "speed/wps": 12911.379464434158, "speed/FLOPS": 202790956583561.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07925283908843994, "optim/lr": 0.0029442, "optim/total_tokens": 2573205504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.117861270904541, "created_at": "2025-01-15T19:24:42.775704+00:00"} {"global_step": 4909, "acc_step": 0, "speed/wps": 12909.810638771623, "speed/FLOPS": 202766316020738.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08423972874879837, "optim/lr": 0.0029448, "optim/total_tokens": 2573729792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.1844592094421387, "created_at": "2025-01-15T19:24:52.935381+00:00"} {"global_step": 4910, "acc_step": 0, "speed/wps": 12908.544089377508, "speed/FLOPS": 202746423122082.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10183563828468323, "optim/lr": 0.0029454, "optim/total_tokens": 2574254080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.1664083003997803, "created_at": "2025-01-15T19:25:03.094527+00:00"} {"global_step": 4911, "acc_step": 0, "speed/wps": 12904.764860880068, "speed/FLOPS": 202687065145326.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0899161845445633, "optim/lr": 0.002946, "optim/total_tokens": 2574778368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.2028775215148926, "created_at": "2025-01-15T19:25:13.255004+00:00"} {"global_step": 4912, "acc_step": 0, "speed/wps": 12912.475165322, "speed/FLOPS": 202808166071655.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08667266368865967, "optim/lr": 0.0029465999999999997, "optim/total_tokens": 2575302656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.168217182159424, "created_at": "2025-01-15T19:25:23.409553+00:00"} {"global_step": 4913, "acc_step": 0, "speed/wps": 12912.64952741038, "speed/FLOPS": 202810904667852.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09233514964580536, "optim/lr": 0.0029472, "optim/total_tokens": 2575826944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0034539699554443, "created_at": "2025-01-15T19:25:33.565021+00:00"} {"global_step": 4914, "acc_step": 0, "speed/wps": 12909.1353989361, "speed/FLOPS": 202755710451245.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07587841898202896, "optim/lr": 0.0029478, "optim/total_tokens": 2576351232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.169362783432007, "created_at": "2025-01-15T19:25:43.722438+00:00"} {"global_step": 4915, "acc_step": 0, "speed/wps": 12912.281054781315, "speed/FLOPS": 202805117298878.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09008108079433441, "optim/lr": 0.0029484, "optim/total_tokens": 2576875520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483427, "loss/out": 3.1549715995788574, "created_at": "2025-01-15T19:25:53.878493+00:00"} {"global_step": 4916, "acc_step": 0, "speed/wps": 12911.033987853134, "speed/FLOPS": 202785530399121.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11122281849384308, "optim/lr": 0.0029490000000000002, "optim/total_tokens": 2577399808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.271658420562744, "created_at": "2025-01-15T19:26:04.038691+00:00"} {"global_step": 4917, "acc_step": 0, "speed/wps": 12910.06149437297, "speed/FLOPS": 202770256052669.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10819622129201889, "optim/lr": 0.0029496, "optim/total_tokens": 2577924096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.2777247428894043, "created_at": "2025-01-15T19:26:14.198580+00:00"} {"global_step": 4918, "acc_step": 0, "speed/wps": 12907.46305135808, "speed/FLOPS": 202729443934485.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09054266661405563, "optim/lr": 0.0029502, "optim/total_tokens": 2578448384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333121, "loss/out": 3.210144281387329, "created_at": "2025-01-15T19:26:24.360933+00:00"} {"global_step": 4919, "acc_step": 0, "speed/wps": 12910.257474508058, "speed/FLOPS": 202773334189996.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09450271725654602, "optim/lr": 0.0029508, "optim/total_tokens": 2578972672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.1289217472076416, "created_at": "2025-01-15T19:26:34.524082+00:00"} {"global_step": 4920, "acc_step": 0, "speed/wps": 12909.579879980307, "speed/FLOPS": 202762691636825.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09425358474254608, "optim/lr": 0.0029514000000000003, "optim/total_tokens": 2579496960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.180637836456299, "created_at": "2025-01-15T19:26:44.682301+00:00"} {"global_step": 4921, "acc_step": 0, "speed/wps": 12908.80997672863, "speed/FLOPS": 202750599248299.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08831463754177094, "optim/lr": 0.002952, "optim/total_tokens": 2580021248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.293304681777954, "created_at": "2025-01-15T19:26:54.842638+00:00"} {"global_step": 4922, "acc_step": 0, "speed/wps": 12908.711351354961, "speed/FLOPS": 202749050201280.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07963038235902786, "optim/lr": 0.0029526, "optim/total_tokens": 2580545536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.099381923675537, "created_at": "2025-01-15T19:27:05.007051+00:00"} {"global_step": 4923, "acc_step": 0, "speed/wps": 12907.24591493605, "speed/FLOPS": 202726033508601.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08515261858701706, "optim/lr": 0.0029532, "optim/total_tokens": 2581069824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.2151663303375244, "created_at": "2025-01-15T19:27:15.167791+00:00"} {"global_step": 4924, "acc_step": 0, "speed/wps": 12908.868199286777, "speed/FLOPS": 202751513713581.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08684779703617096, "optim/lr": 0.0029538000000000003, "optim/total_tokens": 2581594112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.180896282196045, "created_at": "2025-01-15T19:27:25.325212+00:00"} {"global_step": 4925, "acc_step": 0, "speed/wps": 12907.310720448093, "speed/FLOPS": 202727051368219.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08238117396831512, "optim/lr": 0.0029544000000000003, "optim/total_tokens": 2582118400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373078, "loss/out": 3.0325734615325928, "created_at": "2025-01-15T19:27:35.483932+00:00"} {"global_step": 4926, "acc_step": 0, "speed/wps": 12911.550442335227, "speed/FLOPS": 202793642026445.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10111629962921143, "optim/lr": 0.002955, "optim/total_tokens": 2582642688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.1722350120544434, "created_at": "2025-01-15T19:27:45.642253+00:00"} {"global_step": 4927, "acc_step": 0, "speed/wps": 12901.575651996238, "speed/FLOPS": 202636974237373.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08177027851343155, "optim/lr": 0.0029556, "optim/total_tokens": 2583166976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.0906076431274414, "created_at": "2025-01-15T19:27:55.806424+00:00"} {"global_step": 4928, "acc_step": 0, "speed/wps": 12911.107187707545, "speed/FLOPS": 202786680103422.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0758020207285881, "optim/lr": 0.0029562000000000004, "optim/total_tokens": 2583691264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.076683521270752, "created_at": "2025-01-15T19:28:05.961830+00:00"} {"global_step": 4929, "acc_step": 0, "speed/wps": 12907.94283500727, "speed/FLOPS": 202736979595988.25, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08336220681667328, "optim/lr": 0.0029568000000000003, "optim/total_tokens": 2584215552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.2460131645202637, "created_at": "2025-01-15T19:28:16.120853+00:00"} {"global_step": 4930, "acc_step": 0, "speed/wps": 12908.71964299402, "speed/FLOPS": 202749180432866.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07503826916217804, "optim/lr": 0.0029574000000000002, "optim/total_tokens": 2584739840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.0575075149536133, "created_at": "2025-01-15T19:28:26.279937+00:00"} {"global_step": 4931, "acc_step": 0, "speed/wps": 12907.732179309756, "speed/FLOPS": 202733670958787.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07129046320915222, "optim/lr": 0.002958, "optim/total_tokens": 2585264128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 3.191216468811035, "created_at": "2025-01-15T19:28:36.437967+00:00"} {"global_step": 4932, "acc_step": 0, "speed/wps": 12911.261177677465, "speed/FLOPS": 202789098727504.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0694081038236618, "optim/lr": 0.0029586, "optim/total_tokens": 2585788416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 2.960413932800293, "created_at": "2025-01-15T19:28:46.594733+00:00"} {"global_step": 4933, "acc_step": 0, "speed/wps": 12912.74502361804, "speed/FLOPS": 202812404567019.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07622184604406357, "optim/lr": 0.0029592000000000004, "optim/total_tokens": 2586312704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 3.188143253326416, "created_at": "2025-01-15T19:28:56.748903+00:00"} {"global_step": 4934, "acc_step": 0, "speed/wps": 12906.585699555199, "speed/FLOPS": 202715663918813.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07648753374814987, "optim/lr": 0.0029598000000000003, "optim/total_tokens": 2586836992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 3.1889023780822754, "created_at": "2025-01-15T19:29:06.908546+00:00"} {"global_step": 4935, "acc_step": 0, "speed/wps": 12911.78601106607, "speed/FLOPS": 202797341957069.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08469294011592865, "optim/lr": 0.0029604, "optim/total_tokens": 2587361280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.20882248878479, "created_at": "2025-01-15T19:29:17.064563+00:00"} {"global_step": 4936, "acc_step": 0, "speed/wps": 12912.597566361528, "speed/FLOPS": 202810088548178.03, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08672569692134857, "optim/lr": 0.002961, "optim/total_tokens": 2587885568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466183, "loss/out": 3.1992580890655518, "created_at": "2025-01-15T19:29:27.219044+00:00"} {"global_step": 4937, "acc_step": 0, "speed/wps": 12910.973209898868, "speed/FLOPS": 202784575797832.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08806483447551727, "optim/lr": 0.0029616, "optim/total_tokens": 2588409856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.2540626525878906, "created_at": "2025-01-15T19:29:37.377593+00:00"} {"global_step": 4938, "acc_step": 0, "speed/wps": 12906.9785887201, "speed/FLOPS": 202721834783034.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09722460061311722, "optim/lr": 0.0029622000000000003, "optim/total_tokens": 2588934144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.1961262226104736, "created_at": "2025-01-15T19:29:47.536200+00:00"} {"global_step": 4939, "acc_step": 0, "speed/wps": 12909.792842438128, "speed/FLOPS": 202766036504866.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09269040822982788, "optim/lr": 0.0029628000000000002, "optim/total_tokens": 2589458432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 3.08847975730896, "created_at": "2025-01-15T19:29:57.695683+00:00"} {"global_step": 4940, "acc_step": 0, "speed/wps": 12912.597795013397, "speed/FLOPS": 202810092139469.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08796984702348709, "optim/lr": 0.0029634, "optim/total_tokens": 2589982720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.99200701713562, "created_at": "2025-01-15T19:30:07.855617+00:00"} {"global_step": 4941, "acc_step": 0, "speed/wps": 12909.628531829296, "speed/FLOPS": 202763455780968.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11384406685829163, "optim/lr": 0.002964, "optim/total_tokens": 2590507008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332493, "loss/out": 3.0543394088745117, "created_at": "2025-01-15T19:30:18.015535+00:00"} {"global_step": 4942, "acc_step": 0, "speed/wps": 12907.057552234739, "speed/FLOPS": 202723075013539.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10859265923500061, "optim/lr": 0.0029646, "optim/total_tokens": 2591031296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.21999454498291, "created_at": "2025-01-15T19:30:28.174318+00:00"} {"global_step": 4943, "acc_step": 0, "speed/wps": 12898.45905413279, "speed/FLOPS": 202588023785273.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09002063423395157, "optim/lr": 0.0029652, "optim/total_tokens": 2591555584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.144493579864502, "created_at": "2025-01-15T19:30:38.340929+00:00"} {"global_step": 4944, "acc_step": 0, "speed/wps": 12907.079671046145, "speed/FLOPS": 202723422419867.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09638812392950058, "optim/lr": 0.0029658, "optim/total_tokens": 2592079872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.144810676574707, "created_at": "2025-01-15T19:30:48.500056+00:00"} {"global_step": 4945, "acc_step": 0, "speed/wps": 12913.487141441748, "speed/FLOPS": 202824060547217.16, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08469602465629578, "optim/lr": 0.0029664, "optim/total_tokens": 2592604160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.1501574516296387, "created_at": "2025-01-15T19:30:58.653697+00:00"} {"global_step": 4946, "acc_step": 0, "speed/wps": 12911.221497737268, "speed/FLOPS": 202788475499517.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08378352224826813, "optim/lr": 0.002967, "optim/total_tokens": 2593128448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 2.9792380332946777, "created_at": "2025-01-15T19:31:08.809688+00:00"} {"global_step": 4947, "acc_step": 0, "speed/wps": 12911.611724417146, "speed/FLOPS": 202794604545752.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0847957581281662, "optim/lr": 0.0029676, "optim/total_tokens": 2593652736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499445, "loss/out": 3.193669319152832, "created_at": "2025-01-15T19:31:18.964668+00:00"} {"global_step": 4948, "acc_step": 0, "speed/wps": 12909.187192401023, "speed/FLOPS": 202756523938783.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07533132284879684, "optim/lr": 0.0029682, "optim/total_tokens": 2594177024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 3.071216344833374, "created_at": "2025-01-15T19:31:29.121886+00:00"} {"global_step": 4949, "acc_step": 0, "speed/wps": 12906.308612547009, "speed/FLOPS": 202711311886592.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08799906820058823, "optim/lr": 0.0029688, "optim/total_tokens": 2594701312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.978180170059204, "created_at": "2025-01-15T19:31:39.282412+00:00"} {"global_step": 4950, "acc_step": 0, "speed/wps": 12911.383285636064, "speed/FLOPS": 202791016600788.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10925135761499405, "optim/lr": 0.0029694, "optim/total_tokens": 2595225600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 3.103022575378418, "created_at": "2025-01-15T19:31:49.438480+00:00"} {"global_step": 4951, "acc_step": 0, "speed/wps": 12910.281757344863, "speed/FLOPS": 202773715585316.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0986117422580719, "optim/lr": 0.00297, "optim/total_tokens": 2595749888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.1953554153442383, "created_at": "2025-01-15T19:31:59.595786+00:00"} {"global_step": 4952, "acc_step": 0, "speed/wps": 12912.5803594432, "speed/FLOPS": 202809818289880.9, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09113252907991409, "optim/lr": 0.0029706, "optim/total_tokens": 2596274176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490867, "loss/out": 3.107962131500244, "created_at": "2025-01-15T19:32:09.753676+00:00"} {"global_step": 4953, "acc_step": 0, "speed/wps": 12911.522894905207, "speed/FLOPS": 202793209356203.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08024102449417114, "optim/lr": 0.0029712, "optim/total_tokens": 2596798464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 3.1408746242523193, "created_at": "2025-01-15T19:32:19.913019+00:00"} {"global_step": 4954, "acc_step": 0, "speed/wps": 12907.908093231346, "speed/FLOPS": 202736433928657.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08973245322704315, "optim/lr": 0.0029718, "optim/total_tokens": 2597322752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.1218883991241455, "created_at": "2025-01-15T19:32:30.074107+00:00"} {"global_step": 4955, "acc_step": 0, "speed/wps": 12911.58578059127, "speed/FLOPS": 202794197062316.78, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08028768002986908, "optim/lr": 0.0029724, "optim/total_tokens": 2597847040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.1208863258361816, "created_at": "2025-01-15T19:32:40.230043+00:00"} {"global_step": 4956, "acc_step": 0, "speed/wps": 12910.183224356044, "speed/FLOPS": 202772167989328.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08032286912202835, "optim/lr": 0.002973, "optim/total_tokens": 2598371328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.0469472408294678, "created_at": "2025-01-15T19:32:50.389497+00:00"} {"global_step": 4957, "acc_step": 0, "speed/wps": 12911.851602543979, "speed/FLOPS": 202798372161361.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09347740560770035, "optim/lr": 0.0029736, "optim/total_tokens": 2598895616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.205613136291504, "created_at": "2025-01-15T19:33:00.544576+00:00"} {"global_step": 4958, "acc_step": 0, "speed/wps": 12911.473285900116, "speed/FLOPS": 202792430178608.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09299233555793762, "optim/lr": 0.0029741999999999998, "optim/total_tokens": 2599419904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.1346333026885986, "created_at": "2025-01-15T19:33:10.702073+00:00"} {"global_step": 4959, "acc_step": 0, "speed/wps": 12910.065532698649, "speed/FLOPS": 202770319480123.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10755433887243271, "optim/lr": 0.0029748, "optim/total_tokens": 2599944192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 3.140789747238159, "created_at": "2025-01-15T19:33:20.859396+00:00"} {"global_step": 4960, "acc_step": 0, "speed/wps": 12907.878508529728, "speed/FLOPS": 202735969260264.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1120232343673706, "optim/lr": 0.0029754, "optim/total_tokens": 2600468480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.168315887451172, "created_at": "2025-01-15T19:33:31.018219+00:00"} {"global_step": 4961, "acc_step": 0, "speed/wps": 12908.247609588027, "speed/FLOPS": 202741766499583.2, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08606814593076706, "optim/lr": 0.002976, "optim/total_tokens": 2600992768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 3.133817434310913, "created_at": "2025-01-15T19:33:41.178846+00:00"} {"global_step": 4962, "acc_step": 0, "speed/wps": 12912.644848896658, "speed/FLOPS": 202810831185365.25, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08682315051555634, "optim/lr": 0.0029766, "optim/total_tokens": 2601517056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 3.1634745597839355, "created_at": "2025-01-15T19:33:51.337765+00:00"} {"global_step": 4963, "acc_step": 0, "speed/wps": 12912.25453370458, "speed/FLOPS": 202804700748924.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07305285334587097, "optim/lr": 0.0029771999999999997, "optim/total_tokens": 2602041344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 3.0636448860168457, "created_at": "2025-01-15T19:34:01.500448+00:00"} {"global_step": 4964, "acc_step": 0, "speed/wps": 12912.265214623618, "speed/FLOPS": 202804868507435.0, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0832165777683258, "optim/lr": 0.0029778, "optim/total_tokens": 2602565632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.1443872451782227, "created_at": "2025-01-15T19:34:11.654885+00:00"} {"global_step": 4965, "acc_step": 0, "speed/wps": 12908.964432891256, "speed/FLOPS": 202753025194578.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08164657652378082, "optim/lr": 0.0029784, "optim/total_tokens": 2603089920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.118377208709717, "created_at": "2025-01-15T19:34:21.811940+00:00"} {"global_step": 4966, "acc_step": 0, "speed/wps": 12912.51508043258, "speed/FLOPS": 202808792993316.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09135032445192337, "optim/lr": 0.002979, "optim/total_tokens": 2603614208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335820, "loss/out": 3.237441062927246, "created_at": "2025-01-15T19:34:31.966675+00:00"} {"global_step": 4967, "acc_step": 0, "speed/wps": 12912.662359279415, "speed/FLOPS": 202811106209988.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07354789972305298, "optim/lr": 0.0029796, "optim/total_tokens": 2604138496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 3.192124605178833, "created_at": "2025-01-15T19:34:42.123521+00:00"} {"global_step": 4968, "acc_step": 0, "speed/wps": 12911.470228658889, "speed/FLOPS": 202792382160433.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0907163992524147, "optim/lr": 0.0029802, "optim/total_tokens": 2604662784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.0828399658203125, "created_at": "2025-01-15T19:34:52.278599+00:00"} {"global_step": 4969, "acc_step": 0, "speed/wps": 12903.232321903079, "speed/FLOPS": 202662994514763.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08821850270032883, "optim/lr": 0.0029808, "optim/total_tokens": 2605187072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.1910271644592285, "created_at": "2025-01-15T19:35:02.451267+00:00"} {"global_step": 4970, "acc_step": 0, "speed/wps": 12912.996643725844, "speed/FLOPS": 202816356606574.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08533234149217606, "optim/lr": 0.0029814, "optim/total_tokens": 2605711360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 3.0608150959014893, "created_at": "2025-01-15T19:35:12.606930+00:00"} {"global_step": 4971, "acc_step": 0, "speed/wps": 12911.272288104426, "speed/FLOPS": 202789273232027.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07540588825941086, "optim/lr": 0.002982, "optim/total_tokens": 2606235648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.1392338275909424, "created_at": "2025-01-15T19:35:22.764928+00:00"} {"global_step": 4972, "acc_step": 0, "speed/wps": 12911.556523728163, "speed/FLOPS": 202793737543078.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0955556333065033, "optim/lr": 0.0029826, "optim/total_tokens": 2606759936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371642, "loss/out": 3.1091275215148926, "created_at": "2025-01-15T19:35:32.920132+00:00"} {"global_step": 4973, "acc_step": 0, "speed/wps": 12907.308872607196, "speed/FLOPS": 202727022345339.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08025062829256058, "optim/lr": 0.0029832, "optim/total_tokens": 2607284224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.1684048175811768, "created_at": "2025-01-15T19:35:43.079428+00:00"} {"global_step": 4974, "acc_step": 0, "speed/wps": 12910.087255618218, "speed/FLOPS": 202770660668425.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09948959201574326, "optim/lr": 0.0029838, "optim/total_tokens": 2607808512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 3.165133476257324, "created_at": "2025-01-15T19:35:53.240113+00:00"} {"global_step": 4975, "acc_step": 0, "speed/wps": 12909.979846127822, "speed/FLOPS": 202768973654783.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10819639265537262, "optim/lr": 0.0029844, "optim/total_tokens": 2608332800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.1493639945983887, "created_at": "2025-01-15T19:36:03.398652+00:00"} {"global_step": 4976, "acc_step": 0, "speed/wps": 12904.079355946504, "speed/FLOPS": 202676298348362.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09194062650203705, "optim/lr": 0.0029850000000000002, "optim/total_tokens": 2608857088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424340, "loss/out": 3.208454132080078, "created_at": "2025-01-15T19:36:13.561004+00:00"} {"global_step": 4977, "acc_step": 0, "speed/wps": 12912.055880834516, "speed/FLOPS": 202801580632622.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08142948895692825, "optim/lr": 0.0029856, "optim/total_tokens": 2609381376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.099392890930176, "created_at": "2025-01-15T19:36:23.719362+00:00"} {"global_step": 4978, "acc_step": 0, "speed/wps": 12909.448520043477, "speed/FLOPS": 202760628448509.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11123787611722946, "optim/lr": 0.0029862, "optim/total_tokens": 2609905664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 3.120631694793701, "created_at": "2025-01-15T19:36:33.876192+00:00"} {"global_step": 4979, "acc_step": 0, "speed/wps": 12911.19642132531, "speed/FLOPS": 202788081640009.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09748286753892899, "optim/lr": 0.0029868000000000004, "optim/total_tokens": 2610429952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.1487388610839844, "created_at": "2025-01-15T19:36:44.034487+00:00"} {"global_step": 4980, "acc_step": 0, "speed/wps": 12911.927362116696, "speed/FLOPS": 202799562069555.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.087861567735672, "optim/lr": 0.0029874000000000003, "optim/total_tokens": 2610954240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.091974973678589, "created_at": "2025-01-15T19:36:54.191585+00:00"} {"global_step": 4981, "acc_step": 0, "speed/wps": 12907.580542859441, "speed/FLOPS": 202731289299965.0, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10200003534555435, "optim/lr": 0.002988, "optim/total_tokens": 2611478528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.0792531967163086, "created_at": "2025-01-15T19:37:04.351671+00:00"} {"global_step": 4982, "acc_step": 0, "speed/wps": 12908.368348736156, "speed/FLOPS": 202743662873817.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09518489986658096, "optim/lr": 0.0029886, "optim/total_tokens": 2612002816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.2478106021881104, "created_at": "2025-01-15T19:37:14.511396+00:00"} {"global_step": 4983, "acc_step": 0, "speed/wps": 12909.941867522284, "speed/FLOPS": 202768377148595.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09785433858633041, "optim/lr": 0.0029892, "optim/total_tokens": 2612527104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.2002077102661133, "created_at": "2025-01-15T19:37:24.667890+00:00"} {"global_step": 4984, "acc_step": 0, "speed/wps": 12915.403711909312, "speed/FLOPS": 202854162920054.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09760482609272003, "optim/lr": 0.0029898000000000004, "optim/total_tokens": 2613051392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.233426570892334, "created_at": "2025-01-15T19:37:34.822054+00:00"} {"global_step": 4985, "acc_step": 0, "speed/wps": 12915.237755599168, "speed/FLOPS": 202851556348155.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09118739515542984, "optim/lr": 0.0029904000000000003, "optim/total_tokens": 2613575680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 3.027911424636841, "created_at": "2025-01-15T19:37:44.974449+00:00"} {"global_step": 4986, "acc_step": 0, "speed/wps": 12904.251174669898, "speed/FLOPS": 202678996997516.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.10210234671831131, "optim/lr": 0.002991, "optim/total_tokens": 2614099968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 3.1100406646728516, "created_at": "2025-01-15T19:37:55.140537+00:00"} {"global_step": 4987, "acc_step": 0, "speed/wps": 12903.803392835056, "speed/FLOPS": 202671963968484.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.097396120429039, "optim/lr": 0.0029916, "optim/total_tokens": 2614624256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 3.1509461402893066, "created_at": "2025-01-15T19:38:05.304947+00:00"} {"global_step": 4988, "acc_step": 0, "speed/wps": 12905.236120097656, "speed/FLOPS": 202694466918916.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08261218667030334, "optim/lr": 0.0029922, "optim/total_tokens": 2615148544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474863, "loss/out": 3.1106014251708984, "created_at": "2025-01-15T19:38:15.468285+00:00"} {"global_step": 4989, "acc_step": 0, "speed/wps": 12909.346102806152, "speed/FLOPS": 202759019845061.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08239512890577316, "optim/lr": 0.0029928000000000003, "optim/total_tokens": 2615672832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405144, "loss/out": 3.19307804107666, "created_at": "2025-01-15T19:38:25.632079+00:00"} {"global_step": 4990, "acc_step": 0, "speed/wps": 12915.333189349301, "speed/FLOPS": 202853055266340.94, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0811312347650528, "optim/lr": 0.0029934000000000002, "optim/total_tokens": 2616197120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.433685779571533, "created_at": "2025-01-15T19:38:35.785924+00:00"} {"global_step": 4991, "acc_step": 0, "speed/wps": 12908.660720846472, "speed/FLOPS": 202748254979571.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09619496762752533, "optim/lr": 0.002994, "optim/total_tokens": 2616721408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.070544719696045, "created_at": "2025-01-15T19:38:45.956717+00:00"} {"global_step": 4992, "acc_step": 0, "speed/wps": 12908.765549716754, "speed/FLOPS": 202749901461029.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07869762182235718, "optim/lr": 0.0029946, "optim/total_tokens": 2617245696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 3.2214956283569336, "created_at": "2025-01-15T19:38:56.121276+00:00"} {"global_step": 4993, "acc_step": 0, "speed/wps": 12910.627704130326, "speed/FLOPS": 202779149154962.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08476278930902481, "optim/lr": 0.0029952, "optim/total_tokens": 2617769984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.0808677673339844, "created_at": "2025-01-15T19:39:06.282181+00:00"} {"global_step": 4994, "acc_step": 0, "speed/wps": 12911.753173524789, "speed/FLOPS": 202796826198359.7, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06788159906864166, "optim/lr": 0.0029958000000000003, "optim/total_tokens": 2618294272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 3.004021644592285, "created_at": "2025-01-15T19:39:16.437912+00:00"} {"global_step": 4995, "acc_step": 0, "speed/wps": 12907.642868086019, "speed/FLOPS": 202732268203292.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07454264909029007, "optim/lr": 0.0029964, "optim/total_tokens": 2618818560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.0544114112854004, "created_at": "2025-01-15T19:39:26.598184+00:00"} {"global_step": 4996, "acc_step": 0, "speed/wps": 12907.050726976226, "speed/FLOPS": 202722967813475.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07229875028133392, "optim/lr": 0.002997, "optim/total_tokens": 2619342848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.0823264122009277, "created_at": "2025-01-15T19:39:36.760823+00:00"} {"global_step": 4997, "acc_step": 0, "speed/wps": 12910.787515793576, "speed/FLOPS": 202781659216738.22, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09757252782583237, "optim/lr": 0.0029976, "optim/total_tokens": 2619867136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.1768555641174316, "created_at": "2025-01-15T19:39:46.916577+00:00"} {"global_step": 4998, "acc_step": 0, "speed/wps": 12910.534363953835, "speed/FLOPS": 202777683119228.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11975511163473129, "optim/lr": 0.0029982, "optim/total_tokens": 2620391424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.1062371730804443, "created_at": "2025-01-15T19:39:57.076828+00:00"} {"global_step": 4999, "acc_step": 0, "speed/wps": 12910.146445408584, "speed/FLOPS": 202771590325418.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10561352968215942, "optim/lr": 0.0029988000000000003, "optim/total_tokens": 2620915712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301523, "loss/out": 3.2710208892822266, "created_at": "2025-01-15T19:40:07.233332+00:00"} {"global_step": 5000, "acc_step": 0, "speed/wps": 12913.619057546572, "speed/FLOPS": 202826132470915.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08740748465061188, "optim/lr": 0.0029994, "optim/total_tokens": 2621440000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.143354654312134, "created_at": "2025-01-15T19:40:17.386719+00:00"} {"global_step": 5001, "acc_step": 0, "speed/wps": 5472.902895830969, "speed/FLOPS": 85959460535702.27, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.242, "optim/grad_norm": 0.08718252927064896, "optim/lr": 0.003, "optim/total_tokens": 2621964288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 3.1292707920074463, "created_at": "2025-01-15T19:40:41.339904+00:00"} {"global_step": 5002, "acc_step": 0, "speed/wps": 12942.865064732641, "speed/FLOPS": 203285481201997.66, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07269799709320068, "optim/lr": 0.002999999997552993, "optim/total_tokens": 2622488576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.2038893699645996, "created_at": "2025-01-15T19:40:51.476235+00:00"} {"global_step": 5003, "acc_step": 0, "speed/wps": 12933.405900267431, "speed/FLOPS": 203136912025817.44, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08723381906747818, "optim/lr": 0.0029999999902119725, "optim/total_tokens": 2623012864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.033292531967163, "created_at": "2025-01-15T19:41:01.618974+00:00"} {"global_step": 5004, "acc_step": 0, "speed/wps": 12922.636847711006, "speed/FLOPS": 202967769257189.8, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11316318064928055, "optim/lr": 0.002999999977976938, "optim/total_tokens": 2623537152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.0843708515167236, "created_at": "2025-01-15T19:41:11.767127+00:00"} {"global_step": 5005, "acc_step": 0, "speed/wps": 12921.468234154721, "speed/FLOPS": 202949414575445.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09441002458333969, "optim/lr": 0.0029999999608478897, "optim/total_tokens": 2624061440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.207540512084961, "created_at": "2025-01-15T19:41:21.919679+00:00"} {"global_step": 5006, "acc_step": 0, "speed/wps": 12925.750074698924, "speed/FLOPS": 203016666765057.25, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07703378796577454, "optim/lr": 0.002999999938824828, "optim/total_tokens": 2624585728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 3.347388982772827, "created_at": "2025-01-15T19:41:32.066783+00:00"} {"global_step": 5007, "acc_step": 0, "speed/wps": 12919.196227933191, "speed/FLOPS": 202913729595672.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08172764629125595, "optim/lr": 0.0029999999119077517, "optim/total_tokens": 2625110016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 3.2418417930603027, "created_at": "2025-01-15T19:41:42.218113+00:00"} {"global_step": 5008, "acc_step": 0, "speed/wps": 12915.236784912904, "speed/FLOPS": 202851541102193.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08492960780858994, "optim/lr": 0.0029999998800966632, "optim/total_tokens": 2625634304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350126, "loss/out": 3.08921217918396, "created_at": "2025-01-15T19:41:52.370316+00:00"} {"global_step": 5009, "acc_step": 0, "speed/wps": 12915.650130306823, "speed/FLOPS": 202858033259601.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07634128630161285, "optim/lr": 0.0029999998433915603, "optim/total_tokens": 2626158592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.139566659927368, "created_at": "2025-01-15T19:42:02.525164+00:00"} {"global_step": 5010, "acc_step": 0, "speed/wps": 12914.999804346558, "speed/FLOPS": 202847818996753.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07956798374652863, "optim/lr": 0.0029999998017924447, "optim/total_tokens": 2626682880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 3.081291913986206, "created_at": "2025-01-15T19:42:12.679800+00:00"} {"global_step": 5011, "acc_step": 0, "speed/wps": 12919.189544002267, "speed/FLOPS": 202913624615352.38, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08245090395212173, "optim/lr": 0.002999999755299316, "optim/total_tokens": 2627207168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 3.109532594680786, "created_at": "2025-01-15T19:42:22.832561+00:00"} {"global_step": 5012, "acc_step": 0, "speed/wps": 12913.79853512664, "speed/FLOPS": 202828951412938.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07839009165763855, "optim/lr": 0.002999999703912174, "optim/total_tokens": 2627731456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355503, "loss/out": 3.2109341621398926, "created_at": "2025-01-15T19:42:32.993641+00:00"} {"global_step": 5013, "acc_step": 0, "speed/wps": 12911.894063048738, "speed/FLOPS": 202799039061935.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07562094181776047, "optim/lr": 0.0029999996476310193, "optim/total_tokens": 2628255744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.13930082321167, "created_at": "2025-01-15T19:42:43.149914+00:00"} {"global_step": 5014, "acc_step": 0, "speed/wps": 12916.886854070908, "speed/FLOPS": 202877457705752.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07482349872589111, "optim/lr": 0.002999999586455852, "optim/total_tokens": 2628780032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389901, "loss/out": 3.11510968208313, "created_at": "2025-01-15T19:42:53.300823+00:00"} {"global_step": 5015, "acc_step": 0, "speed/wps": 12918.771904540015, "speed/FLOPS": 202907065013700.94, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07596231997013092, "optim/lr": 0.002999999520386672, "optim/total_tokens": 2629304320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.086069107055664, "created_at": "2025-01-15T19:43:03.450797+00:00"} {"global_step": 5016, "acc_step": 0, "speed/wps": 12911.351622794184, "speed/FLOPS": 202790519292347.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08268126845359802, "optim/lr": 0.0029999994494234795, "optim/total_tokens": 2629828608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477309, "loss/out": 3.174870729446411, "created_at": "2025-01-15T19:43:13.608174+00:00"} {"global_step": 5017, "acc_step": 0, "speed/wps": 12907.710631234668, "speed/FLOPS": 202733332516658.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07931844145059586, "optim/lr": 0.0029999993735662753, "optim/total_tokens": 2630352896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.1468799114227295, "created_at": "2025-01-15T19:43:23.768786+00:00"} {"global_step": 5018, "acc_step": 0, "speed/wps": 12916.068512681279, "speed/FLOPS": 202864604529709.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08045583218336105, "optim/lr": 0.0029999992928150592, "optim/total_tokens": 2630877184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.0483105182647705, "created_at": "2025-01-15T19:43:33.922032+00:00"} {"global_step": 5019, "acc_step": 0, "speed/wps": 12914.662217584688, "speed/FLOPS": 202842516732765.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08768526464700699, "optim/lr": 0.0029999992071698314, "optim/total_tokens": 2631401472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.1819496154785156, "created_at": "2025-01-15T19:43:44.078274+00:00"} {"global_step": 5020, "acc_step": 0, "speed/wps": 12911.646333325714, "speed/FLOPS": 202795148126219.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08354640007019043, "optim/lr": 0.0029999991166305925, "optim/total_tokens": 2631925760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.169146776199341, "created_at": "2025-01-15T19:43:54.233376+00:00"} {"global_step": 5021, "acc_step": 0, "speed/wps": 12908.919327539688, "speed/FLOPS": 202752316753051.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0782880112528801, "optim/lr": 0.002999999021197343, "optim/total_tokens": 2632450048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.1438820362091064, "created_at": "2025-01-15T19:44:04.396209+00:00"} {"global_step": 5022, "acc_step": 0, "speed/wps": 12909.596482079112, "speed/FLOPS": 202762952395600.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10301103442907333, "optim/lr": 0.0029999989208700827, "optim/total_tokens": 2632974336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.232384204864502, "created_at": "2025-01-15T19:44:14.552875+00:00"} {"global_step": 5023, "acc_step": 0, "speed/wps": 12912.227589814302, "speed/FLOPS": 202804277558102.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12456060200929642, "optim/lr": 0.002999998815648812, "optim/total_tokens": 2633498624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 3.0904552936553955, "created_at": "2025-01-15T19:44:24.711061+00:00"} {"global_step": 5024, "acc_step": 0, "speed/wps": 12910.632923621235, "speed/FLOPS": 202779231134239.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10333772003650665, "optim/lr": 0.0029999987055335316, "optim/total_tokens": 2634022912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.070213794708252, "created_at": "2025-01-15T19:44:34.869330+00:00"} {"global_step": 5025, "acc_step": 0, "speed/wps": 12910.543890841014, "speed/FLOPS": 202777832752082.47, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08656082302331924, "optim/lr": 0.0029999985905242413, "optim/total_tokens": 2634547200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 3.1539623737335205, "created_at": "2025-01-15T19:44:45.025587+00:00"} {"global_step": 5026, "acc_step": 0, "speed/wps": 12904.777384810533, "speed/FLOPS": 202687261850864.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09069550782442093, "optim/lr": 0.002999998470620942, "optim/total_tokens": 2635071488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.108940362930298, "created_at": "2025-01-15T19:44:55.187963+00:00"} {"global_step": 5027, "acc_step": 0, "speed/wps": 12911.069877602149, "speed/FLOPS": 202786094096947.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07431396842002869, "optim/lr": 0.002999998345823634, "optim/total_tokens": 2635595776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 3.1576175689697266, "created_at": "2025-01-15T19:45:05.344542+00:00"} {"global_step": 5028, "acc_step": 0, "speed/wps": 12912.268370568472, "speed/FLOPS": 202804918075885.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07194796204566956, "optim/lr": 0.002999998216132317, "optim/total_tokens": 2636120064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 3.105794668197632, "created_at": "2025-01-15T19:45:15.499861+00:00"} {"global_step": 5029, "acc_step": 0, "speed/wps": 12917.664709334893, "speed/FLOPS": 202889674991558.84, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07222796976566315, "optim/lr": 0.0029999980815469925, "optim/total_tokens": 2636644352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372174, "loss/out": 3.0622310638427734, "created_at": "2025-01-15T19:45:25.651292+00:00"} {"global_step": 5030, "acc_step": 0, "speed/wps": 12908.68941520801, "speed/FLOPS": 202748705663949.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08099013566970825, "optim/lr": 0.0029999979420676603, "optim/total_tokens": 2637168640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.250777006149292, "created_at": "2025-01-15T19:45:35.811603+00:00"} {"global_step": 5031, "acc_step": 0, "speed/wps": 12910.691392393372, "speed/FLOPS": 202780149466650.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07298317551612854, "optim/lr": 0.002999997797694321, "optim/total_tokens": 2637692928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.1122665405273438, "created_at": "2025-01-15T19:45:45.968893+00:00"} {"global_step": 5032, "acc_step": 0, "speed/wps": 12910.384849094427, "speed/FLOPS": 202775334782906.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08026675879955292, "optim/lr": 0.002999997648426975, "optim/total_tokens": 2638217216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.185561180114746, "created_at": "2025-01-15T19:45:56.128695+00:00"} {"global_step": 5033, "acc_step": 0, "speed/wps": 12912.153915244186, "speed/FLOPS": 202803120397739.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07838554680347443, "optim/lr": 0.002999997494265623, "optim/total_tokens": 2638741504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 3.2486095428466797, "created_at": "2025-01-15T19:46:06.285533+00:00"} {"global_step": 5034, "acc_step": 0, "speed/wps": 12911.662132874768, "speed/FLOPS": 202795396279848.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07317046076059341, "optim/lr": 0.0029999973352102657, "optim/total_tokens": 2639265792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.1747283935546875, "created_at": "2025-01-15T19:46:16.441192+00:00"} {"global_step": 5035, "acc_step": 0, "speed/wps": 12899.926919228337, "speed/FLOPS": 202611078623657.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07730931043624878, "optim/lr": 0.002999997171260903, "optim/total_tokens": 2639790080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.201848030090332, "created_at": "2025-01-15T19:46:26.609387+00:00"} {"global_step": 5036, "acc_step": 0, "speed/wps": 12909.257184870195, "speed/FLOPS": 202757623266691.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.086253322660923, "optim/lr": 0.0029999970024175353, "optim/total_tokens": 2640314368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.1711981296539307, "created_at": "2025-01-15T19:46:36.768202+00:00"} {"global_step": 5037, "acc_step": 0, "speed/wps": 12913.11253932815, "speed/FLOPS": 202818176906265.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09555291384458542, "optim/lr": 0.002999996828680164, "optim/total_tokens": 2640838656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.1721181869506836, "created_at": "2025-01-15T19:46:46.922083+00:00"} {"global_step": 5038, "acc_step": 0, "speed/wps": 12913.102775240677, "speed/FLOPS": 202818023547854.8, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1107323095202446, "optim/lr": 0.002999996650048789, "optim/total_tokens": 2641362944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.2684807777404785, "created_at": "2025-01-15T19:46:57.075956+00:00"} {"global_step": 5039, "acc_step": 0, "speed/wps": 12910.092803897844, "speed/FLOPS": 202770747811781.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08868364989757538, "optim/lr": 0.002999996466523411, "optim/total_tokens": 2641887232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381819, "loss/out": 3.1687545776367188, "created_at": "2025-01-15T19:47:07.236542+00:00"} {"global_step": 5040, "acc_step": 0, "speed/wps": 12905.314270606186, "speed/FLOPS": 202695694380043.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09048444032669067, "optim/lr": 0.002999996278104031, "optim/total_tokens": 2642411520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0510473251342773, "created_at": "2025-01-15T19:47:17.398555+00:00"} {"global_step": 5041, "acc_step": 0, "speed/wps": 12910.758614310998, "speed/FLOPS": 202781205279239.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08377400040626526, "optim/lr": 0.0029999960847906498, "optim/total_tokens": 2642935808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.1286797523498535, "created_at": "2025-01-15T19:47:27.554204+00:00"} {"global_step": 5042, "acc_step": 0, "speed/wps": 12916.512054692439, "speed/FLOPS": 202871570966485.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06868124008178711, "optim/lr": 0.002999995886583267, "optim/total_tokens": 2643460096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.1960012912750244, "created_at": "2025-01-15T19:47:37.707961+00:00"} {"global_step": 5043, "acc_step": 0, "speed/wps": 12907.484861407196, "speed/FLOPS": 202729786491277.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07146380841732025, "optim/lr": 0.0029999956834818836, "optim/total_tokens": 2643984384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.1625771522521973, "created_at": "2025-01-15T19:47:47.867989+00:00"} {"global_step": 5044, "acc_step": 0, "speed/wps": 12911.380200009426, "speed/FLOPS": 202790968136781.6, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07538361847400665, "optim/lr": 0.0029999954754865, "optim/total_tokens": 2644508672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.1563384532928467, "created_at": "2025-01-15T19:47:58.024644+00:00"} {"global_step": 5045, "acc_step": 0, "speed/wps": 12912.597152892627, "speed/FLOPS": 202810082054080.78, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0874120444059372, "optim/lr": 0.0029999952625971186, "optim/total_tokens": 2645032960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.1718006134033203, "created_at": "2025-01-15T19:48:08.180007+00:00"} {"global_step": 5046, "acc_step": 0, "speed/wps": 12909.948540703552, "speed/FLOPS": 202768481960076.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09364770352840424, "optim/lr": 0.002999995044813738, "optim/total_tokens": 2645557248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.179339647293091, "created_at": "2025-01-15T19:48:18.337834+00:00"} {"global_step": 5047, "acc_step": 0, "speed/wps": 12912.483878812707, "speed/FLOPS": 202808302929000.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0851828008890152, "optim/lr": 0.00299999482213636, "optim/total_tokens": 2646081536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.1605401039123535, "created_at": "2025-01-15T19:48:28.494633+00:00"} {"global_step": 5048, "acc_step": 0, "speed/wps": 12912.766727121621, "speed/FLOPS": 202812745450366.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11227216571569443, "optim/lr": 0.0029999945945649855, "optim/total_tokens": 2646605824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 3.245327949523926, "created_at": "2025-01-15T19:48:38.655576+00:00"} {"global_step": 5049, "acc_step": 0, "speed/wps": 12906.909378513119, "speed/FLOPS": 202720747741629.56, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0965740978717804, "optim/lr": 0.0029999943620996144, "optim/total_tokens": 2647130112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0994513034820557, "created_at": "2025-01-15T19:48:48.816437+00:00"} {"global_step": 5050, "acc_step": 0, "speed/wps": 12912.347512550376, "speed/FLOPS": 202806161109463.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07950489223003387, "optim/lr": 0.002999994124740248, "optim/total_tokens": 2647654400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.1931118965148926, "created_at": "2025-01-15T19:48:58.973218+00:00"} {"global_step": 5051, "acc_step": 0, "speed/wps": 12915.371527991747, "speed/FLOPS": 202853657427403.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09892287105321884, "optim/lr": 0.002999993882486887, "optim/total_tokens": 2648178688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1265244483947754, "created_at": "2025-01-15T19:49:09.125936+00:00"} {"global_step": 5052, "acc_step": 0, "speed/wps": 12914.754222570858, "speed/FLOPS": 202843961797501.75, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08237981051206589, "optim/lr": 0.0029999936353395317, "optim/total_tokens": 2648702976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 3.187757730484009, "created_at": "2025-01-15T19:49:19.279593+00:00"} {"global_step": 5053, "acc_step": 0, "speed/wps": 12907.792453216442, "speed/FLOPS": 202734617643317.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1071913093328476, "optim/lr": 0.0029999933832981834, "optim/total_tokens": 2649227264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 3.2841482162475586, "created_at": "2025-01-15T19:49:29.437749+00:00"} {"global_step": 5054, "acc_step": 0, "speed/wps": 12909.917823994881, "speed/FLOPS": 202767999511957.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10936429351568222, "optim/lr": 0.002999993126362843, "optim/total_tokens": 2649751552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 3.157095193862915, "created_at": "2025-01-15T19:49:39.594089+00:00"} {"global_step": 5055, "acc_step": 0, "speed/wps": 12912.340515841195, "speed/FLOPS": 202806051216531.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08351356536149979, "optim/lr": 0.0029999928645335116, "optim/total_tokens": 2650275840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.1436710357666016, "created_at": "2025-01-15T19:49:49.750817+00:00"} {"global_step": 5056, "acc_step": 0, "speed/wps": 12908.628979004634, "speed/FLOPS": 202747756430327.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0802491307258606, "optim/lr": 0.0029999925978101897, "optim/total_tokens": 2650800128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.206043243408203, "created_at": "2025-01-15T19:49:59.910385+00:00"} {"global_step": 5057, "acc_step": 0, "speed/wps": 12906.83259475848, "speed/FLOPS": 202719541747251.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09071345627307892, "optim/lr": 0.0029999923261928778, "optim/total_tokens": 2651324416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 3.119819164276123, "created_at": "2025-01-15T19:50:10.070052+00:00"} {"global_step": 5058, "acc_step": 0, "speed/wps": 12910.837904787357, "speed/FLOPS": 202782450645127.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07915085554122925, "optim/lr": 0.002999992049681577, "optim/total_tokens": 2651848704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.234992742538452, "created_at": "2025-01-15T19:50:20.228017+00:00"} {"global_step": 5059, "acc_step": 0, "speed/wps": 12913.52714930639, "speed/FLOPS": 202824688925707.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0911569595336914, "optim/lr": 0.0029999917682762885, "optim/total_tokens": 2652372992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 3.251410722732544, "created_at": "2025-01-15T19:50:30.381736+00:00"} {"global_step": 5060, "acc_step": 0, "speed/wps": 12907.426332891084, "speed/FLOPS": 202728867220505.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09033805876970291, "optim/lr": 0.002999991481977013, "optim/total_tokens": 2652897280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 3.2141876220703125, "created_at": "2025-01-15T19:50:40.541176+00:00"} {"global_step": 5061, "acc_step": 0, "speed/wps": 12912.27846628531, "speed/FLOPS": 202805076642991.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10950962454080582, "optim/lr": 0.0029999911907837517, "optim/total_tokens": 2653421568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.2336275577545166, "created_at": "2025-01-15T19:50:50.695675+00:00"} {"global_step": 5062, "acc_step": 0, "speed/wps": 12909.338570571606, "speed/FLOPS": 202758901540967.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11457604914903641, "optim/lr": 0.002999990894696505, "optim/total_tokens": 2653945856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.152388334274292, "created_at": "2025-01-15T19:51:00.857069+00:00"} {"global_step": 5063, "acc_step": 0, "speed/wps": 12912.973353010502, "speed/FLOPS": 202815990793885.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.08349348604679108, "optim/lr": 0.002999990593715274, "optim/total_tokens": 2654470144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299938, "loss/out": 3.137251377105713, "created_at": "2025-01-15T19:51:11.011381+00:00"} {"global_step": 5064, "acc_step": 0, "speed/wps": 12910.164836040398, "speed/FLOPS": 202771879175562.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08813749998807907, "optim/lr": 0.0029999902878400605, "optim/total_tokens": 2654994432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.1422975063323975, "created_at": "2025-01-15T19:51:21.171869+00:00"} {"global_step": 5065, "acc_step": 0, "speed/wps": 12908.653546980018, "speed/FLOPS": 202748142304140.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07579563558101654, "optim/lr": 0.0029999899770708645, "optim/total_tokens": 2655518720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294596, "loss/out": 3.1373214721679688, "created_at": "2025-01-15T19:51:31.330184+00:00"} {"global_step": 5066, "acc_step": 0, "speed/wps": 12906.85295623119, "speed/FLOPS": 202719861552159.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08681363612413406, "optim/lr": 0.0029999896614076875, "optim/total_tokens": 2656043008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.2457175254821777, "created_at": "2025-01-15T19:51:41.490914+00:00"} {"global_step": 5067, "acc_step": 0, "speed/wps": 12908.372840639278, "speed/FLOPS": 202743733425328.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09653623402118683, "optim/lr": 0.00299998934085053, "optim/total_tokens": 2656567296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337692, "loss/out": 3.1509697437286377, "created_at": "2025-01-15T19:51:51.650903+00:00"} {"global_step": 5068, "acc_step": 0, "speed/wps": 12913.931192274573, "speed/FLOPS": 202831034975736.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.1085410863161087, "optim/lr": 0.0029999890153993938, "optim/total_tokens": 2657091584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.190441608428955, "created_at": "2025-01-15T19:52:01.805114+00:00"} {"global_step": 5069, "acc_step": 0, "speed/wps": 12911.422861149398, "speed/FLOPS": 202791638188607.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07733520865440369, "optim/lr": 0.0029999886850542797, "optim/total_tokens": 2657615872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.1903023719787598, "created_at": "2025-01-15T19:52:11.960545+00:00"} {"global_step": 5070, "acc_step": 0, "speed/wps": 12913.956970150619, "speed/FLOPS": 202831439852702.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0750337466597557, "optim/lr": 0.002999988349815189, "optim/total_tokens": 2658140160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.0727453231811523, "created_at": "2025-01-15T19:52:22.117695+00:00"} {"global_step": 5071, "acc_step": 0, "speed/wps": 12915.080263365291, "speed/FLOPS": 202849082716205.3, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07940947264432907, "optim/lr": 0.002999988009682122, "optim/total_tokens": 2658664448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 3.1827099323272705, "created_at": "2025-01-15T19:52:32.270756+00:00"} {"global_step": 5072, "acc_step": 0, "speed/wps": 12913.657676318218, "speed/FLOPS": 202826739031791.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08750128746032715, "optim/lr": 0.0029999876646550805, "optim/total_tokens": 2659188736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482614, "loss/out": 3.134127140045166, "created_at": "2025-01-15T19:52:42.426882+00:00"} {"global_step": 5073, "acc_step": 0, "speed/wps": 12902.375903127495, "speed/FLOPS": 202649543281050.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09065292775630951, "optim/lr": 0.002999987314734065, "optim/total_tokens": 2659713024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1111536026000977, "created_at": "2025-01-15T19:52:52.589159+00:00"} {"global_step": 5074, "acc_step": 0, "speed/wps": 12912.69502254927, "speed/FLOPS": 202811619231525.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09189210087060928, "optim/lr": 0.002999986959919078, "optim/total_tokens": 2660237312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.0727651119232178, "created_at": "2025-01-15T19:53:02.744554+00:00"} {"global_step": 5075, "acc_step": 0, "speed/wps": 12911.69527675378, "speed/FLOPS": 202795916850012.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0893683061003685, "optim/lr": 0.002999986600210119, "optim/total_tokens": 2660761600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.1698384284973145, "created_at": "2025-01-15T19:53:12.902421+00:00"} {"global_step": 5076, "acc_step": 0, "speed/wps": 12909.16915149444, "speed/FLOPS": 202756240581554.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09790613502264023, "optim/lr": 0.00299998623560719, "optim/total_tokens": 2661285888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.2329561710357666, "created_at": "2025-01-15T19:53:23.059579+00:00"} {"global_step": 5077, "acc_step": 0, "speed/wps": 12909.919761400362, "speed/FLOPS": 202768029941573.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07884754985570908, "optim/lr": 0.002999985866110292, "optim/total_tokens": 2661810176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 3.1044223308563232, "created_at": "2025-01-15T19:53:33.219320+00:00"} {"global_step": 5078, "acc_step": 0, "speed/wps": 12910.185197369645, "speed/FLOPS": 202772198978218.12, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08949372917413712, "optim/lr": 0.0029999854917194266, "optim/total_tokens": 2662334464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.1197917461395264, "created_at": "2025-01-15T19:53:43.375478+00:00"} {"global_step": 5079, "acc_step": 0, "speed/wps": 12914.370219407083, "speed/FLOPS": 202837930500139.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07612913101911545, "optim/lr": 0.0029999851124345943, "optim/total_tokens": 2662858752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499296, "loss/out": 3.145968437194824, "created_at": "2025-01-15T19:53:53.529158+00:00"} {"global_step": 5080, "acc_step": 0, "speed/wps": 12911.994274690574, "speed/FLOPS": 202800613023476.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09059303998947144, "optim/lr": 0.0029999847282557975, "optim/total_tokens": 2663383040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.195608615875244, "created_at": "2025-01-15T19:54:03.687536+00:00"} {"global_step": 5081, "acc_step": 0, "speed/wps": 12906.704409624754, "speed/FLOPS": 202717528423581.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08429183065891266, "optim/lr": 0.002999984339183036, "optim/total_tokens": 2663907328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.165335178375244, "created_at": "2025-01-15T19:54:13.848199+00:00"} {"global_step": 5082, "acc_step": 0, "speed/wps": 12910.780211648784, "speed/FLOPS": 202781544495107.3, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09467220306396484, "optim/lr": 0.0029999839452163122, "optim/total_tokens": 2664431616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 3.135718822479248, "created_at": "2025-01-15T19:54:24.003941+00:00"} {"global_step": 5083, "acc_step": 0, "speed/wps": 12909.784265210274, "speed/FLOPS": 202765901787716.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08931455761194229, "optim/lr": 0.0029999835463556264, "optim/total_tokens": 2664955904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0288467407226562, "created_at": "2025-01-15T19:54:34.164205+00:00"} {"global_step": 5084, "acc_step": 0, "speed/wps": 12911.13910400883, "speed/FLOPS": 202787181392791.6, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07499069720506668, "optim/lr": 0.0029999831426009813, "optim/total_tokens": 2665480192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1681246757507324, "created_at": "2025-01-15T19:54:44.319543+00:00"} {"global_step": 5085, "acc_step": 0, "speed/wps": 12912.499643685673, "speed/FLOPS": 202808550537993.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09040677547454834, "optim/lr": 0.002999982733952377, "optim/total_tokens": 2666004480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.181645393371582, "created_at": "2025-01-15T19:54:54.478959+00:00"} {"global_step": 5086, "acc_step": 0, "speed/wps": 12907.424193808289, "speed/FLOPS": 202728833623270.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08279462158679962, "optim/lr": 0.0029999823204098154, "optim/total_tokens": 2666528768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.973311424255371, "created_at": "2025-01-15T19:55:04.638467+00:00"} {"global_step": 5087, "acc_step": 0, "speed/wps": 12917.123749772645, "speed/FLOPS": 202881178478277.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07715348899364471, "optim/lr": 0.0029999819019732974, "optim/total_tokens": 2667053056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.129892349243164, "created_at": "2025-01-15T19:55:14.789339+00:00"} {"global_step": 5088, "acc_step": 0, "speed/wps": 12908.898856534759, "speed/FLOPS": 202751995227788.72, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07491849362850189, "optim/lr": 0.0029999814786428242, "optim/total_tokens": 2667577344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454201, "loss/out": 3.1131837368011475, "created_at": "2025-01-15T19:55:24.946555+00:00"} {"global_step": 5089, "acc_step": 0, "speed/wps": 12915.807756156208, "speed/FLOPS": 202860508990166.53, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07504337280988693, "optim/lr": 0.0029999810504183983, "optim/total_tokens": 2668101632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.0803325176239014, "created_at": "2025-01-15T19:55:35.102657+00:00"} {"global_step": 5090, "acc_step": 0, "speed/wps": 12910.969269312669, "speed/FLOPS": 202784513905511.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08218415826559067, "optim/lr": 0.0029999806173000203, "optim/total_tokens": 2668625920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.136357545852661, "created_at": "2025-01-15T19:55:45.258711+00:00"} {"global_step": 5091, "acc_step": 0, "speed/wps": 12911.01333010682, "speed/FLOPS": 202785205940829.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0875791683793068, "optim/lr": 0.0029999801792876916, "optim/total_tokens": 2669150208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.127514600753784, "created_at": "2025-01-15T19:55:55.415233+00:00"} {"global_step": 5092, "acc_step": 0, "speed/wps": 12907.007224704712, "speed/FLOPS": 202722284550523.28, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07942792773246765, "optim/lr": 0.0029999797363814136, "optim/total_tokens": 2669674496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.113874673843384, "created_at": "2025-01-15T19:56:05.574233+00:00"} {"global_step": 5093, "acc_step": 0, "speed/wps": 12910.267610899615, "speed/FLOPS": 202773493395954.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07775868475437164, "optim/lr": 0.002999979288581188, "optim/total_tokens": 2670198784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.119680881500244, "created_at": "2025-01-15T19:56:15.735808+00:00"} {"global_step": 5094, "acc_step": 0, "speed/wps": 12911.372503124532, "speed/FLOPS": 202790847246627.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07022319734096527, "optim/lr": 0.0029999788358870162, "optim/total_tokens": 2670723072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.110830307006836, "created_at": "2025-01-15T19:56:25.891086+00:00"} {"global_step": 5095, "acc_step": 0, "speed/wps": 12909.267121944094, "speed/FLOPS": 202757779342092.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08127656579017639, "optim/lr": 0.0029999783782989, "optim/total_tokens": 2671247360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.04766845703125, "created_at": "2025-01-15T19:56:36.048781+00:00"} {"global_step": 5096, "acc_step": 0, "speed/wps": 12912.196506864242, "speed/FLOPS": 202803789357659.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09217551350593567, "optim/lr": 0.0029999779158168404, "optim/total_tokens": 2671771648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.154686450958252, "created_at": "2025-01-15T19:56:46.207095+00:00"} {"global_step": 5097, "acc_step": 0, "speed/wps": 12917.546167482293, "speed/FLOPS": 202887813128870.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10436282306909561, "optim/lr": 0.0029999774484408383, "optim/total_tokens": 2672295936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 3.2356085777282715, "created_at": "2025-01-15T19:56:56.357374+00:00"} {"global_step": 5098, "acc_step": 0, "speed/wps": 12910.1231649737, "speed/FLOPS": 202771224674198.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09948541969060898, "optim/lr": 0.0029999769761708964, "optim/total_tokens": 2672820224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.2165591716766357, "created_at": "2025-01-15T19:57:06.514069+00:00"} {"global_step": 5099, "acc_step": 0, "speed/wps": 12903.852682631843, "speed/FLOPS": 202672738132473.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09112534672021866, "optim/lr": 0.002999976499007016, "optim/total_tokens": 2673344512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.174591064453125, "created_at": "2025-01-15T19:57:16.676695+00:00"} {"global_step": 5100, "acc_step": 0, "speed/wps": 12912.185739579487, "speed/FLOPS": 202803620242656.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07536740601062775, "optim/lr": 0.0029999760169491982, "optim/total_tokens": 2673868800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0075252056121826, "created_at": "2025-01-15T19:57:26.834037+00:00"} {"global_step": 5101, "acc_step": 0, "speed/wps": 12910.514890964385, "speed/FLOPS": 202777377269170.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07538463175296783, "optim/lr": 0.0029999755299974454, "optim/total_tokens": 2674393088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.1215481758117676, "created_at": "2025-01-15T19:57:36.989970+00:00"} {"global_step": 5102, "acc_step": 0, "speed/wps": 12909.604858931765, "speed/FLOPS": 202763083965582.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07192099094390869, "optim/lr": 0.0029999750381517584, "optim/total_tokens": 2674917376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.134162664413452, "created_at": "2025-01-15T19:57:47.150640+00:00"} {"global_step": 5103, "acc_step": 0, "speed/wps": 12910.864279100655, "speed/FLOPS": 202782864889959.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.06662942469120026, "optim/lr": 0.0029999745414121385, "optim/total_tokens": 2675441664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 3.194953441619873, "created_at": "2025-01-15T19:57:57.306151+00:00"} {"global_step": 5104, "acc_step": 0, "speed/wps": 12915.335162752088, "speed/FLOPS": 202853086261343.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07912999391555786, "optim/lr": 0.0029999740397785883, "optim/total_tokens": 2675965952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.0834951400756836, "created_at": "2025-01-15T19:58:07.458209+00:00"} {"global_step": 5105, "acc_step": 0, "speed/wps": 12911.27520311407, "speed/FLOPS": 202789319016259.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07372475415468216, "optim/lr": 0.002999973533251109, "optim/total_tokens": 2676490240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 3.1179656982421875, "created_at": "2025-01-15T19:58:17.613578+00:00"} {"global_step": 5106, "acc_step": 0, "speed/wps": 12911.037042518898, "speed/FLOPS": 202785578376845.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06918821483850479, "optim/lr": 0.002999973021829702, "optim/total_tokens": 2677014528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.3267323970794678, "created_at": "2025-01-15T19:58:27.777738+00:00"} {"global_step": 5107, "acc_step": 0, "speed/wps": 12910.600868883808, "speed/FLOPS": 202778727670540.6, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0747627392411232, "optim/lr": 0.002999972505514369, "optim/total_tokens": 2677538816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 3.1605377197265625, "created_at": "2025-01-15T19:58:37.939857+00:00"} {"global_step": 5108, "acc_step": 0, "speed/wps": 12904.951321336079, "speed/FLOPS": 202689993763011.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07096688449382782, "optim/lr": 0.002999971984305112, "optim/total_tokens": 2678063104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.2458603382110596, "created_at": "2025-01-15T19:58:48.100973+00:00"} {"global_step": 5109, "acc_step": 0, "speed/wps": 12903.378775073914, "speed/FLOPS": 202665294763059.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07222889363765717, "optim/lr": 0.002999971458201933, "optim/total_tokens": 2678587392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.1685824394226074, "created_at": "2025-01-15T19:58:58.264286+00:00"} {"global_step": 5110, "acc_step": 0, "speed/wps": 12902.320310611283, "speed/FLOPS": 202648670124191.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08861729502677917, "optim/lr": 0.0029999709272048336, "optim/total_tokens": 2679111680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.0664772987365723, "created_at": "2025-01-15T19:59:08.427689+00:00"} {"global_step": 5111, "acc_step": 0, "speed/wps": 12911.29277374457, "speed/FLOPS": 202789594987155.75, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09193255007266998, "optim/lr": 0.002999970391313814, "optim/total_tokens": 2679635968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.1880805492401123, "created_at": "2025-01-15T19:59:18.582950+00:00"} {"global_step": 5112, "acc_step": 0, "speed/wps": 12908.69346688345, "speed/FLOPS": 202748769301079.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07720933854579926, "optim/lr": 0.0029999698505288786, "optim/total_tokens": 2680160256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.0502281188964844, "created_at": "2025-01-15T19:59:28.745656+00:00"} {"global_step": 5113, "acc_step": 0, "speed/wps": 12911.463039824217, "speed/FLOPS": 202792269249906.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08710270375013351, "optim/lr": 0.0029999693048500266, "optim/total_tokens": 2680684544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.1616623401641846, "created_at": "2025-01-15T19:59:38.900929+00:00"} {"global_step": 5114, "acc_step": 0, "speed/wps": 12908.697555265715, "speed/FLOPS": 202748833514741.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08398321270942688, "optim/lr": 0.0029999687542772617, "optim/total_tokens": 2681208832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 3.1736912727355957, "created_at": "2025-01-15T19:59:49.059254+00:00"} {"global_step": 5115, "acc_step": 0, "speed/wps": 12913.758096632959, "speed/FLOPS": 202828316270826.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07488042116165161, "optim/lr": 0.0029999681988105847, "optim/total_tokens": 2681733120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.146958351135254, "created_at": "2025-01-15T19:59:59.212627+00:00"} {"global_step": 5116, "acc_step": 0, "speed/wps": 12908.707182451215, "speed/FLOPS": 202748984722918.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08290743082761765, "optim/lr": 0.0029999676384499974, "optim/total_tokens": 2682257408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294366, "loss/out": 3.195176601409912, "created_at": "2025-01-15T20:00:09.369969+00:00"} {"global_step": 5117, "acc_step": 0, "speed/wps": 12910.352183299145, "speed/FLOPS": 202774821721703.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07726436108350754, "optim/lr": 0.0029999670731955023, "optim/total_tokens": 2682781696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.1651082038879395, "created_at": "2025-01-15T20:00:19.526280+00:00"} {"global_step": 5118, "acc_step": 0, "speed/wps": 12915.247097424855, "speed/FLOPS": 202851703074364.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07400576025247574, "optim/lr": 0.0029999665030471002, "optim/total_tokens": 2683305984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 3.08827543258667, "created_at": "2025-01-15T20:00:29.678358+00:00"} {"global_step": 5119, "acc_step": 0, "speed/wps": 12906.452397897896, "speed/FLOPS": 202713570233110.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07285056263208389, "optim/lr": 0.0029999659280047944, "optim/total_tokens": 2683830272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.2208709716796875, "created_at": "2025-01-15T20:00:39.839531+00:00"} {"global_step": 5120, "acc_step": 0, "speed/wps": 12913.497716575113, "speed/FLOPS": 202824226644218.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11356157809495926, "optim/lr": 0.0029999653480685855, "optim/total_tokens": 2684354560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.1290810108184814, "created_at": "2025-01-15T20:00:49.993252+00:00"} {"global_step": 5121, "acc_step": 0, "speed/wps": 12908.605861513195, "speed/FLOPS": 202747393338357.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11442127078771591, "optim/lr": 0.002999964763238476, "optim/total_tokens": 2684878848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0746829509735107, "created_at": "2025-01-15T20:01:00.153323+00:00"} {"global_step": 5122, "acc_step": 0, "speed/wps": 12911.434061877475, "speed/FLOPS": 202791814111433.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08040317893028259, "optim/lr": 0.002999964173514467, "optim/total_tokens": 2685403136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354398, "loss/out": 3.276113510131836, "created_at": "2025-01-15T20:01:10.308632+00:00"} {"global_step": 5123, "acc_step": 0, "speed/wps": 12915.041359939238, "speed/FLOPS": 202848471684440.25, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09439948946237564, "optim/lr": 0.002999963578896562, "optim/total_tokens": 2685927424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.093446731567383, "created_at": "2025-01-15T20:01:20.464784+00:00"} {"global_step": 5124, "acc_step": 0, "speed/wps": 12916.068042092707, "speed/FLOPS": 202864597138469.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10942323505878448, "optim/lr": 0.002999962979384761, "optim/total_tokens": 2686451712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412087, "loss/out": 3.1514246463775635, "created_at": "2025-01-15T20:01:30.616962+00:00"} {"global_step": 5125, "acc_step": 0, "speed/wps": 12910.943050914006, "speed/FLOPS": 202784102109532.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09756191819906235, "optim/lr": 0.0029999623749790674, "optim/total_tokens": 2686976000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.165752410888672, "created_at": "2025-01-15T20:01:40.772892+00:00"} {"global_step": 5126, "acc_step": 0, "speed/wps": 12904.88155304714, "speed/FLOPS": 202688897956162.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11716647446155548, "optim/lr": 0.0029999617656794836, "optim/total_tokens": 2687500288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 3.191831111907959, "created_at": "2025-01-15T20:01:50.935992+00:00"} {"global_step": 5127, "acc_step": 0, "speed/wps": 12903.097438061895, "speed/FLOPS": 202660875978688.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.09226015210151672, "optim/lr": 0.0029999611514860097, "optim/total_tokens": 2688024576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.1398348808288574, "created_at": "2025-01-15T20:02:01.097760+00:00"} {"global_step": 5128, "acc_step": 0, "speed/wps": 12912.714998524538, "speed/FLOPS": 202811932981666.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07760732620954514, "optim/lr": 0.002999960532398649, "optim/total_tokens": 2688548864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.0977892875671387, "created_at": "2025-01-15T20:02:11.259640+00:00"} {"global_step": 5129, "acc_step": 0, "speed/wps": 12911.778391879241, "speed/FLOPS": 202797222287270.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08844923228025436, "optim/lr": 0.0029999599084174033, "optim/total_tokens": 2689073152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.3402485847473145, "created_at": "2025-01-15T20:02:21.414585+00:00"} {"global_step": 5130, "acc_step": 0, "speed/wps": 12907.70278719443, "speed/FLOPS": 202733209315227.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0722527876496315, "optim/lr": 0.0029999592795422747, "optim/total_tokens": 2689597440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.974865436553955, "created_at": "2025-01-15T20:02:31.575487+00:00"} {"global_step": 5131, "acc_step": 0, "speed/wps": 12915.753541314947, "speed/FLOPS": 202859657471585.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07169824838638306, "optim/lr": 0.002999958645773265, "optim/total_tokens": 2690121728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 3.0874826908111572, "created_at": "2025-01-15T20:02:41.730067+00:00"} {"global_step": 5132, "acc_step": 0, "speed/wps": 12909.031216481579, "speed/FLOPS": 202754074122635.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07682102173566818, "optim/lr": 0.0029999580071103766, "optim/total_tokens": 2690646016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.077232599258423, "created_at": "2025-01-15T20:02:51.887155+00:00"} {"global_step": 5133, "acc_step": 0, "speed/wps": 12912.163989458433, "speed/FLOPS": 202803278627117.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08538106828927994, "optim/lr": 0.0029999573635536114, "optim/total_tokens": 2691170304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 3.240955114364624, "created_at": "2025-01-15T20:03:02.043225+00:00"} {"global_step": 5134, "acc_step": 0, "speed/wps": 12915.918351623903, "speed/FLOPS": 202862246043960.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0743025541305542, "optim/lr": 0.002999956715102971, "optim/total_tokens": 2691694592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.035717487335205, "created_at": "2025-01-15T20:03:12.200502+00:00"} {"global_step": 5135, "acc_step": 0, "speed/wps": 12904.656866654152, "speed/FLOPS": 202685368947608.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07422348111867905, "optim/lr": 0.0029999560617584584, "optim/total_tokens": 2692218880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371444, "loss/out": 3.1965510845184326, "created_at": "2025-01-15T20:03:22.363081+00:00"} {"global_step": 5136, "acc_step": 0, "speed/wps": 12912.590989958177, "speed/FLOPS": 202809985256726.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0817226991057396, "optim/lr": 0.0029999554035200753, "optim/total_tokens": 2692743168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.139277696609497, "created_at": "2025-01-15T20:03:32.519911+00:00"} {"global_step": 5137, "acc_step": 0, "speed/wps": 12916.950481993053, "speed/FLOPS": 202878457069704.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09070312976837158, "optim/lr": 0.002999954740387824, "optim/total_tokens": 2693267456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468042, "loss/out": 3.0898962020874023, "created_at": "2025-01-15T20:03:42.672279+00:00"} {"global_step": 5138, "acc_step": 0, "speed/wps": 12911.178646044726, "speed/FLOPS": 202787802454801.8, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08542120456695557, "optim/lr": 0.002999954072361706, "optim/total_tokens": 2693791744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.0768795013427734, "created_at": "2025-01-15T20:03:52.829131+00:00"} {"global_step": 5139, "acc_step": 0, "speed/wps": 12905.669502867511, "speed/FLOPS": 202701273790847.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.071153424680233, "optim/lr": 0.0029999533994417246, "optim/total_tokens": 2694316032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 3.1786954402923584, "created_at": "2025-01-15T20:04:02.993976+00:00"} {"global_step": 5140, "acc_step": 0, "speed/wps": 12915.436137777378, "speed/FLOPS": 202854672212869.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07068872451782227, "optim/lr": 0.0029999527216278813, "optim/total_tokens": 2694840320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.131420612335205, "created_at": "2025-01-15T20:04:13.150013+00:00"} {"global_step": 5141, "acc_step": 0, "speed/wps": 12908.467096737066, "speed/FLOPS": 202745213846865.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08656005561351776, "optim/lr": 0.002999952038920178, "optim/total_tokens": 2695364608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.2328896522521973, "created_at": "2025-01-15T20:04:23.309302+00:00"} {"global_step": 5142, "acc_step": 0, "speed/wps": 12907.497588310445, "speed/FLOPS": 202729986384781.53, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07313135266304016, "optim/lr": 0.002999951351318618, "optim/total_tokens": 2695888896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.24450945854187, "created_at": "2025-01-15T20:04:33.467802+00:00"} {"global_step": 5143, "acc_step": 0, "speed/wps": 12900.617150682656, "speed/FLOPS": 202621919657125.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07747454941272736, "optim/lr": 0.0029999506588232025, "optim/total_tokens": 2696413184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.180108070373535, "created_at": "2025-01-15T20:04:43.634097+00:00"} {"global_step": 5144, "acc_step": 0, "speed/wps": 12912.390401619292, "speed/FLOPS": 202806834741227.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07555708289146423, "optim/lr": 0.0029999499614339347, "optim/total_tokens": 2696937472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 3.110912561416626, "created_at": "2025-01-15T20:04:53.789488+00:00"} {"global_step": 5145, "acc_step": 0, "speed/wps": 12909.624838373158, "speed/FLOPS": 202763397770164.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09308114647865295, "optim/lr": 0.002999949259150816, "optim/total_tokens": 2697461760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.167897939682007, "created_at": "2025-01-15T20:05:03.955115+00:00"} {"global_step": 5146, "acc_step": 0, "speed/wps": 12916.440887008404, "speed/FLOPS": 202870453180213.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0812038779258728, "optim/lr": 0.002999948551973849, "optim/total_tokens": 2697986048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375623, "loss/out": 3.1265952587127686, "created_at": "2025-01-15T20:05:14.110714+00:00"} {"global_step": 5147, "acc_step": 0, "speed/wps": 12907.277331548048, "speed/FLOPS": 202726526949663.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07629916071891785, "optim/lr": 0.002999947839903036, "optim/total_tokens": 2698510336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.2739462852478027, "created_at": "2025-01-15T20:05:24.271946+00:00"} {"global_step": 5148, "acc_step": 0, "speed/wps": 12909.634431416394, "speed/FLOPS": 202763548442090.6, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09060537070035934, "optim/lr": 0.0029999471229383793, "optim/total_tokens": 2699034624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.1321041584014893, "created_at": "2025-01-15T20:05:34.428848+00:00"} {"global_step": 5149, "acc_step": 0, "speed/wps": 12909.695464394212, "speed/FLOPS": 202764507048875.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08087626099586487, "optim/lr": 0.002999946401079882, "optim/total_tokens": 2699558912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.3004395961761475, "created_at": "2025-01-15T20:05:44.589214+00:00"} {"global_step": 5150, "acc_step": 0, "speed/wps": 12912.038116218202, "speed/FLOPS": 202801301614912.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07325904071331024, "optim/lr": 0.0029999456743275447, "optim/total_tokens": 2700083200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.1013827323913574, "created_at": "2025-01-15T20:05:54.746670+00:00"} {"global_step": 5151, "acc_step": 0, "speed/wps": 12917.103602476003, "speed/FLOPS": 202880862037298.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.09005347639322281, "optim/lr": 0.0029999449426813714, "optim/total_tokens": 2700607488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 3.1821842193603516, "created_at": "2025-01-15T20:06:04.898856+00:00"} {"global_step": 5152, "acc_step": 0, "speed/wps": 12907.958514209808, "speed/FLOPS": 202737225859409.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08057976514101028, "optim/lr": 0.002999944206141364, "optim/total_tokens": 2701131776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499296, "loss/out": 3.1067309379577637, "created_at": "2025-01-15T20:06:15.056812+00:00"} {"global_step": 5153, "acc_step": 0, "speed/wps": 12914.158999552783, "speed/FLOPS": 202834613002080.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07785660028457642, "optim/lr": 0.0029999434647075243, "optim/total_tokens": 2701656064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.1653809547424316, "created_at": "2025-01-15T20:06:25.212519+00:00"} {"global_step": 5154, "acc_step": 0, "speed/wps": 12910.492728400437, "speed/FLOPS": 202777029175648.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07853370904922485, "optim/lr": 0.0029999427183798554, "optim/total_tokens": 2702180352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.2180943489074707, "created_at": "2025-01-15T20:06:35.369278+00:00"} {"global_step": 5155, "acc_step": 0, "speed/wps": 12908.875487059824, "speed/FLOPS": 202751628178071.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08872419595718384, "optim/lr": 0.0029999419671583607, "optim/total_tokens": 2702704640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 3.149254083633423, "created_at": "2025-01-15T20:06:45.526369+00:00"} {"global_step": 5156, "acc_step": 0, "speed/wps": 12906.766120359332, "speed/FLOPS": 202718497675467.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07993818819522858, "optim/lr": 0.00299994121104304, "optim/total_tokens": 2703228928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.1388349533081055, "created_at": "2025-01-15T20:06:55.686399+00:00"} {"global_step": 5157, "acc_step": 0, "speed/wps": 12915.959436620515, "speed/FLOPS": 202862891340289.25, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06958911567926407, "optim/lr": 0.002999940450033898, "optim/total_tokens": 2703753216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390425, "loss/out": 3.185941457748413, "created_at": "2025-01-15T20:07:05.841068+00:00"} {"global_step": 5158, "acc_step": 0, "speed/wps": 12913.856930668559, "speed/FLOPS": 202829868595167.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08180759847164154, "optim/lr": 0.002999939684130937, "optim/total_tokens": 2704277504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.1542656421661377, "created_at": "2025-01-15T20:07:15.997581+00:00"} {"global_step": 5159, "acc_step": 0, "speed/wps": 12902.923754974167, "speed/FLOPS": 202658148047129.62, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06595651060342789, "optim/lr": 0.002999938913334158, "optim/total_tokens": 2704801792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 3.1409058570861816, "created_at": "2025-01-15T20:07:26.160793+00:00"} {"global_step": 5160, "acc_step": 0, "speed/wps": 12908.896402007726, "speed/FLOPS": 202751956676068.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06464313715696335, "optim/lr": 0.002999938137643565, "optim/total_tokens": 2705326080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.201712131500244, "created_at": "2025-01-15T20:07:36.318493+00:00"} {"global_step": 5161, "acc_step": 0, "speed/wps": 12913.013074524191, "speed/FLOPS": 202816614674840.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.080654576420784, "optim/lr": 0.00299993735705916, "optim/total_tokens": 2705850368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 3.0754613876342773, "created_at": "2025-01-15T20:07:46.473863+00:00"} {"global_step": 5162, "acc_step": 0, "speed/wps": 12907.41806067002, "speed/FLOPS": 202728737293906.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07247794419527054, "optim/lr": 0.002999936571580945, "optim/total_tokens": 2706374656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.049964189529419, "created_at": "2025-01-15T20:07:56.632226+00:00"} {"global_step": 5163, "acc_step": 0, "speed/wps": 12907.335591183268, "speed/FLOPS": 202727441997291.84, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07138627022504807, "optim/lr": 0.002999935781208924, "optim/total_tokens": 2706898944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.17071270942688, "created_at": "2025-01-15T20:08:06.799643+00:00"} {"global_step": 5164, "acc_step": 0, "speed/wps": 12912.30870969769, "speed/FLOPS": 202805551657341.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08978396654129028, "optim/lr": 0.002999934985943098, "optim/total_tokens": 2707423232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 3.129396438598633, "created_at": "2025-01-15T20:08:16.955004+00:00"} {"global_step": 5165, "acc_step": 0, "speed/wps": 12911.439481018435, "speed/FLOPS": 202791899226488.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07582676410675049, "optim/lr": 0.0029999341857834703, "optim/total_tokens": 2707947520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1219372749328613, "created_at": "2025-01-15T20:08:27.112930+00:00"} {"global_step": 5166, "acc_step": 0, "speed/wps": 12911.713020289175, "speed/FLOPS": 202796195536618.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08615937829017639, "optim/lr": 0.002999933380730044, "optim/total_tokens": 2708471808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.145402669906616, "created_at": "2025-01-15T20:08:37.267984+00:00"} {"global_step": 5167, "acc_step": 0, "speed/wps": 12911.567877493255, "speed/FLOPS": 202793915869561.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0937662273645401, "optim/lr": 0.002999932570782821, "optim/total_tokens": 2708996096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.197495937347412, "created_at": "2025-01-15T20:08:47.427259+00:00"} {"global_step": 5168, "acc_step": 0, "speed/wps": 12911.257940495894, "speed/FLOPS": 202789047883119.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09119579941034317, "optim/lr": 0.0029999317559418043, "optim/total_tokens": 2709520384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 3.183426856994629, "created_at": "2025-01-15T20:08:57.582892+00:00"} {"global_step": 5169, "acc_step": 0, "speed/wps": 12908.110861594576, "speed/FLOPS": 202739618684434.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0883466899394989, "optim/lr": 0.0029999309362069965, "optim/total_tokens": 2710044672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.0157101154327393, "created_at": "2025-01-15T20:09:07.746490+00:00"} {"global_step": 5170, "acc_step": 0, "speed/wps": 12911.439533136972, "speed/FLOPS": 202791900045081.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09871551394462585, "optim/lr": 0.0029999301115783998, "optim/total_tokens": 2710568960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 3.243677854537964, "created_at": "2025-01-15T20:09:17.901760+00:00"} {"global_step": 5171, "acc_step": 0, "speed/wps": 12908.938551801248, "speed/FLOPS": 202752618696495.94, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10205560177564621, "optim/lr": 0.002999929282056018, "optim/total_tokens": 2711093248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.12467622756958, "created_at": "2025-01-15T20:09:28.067849+00:00"} {"global_step": 5172, "acc_step": 0, "speed/wps": 12911.351749535264, "speed/FLOPS": 202790521282989.84, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09791892766952515, "optim/lr": 0.0029999284476398525, "optim/total_tokens": 2711617536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 3.0869698524475098, "created_at": "2025-01-15T20:09:38.223982+00:00"} {"global_step": 5173, "acc_step": 0, "speed/wps": 12910.87275590458, "speed/FLOPS": 202782998029813.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08019018918275833, "optim/lr": 0.002999927608329907, "optim/total_tokens": 2712141824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467434, "loss/out": 3.0870063304901123, "created_at": "2025-01-15T20:09:48.383338+00:00"} {"global_step": 5174, "acc_step": 0, "speed/wps": 12913.411235597856, "speed/FLOPS": 202822868341633.25, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08091484010219574, "optim/lr": 0.0029999267641261838, "optim/total_tokens": 2712666112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 3.29162335395813, "created_at": "2025-01-15T20:09:58.537448+00:00"} {"global_step": 5175, "acc_step": 0, "speed/wps": 12916.042921958642, "speed/FLOPS": 202864202592245.44, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07075046747922897, "optim/lr": 0.0029999259150286853, "optim/total_tokens": 2713190400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.2417123317718506, "created_at": "2025-01-15T20:10:08.690536+00:00"} {"global_step": 5176, "acc_step": 0, "speed/wps": 12910.49807924389, "speed/FLOPS": 202777113217997.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07263896614313126, "optim/lr": 0.002999925061037415, "optim/total_tokens": 2713714688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.0005714893341064, "created_at": "2025-01-15T20:10:18.847610+00:00"} {"global_step": 5177, "acc_step": 0, "speed/wps": 12901.705340235521, "speed/FLOPS": 202639011169381.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07003223150968552, "optim/lr": 0.0029999242021523757, "optim/total_tokens": 2714238976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.1247050762176514, "created_at": "2025-01-15T20:10:29.011190+00:00"} {"global_step": 5178, "acc_step": 0, "speed/wps": 12910.436197803823, "speed/FLOPS": 202776141284947.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06894460320472717, "optim/lr": 0.0029999233383735694, "optim/total_tokens": 2714763264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 3.0458076000213623, "created_at": "2025-01-15T20:10:39.168816+00:00"} {"global_step": 5179, "acc_step": 0, "speed/wps": 12907.898377264162, "speed/FLOPS": 202736281326041.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06917302310466766, "optim/lr": 0.0029999224697009994, "optim/total_tokens": 2715287552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.142514944076538, "created_at": "2025-01-15T20:10:49.328554+00:00"} {"global_step": 5180, "acc_step": 0, "speed/wps": 12917.682877147086, "speed/FLOPS": 202889960342014.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08327141404151917, "optim/lr": 0.0029999215961346694, "optim/total_tokens": 2715811840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432861, "loss/out": 3.317603826522827, "created_at": "2025-01-15T20:10:59.478991+00:00"} {"global_step": 5181, "acc_step": 0, "speed/wps": 12913.656074307635, "speed/FLOPS": 202826713870013.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07178357988595963, "optim/lr": 0.0029999207176745805, "optim/total_tokens": 2716336128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.2278053760528564, "created_at": "2025-01-15T20:11:09.633134+00:00"} {"global_step": 5182, "acc_step": 0, "speed/wps": 12909.136686041473, "speed/FLOPS": 202755730667003.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08176225423812866, "optim/lr": 0.002999919834320737, "optim/total_tokens": 2716860416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.064587116241455, "created_at": "2025-01-15T20:11:19.790131+00:00"} {"global_step": 5183, "acc_step": 0, "speed/wps": 12910.343039227339, "speed/FLOPS": 202774678101490.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08305229246616364, "optim/lr": 0.0029999189460731407, "optim/total_tokens": 2717384704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.1238317489624023, "created_at": "2025-01-15T20:11:29.946224+00:00"} {"global_step": 5184, "acc_step": 0, "speed/wps": 12914.967244376045, "speed/FLOPS": 202847307597674.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07103557139635086, "optim/lr": 0.0029999180529317955, "optim/total_tokens": 2717908992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.180509090423584, "created_at": "2025-01-15T20:11:40.101486+00:00"} {"global_step": 5185, "acc_step": 0, "speed/wps": 12915.045615897447, "speed/FLOPS": 202848538530112.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06828999519348145, "optim/lr": 0.002999917154896703, "optim/total_tokens": 2718433280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.117664098739624, "created_at": "2025-01-15T20:11:50.255779+00:00"} {"global_step": 5186, "acc_step": 0, "speed/wps": 12907.546656552027, "speed/FLOPS": 202730757068942.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06478668004274368, "optim/lr": 0.002999916251967868, "optim/total_tokens": 2718957568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.206084966659546, "created_at": "2025-01-15T20:12:00.417122+00:00"} {"global_step": 5187, "acc_step": 0, "speed/wps": 12914.779572347208, "speed/FLOPS": 202844359950573.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08123939484357834, "optim/lr": 0.0029999153441452923, "optim/total_tokens": 2719481856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.096679210662842, "created_at": "2025-01-15T20:12:10.573565+00:00"} {"global_step": 5188, "acc_step": 0, "speed/wps": 12911.99297872301, "speed/FLOPS": 202800592668524.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0776289775967598, "optim/lr": 0.0029999144314289787, "optim/total_tokens": 2720006144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.2012505531311035, "created_at": "2025-01-15T20:12:20.729815+00:00"} {"global_step": 5189, "acc_step": 0, "speed/wps": 12913.411107631466, "speed/FLOPS": 202822866331745.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07246677577495575, "optim/lr": 0.0029999135138189304, "optim/total_tokens": 2720530432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.1494150161743164, "created_at": "2025-01-15T20:12:30.886221+00:00"} {"global_step": 5190, "acc_step": 0, "speed/wps": 12912.046630102303, "speed/FLOPS": 202801435337161.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08145340532064438, "optim/lr": 0.0029999125913151505, "optim/total_tokens": 2721054720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.2364041805267334, "created_at": "2025-01-15T20:12:41.041988+00:00"} {"global_step": 5191, "acc_step": 0, "speed/wps": 12916.938951601096, "speed/FLOPS": 202878275969053.94, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07826780527830124, "optim/lr": 0.002999911663917642, "optim/total_tokens": 2721579008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.0639121532440186, "created_at": "2025-01-15T20:12:51.192997+00:00"} {"global_step": 5192, "acc_step": 0, "speed/wps": 12913.75182950079, "speed/FLOPS": 202828217836903.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08196128904819489, "optim/lr": 0.002999910731626408, "optim/total_tokens": 2722103296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480757, "loss/out": 3.0173912048339844, "created_at": "2025-01-15T20:13:01.348562+00:00"} {"global_step": 5193, "acc_step": 0, "speed/wps": 12909.322511464827, "speed/FLOPS": 202758649310628.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08537758141756058, "optim/lr": 0.0029999097944414512, "optim/total_tokens": 2722627584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.12198805809021, "created_at": "2025-01-15T20:13:11.506884+00:00"} {"global_step": 5194, "acc_step": 0, "speed/wps": 12910.674769712701, "speed/FLOPS": 202779888384608.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09830616414546967, "optim/lr": 0.0029999088523627758, "optim/total_tokens": 2723151872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.1969213485717773, "created_at": "2025-01-15T20:13:21.663648+00:00"} {"global_step": 5195, "acc_step": 0, "speed/wps": 12912.82960241391, "speed/FLOPS": 202813732993231.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08679431676864624, "optim/lr": 0.002999907905390383, "optim/total_tokens": 2723676160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.076930046081543, "created_at": "2025-01-15T20:13:31.820555+00:00"} {"global_step": 5196, "acc_step": 0, "speed/wps": 12882.582680687987, "speed/FLOPS": 202338663523900.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07901064306497574, "optim/lr": 0.002999906953524277, "optim/total_tokens": 2724200448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1072616577148438, "created_at": "2025-01-15T20:13:42.000285+00:00"} {"global_step": 5197, "acc_step": 0, "speed/wps": 12904.062694679149, "speed/FLOPS": 202676036660263.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0724789947271347, "optim/lr": 0.002999905996764461, "optim/total_tokens": 2724724736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.075437307357788, "created_at": "2025-01-15T20:13:52.163777+00:00"} {"global_step": 5198, "acc_step": 0, "speed/wps": 12908.691725207344, "speed/FLOPS": 202748741945663.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.08772002905607224, "optim/lr": 0.002999905035110938, "optim/total_tokens": 2725249024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.145139694213867, "created_at": "2025-01-15T20:14:02.325564+00:00"} {"global_step": 5199, "acc_step": 0, "speed/wps": 12906.046723200547, "speed/FLOPS": 202707198554533.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.0805853009223938, "optim/lr": 0.0029999040685637106, "optim/total_tokens": 2725773312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.047954797744751, "created_at": "2025-01-15T20:14:12.485265+00:00"} {"global_step": 5200, "acc_step": 0, "speed/wps": 12909.480443538801, "speed/FLOPS": 202761129850871.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07939494401216507, "optim/lr": 0.002999903097122782, "optim/total_tokens": 2726297600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.1257071495056152, "created_at": "2025-01-15T20:14:22.649370+00:00"} {"global_step": 5201, "acc_step": 0, "speed/wps": 12908.211730792082, "speed/FLOPS": 202741202973790.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0798245519399643, "optim/lr": 0.002999902120788157, "optim/total_tokens": 2726821888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.160653591156006, "created_at": "2025-01-15T20:14:32.811005+00:00"} {"global_step": 5202, "acc_step": 0, "speed/wps": 12908.880516871268, "speed/FLOPS": 202751707178172.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06530766934156418, "optim/lr": 0.002999901139559836, "optim/total_tokens": 2727346176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 3.227107048034668, "created_at": "2025-01-15T20:14:42.973750+00:00"} {"global_step": 5203, "acc_step": 0, "speed/wps": 12911.64799288067, "speed/FLOPS": 202795174191810.44, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07504371553659439, "optim/lr": 0.002999900153437825, "optim/total_tokens": 2727870464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.1918845176696777, "created_at": "2025-01-15T20:14:53.129967+00:00"} {"global_step": 5204, "acc_step": 0, "speed/wps": 12910.933698731791, "speed/FLOPS": 202783955220659.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07623200118541718, "optim/lr": 0.0029998991624221255, "optim/total_tokens": 2728394752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456346, "loss/out": 3.15260648727417, "created_at": "2025-01-15T20:15:03.285494+00:00"} {"global_step": 5205, "acc_step": 0, "speed/wps": 12910.63760423024, "speed/FLOPS": 202779304649635.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07140704989433289, "optim/lr": 0.002999898166512741, "optim/total_tokens": 2728919040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450742, "loss/out": 3.1134119033813477, "created_at": "2025-01-15T20:15:13.447057+00:00"} {"global_step": 5206, "acc_step": 0, "speed/wps": 12905.649590105802, "speed/FLOPS": 202700961033561.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07152468711137772, "optim/lr": 0.002999897165709675, "optim/total_tokens": 2729443328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.1658775806427, "created_at": "2025-01-15T20:15:23.607445+00:00"} {"global_step": 5207, "acc_step": 0, "speed/wps": 12905.131288191436, "speed/FLOPS": 202692820389775.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06673819571733475, "optim/lr": 0.002999896160012931, "optim/total_tokens": 2729967616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.157022714614868, "created_at": "2025-01-15T20:15:33.769475+00:00"} {"global_step": 5208, "acc_step": 0, "speed/wps": 12906.930597215996, "speed/FLOPS": 202721081010515.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07637815177440643, "optim/lr": 0.002999895149422512, "optim/total_tokens": 2730491904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.2247729301452637, "created_at": "2025-01-15T20:15:43.929916+00:00"} {"global_step": 5209, "acc_step": 0, "speed/wps": 12911.860637401456, "speed/FLOPS": 202798514066213.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07362707704305649, "optim/lr": 0.0029998941339384215, "optim/total_tokens": 2731016192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 3.200535297393799, "created_at": "2025-01-15T20:15:54.085920+00:00"} {"global_step": 5210, "acc_step": 0, "speed/wps": 12906.596372281481, "speed/FLOPS": 202715831548646.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07812918722629547, "optim/lr": 0.002999893113560662, "optim/total_tokens": 2731540480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.1125383377075195, "created_at": "2025-01-15T20:16:04.246737+00:00"} {"global_step": 5211, "acc_step": 0, "speed/wps": 12900.744428534774, "speed/FLOPS": 202623918730690.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07679621130228043, "optim/lr": 0.002999892088289237, "optim/total_tokens": 2732064768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 3.11181902885437, "created_at": "2025-01-15T20:16:14.410416+00:00"} {"global_step": 5212, "acc_step": 0, "speed/wps": 12907.589598968232, "speed/FLOPS": 202731431538597.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07188218086957932, "optim/lr": 0.0029998910581241515, "optim/total_tokens": 2732589056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.1855969429016113, "created_at": "2025-01-15T20:16:24.568833+00:00"} {"global_step": 5213, "acc_step": 0, "speed/wps": 12906.417756464627, "speed/FLOPS": 202713026141798.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08281245827674866, "optim/lr": 0.0029998900230654066, "optim/total_tokens": 2733113344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 3.236206531524658, "created_at": "2025-01-15T20:16:34.736206+00:00"} {"global_step": 5214, "acc_step": 0, "speed/wps": 12912.146364332384, "speed/FLOPS": 202803001800293.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08377619087696075, "optim/lr": 0.0029998889831130073, "optim/total_tokens": 2733637632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.1450552940368652, "created_at": "2025-01-15T20:16:44.891424+00:00"} {"global_step": 5215, "acc_step": 0, "speed/wps": 12906.394684767238, "speed/FLOPS": 202712663769087.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08399289101362228, "optim/lr": 0.002999887938266956, "optim/total_tokens": 2734161920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.2291347980499268, "created_at": "2025-01-15T20:16:55.054641+00:00"} {"global_step": 5216, "acc_step": 0, "speed/wps": 12909.79429546405, "speed/FLOPS": 202766059326635.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07589930295944214, "optim/lr": 0.0029998868885272568, "optim/total_tokens": 2734686208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.3489975929260254, "created_at": "2025-01-15T20:17:05.214358+00:00"} {"global_step": 5217, "acc_step": 0, "speed/wps": 12914.643972940967, "speed/FLOPS": 202842230175565.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08601081371307373, "optim/lr": 0.002999885833893913, "optim/total_tokens": 2735210496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.110076904296875, "created_at": "2025-01-15T20:17:15.367783+00:00"} {"global_step": 5218, "acc_step": 0, "speed/wps": 12909.082868205009, "speed/FLOPS": 202754885383928.5, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08823061734437943, "optim/lr": 0.0029998847743669278, "optim/total_tokens": 2735734784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.1028127670288086, "created_at": "2025-01-15T20:17:25.524775+00:00"} {"global_step": 5219, "acc_step": 0, "speed/wps": 12913.694812978212, "speed/FLOPS": 202827322314067.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08507279306650162, "optim/lr": 0.002999883709946304, "optim/total_tokens": 2736259072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.115556240081787, "created_at": "2025-01-15T20:17:35.679541+00:00"} {"global_step": 5220, "acc_step": 0, "speed/wps": 12914.618667545474, "speed/FLOPS": 202841832719555.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08703483641147614, "optim/lr": 0.0029998826406320463, "optim/total_tokens": 2736783360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.174783229827881, "created_at": "2025-01-15T20:17:45.837463+00:00"} {"global_step": 5221, "acc_step": 0, "speed/wps": 12911.121028034806, "speed/FLOPS": 202786897484780.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09811906516551971, "optim/lr": 0.002999881566424158, "optim/total_tokens": 2737307648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.2029411792755127, "created_at": "2025-01-15T20:17:55.995259+00:00"} {"global_step": 5222, "acc_step": 0, "speed/wps": 12911.583461264274, "speed/FLOPS": 202794160634099.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0887250155210495, "optim/lr": 0.002999880487322642, "optim/total_tokens": 2737831936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 3.17456316947937, "created_at": "2025-01-15T20:18:06.150478+00:00"} {"global_step": 5223, "acc_step": 0, "speed/wps": 12911.21054019437, "speed/FLOPS": 202788303396249.12, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09615468233823776, "optim/lr": 0.002999879403327502, "optim/total_tokens": 2738356224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.0886759757995605, "created_at": "2025-01-15T20:18:16.310064+00:00"} {"global_step": 5224, "acc_step": 0, "speed/wps": 12912.2782080281, "speed/FLOPS": 202805072586707.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07542425394058228, "optim/lr": 0.0029998783144387416, "optim/total_tokens": 2738880512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 3.1767983436584473, "created_at": "2025-01-15T20:18:26.468700+00:00"} {"global_step": 5225, "acc_step": 0, "speed/wps": 12910.589668416926, "speed/FLOPS": 202778551751817.12, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0862271711230278, "optim/lr": 0.002999877220656365, "optim/total_tokens": 2739404800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 3.0138254165649414, "created_at": "2025-01-15T20:18:36.624910+00:00"} {"global_step": 5226, "acc_step": 0, "speed/wps": 12911.234365855928, "speed/FLOPS": 202788677611003.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07516440749168396, "optim/lr": 0.0029998761219803744, "optim/total_tokens": 2739929088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.0965828895568848, "created_at": "2025-01-15T20:18:46.780170+00:00"} {"global_step": 5227, "acc_step": 0, "speed/wps": 12911.601552747667, "speed/FLOPS": 202794444785706.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08285347372293472, "optim/lr": 0.002999875018410775, "optim/total_tokens": 2740453376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.19195556640625, "created_at": "2025-01-15T20:18:56.940280+00:00"} {"global_step": 5228, "acc_step": 0, "speed/wps": 12914.960814888665, "speed/FLOPS": 202847206613740.5, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08597120642662048, "optim/lr": 0.002999873909947569, "optim/total_tokens": 2740977664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 3.148466110229492, "created_at": "2025-01-15T20:19:07.092696+00:00"} {"global_step": 5229, "acc_step": 0, "speed/wps": 12908.781608621874, "speed/FLOPS": 202750153688201.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07174891233444214, "optim/lr": 0.0029998727965907606, "optim/total_tokens": 2741501952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0963597297668457, "created_at": "2025-01-15T20:19:17.251921+00:00"} {"global_step": 5230, "acc_step": 0, "speed/wps": 12918.25545322329, "speed/FLOPS": 202898953436093.28, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12290480732917786, "optim/lr": 0.0029998716783403534, "optim/total_tokens": 2742026240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 3.0332322120666504, "created_at": "2025-01-15T20:19:27.402069+00:00"} {"global_step": 5231, "acc_step": 0, "speed/wps": 12910.639346431473, "speed/FLOPS": 202779332013300.34, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10357014834880829, "optim/lr": 0.002999870555196351, "optim/total_tokens": 2742550528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.0819859504699707, "created_at": "2025-01-15T20:19:37.557805+00:00"} {"global_step": 5232, "acc_step": 0, "speed/wps": 12911.739445560665, "speed/FLOPS": 202796610581818.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0860542580485344, "optim/lr": 0.0029998694271587576, "optim/total_tokens": 2743074816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.1453797817230225, "created_at": "2025-01-15T20:19:47.712935+00:00"} {"global_step": 5233, "acc_step": 0, "speed/wps": 12909.1192704737, "speed/FLOPS": 202755457131580.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10079707205295563, "optim/lr": 0.002999868294227576, "optim/total_tokens": 2743599104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.0386202335357666, "created_at": "2025-01-15T20:19:57.873096+00:00"} {"global_step": 5234, "acc_step": 0, "speed/wps": 12911.848681351123, "speed/FLOPS": 202798326280013.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07919226586818695, "optim/lr": 0.0029998671564028102, "optim/total_tokens": 2744123392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.020972728729248, "created_at": "2025-01-15T20:20:08.027951+00:00"} {"global_step": 5235, "acc_step": 0, "speed/wps": 12908.575050554604, "speed/FLOPS": 202746909409914.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08527779579162598, "optim/lr": 0.002999866013684464, "optim/total_tokens": 2744647680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.159133195877075, "created_at": "2025-01-15T20:20:18.187278+00:00"} {"global_step": 5236, "acc_step": 0, "speed/wps": 12909.770022750374, "speed/FLOPS": 202765678090312.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09502138197422028, "optim/lr": 0.0029998648660725415, "optim/total_tokens": 2745171968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304284, "loss/out": 3.19920015335083, "created_at": "2025-01-15T20:20:28.345946+00:00"} {"global_step": 5237, "acc_step": 0, "speed/wps": 12909.003611124495, "speed/FLOPS": 202753640542568.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08086689561605453, "optim/lr": 0.002999863713567046, "optim/total_tokens": 2745696256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.0427823066711426, "created_at": "2025-01-15T20:20:38.506723+00:00"} {"global_step": 5238, "acc_step": 0, "speed/wps": 12913.810923739202, "speed/FLOPS": 202829145993122.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08183358609676361, "optim/lr": 0.0029998625561679813, "optim/total_tokens": 2746220544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 3.075643539428711, "created_at": "2025-01-15T20:20:48.659995+00:00"} {"global_step": 5239, "acc_step": 0, "speed/wps": 12908.918039293609, "speed/FLOPS": 202752296519376.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08298411220312119, "optim/lr": 0.0029998613938753507, "optim/total_tokens": 2746744832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.0539278984069824, "created_at": "2025-01-15T20:20:58.818834+00:00"} {"global_step": 5240, "acc_step": 0, "speed/wps": 12911.297387316323, "speed/FLOPS": 202789667449639.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07951710373163223, "optim/lr": 0.0029998602266891592, "optim/total_tokens": 2747269120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 3.130669593811035, "created_at": "2025-01-15T20:21:08.974427+00:00"} {"global_step": 5241, "acc_step": 0, "speed/wps": 12908.06333996262, "speed/FLOPS": 202738872291903.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07200706750154495, "optim/lr": 0.0029998590546094096, "optim/total_tokens": 2747793408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.1162822246551514, "created_at": "2025-01-15T20:21:19.136106+00:00"} {"global_step": 5242, "acc_step": 0, "speed/wps": 12912.42636406789, "speed/FLOPS": 202807399580899.8, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0869981124997139, "optim/lr": 0.002999857877636106, "optim/total_tokens": 2748317696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369865, "loss/out": 3.1223292350769043, "created_at": "2025-01-15T20:21:29.290531+00:00"} {"global_step": 5243, "acc_step": 0, "speed/wps": 12908.530357580617, "speed/FLOPS": 202746207445343.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08933036029338837, "optim/lr": 0.0029998566957692525, "optim/total_tokens": 2748841984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.1338515281677246, "created_at": "2025-01-15T20:21:39.450911+00:00"} {"global_step": 5244, "acc_step": 0, "speed/wps": 12911.535391682293, "speed/FLOPS": 202793405635259.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07383079826831818, "optim/lr": 0.0029998555090088525, "optim/total_tokens": 2749366272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 3.165872573852539, "created_at": "2025-01-15T20:21:49.605941+00:00"} {"global_step": 5245, "acc_step": 0, "speed/wps": 12908.34268196224, "speed/FLOPS": 202743259741863.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07212970405817032, "optim/lr": 0.0029998543173549103, "optim/total_tokens": 2749890560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.1342878341674805, "created_at": "2025-01-15T20:21:59.768080+00:00"} {"global_step": 5246, "acc_step": 0, "speed/wps": 12912.281855615987, "speed/FLOPS": 202805129877087.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06598305702209473, "optim/lr": 0.0029998531208074294, "optim/total_tokens": 2750414848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 3.0844950675964355, "created_at": "2025-01-15T20:22:09.923117+00:00"} {"global_step": 5247, "acc_step": 0, "speed/wps": 12908.821743627142, "speed/FLOPS": 202750784063610.34, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0730658695101738, "optim/lr": 0.002999851919366414, "optim/total_tokens": 2750939136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.115739107131958, "created_at": "2025-01-15T20:22:20.081656+00:00"} {"global_step": 5248, "acc_step": 0, "speed/wps": 12913.793231314696, "speed/FLOPS": 202828868109283.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07524044066667557, "optim/lr": 0.0029998507130318684, "optim/total_tokens": 2751463424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.1133008003234863, "created_at": "2025-01-15T20:22:30.239873+00:00"} {"global_step": 5249, "acc_step": 0, "speed/wps": 12905.721492199233, "speed/FLOPS": 202702090354743.12, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08281723409891129, "optim/lr": 0.0029998495018037953, "optim/total_tokens": 2751987712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.039534091949463, "created_at": "2025-01-15T20:22:40.403272+00:00"} {"global_step": 5250, "acc_step": 0, "speed/wps": 12904.891391087514, "speed/FLOPS": 202689052476105.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10513082146644592, "optim/lr": 0.0029998482856821995, "optim/total_tokens": 2752512000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.105259418487549, "created_at": "2025-01-15T20:22:50.566775+00:00"} {"global_step": 5251, "acc_step": 0, "speed/wps": 12909.969421232963, "speed/FLOPS": 202768809917484.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07872620970010757, "optim/lr": 0.0029998470646670854, "optim/total_tokens": 2753036288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.1729583740234375, "created_at": "2025-01-15T20:23:00.724154+00:00"} {"global_step": 5252, "acc_step": 0, "speed/wps": 12909.159560326776, "speed/FLOPS": 202756089939086.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10628620535135269, "optim/lr": 0.0029998458387584565, "optim/total_tokens": 2753560576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.0969114303588867, "created_at": "2025-01-15T20:23:10.887072+00:00"} {"global_step": 5253, "acc_step": 0, "speed/wps": 12914.898481784836, "speed/FLOPS": 202846227586690.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10994663089513779, "optim/lr": 0.002999844607956316, "optim/total_tokens": 2754084864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 3.1824586391448975, "created_at": "2025-01-15T20:23:21.041754+00:00"} {"global_step": 5254, "acc_step": 0, "speed/wps": 12903.193670821009, "speed/FLOPS": 202662387446407.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07399725914001465, "optim/lr": 0.0029998433722606696, "optim/total_tokens": 2754609152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 3.1808886528015137, "created_at": "2025-01-15T20:23:31.205521+00:00"} {"global_step": 5255, "acc_step": 0, "speed/wps": 12907.812727745402, "speed/FLOPS": 202734936082654.5, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08609339594841003, "optim/lr": 0.0029998421316715196, "optim/total_tokens": 2755133440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491917, "loss/out": 3.0526692867279053, "created_at": "2025-01-15T20:23:41.363673+00:00"} {"global_step": 5256, "acc_step": 0, "speed/wps": 12906.566281085732, "speed/FLOPS": 202715358925067.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.07985734939575195, "optim/lr": 0.002999840886188871, "optim/total_tokens": 2755657728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0380728244781494, "created_at": "2025-01-15T20:23:51.524371+00:00"} {"global_step": 5257, "acc_step": 0, "speed/wps": 12908.494769653911, "speed/FLOPS": 202745648488051.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07894714921712875, "optim/lr": 0.0029998396358127282, "optim/total_tokens": 2756182016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.1833510398864746, "created_at": "2025-01-15T20:24:01.682973+00:00"} {"global_step": 5258, "acc_step": 0, "speed/wps": 12910.106896667958, "speed/FLOPS": 202770969158101.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06918293237686157, "optim/lr": 0.0029998383805430948, "optim/total_tokens": 2756706304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.085000514984131, "created_at": "2025-01-15T20:24:11.841303+00:00"} {"global_step": 5259, "acc_step": 0, "speed/wps": 12907.514814831315, "speed/FLOPS": 202730256950963.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0756373330950737, "optim/lr": 0.0029998371203799746, "optim/total_tokens": 2757230592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.110185384750366, "created_at": "2025-01-15T20:24:22.000860+00:00"} {"global_step": 5260, "acc_step": 0, "speed/wps": 12913.528196753889, "speed/FLOPS": 202824705377310.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07924803346395493, "optim/lr": 0.0029998358553233718, "optim/total_tokens": 2757754880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.1968231201171875, "created_at": "2025-01-15T20:24:32.155447+00:00"} {"global_step": 5261, "acc_step": 0, "speed/wps": 12911.662362678062, "speed/FLOPS": 202795399889224.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10353703051805496, "optim/lr": 0.002999834585373291, "optim/total_tokens": 2758279168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0272390842437744, "created_at": "2025-01-15T20:24:42.310403+00:00"} {"global_step": 5262, "acc_step": 0, "speed/wps": 12905.622087968084, "speed/FLOPS": 202700529074697.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08538254350423813, "optim/lr": 0.0029998333105297364, "optim/total_tokens": 2758803456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 3.066819667816162, "created_at": "2025-01-15T20:24:52.470011+00:00"} {"global_step": 5263, "acc_step": 0, "speed/wps": 12908.504195261414, "speed/FLOPS": 202745796530169.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07559536397457123, "optim/lr": 0.0029998320307927113, "optim/total_tokens": 2759327744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.175483465194702, "created_at": "2025-01-15T20:25:02.629902+00:00"} {"global_step": 5264, "acc_step": 0, "speed/wps": 12910.809427093143, "speed/FLOPS": 202782003363807.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07710415869951248, "optim/lr": 0.0029998307461622203, "optim/total_tokens": 2759852032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 3.1777005195617676, "created_at": "2025-01-15T20:25:12.787183+00:00"} {"global_step": 5265, "acc_step": 0, "speed/wps": 12909.67326911783, "speed/FLOPS": 202764158441560.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07438184320926666, "optim/lr": 0.002999829456638268, "optim/total_tokens": 2760376320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.1547045707702637, "created_at": "2025-01-15T20:25:22.949637+00:00"} {"global_step": 5266, "acc_step": 0, "speed/wps": 12912.093752231287, "speed/FLOPS": 202802175454949.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0636405348777771, "optim/lr": 0.0029998281622208582, "optim/total_tokens": 2760900608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 3.1648337841033936, "created_at": "2025-01-15T20:25:33.111959+00:00"} {"global_step": 5267, "acc_step": 0, "speed/wps": 12910.733847636546, "speed/FLOPS": 202780816284584.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07272512465715408, "optim/lr": 0.0029998268629099953, "optim/total_tokens": 2761424896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.086348533630371, "created_at": "2025-01-15T20:25:43.270212+00:00"} {"global_step": 5268, "acc_step": 0, "speed/wps": 12910.228169187018, "speed/FLOPS": 202772873909659.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08012450486421585, "optim/lr": 0.0029998255587056835, "optim/total_tokens": 2761949184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.1007604598999023, "created_at": "2025-01-15T20:25:53.429477+00:00"} {"global_step": 5269, "acc_step": 0, "speed/wps": 12909.98037785357, "speed/FLOPS": 202768982006266.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06773442775011063, "optim/lr": 0.0029998242496079275, "optim/total_tokens": 2762473472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.1338422298431396, "created_at": "2025-01-15T20:26:03.585740+00:00"} {"global_step": 5270, "acc_step": 0, "speed/wps": 12912.532935214813, "speed/FLOPS": 202809073427206.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09266320616006851, "optim/lr": 0.0029998229356167304, "optim/total_tokens": 2762997760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.1773407459259033, "created_at": "2025-01-15T20:26:13.741117+00:00"} {"global_step": 5271, "acc_step": 0, "speed/wps": 12911.554647427121, "speed/FLOPS": 202793708073192.38, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0819452553987503, "optim/lr": 0.0029998216167320974, "optim/total_tokens": 2763522048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.0663273334503174, "created_at": "2025-01-15T20:26:23.897275+00:00"} {"global_step": 5272, "acc_step": 0, "speed/wps": 12906.007914417443, "speed/FLOPS": 202706589009265.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07895174622535706, "optim/lr": 0.002999820292954033, "optim/total_tokens": 2764046336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.044301986694336, "created_at": "2025-01-15T20:26:34.060164+00:00"} {"global_step": 5273, "acc_step": 0, "speed/wps": 12908.927807710093, "speed/FLOPS": 202752449945780.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08108772337436676, "optim/lr": 0.0029998189642825406, "optim/total_tokens": 2764570624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.1124908924102783, "created_at": "2025-01-15T20:26:44.218835+00:00"} {"global_step": 5274, "acc_step": 0, "speed/wps": 12908.396838157807, "speed/FLOPS": 202744110339333.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07203438133001328, "optim/lr": 0.0029998176307176254, "optim/total_tokens": 2765094912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.1085615158081055, "created_at": "2025-01-15T20:26:54.379440+00:00"} {"global_step": 5275, "acc_step": 0, "speed/wps": 12912.070227847267, "speed/FLOPS": 202801805972172.5, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09662138670682907, "optim/lr": 0.002999816292259291, "optim/total_tokens": 2765619200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 3.1444640159606934, "created_at": "2025-01-15T20:27:04.537902+00:00"} {"global_step": 5276, "acc_step": 0, "speed/wps": 12912.063422163475, "speed/FLOPS": 202801699079556.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0809110552072525, "optim/lr": 0.0029998149489075423, "optim/total_tokens": 2766143488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 3.0804126262664795, "created_at": "2025-01-15T20:27:14.693868+00:00"} {"global_step": 5277, "acc_step": 0, "speed/wps": 12910.723692747139, "speed/FLOPS": 202780656788091.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06471005827188492, "optim/lr": 0.0029998136006623836, "optim/total_tokens": 2766667776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 3.1544957160949707, "created_at": "2025-01-15T20:27:24.850871+00:00"} {"global_step": 5278, "acc_step": 0, "speed/wps": 12909.185267056397, "speed/FLOPS": 202756493698600.3, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07183745503425598, "optim/lr": 0.0029998122475238197, "optim/total_tokens": 2767192064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.133904457092285, "created_at": "2025-01-15T20:27:35.011308+00:00"} {"global_step": 5279, "acc_step": 0, "speed/wps": 12911.548785174415, "speed/FLOPS": 202793615998458.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07244188338518143, "optim/lr": 0.0029998108894918538, "optim/total_tokens": 2767716352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.9775466918945312, "created_at": "2025-01-15T20:27:45.167669+00:00"} {"global_step": 5280, "acc_step": 0, "speed/wps": 12909.679090589329, "speed/FLOPS": 202764249875769.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06907155364751816, "optim/lr": 0.002999809526566491, "optim/total_tokens": 2768240640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 3.0920310020446777, "created_at": "2025-01-15T20:27:55.324661+00:00"} {"global_step": 5281, "acc_step": 0, "speed/wps": 12906.611826874938, "speed/FLOPS": 202716074284273.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06704925000667572, "optim/lr": 0.0029998081587477364, "optim/total_tokens": 2768764928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447797, "loss/out": 3.0938544273376465, "created_at": "2025-01-15T20:28:05.488012+00:00"} {"global_step": 5282, "acc_step": 0, "speed/wps": 12907.852080117844, "speed/FLOPS": 202735554165739.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07291604578495026, "optim/lr": 0.0029998067860355936, "optim/total_tokens": 2769289216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.211822509765625, "created_at": "2025-01-15T20:28:15.646333+00:00"} {"global_step": 5283, "acc_step": 0, "speed/wps": 12905.99070020318, "speed/FLOPS": 202706318636375.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08955623209476471, "optim/lr": 0.0029998054084300675, "optim/total_tokens": 2769813504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.1643118858337402, "created_at": "2025-01-15T20:28:25.809730+00:00"} {"global_step": 5284, "acc_step": 0, "speed/wps": 12919.468460497385, "speed/FLOPS": 202918005382179.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09753701835870743, "optim/lr": 0.0029998040259311626, "optim/total_tokens": 2770337792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.0698657035827637, "created_at": "2025-01-15T20:28:35.967187+00:00"} {"global_step": 5285, "acc_step": 0, "speed/wps": 12913.796570488017, "speed/FLOPS": 202828920555588.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08806537091732025, "optim/lr": 0.002999802638538883, "optim/total_tokens": 2770862080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.1849565505981445, "created_at": "2025-01-15T20:28:46.122177+00:00"} {"global_step": 5286, "acc_step": 0, "speed/wps": 12915.93662238987, "speed/FLOPS": 202862533011446.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08632297813892365, "optim/lr": 0.002999801246253234, "optim/total_tokens": 2771386368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.9957003593444824, "created_at": "2025-01-15T20:28:56.274824+00:00"} {"global_step": 5287, "acc_step": 0, "speed/wps": 12908.597670659527, "speed/FLOPS": 202747264689745.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09635356068611145, "optim/lr": 0.0029997998490742194, "optim/total_tokens": 2771910656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.16099214553833, "created_at": "2025-01-15T20:29:06.432366+00:00"} {"global_step": 5288, "acc_step": 0, "speed/wps": 12907.985156155653, "speed/FLOPS": 202737644307778.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0940924808382988, "optim/lr": 0.0029997984470018437, "optim/total_tokens": 2772434944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 3.045821189880371, "created_at": "2025-01-15T20:29:16.591549+00:00"} {"global_step": 5289, "acc_step": 0, "speed/wps": 12899.720670447752, "speed/FLOPS": 202607839203142.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08522889763116837, "optim/lr": 0.0029997970400361122, "optim/total_tokens": 2772959232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.2064342498779297, "created_at": "2025-01-15T20:29:26.757521+00:00"} {"global_step": 5290, "acc_step": 0, "speed/wps": 12895.474384820554, "speed/FLOPS": 202541145452359.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06953651458024979, "optim/lr": 0.0029997956281770286, "optim/total_tokens": 2773483520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.137312173843384, "created_at": "2025-01-15T20:29:36.925595+00:00"} {"global_step": 5291, "acc_step": 0, "speed/wps": 12901.170773415522, "speed/FLOPS": 202630615062902.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07957502454519272, "optim/lr": 0.0029997942114245984, "optim/total_tokens": 2774007808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 3.015610694885254, "created_at": "2025-01-15T20:29:47.094603+00:00"} {"global_step": 5292, "acc_step": 0, "speed/wps": 12901.2380889982, "speed/FLOPS": 202631672346629.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07521042227745056, "optim/lr": 0.002999792789778826, "optim/total_tokens": 2774532096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.0912046432495117, "created_at": "2025-01-15T20:29:57.259450+00:00"} {"global_step": 5293, "acc_step": 0, "speed/wps": 12907.656351760614, "speed/FLOPS": 202732479982930.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0729469507932663, "optim/lr": 0.0029997913632397156, "optim/total_tokens": 2775056384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.0963449478149414, "created_at": "2025-01-15T20:30:07.420066+00:00"} {"global_step": 5294, "acc_step": 0, "speed/wps": 12908.764145467274, "speed/FLOPS": 202749879405361.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06304866820573807, "optim/lr": 0.0029997899318072716, "optim/total_tokens": 2775580672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.1552584171295166, "created_at": "2025-01-15T20:30:17.581616+00:00"} {"global_step": 5295, "acc_step": 0, "speed/wps": 12908.406886383025, "speed/FLOPS": 202744268160517.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07428691536188126, "optim/lr": 0.0029997884954814998, "optim/total_tokens": 2776104960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294022, "loss/out": 3.123112678527832, "created_at": "2025-01-15T20:30:27.739187+00:00"} {"global_step": 5296, "acc_step": 0, "speed/wps": 12908.38196179324, "speed/FLOPS": 202743876685585.34, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0874112993478775, "optim/lr": 0.002999787054262404, "optim/total_tokens": 2776629248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.1264243125915527, "created_at": "2025-01-15T20:30:37.903590+00:00"} {"global_step": 5297, "acc_step": 0, "speed/wps": 12904.690812336334, "speed/FLOPS": 202685902111193.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08444786071777344, "optim/lr": 0.002999785608149989, "optim/total_tokens": 2777153536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 3.158421516418457, "created_at": "2025-01-15T20:30:48.064740+00:00"} {"global_step": 5298, "acc_step": 0, "speed/wps": 12910.168671919011, "speed/FLOPS": 202771939423307.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0713912844657898, "optim/lr": 0.00299978415714426, "optim/total_tokens": 2777677824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.084010601043701, "created_at": "2025-01-15T20:30:58.223966+00:00"} {"global_step": 5299, "acc_step": 0, "speed/wps": 12908.15236576352, "speed/FLOPS": 202740270564440.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07634352147579193, "optim/lr": 0.0029997827012452215, "optim/total_tokens": 2778202112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 3.1257388591766357, "created_at": "2025-01-15T20:31:08.384548+00:00"} {"global_step": 5300, "acc_step": 0, "speed/wps": 12909.157164905275, "speed/FLOPS": 202756052315700.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06071896851062775, "optim/lr": 0.0029997812404528776, "optim/total_tokens": 2778726400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.0709335803985596, "created_at": "2025-01-15T20:31:18.544395+00:00"} {"global_step": 5301, "acc_step": 0, "speed/wps": 12912.243240334137, "speed/FLOPS": 202804523371022.88, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07971159368753433, "optim/lr": 0.0029997797747672345, "optim/total_tokens": 2779250688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.168544054031372, "created_at": "2025-01-15T20:31:28.698866+00:00"} {"global_step": 5302, "acc_step": 0, "speed/wps": 12906.843765034095, "speed/FLOPS": 202719717191779.12, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07175862044095993, "optim/lr": 0.0029997783041882955, "optim/total_tokens": 2779774976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 3.1261048316955566, "created_at": "2025-01-15T20:31:38.859185+00:00"} {"global_step": 5303, "acc_step": 0, "speed/wps": 12905.99397261645, "speed/FLOPS": 202706370034122.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08284009248018265, "optim/lr": 0.002999776828716066, "optim/total_tokens": 2780299264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.0816650390625, "created_at": "2025-01-15T20:31:49.018707+00:00"} {"global_step": 5304, "acc_step": 0, "speed/wps": 12904.234357960886, "speed/FLOPS": 202678732867993.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09338963776826859, "optim/lr": 0.002999775348350551, "optim/total_tokens": 2780823552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.1581666469573975, "created_at": "2025-01-15T20:31:59.182925+00:00"} {"global_step": 5305, "acc_step": 0, "speed/wps": 12911.207172746537, "speed/FLOPS": 202788250505853.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07882092893123627, "optim/lr": 0.002999773863091755, "optim/total_tokens": 2781347840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487913, "loss/out": 3.0223138332366943, "created_at": "2025-01-15T20:32:09.338855+00:00"} {"global_step": 5306, "acc_step": 0, "speed/wps": 12914.146962220037, "speed/FLOPS": 202834423939228.94, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06833843141794205, "optim/lr": 0.0029997723729396834, "optim/total_tokens": 2781872128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.937009811401367, "created_at": "2025-01-15T20:32:19.491790+00:00"} {"global_step": 5307, "acc_step": 0, "speed/wps": 12911.419498328845, "speed/FLOPS": 202791585370889.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07916691154241562, "optim/lr": 0.0029997708778943402, "optim/total_tokens": 2782396416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.1706857681274414, "created_at": "2025-01-15T20:32:29.647190+00:00"} {"global_step": 5308, "acc_step": 0, "speed/wps": 12911.482026467196, "speed/FLOPS": 202792567461225.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07232166826725006, "optim/lr": 0.0029997693779557303, "optim/total_tokens": 2782920704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.1392385959625244, "created_at": "2025-01-15T20:32:39.802372+00:00"} {"global_step": 5309, "acc_step": 0, "speed/wps": 12906.633028028662, "speed/FLOPS": 202716407277525.8, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07161219418048859, "optim/lr": 0.00299976787312386, "optim/total_tokens": 2783444992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.1224565505981445, "created_at": "2025-01-15T20:32:49.961701+00:00"} {"global_step": 5310, "acc_step": 0, "speed/wps": 12909.318712796836, "speed/FLOPS": 202758589647327.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06450535356998444, "optim/lr": 0.0029997663633987326, "optim/total_tokens": 2783969280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0516371726989746, "created_at": "2025-01-15T20:33:00.120776+00:00"} {"global_step": 5311, "acc_step": 0, "speed/wps": 12909.756762026489, "speed/FLOPS": 202765469812422.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05867846682667732, "optim/lr": 0.002999764848780354, "optim/total_tokens": 2784493568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.104435920715332, "created_at": "2025-01-15T20:33:10.277303+00:00"} {"global_step": 5312, "acc_step": 0, "speed/wps": 12898.833359807499, "speed/FLOPS": 202593902770245.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0733499526977539, "optim/lr": 0.002999763329268729, "optim/total_tokens": 2785017856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 3.184567451477051, "created_at": "2025-01-15T20:33:20.445894+00:00"} {"global_step": 5313, "acc_step": 0, "speed/wps": 12910.382735085923, "speed/FLOPS": 202775301579497.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08108942210674286, "optim/lr": 0.0029997618048638623, "optim/total_tokens": 2785542144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.0811338424682617, "created_at": "2025-01-15T20:33:30.602547+00:00"} {"global_step": 5314, "acc_step": 0, "speed/wps": 12911.497713084465, "speed/FLOPS": 202792813841104.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0858822613954544, "optim/lr": 0.0029997602755657583, "optim/total_tokens": 2786066432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.071179151535034, "created_at": "2025-01-15T20:33:40.762261+00:00"} {"global_step": 5315, "acc_step": 0, "speed/wps": 12911.476443826903, "speed/FLOPS": 202792479778187.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08677257597446442, "optim/lr": 0.0029997587413744236, "optim/total_tokens": 2786590720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 3.174290418624878, "created_at": "2025-01-15T20:33:50.920899+00:00"} {"global_step": 5316, "acc_step": 0, "speed/wps": 12908.131425996693, "speed/FLOPS": 202739941676628.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10182829946279526, "optim/lr": 0.002999757202289861, "optim/total_tokens": 2787115008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 3.246046781539917, "created_at": "2025-01-15T20:34:01.082282+00:00"} {"global_step": 5317, "acc_step": 0, "speed/wps": 12913.67557573273, "speed/FLOPS": 202827020166692.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09784039109945297, "optim/lr": 0.002999755658312078, "optim/total_tokens": 2787639296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.1550326347351074, "created_at": "2025-01-15T20:34:11.235852+00:00"} {"global_step": 5318, "acc_step": 0, "speed/wps": 12912.333444504997, "speed/FLOPS": 202805940151479.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08300703018903732, "optim/lr": 0.002999754109441077, "optim/total_tokens": 2788163584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 3.030655860900879, "created_at": "2025-01-15T20:34:21.390348+00:00"} {"global_step": 5319, "acc_step": 0, "speed/wps": 12909.389857672431, "speed/FLOPS": 202759707075362.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.07173576951026917, "optim/lr": 0.0029997525556768657, "optim/total_tokens": 2788687872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.182723045349121, "created_at": "2025-01-15T20:34:31.548854+00:00"} {"global_step": 5320, "acc_step": 0, "speed/wps": 12913.631661469206, "speed/FLOPS": 202826330432839.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07971218973398209, "optim/lr": 0.0029997509970194473, "optim/total_tokens": 2789212160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417398, "loss/out": 3.2558605670928955, "created_at": "2025-01-15T20:34:41.703316+00:00"} {"global_step": 5321, "acc_step": 0, "speed/wps": 12909.338992120736, "speed/FLOPS": 202758908161976.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07471390813589096, "optim/lr": 0.0029997494334688276, "optim/total_tokens": 2789736448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.128397226333618, "created_at": "2025-01-15T20:34:51.863015+00:00"} {"global_step": 5322, "acc_step": 0, "speed/wps": 12908.286401059713, "speed/FLOPS": 202742375772951.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0885576382279396, "optim/lr": 0.0029997478650250118, "optim/total_tokens": 2790260736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1759042739868164, "created_at": "2025-01-15T20:35:02.028823+00:00"} {"global_step": 5323, "acc_step": 0, "speed/wps": 12914.400496320079, "speed/FLOPS": 202838406040662.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08399413526058197, "optim/lr": 0.002999746291688005, "optim/total_tokens": 2790785024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 3.069035530090332, "created_at": "2025-01-15T20:35:12.186028+00:00"} {"global_step": 5324, "acc_step": 0, "speed/wps": 12911.25956915455, "speed/FLOPS": 202789073463441.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0768001526594162, "optim/lr": 0.002999744713457812, "optim/total_tokens": 2791309312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 3.0239179134368896, "created_at": "2025-01-15T20:35:22.341486+00:00"} {"global_step": 5325, "acc_step": 0, "speed/wps": 12908.965167007318, "speed/FLOPS": 202753036724880.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08045203238725662, "optim/lr": 0.0029997431303344377, "optim/total_tokens": 2791833600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.116248607635498, "created_at": "2025-01-15T20:35:32.502046+00:00"} {"global_step": 5326, "acc_step": 0, "speed/wps": 12908.989028192922, "speed/FLOPS": 202753411497589.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.08702481538057327, "optim/lr": 0.002999741542317888, "optim/total_tokens": 2792357888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.251892328262329, "created_at": "2025-01-15T20:35:42.660268+00:00"} {"global_step": 5327, "acc_step": 0, "speed/wps": 12905.079238530743, "speed/FLOPS": 202692002878330.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08516878634691238, "optim/lr": 0.002999739949408168, "optim/total_tokens": 2792882176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 3.1677486896514893, "created_at": "2025-01-15T20:35:52.824502+00:00"} {"global_step": 5328, "acc_step": 0, "speed/wps": 12910.902969044068, "speed/FLOPS": 202783472568686.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11250393092632294, "optim/lr": 0.002999738351605282, "optim/total_tokens": 2793406464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.0835347175598145, "created_at": "2025-01-15T20:36:02.980144+00:00"} {"global_step": 5329, "acc_step": 0, "speed/wps": 12916.475501658075, "speed/FLOPS": 202870996850851.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0888512060046196, "optim/lr": 0.0029997367489092365, "optim/total_tokens": 2793930752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 2.951946973800659, "created_at": "2025-01-15T20:36:13.132200+00:00"} {"global_step": 5330, "acc_step": 0, "speed/wps": 12910.907022109172, "speed/FLOPS": 202783536227643.53, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07437096536159515, "optim/lr": 0.0029997351413200354, "optim/total_tokens": 2794455040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.0297021865844727, "created_at": "2025-01-15T20:36:23.288805+00:00"} {"global_step": 5331, "acc_step": 0, "speed/wps": 12903.14318059126, "speed/FLOPS": 202661594427968.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06475397944450378, "optim/lr": 0.002999733528837685, "optim/total_tokens": 2794979328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 3.1620562076568604, "created_at": "2025-01-15T20:36:33.452535+00:00"} {"global_step": 5332, "acc_step": 0, "speed/wps": 12909.924243725172, "speed/FLOPS": 202768100342643.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07369422912597656, "optim/lr": 0.0029997319114621903, "optim/total_tokens": 2795503616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 3.125479221343994, "created_at": "2025-01-15T20:36:43.610946+00:00"} {"global_step": 5333, "acc_step": 0, "speed/wps": 12915.05636780436, "speed/FLOPS": 202848707403584.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06983397156000137, "optim/lr": 0.002999730289193556, "optim/total_tokens": 2796027904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 3.107154130935669, "created_at": "2025-01-15T20:36:53.768883+00:00"} {"global_step": 5334, "acc_step": 0, "speed/wps": 12912.107053295698, "speed/FLOPS": 202802384366443.6, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06317102164030075, "optim/lr": 0.002999728662031789, "optim/total_tokens": 2796552192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.1899218559265137, "created_at": "2025-01-15T20:37:03.923935+00:00"} {"global_step": 5335, "acc_step": 0, "speed/wps": 12906.11395356977, "speed/FLOPS": 202708254499865.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07621745765209198, "optim/lr": 0.0029997270299768923, "optim/total_tokens": 2797076480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.1677489280700684, "created_at": "2025-01-15T20:37:14.084556+00:00"} {"global_step": 5336, "acc_step": 0, "speed/wps": 12913.030082363899, "speed/FLOPS": 202816881806334.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07251527905464172, "optim/lr": 0.002999725393028873, "optim/total_tokens": 2797600768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.1798653602600098, "created_at": "2025-01-15T20:37:24.239521+00:00"} {"global_step": 5337, "acc_step": 0, "speed/wps": 12913.209735516291, "speed/FLOPS": 202819703505962.16, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07070048898458481, "optim/lr": 0.0029997237511877354, "optim/total_tokens": 2798125056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485528, "loss/out": 3.067626476287842, "created_at": "2025-01-15T20:37:34.396199+00:00"} {"global_step": 5338, "acc_step": 0, "speed/wps": 12909.734895746768, "speed/FLOPS": 202765126372451.56, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09750641137361526, "optim/lr": 0.002999722104453485, "optim/total_tokens": 2798649344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.01995849609375, "created_at": "2025-01-15T20:37:44.571561+00:00"} {"global_step": 5339, "acc_step": 0, "speed/wps": 12905.060260063909, "speed/FLOPS": 202691704795429.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10063333064317703, "optim/lr": 0.0029997204528261286, "optim/total_tokens": 2799173632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0604124069213867, "created_at": "2025-01-15T20:37:54.734541+00:00"} {"global_step": 5340, "acc_step": 0, "speed/wps": 12909.854060468098, "speed/FLOPS": 202766998018149.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09080024063587189, "optim/lr": 0.0029997187963056696, "optim/total_tokens": 2799697920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 3.1567208766937256, "created_at": "2025-01-15T20:38:04.893088+00:00"} {"global_step": 5341, "acc_step": 0, "speed/wps": 12910.330506860364, "speed/FLOPS": 202774481263446.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0814662054181099, "optim/lr": 0.002999717134892115, "optim/total_tokens": 2800222208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.0600175857543945, "created_at": "2025-01-15T20:38:15.052965+00:00"} {"global_step": 5342, "acc_step": 0, "speed/wps": 12909.614640229509, "speed/FLOPS": 202763237594304.62, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08869614452123642, "optim/lr": 0.0029997154685854686, "optim/total_tokens": 2800746496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.120467185974121, "created_at": "2025-01-15T20:38:25.212811+00:00"} {"global_step": 5343, "acc_step": 0, "speed/wps": 12908.13394297983, "speed/FLOPS": 202739981209307.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09360656887292862, "optim/lr": 0.0029997137973857375, "optim/total_tokens": 2801270784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.051201820373535, "created_at": "2025-01-15T20:38:35.374855+00:00"} {"global_step": 5344, "acc_step": 0, "speed/wps": 12908.925421847005, "speed/FLOPS": 202752412472522.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0725381150841713, "optim/lr": 0.0029997121212929257, "optim/total_tokens": 2801795072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423001, "loss/out": 3.09466552734375, "created_at": "2025-01-15T20:38:45.532639+00:00"} {"global_step": 5345, "acc_step": 0, "speed/wps": 12911.870849755254, "speed/FLOPS": 202798674465263.25, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07167071104049683, "optim/lr": 0.0029997104403070396, "optim/total_tokens": 2802319360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0749239921569824, "created_at": "2025-01-15T20:38:55.687433+00:00"} {"global_step": 5346, "acc_step": 0, "speed/wps": 12914.679229768877, "speed/FLOPS": 202842783932495.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07160269469022751, "optim/lr": 0.002999708754428085, "optim/total_tokens": 2802843648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478937, "loss/out": 3.1058051586151123, "created_at": "2025-01-15T20:39:05.845944+00:00"} {"global_step": 5347, "acc_step": 0, "speed/wps": 12904.150853645942, "speed/FLOPS": 202677421317980.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07779233902692795, "optim/lr": 0.002999707063656066, "optim/total_tokens": 2803367936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0606436729431152, "created_at": "2025-01-15T20:39:16.008794+00:00"} {"global_step": 5348, "acc_step": 0, "speed/wps": 12914.02033339888, "speed/FLOPS": 202832435059586.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06504368782043457, "optim/lr": 0.0029997053679909893, "optim/total_tokens": 2803892224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.203118324279785, "created_at": "2025-01-15T20:39:26.164568+00:00"} {"global_step": 5349, "acc_step": 0, "speed/wps": 12909.101929550168, "speed/FLOPS": 202755184768547.28, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07679954171180725, "optim/lr": 0.0029997036674328597, "optim/total_tokens": 2804416512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 3.1910743713378906, "created_at": "2025-01-15T20:39:36.325567+00:00"} {"global_step": 5350, "acc_step": 0, "speed/wps": 12912.206380938607, "speed/FLOPS": 202803944443565.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0792839378118515, "optim/lr": 0.002999701961981683, "optim/total_tokens": 2804940800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 3.0701041221618652, "created_at": "2025-01-15T20:39:46.487417+00:00"} {"global_step": 5351, "acc_step": 0, "speed/wps": 12914.630388151392, "speed/FLOPS": 202842016807777.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07492858916521072, "optim/lr": 0.0029997002516374657, "optim/total_tokens": 2805465088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458694, "loss/out": 3.1091229915618896, "created_at": "2025-01-15T20:39:56.643537+00:00"} {"global_step": 5352, "acc_step": 0, "speed/wps": 12904.719676009496, "speed/FLOPS": 202686355454843.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07692324370145798, "optim/lr": 0.0029996985364002123, "optim/total_tokens": 2805989376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.1004812717437744, "created_at": "2025-01-15T20:40:06.804214+00:00"} {"global_step": 5353, "acc_step": 0, "speed/wps": 12909.135214218148, "speed/FLOPS": 202755707549995.8, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07101832330226898, "optim/lr": 0.0029996968162699284, "optim/total_tokens": 2806513664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332595, "loss/out": 3.1688034534454346, "created_at": "2025-01-15T20:40:16.965760+00:00"} {"global_step": 5354, "acc_step": 0, "speed/wps": 12915.985556862173, "speed/FLOPS": 202863301594577.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07850861549377441, "optim/lr": 0.0029996950912466198, "optim/total_tokens": 2807037952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0161519050598145, "created_at": "2025-01-15T20:40:27.119967+00:00"} {"global_step": 5355, "acc_step": 0, "speed/wps": 12913.554448166089, "speed/FLOPS": 202825117691812.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07123978435993195, "optim/lr": 0.002999693361330293, "optim/total_tokens": 2807562240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474921, "loss/out": 3.137253761291504, "created_at": "2025-01-15T20:40:37.277435+00:00"} {"global_step": 5356, "acc_step": 0, "speed/wps": 12911.760849541268, "speed/FLOPS": 202796946760746.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07042321562767029, "optim/lr": 0.0029996916265209515, "optim/total_tokens": 2808086528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.9552557468414307, "created_at": "2025-01-15T20:40:47.432717+00:00"} {"global_step": 5357, "acc_step": 0, "speed/wps": 12910.127748102834, "speed/FLOPS": 202771296658539.28, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06959499418735504, "optim/lr": 0.0029996898868186036, "optim/total_tokens": 2808610816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435320, "loss/out": 3.186683416366577, "created_at": "2025-01-15T20:40:57.592083+00:00"} {"global_step": 5358, "acc_step": 0, "speed/wps": 12912.173972470584, "speed/FLOPS": 202803435424041.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07387800514698029, "optim/lr": 0.002999688142223253, "optim/total_tokens": 2809135104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.187592029571533, "created_at": "2025-01-15T20:41:07.746612+00:00"} {"global_step": 5359, "acc_step": 0, "speed/wps": 12903.893763074222, "speed/FLOPS": 202673383357271.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07600133121013641, "optim/lr": 0.0029996863927349063, "optim/total_tokens": 2809659392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.1329283714294434, "created_at": "2025-01-15T20:41:17.907739+00:00"} {"global_step": 5360, "acc_step": 0, "speed/wps": 12906.580969804358, "speed/FLOPS": 202715589631577.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07709391415119171, "optim/lr": 0.0029996846383535686, "optim/total_tokens": 2810183680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.096606731414795, "created_at": "2025-01-15T20:41:28.071609+00:00"} {"global_step": 5361, "acc_step": 0, "speed/wps": 12904.68122072245, "speed/FLOPS": 202685751461717.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07196328043937683, "optim/lr": 0.0029996828790792465, "optim/total_tokens": 2810707968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.16907000541687, "created_at": "2025-01-15T20:41:38.233326+00:00"} {"global_step": 5362, "acc_step": 0, "speed/wps": 12905.296494950819, "speed/FLOPS": 202695415188949.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07152155786752701, "optim/lr": 0.002999681114911945, "optim/total_tokens": 2811232256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 3.1496472358703613, "created_at": "2025-01-15T20:41:48.395718+00:00"} {"global_step": 5363, "acc_step": 0, "speed/wps": 12906.462376814585, "speed/FLOPS": 202713726965709.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07128848880529404, "optim/lr": 0.0029996793458516705, "optim/total_tokens": 2811756544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.201324462890625, "created_at": "2025-01-15T20:41:58.554728+00:00"} {"global_step": 5364, "acc_step": 0, "speed/wps": 12912.494580251172, "speed/FLOPS": 202808471009797.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07587581127882004, "optim/lr": 0.002999677571898428, "optim/total_tokens": 2812280832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 3.018643856048584, "created_at": "2025-01-15T20:42:08.711540+00:00"} {"global_step": 5365, "acc_step": 0, "speed/wps": 12904.211671505422, "speed/FLOPS": 202678376546035.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0771070346236229, "optim/lr": 0.002999675793052224, "optim/total_tokens": 2812805120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.1565308570861816, "created_at": "2025-01-15T20:42:18.874129+00:00"} {"global_step": 5366, "acc_step": 0, "speed/wps": 12912.978106404327, "speed/FLOPS": 202816065452467.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08749248832464218, "optim/lr": 0.0029996740093130633, "optim/total_tokens": 2813329408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436366, "loss/out": 3.0990753173828125, "created_at": "2025-01-15T20:42:29.031589+00:00"} {"global_step": 5367, "acc_step": 0, "speed/wps": 12908.698761771142, "speed/FLOPS": 202748852464567.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08590236306190491, "optim/lr": 0.002999672220680953, "optim/total_tokens": 2813853696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.194319248199463, "created_at": "2025-01-15T20:42:39.194335+00:00"} {"global_step": 5368, "acc_step": 0, "speed/wps": 12900.349875485987, "speed/FLOPS": 202617721732886.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08471259474754333, "optim/lr": 0.0029996704271558977, "optim/total_tokens": 2814377984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.0725154876708984, "created_at": "2025-01-15T20:42:49.359508+00:00"} {"global_step": 5369, "acc_step": 0, "speed/wps": 12907.295380188549, "speed/FLOPS": 202726810428364.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06413285434246063, "optim/lr": 0.0029996686287379047, "optim/total_tokens": 2814902272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 3.136648178100586, "created_at": "2025-01-15T20:42:59.518445+00:00"} {"global_step": 5370, "acc_step": 0, "speed/wps": 12908.398438863782, "speed/FLOPS": 202744135480620.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08121345192193985, "optim/lr": 0.0029996668254269783, "optim/total_tokens": 2815426560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.142651081085205, "created_at": "2025-01-15T20:43:09.676077+00:00"} {"global_step": 5371, "acc_step": 0, "speed/wps": 12911.311093019658, "speed/FLOPS": 202789882716544.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0873686745762825, "optim/lr": 0.0029996650172231256, "optim/total_tokens": 2815950848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.039930820465088, "created_at": "2025-01-15T20:43:19.833626+00:00"} {"global_step": 5372, "acc_step": 0, "speed/wps": 12914.549472555515, "speed/FLOPS": 202840745917154.78, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07906793057918549, "optim/lr": 0.0029996632041263517, "optim/total_tokens": 2816475136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.2330362796783447, "created_at": "2025-01-15T20:43:29.990167+00:00"} {"global_step": 5373, "acc_step": 0, "speed/wps": 12905.984903360888, "speed/FLOPS": 202706227589001.78, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06959541887044907, "optim/lr": 0.002999661386136663, "optim/total_tokens": 2816999424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.130298137664795, "created_at": "2025-01-15T20:43:40.149646+00:00"} {"global_step": 5374, "acc_step": 0, "speed/wps": 12902.862357784512, "speed/FLOPS": 202657183719897.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.07916507869958878, "optim/lr": 0.0029996595632540654, "optim/total_tokens": 2817523712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.0989303588867188, "created_at": "2025-01-15T20:43:50.314176+00:00"} {"global_step": 5375, "acc_step": 0, "speed/wps": 12897.45224496108, "speed/FLOPS": 202572210463731.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08402412384748459, "optim/lr": 0.002999657735478565, "optim/total_tokens": 2818048000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475497, "loss/out": 3.0456957817077637, "created_at": "2025-01-15T20:44:00.481159+00:00"} {"global_step": 5376, "acc_step": 0, "speed/wps": 12910.158297643111, "speed/FLOPS": 202771776481048.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.07831898331642151, "optim/lr": 0.0029996559028101664, "optim/total_tokens": 2818572288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.1399707794189453, "created_at": "2025-01-15T20:44:10.637751+00:00"} {"global_step": 5377, "acc_step": 0, "speed/wps": 12913.001545236666, "speed/FLOPS": 202816433591537.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11699546873569489, "optim/lr": 0.0029996540652488775, "optim/total_tokens": 2819096576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337297, "loss/out": 3.100264072418213, "created_at": "2025-01-15T20:44:20.794575+00:00"} {"global_step": 5378, "acc_step": 0, "speed/wps": 12907.08375130397, "speed/FLOPS": 202723486505923.2, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10688957571983337, "optim/lr": 0.0029996522227947035, "optim/total_tokens": 2819620864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.183692216873169, "created_at": "2025-01-15T20:44:30.955643+00:00"} {"global_step": 5379, "acc_step": 0, "speed/wps": 12906.45048402132, "speed/FLOPS": 202713540173048.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0709400475025177, "optim/lr": 0.00299965037544765, "optim/total_tokens": 2820145152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0056633949279785, "created_at": "2025-01-15T20:44:41.119032+00:00"} {"global_step": 5380, "acc_step": 0, "speed/wps": 12913.4714547296, "speed/FLOPS": 202823814165847.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09031813591718674, "optim/lr": 0.002999648523207724, "optim/total_tokens": 2820669440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.0709102153778076, "created_at": "2025-01-15T20:44:51.276218+00:00"} {"global_step": 5381, "acc_step": 0, "speed/wps": 12911.435893132033, "speed/FLOPS": 202791842873802.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09254109114408493, "optim/lr": 0.00299964666607493, "optim/total_tokens": 2821193728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.96795654296875, "created_at": "2025-01-15T20:45:01.433186+00:00"} {"global_step": 5382, "acc_step": 0, "speed/wps": 12910.241083881712, "speed/FLOPS": 202773076752686.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07811106741428375, "optim/lr": 0.0029996448040492756, "optim/total_tokens": 2821718016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.1452174186706543, "created_at": "2025-01-15T20:45:11.590163+00:00"} {"global_step": 5383, "acc_step": 0, "speed/wps": 12911.180664372248, "speed/FLOPS": 202787834155409.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08520259708166122, "optim/lr": 0.0029996429371307665, "optim/total_tokens": 2822242304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.977860927581787, "created_at": "2025-01-15T20:45:21.746620+00:00"} {"global_step": 5384, "acc_step": 0, "speed/wps": 12908.571747232647, "speed/FLOPS": 202746857526703.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08023244887590408, "optim/lr": 0.002999641065319408, "optim/total_tokens": 2822766592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.0824174880981445, "created_at": "2025-01-15T20:45:31.905322+00:00"} {"global_step": 5385, "acc_step": 0, "speed/wps": 12908.189599813084, "speed/FLOPS": 202740855376353.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09125430136919022, "optim/lr": 0.002999639188615207, "optim/total_tokens": 2823290880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1886935234069824, "created_at": "2025-01-15T20:45:42.069671+00:00"} {"global_step": 5386, "acc_step": 0, "speed/wps": 12911.034337262288, "speed/FLOPS": 202785535887072.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08204322308301926, "optim/lr": 0.0029996373070181694, "optim/total_tokens": 2823815168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 3.1025543212890625, "created_at": "2025-01-15T20:45:52.225347+00:00"} {"global_step": 5387, "acc_step": 0, "speed/wps": 12914.67843930276, "speed/FLOPS": 202842771517139.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07809367775917053, "optim/lr": 0.002999635420528301, "optim/total_tokens": 2824339456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.1916441917419434, "created_at": "2025-01-15T20:46:02.377924+00:00"} {"global_step": 5388, "acc_step": 0, "speed/wps": 12910.65821103075, "speed/FLOPS": 202779628307754.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08052534610033035, "optim/lr": 0.002999633529145609, "optim/total_tokens": 2824863744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 3.1399240493774414, "created_at": "2025-01-15T20:46:12.535631+00:00"} {"global_step": 5389, "acc_step": 0, "speed/wps": 12912.79180489468, "speed/FLOPS": 202813139331252.88, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07375909388065338, "optim/lr": 0.0029996316328700983, "optim/total_tokens": 2825388032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.134228229522705, "created_at": "2025-01-15T20:46:22.689921+00:00"} {"global_step": 5390, "acc_step": 0, "speed/wps": 12913.551669575545, "speed/FLOPS": 202825074050229.38, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06439295411109924, "optim/lr": 0.002999629731701776, "optim/total_tokens": 2825912320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 3.172051429748535, "created_at": "2025-01-15T20:46:32.843385+00:00"} {"global_step": 5391, "acc_step": 0, "speed/wps": 12905.445989016398, "speed/FLOPS": 202697763198675.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.060762807726860046, "optim/lr": 0.002999627825640648, "optim/total_tokens": 2826436608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474336, "loss/out": 3.104177236557007, "created_at": "2025-01-15T20:46:43.022200+00:00"} {"global_step": 5392, "acc_step": 0, "speed/wps": 12911.051602835869, "speed/FLOPS": 202785807066631.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06523451209068298, "optim/lr": 0.00299962591468672, "optim/total_tokens": 2826960896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.13187313079834, "created_at": "2025-01-15T20:46:53.181386+00:00"} {"global_step": 5393, "acc_step": 0, "speed/wps": 12910.23416524976, "speed/FLOPS": 202772968086064.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06805168092250824, "optim/lr": 0.002999623998839999, "optim/total_tokens": 2827485184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.1727190017700195, "created_at": "2025-01-15T20:47:03.338485+00:00"} {"global_step": 5394, "acc_step": 0, "speed/wps": 12909.80203309904, "speed/FLOPS": 202766180856825.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06259678304195404, "optim/lr": 0.0029996220781004913, "optim/total_tokens": 2828009472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.0596413612365723, "created_at": "2025-01-15T20:47:13.497253+00:00"} {"global_step": 5395, "acc_step": 0, "speed/wps": 12912.695273714698, "speed/FLOPS": 202811623176423.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05882943421602249, "optim/lr": 0.002999620152468202, "optim/total_tokens": 2828533760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 3.2026262283325195, "created_at": "2025-01-15T20:47:23.653265+00:00"} {"global_step": 5396, "acc_step": 0, "speed/wps": 12914.258183383557, "speed/FLOPS": 202836170820435.9, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06826235353946686, "optim/lr": 0.0029996182219431387, "optim/total_tokens": 2829058048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.012472152709961, "created_at": "2025-01-15T20:47:33.806116+00:00"} {"global_step": 5397, "acc_step": 0, "speed/wps": 12909.204419879821, "speed/FLOPS": 202756794520010.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0740155577659607, "optim/lr": 0.002999616286525307, "optim/total_tokens": 2829582336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.1349825859069824, "created_at": "2025-01-15T20:47:43.985709+00:00"} {"global_step": 5398, "acc_step": 0, "speed/wps": 12905.439877875922, "speed/FLOPS": 202697667214816.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07083562761545181, "optim/lr": 0.002999614346214714, "optim/total_tokens": 2830106624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.0820136070251465, "created_at": "2025-01-15T20:47:54.146114+00:00"} {"global_step": 5399, "acc_step": 0, "speed/wps": 12907.28068510877, "speed/FLOPS": 202726579621943.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06263242661952972, "optim/lr": 0.0029996124010113648, "optim/total_tokens": 2830630912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464846, "loss/out": 3.1243393421173096, "created_at": "2025-01-15T20:48:04.304488+00:00"} {"global_step": 5400, "acc_step": 0, "speed/wps": 12911.300744150833, "speed/FLOPS": 202789720173338.6, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0771859809756279, "optim/lr": 0.0029996104509152667, "optim/total_tokens": 2831155200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.2130401134490967, "created_at": "2025-01-15T20:48:14.464594+00:00"} {"global_step": 5401, "acc_step": 0, "speed/wps": 12912.554838222595, "speed/FLOPS": 202809417444041.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0894777923822403, "optim/lr": 0.0029996084959264258, "optim/total_tokens": 2831679488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.2210912704467773, "created_at": "2025-01-15T20:48:24.618951+00:00"} {"global_step": 5402, "acc_step": 0, "speed/wps": 12908.573671210284, "speed/FLOPS": 202746887745416.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0688251480460167, "optim/lr": 0.002999606536044848, "optim/total_tokens": 2832203776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 3.101856231689453, "created_at": "2025-01-15T20:48:34.781217+00:00"} {"global_step": 5403, "acc_step": 0, "speed/wps": 12915.519644796954, "speed/FLOPS": 202855983805362.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08215980976819992, "optim/lr": 0.0029996045712705408, "optim/total_tokens": 2832728064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.0076818466186523, "created_at": "2025-01-15T20:48:44.934310+00:00"} {"global_step": 5404, "acc_step": 0, "speed/wps": 12910.849456268508, "speed/FLOPS": 202782632077012.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0777134820818901, "optim/lr": 0.0029996026016035093, "optim/total_tokens": 2833252352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.058995246887207, "created_at": "2025-01-15T20:48:55.091791+00:00"} {"global_step": 5405, "acc_step": 0, "speed/wps": 12919.981888231354, "speed/FLOPS": 202926069470265.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08321484923362732, "optim/lr": 0.002999600627043761, "optim/total_tokens": 2833776640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.134284257888794, "created_at": "2025-01-15T20:49:05.241508+00:00"} {"global_step": 5406, "acc_step": 0, "speed/wps": 12906.33655316678, "speed/FLOPS": 202711750732420.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07341151684522629, "optim/lr": 0.002999598647591302, "optim/total_tokens": 2834300928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.192722797393799, "created_at": "2025-01-15T20:49:15.400772+00:00"} {"global_step": 5407, "acc_step": 0, "speed/wps": 12911.243279033051, "speed/FLOPS": 202788817604698.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09066317230463028, "optim/lr": 0.002999596663246138, "optim/total_tokens": 2834825216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.189469814300537, "created_at": "2025-01-15T20:49:25.559082+00:00"} {"global_step": 5408, "acc_step": 0, "speed/wps": 12910.787929146594, "speed/FLOPS": 202781665709015.38, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09339553117752075, "optim/lr": 0.0029995946740082767, "optim/total_tokens": 2835349504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 3.06966233253479, "created_at": "2025-01-15T20:49:35.723703+00:00"} {"global_step": 5409, "acc_step": 0, "speed/wps": 12913.42003684787, "speed/FLOPS": 202823006577358.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07666849344968796, "optim/lr": 0.002999592679877724, "optim/total_tokens": 2835873792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.057827949523926, "created_at": "2025-01-15T20:49:45.877274+00:00"} {"global_step": 5410, "acc_step": 0, "speed/wps": 12912.254106041708, "speed/FLOPS": 202804694031891.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0779234766960144, "optim/lr": 0.0029995906808544864, "optim/total_tokens": 2836398080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296188, "loss/out": 3.101778030395508, "created_at": "2025-01-15T20:49:56.034213+00:00"} {"global_step": 5411, "acc_step": 0, "speed/wps": 12901.621930149207, "speed/FLOPS": 202637701099358.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.08723777532577515, "optim/lr": 0.00299958867693857, "optim/total_tokens": 2836922368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.1262547969818115, "created_at": "2025-01-15T20:50:06.199129+00:00"} {"global_step": 5412, "acc_step": 0, "speed/wps": 12893.761816679315, "speed/FLOPS": 202514247216386.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08585023880004883, "optim/lr": 0.0029995866681299826, "optim/total_tokens": 2837446656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.1661205291748047, "created_at": "2025-01-15T20:50:16.368784+00:00"} {"global_step": 5413, "acc_step": 0, "speed/wps": 12903.768778634536, "speed/FLOPS": 202671420304899.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06302231550216675, "optim/lr": 0.0029995846544287295, "optim/total_tokens": 2837970944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.118903636932373, "created_at": "2025-01-15T20:50:26.530850+00:00"} {"global_step": 5414, "acc_step": 0, "speed/wps": 12908.550196354334, "speed/FLOPS": 202746519040545.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06752879917621613, "optim/lr": 0.0029995826358348177, "optim/total_tokens": 2838495232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.2327423095703125, "created_at": "2025-01-15T20:50:36.700869+00:00"} {"global_step": 5415, "acc_step": 0, "speed/wps": 12910.484246173843, "speed/FLOPS": 202776895950624.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062463659793138504, "optim/lr": 0.002999580612348254, "optim/total_tokens": 2839019520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402986, "loss/out": 3.040587902069092, "created_at": "2025-01-15T20:50:46.856768+00:00"} {"global_step": 5416, "acc_step": 0, "speed/wps": 12907.038763260067, "speed/FLOPS": 202722779906873.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06570295989513397, "optim/lr": 0.0029995785839690447, "optim/total_tokens": 2839543808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.959777355194092, "created_at": "2025-01-15T20:50:57.018562+00:00"} {"global_step": 5417, "acc_step": 0, "speed/wps": 12912.182828897276, "speed/FLOPS": 202803574526392.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06456699222326279, "optim/lr": 0.002999576550697196, "optim/total_tokens": 2840068096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 3.181431770324707, "created_at": "2025-01-15T20:51:07.173619+00:00"} {"global_step": 5418, "acc_step": 0, "speed/wps": 12911.025964476581, "speed/FLOPS": 202785404380967.62, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0849732905626297, "optim/lr": 0.0029995745125327164, "optim/total_tokens": 2840592384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.1717705726623535, "created_at": "2025-01-15T20:51:17.331704+00:00"} {"global_step": 5419, "acc_step": 0, "speed/wps": 12906.615539912516, "speed/FLOPS": 202716132602630.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09123466908931732, "optim/lr": 0.00299957246947561, "optim/total_tokens": 2841116672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466420, "loss/out": 3.215630292892456, "created_at": "2025-01-15T20:51:27.490574+00:00"} {"global_step": 5420, "acc_step": 0, "speed/wps": 12911.31283895613, "speed/FLOPS": 202789910138875.78, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0860300362110138, "optim/lr": 0.0029995704215258854, "optim/total_tokens": 2841640960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.162595748901367, "created_at": "2025-01-15T20:51:37.646440+00:00"} {"global_step": 5421, "acc_step": 0, "speed/wps": 12910.22408812518, "speed/FLOPS": 202772809810975.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09280085563659668, "optim/lr": 0.002999568368683548, "optim/total_tokens": 2842165248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.175084352493286, "created_at": "2025-01-15T20:51:47.803038+00:00"} {"global_step": 5422, "acc_step": 0, "speed/wps": 12906.537016192, "speed/FLOPS": 202714899279696.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.1014375239610672, "optim/lr": 0.0029995663109486054, "optim/total_tokens": 2842689536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308860, "loss/out": 3.10910701751709, "created_at": "2025-01-15T20:51:57.965187+00:00"} {"global_step": 5423, "acc_step": 0, "speed/wps": 12911.566531860793, "speed/FLOPS": 202793894734554.12, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09572246670722961, "optim/lr": 0.002999564248321064, "optim/total_tokens": 2843213824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.1706278324127197, "created_at": "2025-01-15T20:52:08.120176+00:00"} {"global_step": 5424, "acc_step": 0, "speed/wps": 12908.58683952062, "speed/FLOPS": 202747094571825.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0867818146944046, "optim/lr": 0.0029995621808009305, "optim/total_tokens": 2843738112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.118701219558716, "created_at": "2025-01-15T20:52:18.278487+00:00"} {"global_step": 5425, "acc_step": 0, "speed/wps": 12913.528138694011, "speed/FLOPS": 202824704465400.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08876058459281921, "optim/lr": 0.0029995601083882112, "optim/total_tokens": 2844262400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.0698325634002686, "created_at": "2025-01-15T20:52:28.432018+00:00"} {"global_step": 5426, "acc_step": 0, "speed/wps": 12909.094401152803, "speed/FLOPS": 202755066524721.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08615216612815857, "optim/lr": 0.002999558031082914, "optim/total_tokens": 2844786688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.9640750885009766, "created_at": "2025-01-15T20:52:38.590539+00:00"} {"global_step": 5427, "acc_step": 0, "speed/wps": 12912.789587023735, "speed/FLOPS": 202813104496542.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07830739766359329, "optim/lr": 0.0029995559488850444, "optim/total_tokens": 2845310976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 3.068051338195801, "created_at": "2025-01-15T20:52:48.747406+00:00"} {"global_step": 5428, "acc_step": 0, "speed/wps": 12909.337089229233, "speed/FLOPS": 202758878274450.25, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06978801637887955, "optim/lr": 0.0029995538617946104, "optim/total_tokens": 2845835264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 3.1620893478393555, "created_at": "2025-01-15T20:52:58.908167+00:00"} {"global_step": 5429, "acc_step": 0, "speed/wps": 12914.63778909096, "speed/FLOPS": 202842133049703.6, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0820021703839302, "optim/lr": 0.0029995517698116177, "optim/total_tokens": 2846359552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 3.2036404609680176, "created_at": "2025-01-15T20:53:09.060771+00:00"} {"global_step": 5430, "acc_step": 0, "speed/wps": 12902.731815877174, "speed/FLOPS": 202655133379859.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07848705351352692, "optim/lr": 0.002999549672936073, "optim/total_tokens": 2846883840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 3.142397880554199, "created_at": "2025-01-15T20:53:19.227463+00:00"} {"global_step": 5431, "acc_step": 0, "speed/wps": 12914.662551783904, "speed/FLOPS": 202842521981823.78, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08575884252786636, "optim/lr": 0.002999547571167985, "optim/total_tokens": 2847408128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449165, "loss/out": 3.0696072578430176, "created_at": "2025-01-15T20:53:29.386098+00:00"} {"global_step": 5432, "acc_step": 0, "speed/wps": 12909.064236735863, "speed/FLOPS": 202754592751103.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08583014458417892, "optim/lr": 0.0029995454645073585, "optim/total_tokens": 2847932416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487384, "loss/out": 3.1644816398620605, "created_at": "2025-01-15T20:53:39.545430+00:00"} {"global_step": 5433, "acc_step": 0, "speed/wps": 12915.882612713753, "speed/FLOPS": 202861684715266.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07213075459003448, "optim/lr": 0.0029995433529542014, "optim/total_tokens": 2848456704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.092597007751465, "created_at": "2025-01-15T20:53:49.700066+00:00"} {"global_step": 5434, "acc_step": 0, "speed/wps": 12914.141089318055, "speed/FLOPS": 202834331697233.1, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08073100447654724, "optim/lr": 0.0029995412365085208, "optim/total_tokens": 2848980992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.067443370819092, "created_at": "2025-01-15T20:53:59.857453+00:00"} {"global_step": 5435, "acc_step": 0, "speed/wps": 12911.270583635765, "speed/FLOPS": 202789246461004.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06741626560688019, "optim/lr": 0.0029995391151703228, "optim/total_tokens": 2849505280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.277728796005249, "created_at": "2025-01-15T20:54:10.013137+00:00"} {"global_step": 5436, "acc_step": 0, "speed/wps": 12910.576972130286, "speed/FLOPS": 202778352339188.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06613391637802124, "optim/lr": 0.002999536988939615, "optim/total_tokens": 2850029568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.134986400604248, "created_at": "2025-01-15T20:54:20.172382+00:00"} {"global_step": 5437, "acc_step": 0, "speed/wps": 12915.194228937913, "speed/FLOPS": 202850872702128.47, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06416347622871399, "optim/lr": 0.0029995348578164033, "optim/total_tokens": 2850553856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.001997232437134, "created_at": "2025-01-15T20:54:30.327513+00:00"} {"global_step": 5438, "acc_step": 0, "speed/wps": 12907.371046055741, "speed/FLOPS": 202727998864784.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060680847615003586, "optim/lr": 0.0029995327218006957, "optim/total_tokens": 2851078144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442200, "loss/out": 3.101128339767456, "created_at": "2025-01-15T20:54:40.490623+00:00"} {"global_step": 5439, "acc_step": 0, "speed/wps": 12911.02513418678, "speed/FLOPS": 202785391340125.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0673270896077156, "optim/lr": 0.0029995305808924993, "optim/total_tokens": 2851602432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2636876106262207, "created_at": "2025-01-15T20:54:50.646136+00:00"} {"global_step": 5440, "acc_step": 0, "speed/wps": 12904.61136187646, "speed/FLOPS": 202684654232545.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06773250550031662, "optim/lr": 0.0029995284350918206, "optim/total_tokens": 2852126720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.2125110626220703, "created_at": "2025-01-15T20:55:00.806649+00:00"} {"global_step": 5441, "acc_step": 0, "speed/wps": 12906.57640813103, "speed/FLOPS": 202715517984229.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07766947895288467, "optim/lr": 0.002999526284398666, "optim/total_tokens": 2852651008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.2432422637939453, "created_at": "2025-01-15T20:55:10.965705+00:00"} {"global_step": 5442, "acc_step": 0, "speed/wps": 12912.107407501475, "speed/FLOPS": 202802389929732.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07681553065776825, "optim/lr": 0.0029995241288130435, "optim/total_tokens": 2853175296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.0801687240600586, "created_at": "2025-01-15T20:55:21.121400+00:00"} {"global_step": 5443, "acc_step": 0, "speed/wps": 12913.672074287684, "speed/FLOPS": 202826965171686.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0725456029176712, "optim/lr": 0.00299952196833496, "optim/total_tokens": 2853699584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.1365175247192383, "created_at": "2025-01-15T20:55:31.279178+00:00"} {"global_step": 5444, "acc_step": 0, "speed/wps": 12911.161561353922, "speed/FLOPS": 202787534116255.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08293628692626953, "optim/lr": 0.0029995198029644224, "optim/total_tokens": 2854223872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.880904197692871, "created_at": "2025-01-15T20:55:41.434678+00:00"} {"global_step": 5445, "acc_step": 0, "speed/wps": 12915.1866566918, "speed/FLOPS": 202850753769597.88, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08650479465723038, "optim/lr": 0.0029995176327014374, "optim/total_tokens": 2854748160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.1756911277770996, "created_at": "2025-01-15T20:55:51.588190+00:00"} {"global_step": 5446, "acc_step": 0, "speed/wps": 12908.627149730208, "speed/FLOPS": 202747727699058.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0812641903758049, "optim/lr": 0.0029995154575460127, "optim/total_tokens": 2855272448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.0503876209259033, "created_at": "2025-01-15T20:56:01.749464+00:00"} {"global_step": 5447, "acc_step": 0, "speed/wps": 12907.03324601405, "speed/FLOPS": 202722693250943.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.07367663830518723, "optim/lr": 0.0029995132774981548, "optim/total_tokens": 2855796736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 3.0588178634643555, "created_at": "2025-01-15T20:56:11.908239+00:00"} {"global_step": 5448, "acc_step": 0, "speed/wps": 12904.823713151594, "speed/FLOPS": 202687989501122.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07407572865486145, "optim/lr": 0.0029995110925578714, "optim/total_tokens": 2856321024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.0485315322875977, "created_at": "2025-01-15T20:56:22.069568+00:00"} {"global_step": 5449, "acc_step": 0, "speed/wps": 12905.409690299639, "speed/FLOPS": 202697193077449.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07426871359348297, "optim/lr": 0.002999508902725169, "optim/total_tokens": 2856845312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 3.024224042892456, "created_at": "2025-01-15T20:56:32.231806+00:00"} {"global_step": 5450, "acc_step": 0, "speed/wps": 12906.432730091266, "speed/FLOPS": 202713261323180.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08038103580474854, "optim/lr": 0.002999506708000055, "optim/total_tokens": 2857369600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.0599122047424316, "created_at": "2025-01-15T20:56:42.400615+00:00"} {"global_step": 5451, "acc_step": 0, "speed/wps": 12914.004319521026, "speed/FLOPS": 202832183539629.34, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07307849079370499, "optim/lr": 0.002999504508382537, "optim/total_tokens": 2857893888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.229825019836426, "created_at": "2025-01-15T20:56:52.554249+00:00"} {"global_step": 5452, "acc_step": 0, "speed/wps": 12916.976296820401, "speed/FLOPS": 202878862527040.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08390102535486221, "optim/lr": 0.0029995023038726218, "optim/total_tokens": 2858418176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.104236364364624, "created_at": "2025-01-15T20:57:02.707669+00:00"} {"global_step": 5453, "acc_step": 0, "speed/wps": 12910.470702111037, "speed/FLOPS": 202776683222506.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07124019414186478, "optim/lr": 0.0029995000944703165, "optim/total_tokens": 2858942464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.1316146850585938, "created_at": "2025-01-15T20:57:12.863676+00:00"} {"global_step": 5454, "acc_step": 0, "speed/wps": 12907.55739360904, "speed/FLOPS": 202730925709177.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07691473513841629, "optim/lr": 0.002999497880175629, "optim/total_tokens": 2859466752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.126647472381592, "created_at": "2025-01-15T20:57:23.023093+00:00"} {"global_step": 5455, "acc_step": 0, "speed/wps": 12909.88581662468, "speed/FLOPS": 202767496792226.2, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07863262295722961, "optim/lr": 0.002999495660988565, "optim/total_tokens": 2859991040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.1025993824005127, "created_at": "2025-01-15T20:57:33.190165+00:00"} {"global_step": 5456, "acc_step": 0, "speed/wps": 12913.49918939338, "speed/FLOPS": 202824249776853.16, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07224971801042557, "optim/lr": 0.0029994934369091328, "optim/total_tokens": 2860515328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 3.128293514251709, "created_at": "2025-01-15T20:57:43.346552+00:00"} {"global_step": 5457, "acc_step": 0, "speed/wps": 12910.093715782037, "speed/FLOPS": 202770762134175.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08414295315742493, "optim/lr": 0.00299949120793734, "optim/total_tokens": 2861039616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.0899558067321777, "created_at": "2025-01-15T20:57:53.502976+00:00"} {"global_step": 5458, "acc_step": 0, "speed/wps": 12916.537158869636, "speed/FLOPS": 202871965262085.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07997661083936691, "optim/lr": 0.002999488974073193, "optim/total_tokens": 2861563904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0803017616271973, "created_at": "2025-01-15T20:58:03.654092+00:00"} {"global_step": 5459, "acc_step": 0, "speed/wps": 12907.619973131525, "speed/FLOPS": 202731908606570.94, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.07360439002513885, "optim/lr": 0.0029994867353167, "optim/total_tokens": 2862088192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.1860549449920654, "created_at": "2025-01-15T20:58:13.815995+00:00"} {"global_step": 5460, "acc_step": 0, "speed/wps": 12911.719647919052, "speed/FLOPS": 202796299632652.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06890583783388138, "optim/lr": 0.0029994844916678676, "optim/total_tokens": 2862612480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 3.0743422508239746, "created_at": "2025-01-15T20:58:23.972088+00:00"} {"global_step": 5461, "acc_step": 0, "speed/wps": 12913.13543599939, "speed/FLOPS": 202818536529950.97, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07130749523639679, "optim/lr": 0.002999482243126703, "optim/total_tokens": 2863136768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.0970096588134766, "created_at": "2025-01-15T20:58:34.132475+00:00"} {"global_step": 5462, "acc_step": 0, "speed/wps": 12910.773362317987, "speed/FLOPS": 202781436916955.16, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07834195345640182, "optim/lr": 0.002999479989693214, "optim/total_tokens": 2863661056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370151, "loss/out": 3.056947708129883, "created_at": "2025-01-15T20:58:44.290613+00:00"} {"global_step": 5463, "acc_step": 0, "speed/wps": 12911.901214456286, "speed/FLOPS": 202799151384618.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.08107703179121017, "optim/lr": 0.002999477731367408, "optim/total_tokens": 2864185344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.169367790222168, "created_at": "2025-01-15T20:58:54.454280+00:00"} {"global_step": 5464, "acc_step": 0, "speed/wps": 12912.16622014558, "speed/FLOPS": 202803313663124.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06393149495124817, "optim/lr": 0.002999475468149292, "optim/total_tokens": 2864709632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 3.058384895324707, "created_at": "2025-01-15T20:59:04.610529+00:00"} {"global_step": 5465, "acc_step": 0, "speed/wps": 12910.665519774982, "speed/FLOPS": 202779743101626.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07869402319192886, "optim/lr": 0.002999473200038874, "optim/total_tokens": 2865233920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.0763258934020996, "created_at": "2025-01-15T20:59:14.771915+00:00"} {"global_step": 5466, "acc_step": 0, "speed/wps": 12908.201861642337, "speed/FLOPS": 202741047965231.88, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0788339301943779, "optim/lr": 0.0029994709270361603, "optim/total_tokens": 2865758208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.1289005279541016, "created_at": "2025-01-15T20:59:24.933394+00:00"} {"global_step": 5467, "acc_step": 0, "speed/wps": 12911.96486790012, "speed/FLOPS": 202800151149422.8, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06879006326198578, "optim/lr": 0.0029994686491411587, "optim/total_tokens": 2866282496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.10427188873291, "created_at": "2025-01-15T20:59:35.090128+00:00"} {"global_step": 5468, "acc_step": 0, "speed/wps": 12908.424112963186, "speed/FLOPS": 202744538727631.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07449177652597427, "optim/lr": 0.0029994663663538775, "optim/total_tokens": 2866806784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.9897193908691406, "created_at": "2025-01-15T20:59:45.248794+00:00"} {"global_step": 5469, "acc_step": 0, "speed/wps": 12909.585999778948, "speed/FLOPS": 202762787756672.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07030532509088516, "optim/lr": 0.0029994640786743243, "optim/total_tokens": 2867331072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.178316593170166, "created_at": "2025-01-15T20:59:55.407125+00:00"} {"global_step": 5470, "acc_step": 0, "speed/wps": 12911.695320582467, "speed/FLOPS": 202795917538402.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06934664398431778, "optim/lr": 0.002999461786102505, "optim/total_tokens": 2867855360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9664478302001953, "created_at": "2025-01-15T21:00:05.565496+00:00"} {"global_step": 5471, "acc_step": 0, "speed/wps": 12901.646763590741, "speed/FLOPS": 202638091142682.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07068252563476562, "optim/lr": 0.002999459488638428, "optim/total_tokens": 2868379648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.140036106109619, "created_at": "2025-01-15T21:00:15.731536+00:00"} {"global_step": 5472, "acc_step": 0, "speed/wps": 12895.93918735829, "speed/FLOPS": 202548445814920.62, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08158013969659805, "optim/lr": 0.0029994571862821, "optim/total_tokens": 2868903936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.128805637359619, "created_at": "2025-01-15T21:00:25.900717+00:00"} {"global_step": 5473, "acc_step": 0, "speed/wps": 12912.586582788359, "speed/FLOPS": 202809916036068.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07995405048131943, "optim/lr": 0.0029994548790335303, "optim/total_tokens": 2869428224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 3.053452491760254, "created_at": "2025-01-15T21:00:36.055477+00:00"} {"global_step": 5474, "acc_step": 0, "speed/wps": 12903.0738068556, "speed/FLOPS": 202660504818120.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08187655359506607, "optim/lr": 0.0029994525668927247, "optim/total_tokens": 2869952512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 3.0357718467712402, "created_at": "2025-01-15T21:00:46.223563+00:00"} {"global_step": 5475, "acc_step": 0, "speed/wps": 12907.604870082427, "speed/FLOPS": 202731671392431.1, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07974615693092346, "optim/lr": 0.002999450249859692, "optim/total_tokens": 2870476800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.161975383758545, "created_at": "2025-01-15T21:00:56.386749+00:00"} {"global_step": 5476, "acc_step": 0, "speed/wps": 12912.37452328343, "speed/FLOPS": 202806585350143.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06683699786663055, "optim/lr": 0.002999447927934438, "optim/total_tokens": 2871001088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481564, "loss/out": 3.1266393661499023, "created_at": "2025-01-15T21:01:06.541973+00:00"} {"global_step": 5477, "acc_step": 0, "speed/wps": 12907.725610229027, "speed/FLOPS": 202733567782347.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0650448277592659, "optim/lr": 0.0029994456011169725, "optim/total_tokens": 2871525376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0954761505126953, "created_at": "2025-01-15T21:01:16.702195+00:00"} {"global_step": 5478, "acc_step": 0, "speed/wps": 12908.237524909111, "speed/FLOPS": 202741608105843.03, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05827542021870613, "optim/lr": 0.0029994432694073017, "optim/total_tokens": 2872049664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.0498270988464355, "created_at": "2025-01-15T21:01:26.861552+00:00"} {"global_step": 5479, "acc_step": 0, "speed/wps": 12910.059828120267, "speed/FLOPS": 202770229881881.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06426315009593964, "optim/lr": 0.0029994409328054335, "optim/total_tokens": 2872573952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.093961715698242, "created_at": "2025-01-15T21:01:37.019026+00:00"} {"global_step": 5480, "acc_step": 0, "speed/wps": 12909.522663459396, "speed/FLOPS": 202761792972740.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06283766776323318, "optim/lr": 0.002999438591311376, "optim/total_tokens": 2873098240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 3.0283467769622803, "created_at": "2025-01-15T21:01:47.175962+00:00"} {"global_step": 5481, "acc_step": 0, "speed/wps": 12911.638041480084, "speed/FLOPS": 202795017891389.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060111839324235916, "optim/lr": 0.002999436244925135, "optim/total_tokens": 2873622528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0773468017578125, "created_at": "2025-01-15T21:01:57.332941+00:00"} {"global_step": 5482, "acc_step": 0, "speed/wps": 12910.1661044026, "speed/FLOPS": 202771899096934.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06601449847221375, "optim/lr": 0.002999433893646721, "optim/total_tokens": 2874146816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 3.0910511016845703, "created_at": "2025-01-15T21:02:07.489138+00:00"} {"global_step": 5483, "acc_step": 0, "speed/wps": 12910.904905560828, "speed/FLOPS": 202783502984343.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07845485955476761, "optim/lr": 0.0029994315374761394, "optim/total_tokens": 2874671104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.0579099655151367, "created_at": "2025-01-15T21:02:17.644867+00:00"} {"global_step": 5484, "acc_step": 0, "speed/wps": 12907.267769233753, "speed/FLOPS": 202726376760377.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10326316952705383, "optim/lr": 0.0029994291764133987, "optim/total_tokens": 2875195392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.119199752807617, "created_at": "2025-01-15T21:02:27.804900+00:00"} {"global_step": 5485, "acc_step": 0, "speed/wps": 12912.209714552728, "speed/FLOPS": 202803996802556.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10022362321615219, "optim/lr": 0.0029994268104585075, "optim/total_tokens": 2875719680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.1535532474517822, "created_at": "2025-01-15T21:02:37.961507+00:00"} {"global_step": 5486, "acc_step": 0, "speed/wps": 12906.28318712607, "speed/FLOPS": 202710912545418.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07274552434682846, "optim/lr": 0.002999424439611472, "optim/total_tokens": 2876243968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302164, "loss/out": 3.1171138286590576, "created_at": "2025-01-15T21:02:48.121908+00:00"} {"global_step": 5487, "acc_step": 0, "speed/wps": 12911.1859245821, "speed/FLOPS": 202787916774233.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11667855083942413, "optim/lr": 0.0029994220638723005, "optim/total_tokens": 2876768256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.0340728759765625, "created_at": "2025-01-15T21:02:58.278426+00:00"} {"global_step": 5488, "acc_step": 0, "speed/wps": 12906.428975733113, "speed/FLOPS": 202713202355826.97, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09179063886404037, "optim/lr": 0.0029994196832410004, "optim/total_tokens": 2877292544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.0735576152801514, "created_at": "2025-01-15T21:03:08.437533+00:00"} {"global_step": 5489, "acc_step": 0, "speed/wps": 12912.88402848166, "speed/FLOPS": 202814587829414.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09853368252515793, "optim/lr": 0.0029994172977175805, "optim/total_tokens": 2877816832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.2023167610168457, "created_at": "2025-01-15T21:03:18.593751+00:00"} {"global_step": 5490, "acc_step": 0, "speed/wps": 12908.785601153922, "speed/FLOPS": 202750216396403.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08217126131057739, "optim/lr": 0.0029994149073020475, "optim/total_tokens": 2878341120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 3.129314661026001, "created_at": "2025-01-15T21:03:28.752933+00:00"} {"global_step": 5491, "acc_step": 0, "speed/wps": 12906.606305259409, "speed/FLOPS": 202715987559713.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.08551723510026932, "optim/lr": 0.00299941251199441, "optim/total_tokens": 2878865408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.0961830615997314, "created_at": "2025-01-15T21:03:38.912190+00:00"} {"global_step": 5492, "acc_step": 0, "speed/wps": 12911.066952025225, "speed/FLOPS": 202786048146741.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07683335989713669, "optim/lr": 0.002999410111794675, "optim/total_tokens": 2879389696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.124711751937866, "created_at": "2025-01-15T21:03:49.068550+00:00"} {"global_step": 5493, "acc_step": 0, "speed/wps": 12909.71632042106, "speed/FLOPS": 202764834621436.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09223860502243042, "optim/lr": 0.0029994077067028516, "optim/total_tokens": 2879913984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.084697723388672, "created_at": "2025-01-15T21:03:59.225076+00:00"} {"global_step": 5494, "acc_step": 0, "speed/wps": 12908.655476902048, "speed/FLOPS": 202748172616218.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09096874296665192, "optim/lr": 0.0029994052967189464, "optim/total_tokens": 2880438272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.040052890777588, "created_at": "2025-01-15T21:04:09.384273+00:00"} {"global_step": 5495, "acc_step": 0, "speed/wps": 12912.900507569411, "speed/FLOPS": 202814846656132.7, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08197406679391861, "optim/lr": 0.0029994028818429675, "optim/total_tokens": 2880962560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 3.1033506393432617, "created_at": "2025-01-15T21:04:19.540045+00:00"} {"global_step": 5496, "acc_step": 0, "speed/wps": 12917.26718578259, "speed/FLOPS": 202883431337915.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07613944262266159, "optim/lr": 0.002999400462074923, "optim/total_tokens": 2881486848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.1517863273620605, "created_at": "2025-01-15T21:04:29.697339+00:00"} {"global_step": 5497, "acc_step": 0, "speed/wps": 12911.773985264099, "speed/FLOPS": 202797153075323.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08020729571580887, "optim/lr": 0.0029993980374148216, "optim/total_tokens": 2882011136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.2223174571990967, "created_at": "2025-01-15T21:04:39.855442+00:00"} {"global_step": 5498, "acc_step": 0, "speed/wps": 12910.541653608288, "speed/FLOPS": 202777797613268.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06563178449869156, "optim/lr": 0.0029993956078626697, "optim/total_tokens": 2882535424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.248318672180176, "created_at": "2025-01-15T21:04:50.012189+00:00"} {"global_step": 5499, "acc_step": 0, "speed/wps": 12908.44032374989, "speed/FLOPS": 202744793340312.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08224254846572876, "optim/lr": 0.002999393173418476, "optim/total_tokens": 2883059712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 3.1015539169311523, "created_at": "2025-01-15T21:05:00.173058+00:00"} {"global_step": 5500, "acc_step": 0, "speed/wps": 12911.511288915317, "speed/FLOPS": 202793027068183.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06650097668170929, "optim/lr": 0.0029993907340822483, "optim/total_tokens": 2883584000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.0862183570861816, "created_at": "2025-01-15T21:05:10.328623+00:00"} {"global_step": 5501, "acc_step": 0, "speed/wps": 12910.761945993216, "speed/FLOPS": 202781257607886.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07760969549417496, "optim/lr": 0.0029993882898539945, "optim/total_tokens": 2884108288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349355, "loss/out": 3.059828281402588, "created_at": "2025-01-15T21:05:20.488440+00:00"} {"global_step": 5502, "acc_step": 0, "speed/wps": 12916.07887156532, "speed/FLOPS": 202864767230218.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06511232256889343, "optim/lr": 0.002999385840733723, "optim/total_tokens": 2884632576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.0375304222106934, "created_at": "2025-01-15T21:05:30.641961+00:00"} {"global_step": 5503, "acc_step": 0, "speed/wps": 12909.159145893927, "speed/FLOPS": 202756083429849.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06665606051683426, "optim/lr": 0.0029993833867214413, "optim/total_tokens": 2885156864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.071542263031006, "created_at": "2025-01-15T21:05:40.800138+00:00"} {"global_step": 5504, "acc_step": 0, "speed/wps": 12913.680996755453, "speed/FLOPS": 202827105311304.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06981708854436874, "optim/lr": 0.0029993809278171576, "optim/total_tokens": 2885681152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.195585012435913, "created_at": "2025-01-15T21:05:50.956337+00:00"} {"global_step": 5505, "acc_step": 0, "speed/wps": 12910.00876899311, "speed/FLOPS": 202769427928126.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06600261479616165, "optim/lr": 0.00299937846402088, "optim/total_tokens": 2886205440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 3.1711788177490234, "created_at": "2025-01-15T21:06:01.117313+00:00"} {"global_step": 5506, "acc_step": 0, "speed/wps": 12911.043686036983, "speed/FLOPS": 202785682722425.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0725952535867691, "optim/lr": 0.0029993759953326165, "optim/total_tokens": 2886729728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 3.196835994720459, "created_at": "2025-01-15T21:06:11.278927+00:00"} {"global_step": 5507, "acc_step": 0, "speed/wps": 12911.314914181095, "speed/FLOPS": 202789942733135.53, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0750618577003479, "optim/lr": 0.0029993735217523753, "optim/total_tokens": 2887254016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456346, "loss/out": 3.2046263217926025, "created_at": "2025-01-15T21:06:21.435301+00:00"} {"global_step": 5508, "acc_step": 0, "speed/wps": 12909.289075448052, "speed/FLOPS": 202758124152038.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07727383077144623, "optim/lr": 0.002999371043280164, "optim/total_tokens": 2887778304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1204230785369873, "created_at": "2025-01-15T21:06:31.594780+00:00"} {"global_step": 5509, "acc_step": 0, "speed/wps": 12906.766437578843, "speed/FLOPS": 202718502657835.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.071869857609272, "optim/lr": 0.002999368559915991, "optim/total_tokens": 2888302592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497150, "loss/out": 3.064662218093872, "created_at": "2025-01-15T21:06:41.754820+00:00"} {"global_step": 5510, "acc_step": 0, "speed/wps": 12914.975035619653, "speed/FLOPS": 202847429969862.0, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06447988748550415, "optim/lr": 0.0029993660716598645, "optim/total_tokens": 2888826880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 2.9812910556793213, "created_at": "2025-01-15T21:06:51.911662+00:00"} {"global_step": 5511, "acc_step": 0, "speed/wps": 12911.655730368786, "speed/FLOPS": 202795295719693.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.069778673350811, "optim/lr": 0.0029993635785117924, "optim/total_tokens": 2889351168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 3.110276460647583, "created_at": "2025-01-15T21:07:02.067868+00:00"} {"global_step": 5512, "acc_step": 0, "speed/wps": 12912.070373556502, "speed/FLOPS": 202801808260736.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07764855027198792, "optim/lr": 0.0029993610804717828, "optim/total_tokens": 2889875456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.0515098571777344, "created_at": "2025-01-15T21:07:12.226761+00:00"} {"global_step": 5513, "acc_step": 0, "speed/wps": 12905.564510073802, "speed/FLOPS": 202699624734747.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0772300437092781, "optim/lr": 0.0029993585775398442, "optim/total_tokens": 2890399744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.1106667518615723, "created_at": "2025-01-15T21:07:22.390088+00:00"} {"global_step": 5514, "acc_step": 0, "speed/wps": 12912.36119442029, "speed/FLOPS": 202806376002031.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07314597815275192, "optim/lr": 0.002999356069715984, "optim/total_tokens": 2890924032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 2.969174385070801, "created_at": "2025-01-15T21:07:32.545046+00:00"} {"global_step": 5515, "acc_step": 0, "speed/wps": 12912.487479128617, "speed/FLOPS": 202808359476909.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06637702137231827, "optim/lr": 0.0029993535570002113, "optim/total_tokens": 2891448320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.1724600791931152, "created_at": "2025-01-15T21:07:42.699400+00:00"} {"global_step": 5516, "acc_step": 0, "speed/wps": 12908.512123153974, "speed/FLOPS": 202745921048616.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07740963250398636, "optim/lr": 0.0029993510393925343, "optim/total_tokens": 2891972608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.989299774169922, "created_at": "2025-01-15T21:07:52.856901+00:00"} {"global_step": 5517, "acc_step": 0, "speed/wps": 12909.07995418522, "speed/FLOPS": 202754839615243.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08530095964670181, "optim/lr": 0.0029993485168929605, "optim/total_tokens": 2892496896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.1096739768981934, "created_at": "2025-01-15T21:08:03.016987+00:00"} {"global_step": 5518, "acc_step": 0, "speed/wps": 12907.34725832557, "speed/FLOPS": 202727625245794.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0764392837882042, "optim/lr": 0.002999345989501498, "optim/total_tokens": 2893021184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.169703960418701, "created_at": "2025-01-15T21:08:13.179637+00:00"} {"global_step": 5519, "acc_step": 0, "speed/wps": 12906.353318543575, "speed/FLOPS": 202712014055701.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06900288909673691, "optim/lr": 0.0029993434572181564, "optim/total_tokens": 2893545472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0611541271209717, "created_at": "2025-01-15T21:08:23.338994+00:00"} {"global_step": 5520, "acc_step": 0, "speed/wps": 12910.15117422017, "speed/FLOPS": 202771664597902.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05821220204234123, "optim/lr": 0.0029993409200429425, "optim/total_tokens": 2894069760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 3.185497283935547, "created_at": "2025-01-15T21:08:33.496710+00:00"} {"global_step": 5521, "acc_step": 0, "speed/wps": 12911.274558755873, "speed/FLOPS": 202789308895728.25, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08075196295976639, "optim/lr": 0.0029993383779758653, "optim/total_tokens": 2894594048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.0598955154418945, "created_at": "2025-01-15T21:08:43.656210+00:00"} {"global_step": 5522, "acc_step": 0, "speed/wps": 12916.427360080486, "speed/FLOPS": 202870240721222.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07479649037122726, "optim/lr": 0.0029993358310169333, "optim/total_tokens": 2895118336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 3.2147574424743652, "created_at": "2025-01-15T21:08:53.810334+00:00"} {"global_step": 5523, "acc_step": 0, "speed/wps": 12913.664630612851, "speed/FLOPS": 202826848258545.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07346681505441666, "optim/lr": 0.0029993332791661536, "optim/total_tokens": 2895642624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.121608018875122, "created_at": "2025-01-15T21:09:03.971045+00:00"} {"global_step": 5524, "acc_step": 0, "speed/wps": 12910.007647508754, "speed/FLOPS": 202769410313673.75, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0815713182091713, "optim/lr": 0.0029993307224235365, "optim/total_tokens": 2896166912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436895, "loss/out": 3.1302785873413086, "created_at": "2025-01-15T21:09:14.132773+00:00"} {"global_step": 5525, "acc_step": 0, "speed/wps": 12908.691114259207, "speed/FLOPS": 202748732349883.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08100574463605881, "optim/lr": 0.002999328160789088, "optim/total_tokens": 2896691200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.0158751010894775, "created_at": "2025-01-15T21:09:24.298188+00:00"} {"global_step": 5526, "acc_step": 0, "speed/wps": 12909.00211446615, "speed/FLOPS": 202753617035492.47, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.07059784233570099, "optim/lr": 0.0029993255942628184, "optim/total_tokens": 2897215488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.1339199542999268, "created_at": "2025-01-15T21:09:34.458025+00:00"} {"global_step": 5527, "acc_step": 0, "speed/wps": 12908.452336239045, "speed/FLOPS": 202744982012961.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07157010585069656, "optim/lr": 0.002999323022844735, "optim/total_tokens": 2897739776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 3.098205089569092, "created_at": "2025-01-15T21:09:44.620011+00:00"} {"global_step": 5528, "acc_step": 0, "speed/wps": 12915.11108642283, "speed/FLOPS": 202849566834679.28, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07527543604373932, "optim/lr": 0.002999320446534847, "optim/total_tokens": 2898264064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.1450109481811523, "created_at": "2025-01-15T21:09:54.775995+00:00"} {"global_step": 5529, "acc_step": 0, "speed/wps": 12912.328019874889, "speed/FLOPS": 202805854950209.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07438014447689056, "optim/lr": 0.002999317865333162, "optim/total_tokens": 2898788352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.088319778442383, "created_at": "2025-01-15T21:10:04.931172+00:00"} {"global_step": 5530, "acc_step": 0, "speed/wps": 12908.18495768027, "speed/FLOPS": 202740782465279.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07195088267326355, "optim/lr": 0.0029993152792396885, "optim/total_tokens": 2899312640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.0570173263549805, "created_at": "2025-01-15T21:10:15.091235+00:00"} {"global_step": 5531, "acc_step": 0, "speed/wps": 12910.386327124314, "speed/FLOPS": 202775357997396.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0695696696639061, "optim/lr": 0.002999312688254436, "optim/total_tokens": 2899836928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 3.104753255844116, "created_at": "2025-01-15T21:10:25.248515+00:00"} {"global_step": 5532, "acc_step": 0, "speed/wps": 12913.491153473498, "speed/FLOPS": 202824123561688.88, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07158410549163818, "optim/lr": 0.002999310092377412, "optim/total_tokens": 2900361216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0431013107299805, "created_at": "2025-01-15T21:10:35.403684+00:00"} {"global_step": 5533, "acc_step": 0, "speed/wps": 12907.412731319706, "speed/FLOPS": 202728653589136.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09541701525449753, "optim/lr": 0.0029993074916086246, "optim/total_tokens": 2900885504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376202, "loss/out": 3.0521883964538574, "created_at": "2025-01-15T21:10:45.562099+00:00"} {"global_step": 5534, "acc_step": 0, "speed/wps": 12906.552950007672, "speed/FLOPS": 202715149542167.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08430764824151993, "optim/lr": 0.002999304885948083, "optim/total_tokens": 2901409792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.180605411529541, "created_at": "2025-01-15T21:10:55.721235+00:00"} {"global_step": 5535, "acc_step": 0, "speed/wps": 12911.445249595341, "speed/FLOPS": 202791989829915.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0808042511343956, "optim/lr": 0.0029993022753957957, "optim/total_tokens": 2901934080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.092979907989502, "created_at": "2025-01-15T21:11:05.878863+00:00"} {"global_step": 5536, "acc_step": 0, "speed/wps": 12911.421051219697, "speed/FLOPS": 202791609761174.34, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08525720983743668, "optim/lr": 0.0029992996599517707, "optim/total_tokens": 2902458368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0685079097747803, "created_at": "2025-01-15T21:11:16.037387+00:00"} {"global_step": 5537, "acc_step": 0, "speed/wps": 12912.313013611325, "speed/FLOPS": 202805619256219.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07363500446081161, "optim/lr": 0.0029992970396160177, "optim/total_tokens": 2902982656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373317, "loss/out": 3.063791275024414, "created_at": "2025-01-15T21:11:26.191906+00:00"} {"global_step": 5538, "acc_step": 0, "speed/wps": 12907.102206601558, "speed/FLOPS": 202723776371732.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07836569100618362, "optim/lr": 0.0029992944143885434, "optim/total_tokens": 2903506944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.0039453506469727, "created_at": "2025-01-15T21:11:36.350488+00:00"} {"global_step": 5539, "acc_step": 0, "speed/wps": 12913.177236703583, "speed/FLOPS": 202819193067450.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07680333405733109, "optim/lr": 0.002999291784269358, "optim/total_tokens": 2904031232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.1568093299865723, "created_at": "2025-01-15T21:11:46.508251+00:00"} {"global_step": 5540, "acc_step": 0, "speed/wps": 12909.891059200309, "speed/FLOPS": 202767579134080.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0841679498553276, "optim/lr": 0.0029992891492584693, "optim/total_tokens": 2904555520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.2706546783447266, "created_at": "2025-01-15T21:11:56.672502+00:00"} {"global_step": 5541, "acc_step": 0, "speed/wps": 12911.33585717085, "speed/FLOPS": 202790271671568.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07844242453575134, "optim/lr": 0.002999286509355886, "optim/total_tokens": 2905079808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.1058802604675293, "created_at": "2025-01-15T21:12:06.828309+00:00"} {"global_step": 5542, "acc_step": 0, "speed/wps": 12912.926235124245, "speed/FLOPS": 202815250742734.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07970552146434784, "optim/lr": 0.002999283864561617, "optim/total_tokens": 2905604096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.120518207550049, "created_at": "2025-01-15T21:12:16.985044+00:00"} {"global_step": 5543, "acc_step": 0, "speed/wps": 12914.507238709939, "speed/FLOPS": 202840082576574.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08806561678647995, "optim/lr": 0.0029992812148756706, "optim/total_tokens": 2906128384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.087416172027588, "created_at": "2025-01-15T21:12:27.145873+00:00"} {"global_step": 5544, "acc_step": 0, "speed/wps": 12910.197731823077, "speed/FLOPS": 202772395849033.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0970938429236412, "optim/lr": 0.0029992785602980555, "optim/total_tokens": 2906652672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 3.1670470237731934, "created_at": "2025-01-15T21:12:37.301915+00:00"} {"global_step": 5545, "acc_step": 0, "speed/wps": 12911.439770039415, "speed/FLOPS": 202791903765960.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11524031311273575, "optim/lr": 0.00299927590082878, "optim/total_tokens": 2907176960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.0551722049713135, "created_at": "2025-01-15T21:12:47.470680+00:00"} {"global_step": 5546, "acc_step": 0, "speed/wps": 12911.577608461701, "speed/FLOPS": 202794068707792.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07941149920225143, "optim/lr": 0.002999273236467854, "optim/total_tokens": 2907701248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 3.0062777996063232, "created_at": "2025-01-15T21:12:57.625820+00:00"} {"global_step": 5547, "acc_step": 0, "speed/wps": 12909.292750950293, "speed/FLOPS": 202758181880852.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0894416868686676, "optim/lr": 0.0029992705672152852, "optim/total_tokens": 2908225536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.0315864086151123, "created_at": "2025-01-15T21:13:07.784293+00:00"} {"global_step": 5548, "acc_step": 0, "speed/wps": 12911.8106147516, "speed/FLOPS": 202797728391759.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08537887036800385, "optim/lr": 0.002999267893071082, "optim/total_tokens": 2908749824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.147557497024536, "created_at": "2025-01-15T21:13:17.940957+00:00"} {"global_step": 5549, "acc_step": 0, "speed/wps": 12915.503671055025, "speed/FLOPS": 202855732915795.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0796351283788681, "optim/lr": 0.0029992652140352536, "optim/total_tokens": 2909274112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.204155445098877, "created_at": "2025-01-15T21:13:28.093239+00:00"} {"global_step": 5550, "acc_step": 0, "speed/wps": 12908.066498589385, "speed/FLOPS": 202738921902476.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07387719303369522, "optim/lr": 0.002999262530107809, "optim/total_tokens": 2909798400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 3.0322670936584473, "created_at": "2025-01-15T21:13:38.253102+00:00"} {"global_step": 5551, "acc_step": 0, "speed/wps": 12910.638108769774, "speed/FLOPS": 202779312574122.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08152948319911957, "optim/lr": 0.002999259841288757, "optim/total_tokens": 2910322688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 3.1690120697021484, "created_at": "2025-01-15T21:13:48.411047+00:00"} {"global_step": 5552, "acc_step": 0, "speed/wps": 12909.27147711548, "speed/FLOPS": 202757847746043.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06615851819515228, "optim/lr": 0.002999257147578106, "optim/total_tokens": 2910846976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404279, "loss/out": 3.135432004928589, "created_at": "2025-01-15T21:13:58.576584+00:00"} {"global_step": 5553, "acc_step": 0, "speed/wps": 12908.276261859319, "speed/FLOPS": 202742216522876.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.08230388164520264, "optim/lr": 0.002999254448975864, "optim/total_tokens": 2911371264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.999781370162964, "created_at": "2025-01-15T21:14:08.734570+00:00"} {"global_step": 5554, "acc_step": 0, "speed/wps": 12906.035095098414, "speed/FLOPS": 202707015919210.3, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07344871014356613, "optim/lr": 0.0029992517454820416, "optim/total_tokens": 2911895552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 3.060934543609619, "created_at": "2025-01-15T21:14:18.895334+00:00"} {"global_step": 5555, "acc_step": 0, "speed/wps": 12914.593484313515, "speed/FLOPS": 202841437182292.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08134867250919342, "optim/lr": 0.0029992490370966463, "optim/total_tokens": 2912419840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.1512866020202637, "created_at": "2025-01-15T21:14:29.049310+00:00"} {"global_step": 5556, "acc_step": 0, "speed/wps": 12912.693417223296, "speed/FLOPS": 202811594017674.97, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0989311933517456, "optim/lr": 0.0029992463238196878, "optim/total_tokens": 2912944128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.1714160442352295, "created_at": "2025-01-15T21:14:39.205186+00:00"} {"global_step": 5557, "acc_step": 0, "speed/wps": 12909.251286812001, "speed/FLOPS": 202757530629582.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07227449864149094, "optim/lr": 0.002999243605651174, "optim/total_tokens": 2913468416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.0574076175689697, "created_at": "2025-01-15T21:14:49.363704+00:00"} {"global_step": 5558, "acc_step": 0, "speed/wps": 12911.520329219178, "speed/FLOPS": 202793169058578.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0680326297879219, "optim/lr": 0.0029992408825911146, "optim/total_tokens": 2913992704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.0900063514709473, "created_at": "2025-01-15T21:14:59.519604+00:00"} {"global_step": 5559, "acc_step": 0, "speed/wps": 12911.278337254309, "speed/FLOPS": 202789368242238.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07803349941968918, "optim/lr": 0.002999238154639518, "optim/total_tokens": 2914516992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.0870814323425293, "created_at": "2025-01-15T21:15:09.674854+00:00"} {"global_step": 5560, "acc_step": 0, "speed/wps": 12911.386290718012, "speed/FLOPS": 202791063799729.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08016981184482574, "optim/lr": 0.0029992354217963937, "optim/total_tokens": 2915041280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.1491293907165527, "created_at": "2025-01-15T21:15:19.832625+00:00"} {"global_step": 5561, "acc_step": 0, "speed/wps": 12907.414438321259, "speed/FLOPS": 202728680399941.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06449400633573532, "optim/lr": 0.0029992326840617488, "optim/total_tokens": 2915565568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.9911046028137207, "created_at": "2025-01-15T21:15:29.994248+00:00"} {"global_step": 5562, "acc_step": 0, "speed/wps": 12908.408682444253, "speed/FLOPS": 202744296370127.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06834127008914948, "optim/lr": 0.0029992299414355952, "optim/total_tokens": 2916089856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.236811876296997, "created_at": "2025-01-15T21:15:40.155103+00:00"} {"global_step": 5563, "acc_step": 0, "speed/wps": 12915.20646378163, "speed/FLOPS": 202851064867161.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0685657188296318, "optim/lr": 0.002999227193917939, "optim/total_tokens": 2916614144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.0996227264404297, "created_at": "2025-01-15T21:15:50.307659+00:00"} {"global_step": 5564, "acc_step": 0, "speed/wps": 12906.215992867428, "speed/FLOPS": 202709857167252.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0598917230963707, "optim/lr": 0.002999224441508791, "optim/total_tokens": 2917138432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.111611843109131, "created_at": "2025-01-15T21:16:00.470065+00:00"} {"global_step": 5565, "acc_step": 0, "speed/wps": 12912.725827114153, "speed/FLOPS": 202812103059546.75, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05971387401223183, "optim/lr": 0.0029992216842081596, "optim/total_tokens": 2917662720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.9370980262756348, "created_at": "2025-01-15T21:16:10.625836+00:00"} {"global_step": 5566, "acc_step": 0, "speed/wps": 12908.844171598772, "speed/FLOPS": 202751136325724.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06466034054756165, "optim/lr": 0.002999218922016054, "optim/total_tokens": 2918187008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.117957592010498, "created_at": "2025-01-15T21:16:20.790092+00:00"} {"global_step": 5567, "acc_step": 0, "speed/wps": 12911.929638921569, "speed/FLOPS": 202799597829904.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07304537296295166, "optim/lr": 0.002999216154932482, "optim/total_tokens": 2918711296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489532, "loss/out": 3.085885524749756, "created_at": "2025-01-15T21:16:30.953023+00:00"} {"global_step": 5568, "acc_step": 0, "speed/wps": 12915.968600503396, "speed/FLOPS": 202863035271662.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07946345210075378, "optim/lr": 0.0029992133829574545, "optim/total_tokens": 2919235584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0608179569244385, "created_at": "2025-01-15T21:16:41.104997+00:00"} {"global_step": 5569, "acc_step": 0, "speed/wps": 12913.945496056189, "speed/FLOPS": 202831259636282.2, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07957866042852402, "optim/lr": 0.0029992106060909795, "optim/total_tokens": 2919759872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 3.096613883972168, "created_at": "2025-01-15T21:16:51.262790+00:00"} {"global_step": 5570, "acc_step": 0, "speed/wps": 12910.7090384195, "speed/FLOPS": 202780426621739.12, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06515802443027496, "optim/lr": 0.0029992078243330665, "optim/total_tokens": 2920284160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 3.0596234798431396, "created_at": "2025-01-15T21:17:01.419913+00:00"} {"global_step": 5571, "acc_step": 0, "speed/wps": 12898.004268640723, "speed/FLOPS": 202580880754180.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.068567655980587, "optim/lr": 0.0029992050376837238, "optim/total_tokens": 2920808448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.079359531402588, "created_at": "2025-01-15T21:17:11.588604+00:00"} {"global_step": 5572, "acc_step": 0, "speed/wps": 12906.30920669965, "speed/FLOPS": 202711321218576.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0666709765791893, "optim/lr": 0.002999202246142961, "optim/total_tokens": 2921332736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 3.179138660430908, "created_at": "2025-01-15T21:17:21.747793+00:00"} {"global_step": 5573, "acc_step": 0, "speed/wps": 12911.46520512075, "speed/FLOPS": 202792303258864.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0725851058959961, "optim/lr": 0.0029991994497107874, "optim/total_tokens": 2921857024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.1476097106933594, "created_at": "2025-01-15T21:17:31.902963+00:00"} {"global_step": 5574, "acc_step": 0, "speed/wps": 12911.013979176681, "speed/FLOPS": 202785216135363.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07072455435991287, "optim/lr": 0.002999196648387212, "optim/total_tokens": 2922381312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329132, "loss/out": 3.0864102840423584, "created_at": "2025-01-15T21:17:42.060308+00:00"} {"global_step": 5575, "acc_step": 0, "speed/wps": 12907.579138867764, "speed/FLOPS": 202731267248346.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06780259311199188, "optim/lr": 0.0029991938421722437, "optim/total_tokens": 2922905600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.1144254207611084, "created_at": "2025-01-15T21:17:52.220149+00:00"} {"global_step": 5576, "acc_step": 0, "speed/wps": 12911.299184182002, "speed/FLOPS": 202789695671884.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06265401095151901, "optim/lr": 0.0029991910310658916, "optim/total_tokens": 2923429888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 3.113473415374756, "created_at": "2025-01-15T21:18:02.376910+00:00"} {"global_step": 5577, "acc_step": 0, "speed/wps": 12912.924602490733, "speed/FLOPS": 202815225099981.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08223047852516174, "optim/lr": 0.0029991882150681654, "optim/total_tokens": 2923954176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.0524837970733643, "created_at": "2025-01-15T21:18:12.531583+00:00"} {"global_step": 5578, "acc_step": 0, "speed/wps": 12913.918585397305, "speed/FLOPS": 202830836967405.28, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07391510903835297, "optim/lr": 0.002999185394179074, "optim/total_tokens": 2924478464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.2662765979766846, "created_at": "2025-01-15T21:18:22.687979+00:00"} {"global_step": 5579, "acc_step": 0, "speed/wps": 12912.66609120069, "speed/FLOPS": 202811164824940.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08205556124448776, "optim/lr": 0.0029991825683986268, "optim/total_tokens": 2925002752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 3.0723252296447754, "created_at": "2025-01-15T21:18:32.843083+00:00"} {"global_step": 5580, "acc_step": 0, "speed/wps": 12908.469907475013, "speed/FLOPS": 202745257993367.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08006194978952408, "optim/lr": 0.002999179737726832, "optim/total_tokens": 2925527040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 3.1610944271087646, "created_at": "2025-01-15T21:18:43.000565+00:00"} {"global_step": 5581, "acc_step": 0, "speed/wps": 12906.781864171793, "speed/FLOPS": 202718744953675.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06214244291186333, "optim/lr": 0.0029991769021637, "optim/total_tokens": 2926051328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.1079304218292236, "created_at": "2025-01-15T21:18:53.160694+00:00"} {"global_step": 5582, "acc_step": 0, "speed/wps": 12911.400192053414, "speed/FLOPS": 202791282139304.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06200961768627167, "optim/lr": 0.00299917406170924, "optim/total_tokens": 2926575616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 2.8675148487091064, "created_at": "2025-01-15T21:19:03.315996+00:00"} {"global_step": 5583, "acc_step": 0, "speed/wps": 12910.557471772245, "speed/FLOPS": 202778046059269.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07065260410308838, "optim/lr": 0.0029991712163634607, "optim/total_tokens": 2927099904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 3.0699634552001953, "created_at": "2025-01-15T21:19:13.476705+00:00"} {"global_step": 5584, "acc_step": 0, "speed/wps": 12905.128050532909, "speed/FLOPS": 202692769537899.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07616880536079407, "optim/lr": 0.0029991683661263722, "optim/total_tokens": 2927624192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492439, "loss/out": 3.032453775405884, "created_at": "2025-01-15T21:19:23.638159+00:00"} {"global_step": 5585, "acc_step": 0, "speed/wps": 12914.639891457087, "speed/FLOPS": 202842166070252.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057076700031757355, "optim/lr": 0.0029991655109979825, "optim/total_tokens": 2928148480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.1151318550109863, "created_at": "2025-01-15T21:19:33.794799+00:00"} {"global_step": 5586, "acc_step": 0, "speed/wps": 12913.41833062722, "speed/FLOPS": 202822979778819.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07466167211532593, "optim/lr": 0.002999162650978302, "optim/total_tokens": 2928672768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.031737804412842, "created_at": "2025-01-15T21:19:43.948406+00:00"} {"global_step": 5587, "acc_step": 0, "speed/wps": 12909.090851276122, "speed/FLOPS": 202755010769030.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07984336465597153, "optim/lr": 0.00299915978606734, "optim/total_tokens": 2929197056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 3.06058669090271, "created_at": "2025-01-15T21:19:54.105539+00:00"} {"global_step": 5588, "acc_step": 0, "speed/wps": 12908.90656112712, "speed/FLOPS": 202752116238998.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10117670893669128, "optim/lr": 0.002999156916265105, "optim/total_tokens": 2929721344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.170037269592285, "created_at": "2025-01-15T21:20:04.265727+00:00"} {"global_step": 5589, "acc_step": 0, "speed/wps": 12910.00632588647, "speed/FLOPS": 202769389555779.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10726878046989441, "optim/lr": 0.002999154041571607, "optim/total_tokens": 2930245632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 3.081953763961792, "created_at": "2025-01-15T21:20:14.422581+00:00"} {"global_step": 5590, "acc_step": 0, "speed/wps": 12908.061848257865, "speed/FLOPS": 202738848862630.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08645446598529816, "optim/lr": 0.002999151161986855, "optim/total_tokens": 2930769920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.0164380073547363, "created_at": "2025-01-15T21:20:24.581948+00:00"} {"global_step": 5591, "acc_step": 0, "speed/wps": 12910.646147061047, "speed/FLOPS": 202779438826532.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08482465147972107, "optim/lr": 0.002999148277510859, "optim/total_tokens": 2931294208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 3.2468318939208984, "created_at": "2025-01-15T21:20:34.739808+00:00"} {"global_step": 5592, "acc_step": 0, "speed/wps": 12909.550056702505, "speed/FLOPS": 202762223221265.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09795552492141724, "optim/lr": 0.002999145388143628, "optim/total_tokens": 2931818496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.1245813369750977, "created_at": "2025-01-15T21:20:44.896487+00:00"} {"global_step": 5593, "acc_step": 0, "speed/wps": 12907.984032655093, "speed/FLOPS": 202737626661658.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07795887440443039, "optim/lr": 0.0029991424938851717, "optim/total_tokens": 2932342784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346953, "loss/out": 3.0495598316192627, "created_at": "2025-01-15T21:20:55.054690+00:00"} {"global_step": 5594, "acc_step": 0, "speed/wps": 12905.707511962848, "speed/FLOPS": 202701870775919.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08869285881519318, "optim/lr": 0.0029991395947354996, "optim/total_tokens": 2932867072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.102684497833252, "created_at": "2025-01-15T21:21:05.214264+00:00"} {"global_step": 5595, "acc_step": 0, "speed/wps": 12909.536707618274, "speed/FLOPS": 202762013555554.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07291506975889206, "optim/lr": 0.00299913669069462, "optim/total_tokens": 2933391360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 3.0598349571228027, "created_at": "2025-01-15T21:21:15.391100+00:00"} {"global_step": 5596, "acc_step": 0, "speed/wps": 12908.485025572194, "speed/FLOPS": 202745495443858.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08697651326656342, "optim/lr": 0.002999133781762544, "optim/total_tokens": 2933915648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.018754005432129, "created_at": "2025-01-15T21:21:25.548648+00:00"} {"global_step": 5597, "acc_step": 0, "speed/wps": 12910.3566540811, "speed/FLOPS": 202774891941477.75, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06570527702569962, "optim/lr": 0.00299913086793928, "optim/total_tokens": 2934439936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.068218946456909, "created_at": "2025-01-15T21:21:35.704608+00:00"} {"global_step": 5598, "acc_step": 0, "speed/wps": 12904.718574376342, "speed/FLOPS": 202686338152181.12, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07174210995435715, "optim/lr": 0.002999127949224839, "optim/total_tokens": 2934964224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.062094211578369, "created_at": "2025-01-15T21:21:45.865074+00:00"} {"global_step": 5599, "acc_step": 0, "speed/wps": 12910.98736025934, "speed/FLOPS": 202784798048688.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0698808953166008, "optim/lr": 0.0029991250256192277, "optim/total_tokens": 2935488512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0475692749023438, "created_at": "2025-01-15T21:21:56.022954+00:00"} {"global_step": 5600, "acc_step": 0, "speed/wps": 12909.623191180031, "speed/FLOPS": 202763371898732.8, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0771612599492073, "optim/lr": 0.0029991220971224584, "optim/total_tokens": 2936012800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 3.146317481994629, "created_at": "2025-01-15T21:22:06.180067+00:00"} {"global_step": 5601, "acc_step": 0, "speed/wps": 12911.641871126893, "speed/FLOPS": 202795078041255.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07790350168943405, "optim/lr": 0.002999119163734539, "optim/total_tokens": 2936537088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.1155264377593994, "created_at": "2025-01-15T21:22:16.336964+00:00"} {"global_step": 5602, "acc_step": 0, "speed/wps": 12909.53578989195, "speed/FLOPS": 202761999141401.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06980045884847641, "optim/lr": 0.0029991162254554803, "optim/total_tokens": 2937061376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 2.9835543632507324, "created_at": "2025-01-15T21:22:26.496206+00:00"} {"global_step": 5603, "acc_step": 0, "speed/wps": 12912.05740663206, "speed/FLOPS": 202801604597369.5, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07321876287460327, "optim/lr": 0.002999113282285291, "optim/total_tokens": 2937585664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.0233888626098633, "created_at": "2025-01-15T21:22:36.652270+00:00"} {"global_step": 5604, "acc_step": 0, "speed/wps": 12900.409666596204, "speed/FLOPS": 202618660834434.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0640299841761589, "optim/lr": 0.0029991103342239807, "optim/total_tokens": 2938109952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 2.9288454055786133, "created_at": "2025-01-15T21:22:46.816562+00:00"} {"global_step": 5605, "acc_step": 0, "speed/wps": 12914.228496094733, "speed/FLOPS": 202835704540770.3, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06219986826181412, "optim/lr": 0.00299910738127156, "optim/total_tokens": 2938634240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.108441114425659, "created_at": "2025-01-15T21:22:56.969790+00:00"} {"global_step": 5606, "acc_step": 0, "speed/wps": 12907.537010972932, "speed/FLOPS": 202730605571868.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06670255213975906, "optim/lr": 0.002999104423428037, "optim/total_tokens": 2939158528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 2.9982266426086426, "created_at": "2025-01-15T21:23:07.128320+00:00"} {"global_step": 5607, "acc_step": 0, "speed/wps": 12907.751202335077, "speed/FLOPS": 202733969741540.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07033558189868927, "optim/lr": 0.0029991014606934223, "optim/total_tokens": 2939682816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.034850835800171, "created_at": "2025-01-15T21:23:17.288034+00:00"} {"global_step": 5608, "acc_step": 0, "speed/wps": 12900.627986189455, "speed/FLOPS": 202622089843648.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07199729233980179, "optim/lr": 0.0029990984930677255, "optim/total_tokens": 2940207104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.116880416870117, "created_at": "2025-01-15T21:23:27.454605+00:00"} {"global_step": 5609, "acc_step": 0, "speed/wps": 12909.399117623705, "speed/FLOPS": 202759852515621.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0688481330871582, "optim/lr": 0.0029990955205509563, "optim/total_tokens": 2940731392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.1919636726379395, "created_at": "2025-01-15T21:23:37.612286+00:00"} {"global_step": 5610, "acc_step": 0, "speed/wps": 12902.5060912511, "speed/FLOPS": 202651588064428.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06711608171463013, "optim/lr": 0.0029990925431431243, "optim/total_tokens": 2941255680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 3.0194218158721924, "created_at": "2025-01-15T21:23:47.774768+00:00"} {"global_step": 5611, "acc_step": 0, "speed/wps": 12909.659369132463, "speed/FLOPS": 202763940123189.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06269055604934692, "optim/lr": 0.0029990895608442395, "optim/total_tokens": 2941779968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.1322598457336426, "created_at": "2025-01-15T21:23:57.932272+00:00"} {"global_step": 5612, "acc_step": 0, "speed/wps": 12908.05855111802, "speed/FLOPS": 202738797076518.16, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057808585464954376, "optim/lr": 0.002999086573654311, "optim/total_tokens": 2942304256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 3.031928539276123, "created_at": "2025-01-15T21:24:08.090166+00:00"} {"global_step": 5613, "acc_step": 0, "speed/wps": 12903.266909565316, "speed/FLOPS": 202663537761527.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06147056818008423, "optim/lr": 0.002999083581573349, "optim/total_tokens": 2942828544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.057048797607422, "created_at": "2025-01-15T21:24:18.254338+00:00"} {"global_step": 5614, "acc_step": 0, "speed/wps": 12903.708964664218, "speed/FLOPS": 202670480844303.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08091620355844498, "optim/lr": 0.002999080584601363, "optim/total_tokens": 2943352832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.994717597961426, "created_at": "2025-01-15T21:24:28.417200+00:00"} {"global_step": 5615, "acc_step": 0, "speed/wps": 12901.316457189647, "speed/FLOPS": 202632903226765.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08764589577913284, "optim/lr": 0.0029990775827383634, "optim/total_tokens": 2943877120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392536, "loss/out": 3.0595526695251465, "created_at": "2025-01-15T21:24:38.580355+00:00"} {"global_step": 5616, "acc_step": 0, "speed/wps": 12910.29909314817, "speed/FLOPS": 202773987867929.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07116153091192245, "optim/lr": 0.0029990745759843596, "optim/total_tokens": 2944401408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.1231040954589844, "created_at": "2025-01-15T21:24:48.737942+00:00"} {"global_step": 5617, "acc_step": 0, "speed/wps": 12912.168521912256, "speed/FLOPS": 202803349815533.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06208537518978119, "optim/lr": 0.0029990715643393607, "optim/total_tokens": 2944925696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 3.047318458557129, "created_at": "2025-01-15T21:24:58.897998+00:00"} {"global_step": 5618, "acc_step": 0, "speed/wps": 12906.427683249665, "speed/FLOPS": 202713182055598.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07530032843351364, "optim/lr": 0.0029990685478033774, "optim/total_tokens": 2945449984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.0677285194396973, "created_at": "2025-01-15T21:25:09.059933+00:00"} {"global_step": 5619, "acc_step": 0, "speed/wps": 12909.708683545357, "speed/FLOPS": 202764714673809.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07826432585716248, "optim/lr": 0.002999065526376419, "optim/total_tokens": 2945974272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.880488157272339, "created_at": "2025-01-15T21:25:19.222338+00:00"} {"global_step": 5620, "acc_step": 0, "speed/wps": 12911.160484678789, "speed/FLOPS": 202787517205593.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0777490958571434, "optim/lr": 0.002999062500058496, "optim/total_tokens": 2946498560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0462584495544434, "created_at": "2025-01-15T21:25:29.378263+00:00"} {"global_step": 5621, "acc_step": 0, "speed/wps": 12910.384441688813, "speed/FLOPS": 202775328384041.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06697716563940048, "optim/lr": 0.002999059468849618, "optim/total_tokens": 2947022848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.0653281211853027, "created_at": "2025-01-15T21:25:39.534656+00:00"} {"global_step": 5622, "acc_step": 0, "speed/wps": 12911.441643938042, "speed/FLOPS": 202791933198113.3, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06650952994823456, "optim/lr": 0.002999056432749795, "optim/total_tokens": 2947547136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.986176013946533, "created_at": "2025-01-15T21:25:49.689796+00:00"} {"global_step": 5623, "acc_step": 0, "speed/wps": 12911.275875899904, "speed/FLOPS": 202789329583285.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07133613526821136, "optim/lr": 0.0029990533917590365, "optim/total_tokens": 2948071424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447797, "loss/out": 3.0919346809387207, "created_at": "2025-01-15T21:25:59.845607+00:00"} {"global_step": 5624, "acc_step": 0, "speed/wps": 12914.808648272123, "speed/FLOPS": 202844816627928.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0673549696803093, "optim/lr": 0.002999050345877353, "optim/total_tokens": 2948595712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 3.0565714836120605, "created_at": "2025-01-15T21:26:10.001623+00:00"} {"global_step": 5625, "acc_step": 0, "speed/wps": 12913.918356698658, "speed/FLOPS": 202830833375378.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062164176255464554, "optim/lr": 0.002999047295104754, "optim/total_tokens": 2949120000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.034236192703247, "created_at": "2025-01-15T21:26:20.154804+00:00"} {"global_step": 5626, "acc_step": 0, "speed/wps": 12912.622707452922, "speed/FLOPS": 202810483423565.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06676691025495529, "optim/lr": 0.002999044239441249, "optim/total_tokens": 2949644288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.100494384765625, "created_at": "2025-01-15T21:26:30.309255+00:00"} {"global_step": 5627, "acc_step": 0, "speed/wps": 12907.170080078968, "speed/FLOPS": 202724842417962.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08268514275550842, "optim/lr": 0.002999041178886849, "optim/total_tokens": 2950168576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.108062505722046, "created_at": "2025-01-15T21:26:40.470106+00:00"} {"global_step": 5628, "acc_step": 0, "speed/wps": 12902.96842924353, "speed/FLOPS": 202658849717918.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08224083483219147, "optim/lr": 0.0029990381134415633, "optim/total_tokens": 2950692864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 3.0507302284240723, "created_at": "2025-01-15T21:26:50.634216+00:00"} {"global_step": 5629, "acc_step": 0, "speed/wps": 12913.825453579842, "speed/FLOPS": 202829374204235.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08623265475034714, "optim/lr": 0.002999035043105402, "optim/total_tokens": 2951217152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.964529514312744, "created_at": "2025-01-15T21:27:00.788719+00:00"} {"global_step": 5630, "acc_step": 0, "speed/wps": 12914.083858428583, "speed/FLOPS": 202833432807470.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14635148644447327, "optim/lr": 0.002999031967878376, "optim/total_tokens": 2951741440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.0667569637298584, "created_at": "2025-01-15T21:27:10.943940+00:00"} {"global_step": 5631, "acc_step": 0, "speed/wps": 12908.678143465055, "speed/FLOPS": 202748528625737.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11193256825208664, "optim/lr": 0.002999028887760494, "optim/total_tokens": 2952265728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.104173421859741, "created_at": "2025-01-15T21:27:21.101625+00:00"} {"global_step": 5632, "acc_step": 0, "speed/wps": 12914.63267894571, "speed/FLOPS": 202842052787850.38, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0637083426117897, "optim/lr": 0.0029990258027517668, "optim/total_tokens": 2952790016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.11967396736145, "created_at": "2025-01-15T21:27:31.257293+00:00"} {"global_step": 5633, "acc_step": 0, "speed/wps": 12902.676707650688, "speed/FLOPS": 202654267829435.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08560816943645477, "optim/lr": 0.002999022712852204, "optim/total_tokens": 2953314304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.0563597679138184, "created_at": "2025-01-15T21:27:41.421332+00:00"} {"global_step": 5634, "acc_step": 0, "speed/wps": 12911.96604540116, "speed/FLOPS": 202800169643694.7, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06052581965923309, "optim/lr": 0.0029990196180618156, "optim/total_tokens": 2953838592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 3.1215696334838867, "created_at": "2025-01-15T21:27:51.576429+00:00"} {"global_step": 5635, "acc_step": 0, "speed/wps": 12914.660070177371, "speed/FLOPS": 202842483004783.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07309633493423462, "optim/lr": 0.002999016518380613, "optim/total_tokens": 2954362880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.064167022705078, "created_at": "2025-01-15T21:28:01.729605+00:00"} {"global_step": 5636, "acc_step": 0, "speed/wps": 12913.418838938744, "speed/FLOPS": 202822987762549.97, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08322056382894516, "optim/lr": 0.0029990134138086046, "optim/total_tokens": 2954887168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 3.0119717121124268, "created_at": "2025-01-15T21:28:11.884589+00:00"} {"global_step": 5637, "acc_step": 0, "speed/wps": 12909.825576455942, "speed/FLOPS": 202766550637597.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07050227373838425, "optim/lr": 0.002999010304345802, "optim/total_tokens": 2955411456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.10129976272583, "created_at": "2025-01-15T21:28:22.041066+00:00"} {"global_step": 5638, "acc_step": 0, "speed/wps": 12910.84891025832, "speed/FLOPS": 202782623501171.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07188034057617188, "optim/lr": 0.002999007189992214, "optim/total_tokens": 2955935744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 3.039198398590088, "created_at": "2025-01-15T21:28:32.199831+00:00"} {"global_step": 5639, "acc_step": 0, "speed/wps": 12908.040207922651, "speed/FLOPS": 202738508971428.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0696454867720604, "optim/lr": 0.0029990040707478516, "optim/total_tokens": 2956460032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.073838710784912, "created_at": "2025-01-15T21:28:42.358773+00:00"} {"global_step": 5640, "acc_step": 0, "speed/wps": 12905.753443419517, "speed/FLOPS": 202702592192563.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06932073086500168, "optim/lr": 0.0029990009466127248, "optim/total_tokens": 2956984320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462167, "loss/out": 3.1495718955993652, "created_at": "2025-01-15T21:28:52.518505+00:00"} {"global_step": 5641, "acc_step": 0, "speed/wps": 12906.935530817067, "speed/FLOPS": 202721158499500.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0676063746213913, "optim/lr": 0.0029989978175868435, "optim/total_tokens": 2957508608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.2382314205169678, "created_at": "2025-01-15T21:29:02.678914+00:00"} {"global_step": 5642, "acc_step": 0, "speed/wps": 12908.897528035206, "speed/FLOPS": 202751974361877.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06466959416866302, "optim/lr": 0.0029989946836702186, "optim/total_tokens": 2958032896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.152595043182373, "created_at": "2025-01-15T21:29:12.839490+00:00"} {"global_step": 5643, "acc_step": 0, "speed/wps": 12909.042712630835, "speed/FLOPS": 202754254685456.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06562729179859161, "optim/lr": 0.0029989915448628596, "optim/total_tokens": 2958557184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.2351300716400146, "created_at": "2025-01-15T21:29:22.996594+00:00"} {"global_step": 5644, "acc_step": 0, "speed/wps": 12909.00536472542, "speed/FLOPS": 202753668085280.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07329030334949493, "optim/lr": 0.0029989884011647774, "optim/total_tokens": 2959081472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.1514012813568115, "created_at": "2025-01-15T21:29:33.154659+00:00"} {"global_step": 5645, "acc_step": 0, "speed/wps": 12909.615278501149, "speed/FLOPS": 202763247619237.8, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0766652375459671, "optim/lr": 0.0029989852525759814, "optim/total_tokens": 2959605760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 3.0171725749969482, "created_at": "2025-01-15T21:29:43.314026+00:00"} {"global_step": 5646, "acc_step": 0, "speed/wps": 12905.105402352263, "speed/FLOPS": 202692413817100.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09246876835823059, "optim/lr": 0.002998982099096483, "optim/total_tokens": 2960130048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.1475400924682617, "created_at": "2025-01-15T21:29:53.475298+00:00"} {"global_step": 5647, "acc_step": 0, "speed/wps": 12906.494374313235, "speed/FLOPS": 202714229530394.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07679286599159241, "optim/lr": 0.002998978940726292, "optim/total_tokens": 2960654336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350938, "loss/out": 3.1460044384002686, "created_at": "2025-01-15T21:30:03.636261+00:00"} {"global_step": 5648, "acc_step": 0, "speed/wps": 12895.830033892633, "speed/FLOPS": 202546731409749.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06855767220258713, "optim/lr": 0.0029989757774654185, "optim/total_tokens": 2961178624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.149933099746704, "created_at": "2025-01-15T21:30:13.807377+00:00"} {"global_step": 5649, "acc_step": 0, "speed/wps": 12904.489650567957, "speed/FLOPS": 202682742589196.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06257481127977371, "optim/lr": 0.002998972609313872, "optim/total_tokens": 2961702912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 3.045259475708008, "created_at": "2025-01-15T21:30:23.969975+00:00"} {"global_step": 5650, "acc_step": 0, "speed/wps": 12898.96886503027, "speed/FLOPS": 202596031065972.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06470393389463425, "optim/lr": 0.0029989694362716644, "optim/total_tokens": 2962227200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.062709331512451, "created_at": "2025-01-15T21:30:34.136239+00:00"} {"global_step": 5651, "acc_step": 0, "speed/wps": 12904.42581759501, "speed/FLOPS": 202681740004640.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06443174183368683, "optim/lr": 0.0029989662583388053, "optim/total_tokens": 2962751488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0504140853881836, "created_at": "2025-01-15T21:30:44.301862+00:00"} {"global_step": 5652, "acc_step": 0, "speed/wps": 12899.81433437447, "speed/FLOPS": 202609310323819.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07015608996152878, "optim/lr": 0.002998963075515305, "optim/total_tokens": 2963275776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342920, "loss/out": 3.0597047805786133, "created_at": "2025-01-15T21:30:54.466345+00:00"} {"global_step": 5653, "acc_step": 0, "speed/wps": 12914.297551365984, "speed/FLOPS": 202836789148697.2, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07037635892629623, "optim/lr": 0.0029989598878011744, "optim/total_tokens": 2963800064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.189127206802368, "created_at": "2025-01-15T21:31:04.621837+00:00"} {"global_step": 5654, "acc_step": 0, "speed/wps": 12914.266890985618, "speed/FLOPS": 202836307585291.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06861034035682678, "optim/lr": 0.0029989566951964227, "optim/total_tokens": 2964324352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398487, "loss/out": 3.0686373710632324, "created_at": "2025-01-15T21:31:14.777532+00:00"} {"global_step": 5655, "acc_step": 0, "speed/wps": 12911.68447003, "speed/FLOPS": 202795747115565.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06882986426353455, "optim/lr": 0.002998953497701062, "optim/total_tokens": 2964848640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.010684013366699, "created_at": "2025-01-15T21:31:24.934250+00:00"} {"global_step": 5656, "acc_step": 0, "speed/wps": 12903.945184391705, "speed/FLOPS": 202674190999723.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06678077578544617, "optim/lr": 0.0029989502953151016, "optim/total_tokens": 2965372928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293783, "loss/out": 3.008418083190918, "created_at": "2025-01-15T21:31:35.095853+00:00"} {"global_step": 5657, "acc_step": 0, "speed/wps": 12913.351588131707, "speed/FLOPS": 202821931496213.2, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06327638030052185, "optim/lr": 0.0029989470880385526, "optim/total_tokens": 2965897216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.9976630210876465, "created_at": "2025-01-15T21:31:45.252237+00:00"} {"global_step": 5658, "acc_step": 0, "speed/wps": 12913.870870503448, "speed/FLOPS": 202830087539430.12, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0640859380364418, "optim/lr": 0.0029989438758714244, "optim/total_tokens": 2966421504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388267, "loss/out": 3.138580322265625, "created_at": "2025-01-15T21:31:55.408502+00:00"} {"global_step": 5659, "acc_step": 0, "speed/wps": 12913.826063829125, "speed/FLOPS": 202829383789039.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0704711452126503, "optim/lr": 0.002998940658813729, "optim/total_tokens": 2966945792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.99287748336792, "created_at": "2025-01-15T21:32:05.561823+00:00"} {"global_step": 5660, "acc_step": 0, "speed/wps": 12914.346811157202, "speed/FLOPS": 202837562841408.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06569027900695801, "optim/lr": 0.0029989374368654753, "optim/total_tokens": 2967470080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479744, "loss/out": 3.115346908569336, "created_at": "2025-01-15T21:32:15.722551+00:00"} {"global_step": 5661, "acc_step": 0, "speed/wps": 12911.302189224802, "speed/FLOPS": 202789742870211.03, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0677972286939621, "optim/lr": 0.002998934210026675, "optim/total_tokens": 2967994368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.1966562271118164, "created_at": "2025-01-15T21:32:25.879189+00:00"} {"global_step": 5662, "acc_step": 0, "speed/wps": 12906.775661808191, "speed/FLOPS": 202718647537032.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08755524456501007, "optim/lr": 0.002998930978297338, "optim/total_tokens": 2968518656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.087620258331299, "created_at": "2025-01-15T21:32:36.039543+00:00"} {"global_step": 5663, "acc_step": 0, "speed/wps": 12907.673782322528, "speed/FLOPS": 202732753753857.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07970426976680756, "optim/lr": 0.0029989277416774753, "optim/total_tokens": 2969042944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.981226921081543, "created_at": "2025-01-15T21:32:46.201078+00:00"} {"global_step": 5664, "acc_step": 0, "speed/wps": 12906.762851105446, "speed/FLOPS": 202718446327342.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08131494373083115, "optim/lr": 0.0029989245001670974, "optim/total_tokens": 2969567232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.121640205383301, "created_at": "2025-01-15T21:32:56.361797+00:00"} {"global_step": 5665, "acc_step": 0, "speed/wps": 12910.301040142516, "speed/FLOPS": 202774018448150.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0781906396150589, "optim/lr": 0.002998921253766214, "optim/total_tokens": 2970091520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 3.0470852851867676, "created_at": "2025-01-15T21:33:06.521552+00:00"} {"global_step": 5666, "acc_step": 0, "speed/wps": 12906.11899898248, "speed/FLOPS": 202708333745005.6, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06392797082662582, "optim/lr": 0.002998918002474837, "optim/total_tokens": 2970615808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.155766487121582, "created_at": "2025-01-15T21:33:16.680893+00:00"} {"global_step": 5667, "acc_step": 0, "speed/wps": 12908.821193052323, "speed/FLOPS": 202750775416076.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07175834476947784, "optim/lr": 0.002998914746292976, "optim/total_tokens": 2971140096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458205, "loss/out": 3.1120433807373047, "created_at": "2025-01-15T21:33:26.841975+00:00"} {"global_step": 5668, "acc_step": 0, "speed/wps": 12905.409568408813, "speed/FLOPS": 202697191162986.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06514694541692734, "optim/lr": 0.0029989114852206422, "optim/total_tokens": 2971664384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0929317474365234, "created_at": "2025-01-15T21:33:37.002046+00:00"} {"global_step": 5669, "acc_step": 0, "speed/wps": 12897.507761258064, "speed/FLOPS": 202573082423463.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06380413472652435, "optim/lr": 0.002998908219257846, "optim/total_tokens": 2972188672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.160585641860962, "created_at": "2025-01-15T21:33:47.169334+00:00"} {"global_step": 5670, "acc_step": 0, "speed/wps": 12914.32995864448, "speed/FLOPS": 202837298149538.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06220221519470215, "optim/lr": 0.002998904948404598, "optim/total_tokens": 2972712960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9888923168182373, "created_at": "2025-01-15T21:33:57.325374+00:00"} {"global_step": 5671, "acc_step": 0, "speed/wps": 12915.967884556056, "speed/FLOPS": 202863024026725.62, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06521549075841904, "optim/lr": 0.002998901672660909, "optim/total_tokens": 2973237248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.0161452293395996, "created_at": "2025-01-15T21:34:07.479063+00:00"} {"global_step": 5672, "acc_step": 0, "speed/wps": 12912.11802776064, "speed/FLOPS": 202802556735496.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07397856563329697, "optim/lr": 0.00299889839202679, "optim/total_tokens": 2973761536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 3.0044543743133545, "created_at": "2025-01-15T21:34:17.633768+00:00"} {"global_step": 5673, "acc_step": 0, "speed/wps": 12907.0514206286, "speed/FLOPS": 202722978708239.53, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.071312315762043, "optim/lr": 0.0029988951065022516, "optim/total_tokens": 2974285824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.070997476577759, "created_at": "2025-01-15T21:34:27.796902+00:00"} {"global_step": 5674, "acc_step": 0, "speed/wps": 12910.60123485054, "speed/FLOPS": 202778733418550.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05980280786752701, "optim/lr": 0.0029988918160873033, "optim/total_tokens": 2974810112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.121342658996582, "created_at": "2025-01-15T21:34:37.952682+00:00"} {"global_step": 5675, "acc_step": 0, "speed/wps": 12911.37053803951, "speed/FLOPS": 202790816382267.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06736759096384048, "optim/lr": 0.002998888520781958, "optim/total_tokens": 2975334400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.1366047859191895, "created_at": "2025-01-15T21:34:48.111649+00:00"} {"global_step": 5676, "acc_step": 0, "speed/wps": 12912.044559380605, "speed/FLOPS": 202801402813631.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07216784358024597, "optim/lr": 0.0029988852205862244, "optim/total_tokens": 2975858688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.1445555686950684, "created_at": "2025-01-15T21:34:58.268922+00:00"} {"global_step": 5677, "acc_step": 0, "speed/wps": 12898.201917243734, "speed/FLOPS": 202583985097088.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.0912242978811264, "optim/lr": 0.0029988819155001146, "optim/total_tokens": 2976382976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.0886330604553223, "created_at": "2025-01-15T21:35:08.438432+00:00"} {"global_step": 5678, "acc_step": 0, "speed/wps": 12902.227522246129, "speed/FLOPS": 202647212755411.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08715219795703888, "optim/lr": 0.002998878605523639, "optim/total_tokens": 2976907264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425464, "loss/out": 3.0990381240844727, "created_at": "2025-01-15T21:35:18.601828+00:00"} {"global_step": 5679, "acc_step": 0, "speed/wps": 12914.411802944878, "speed/FLOPS": 202838583626741.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07755358517169952, "optim/lr": 0.002998875290656808, "optim/total_tokens": 2977431552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.0254108905792236, "created_at": "2025-01-15T21:35:28.757499+00:00"} {"global_step": 5680, "acc_step": 0, "speed/wps": 12908.518508334977, "speed/FLOPS": 202746021336657.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09005847573280334, "optim/lr": 0.002998871970899633, "optim/total_tokens": 2977955840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.996190071105957, "created_at": "2025-01-15T21:35:38.916610+00:00"} {"global_step": 5681, "acc_step": 0, "speed/wps": 12914.122788081482, "speed/FLOPS": 202834044251164.2, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08213996142148972, "optim/lr": 0.0029988686462521243, "optim/total_tokens": 2978480128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.130549192428589, "created_at": "2025-01-15T21:35:49.070153+00:00"} {"global_step": 5682, "acc_step": 0, "speed/wps": 12911.116280745911, "speed/FLOPS": 202786822922085.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06738291680812836, "optim/lr": 0.002998865316714293, "optim/total_tokens": 2979004416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.1877501010894775, "created_at": "2025-01-15T21:35:59.225805+00:00"} {"global_step": 5683, "acc_step": 0, "speed/wps": 12910.882457399439, "speed/FLOPS": 202783150405121.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08887357264757156, "optim/lr": 0.0029988619822861507, "optim/total_tokens": 2979528704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.981623649597168, "created_at": "2025-01-15T21:36:09.382762+00:00"} {"global_step": 5684, "acc_step": 0, "speed/wps": 12907.00314686267, "speed/FLOPS": 202722220502410.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09721323847770691, "optim/lr": 0.0029988586429677065, "optim/total_tokens": 2980052992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9746341705322266, "created_at": "2025-01-15T21:36:19.541528+00:00"} {"global_step": 5685, "acc_step": 0, "speed/wps": 12910.33164261252, "speed/FLOPS": 202774499101994.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06565378606319427, "optim/lr": 0.002998855298758973, "optim/total_tokens": 2980577280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.069211006164551, "created_at": "2025-01-15T21:36:29.697988+00:00"} {"global_step": 5686, "acc_step": 0, "speed/wps": 12893.842411749994, "speed/FLOPS": 202515513072720.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06956196576356888, "optim/lr": 0.0029988519496599608, "optim/total_tokens": 2981101568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 3.218148708343506, "created_at": "2025-01-15T21:36:39.868618+00:00"} {"global_step": 5687, "acc_step": 0, "speed/wps": 12899.32345129, "speed/FLOPS": 202601600330434.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0724627673625946, "optim/lr": 0.00299884859567068, "optim/total_tokens": 2981625856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.2083797454833984, "created_at": "2025-01-15T21:36:50.034223+00:00"} {"global_step": 5688, "acc_step": 0, "speed/wps": 12904.471169610213, "speed/FLOPS": 202682452320359.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08387041091918945, "optim/lr": 0.002998845236791142, "optim/total_tokens": 2982150144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.01678729057312, "created_at": "2025-01-15T21:37:00.195718+00:00"} {"global_step": 5689, "acc_step": 0, "speed/wps": 12911.07144225335, "speed/FLOPS": 202786118671944.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09507450461387634, "optim/lr": 0.0029988418730213583, "optim/total_tokens": 2982674432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406724, "loss/out": 3.0493240356445312, "created_at": "2025-01-15T21:37:10.351940+00:00"} {"global_step": 5690, "acc_step": 0, "speed/wps": 12906.51412397674, "speed/FLOPS": 202714539725999.1, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08622200042009354, "optim/lr": 0.002998838504361339, "optim/total_tokens": 2983198720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0946083068847656, "created_at": "2025-01-15T21:37:20.511876+00:00"} {"global_step": 5691, "acc_step": 0, "speed/wps": 12911.278334885345, "speed/FLOPS": 202789368205030.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06999119371175766, "optim/lr": 0.002998835130811095, "optim/total_tokens": 2983723008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 3.068349838256836, "created_at": "2025-01-15T21:37:30.670415+00:00"} {"global_step": 5692, "acc_step": 0, "speed/wps": 12909.248431921684, "speed/FLOPS": 202757485789607.28, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08174297213554382, "optim/lr": 0.0029988317523706384, "optim/total_tokens": 2984247296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.041107654571533, "created_at": "2025-01-15T21:37:40.832855+00:00"} {"global_step": 5693, "acc_step": 0, "speed/wps": 12906.50675009914, "speed/FLOPS": 202714423909118.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0847080796957016, "optim/lr": 0.00299882836903998, "optim/total_tokens": 2984771584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 3.0696802139282227, "created_at": "2025-01-15T21:37:50.996973+00:00"} {"global_step": 5694, "acc_step": 0, "speed/wps": 12909.15539350171, "speed/FLOPS": 202756024493373.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06888899952173233, "optim/lr": 0.0029988249808191297, "optim/total_tokens": 2985295872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419297, "loss/out": 3.094654083251953, "created_at": "2025-01-15T21:38:01.156776+00:00"} {"global_step": 5695, "acc_step": 0, "speed/wps": 12911.381498222987, "speed/FLOPS": 202790988527009.5, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07704407721757889, "optim/lr": 0.0029988215877080992, "optim/total_tokens": 2985820160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.0978736877441406, "created_at": "2025-01-15T21:38:11.311977+00:00"} {"global_step": 5696, "acc_step": 0, "speed/wps": 12909.981521833837, "speed/FLOPS": 202768999974048.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08565500378608704, "optim/lr": 0.0029988181897069002, "optim/total_tokens": 2986344448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 3.089663505554199, "created_at": "2025-01-15T21:38:21.471184+00:00"} {"global_step": 5697, "acc_step": 0, "speed/wps": 12907.13232041962, "speed/FLOPS": 202724249350625.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06678382307291031, "optim/lr": 0.0029988147868155427, "optim/total_tokens": 2986868736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.957291841506958, "created_at": "2025-01-15T21:38:31.630225+00:00"} {"global_step": 5698, "acc_step": 0, "speed/wps": 12911.485043440312, "speed/FLOPS": 202792614846933.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06872999668121338, "optim/lr": 0.002998811379034039, "optim/total_tokens": 2987393024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 3.065211772918701, "created_at": "2025-01-15T21:38:41.792749+00:00"} {"global_step": 5699, "acc_step": 0, "speed/wps": 12904.742960626218, "speed/FLOPS": 202686721171745.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07649379968643188, "optim/lr": 0.0029988079663623994, "optim/total_tokens": 2987917312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.3683419227600098, "created_at": "2025-01-15T21:38:51.960562+00:00"} {"global_step": 5700, "acc_step": 0, "speed/wps": 12908.611348135935, "speed/FLOPS": 202747479513307.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07298080623149872, "optim/lr": 0.002998804548800635, "optim/total_tokens": 2988441600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.198637008666992, "created_at": "2025-01-15T21:39:02.121150+00:00"} {"global_step": 5701, "acc_step": 0, "speed/wps": 12910.384515116566, "speed/FLOPS": 202775329537324.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07195699959993362, "optim/lr": 0.002998801126348757, "optim/total_tokens": 2988965888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 3.0814332962036133, "created_at": "2025-01-15T21:39:12.281534+00:00"} {"global_step": 5702, "acc_step": 0, "speed/wps": 12910.987690715103, "speed/FLOPS": 202784803238949.97, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06543444842100143, "optim/lr": 0.002998797699006777, "optim/total_tokens": 2989490176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.965684652328491, "created_at": "2025-01-15T21:39:22.444808+00:00"} {"global_step": 5703, "acc_step": 0, "speed/wps": 12909.491724972007, "speed/FLOPS": 202761307041282.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06151486933231354, "optim/lr": 0.002998794266774706, "optim/total_tokens": 2990014464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.055351734161377, "created_at": "2025-01-15T21:39:32.601844+00:00"} {"global_step": 5704, "acc_step": 0, "speed/wps": 12905.05912878906, "speed/FLOPS": 202691687027203.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06662303954362869, "optim/lr": 0.002998790829652555, "optim/total_tokens": 2990538752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.2193703651428223, "created_at": "2025-01-15T21:39:42.764254+00:00"} {"global_step": 5705, "acc_step": 0, "speed/wps": 12910.444472709594, "speed/FLOPS": 202776271253713.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07053197920322418, "optim/lr": 0.0029987873876403353, "optim/total_tokens": 2991063040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.9847354888916016, "created_at": "2025-01-15T21:39:52.921467+00:00"} {"global_step": 5706, "acc_step": 0, "speed/wps": 12910.405920526839, "speed/FLOPS": 202775665738707.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05752040073275566, "optim/lr": 0.0029987839407380583, "optim/total_tokens": 2991587328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.065913438796997, "created_at": "2025-01-15T21:40:03.079575+00:00"} {"global_step": 5707, "acc_step": 0, "speed/wps": 12909.588870212818, "speed/FLOPS": 202762832840780.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05903318524360657, "optim/lr": 0.0029987804889457345, "optim/total_tokens": 2992111616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487147, "loss/out": 3.0909111499786377, "created_at": "2025-01-15T21:40:13.243041+00:00"} {"global_step": 5708, "acc_step": 0, "speed/wps": 12907.939124757577, "speed/FLOPS": 202736921321418.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0663350373506546, "optim/lr": 0.0029987770322633766, "optim/total_tokens": 2992635904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.130953311920166, "created_at": "2025-01-15T21:40:23.401495+00:00"} {"global_step": 5709, "acc_step": 0, "speed/wps": 12911.622155518155, "speed/FLOPS": 202794768380527.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08776234090328217, "optim/lr": 0.0029987735706909946, "optim/total_tokens": 2993160192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329371, "loss/out": 3.0586037635803223, "created_at": "2025-01-15T21:40:33.563421+00:00"} {"global_step": 5710, "acc_step": 0, "speed/wps": 12905.244236866722, "speed/FLOPS": 202694594403928.5, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08738314360380173, "optim/lr": 0.0029987701042285997, "optim/total_tokens": 2993684480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.989274501800537, "created_at": "2025-01-15T21:40:43.725405+00:00"} {"global_step": 5711, "acc_step": 0, "speed/wps": 12904.519817274782, "speed/FLOPS": 202683216398780.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07136084884405136, "optim/lr": 0.0029987666328762045, "optim/total_tokens": 2994208768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.1233479976654053, "created_at": "2025-01-15T21:40:53.886869+00:00"} {"global_step": 5712, "acc_step": 0, "speed/wps": 12909.343116435735, "speed/FLOPS": 202758972940010.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0752752423286438, "optim/lr": 0.0029987631566338194, "optim/total_tokens": 2994733056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.14111065864563, "created_at": "2025-01-15T21:41:04.047171+00:00"} {"global_step": 5713, "acc_step": 0, "speed/wps": 12912.320160744266, "speed/FLOPS": 202805731511763.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07189975678920746, "optim/lr": 0.0029987596755014557, "optim/total_tokens": 2995257344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.068448066711426, "created_at": "2025-01-15T21:41:14.202025+00:00"} {"global_step": 5714, "acc_step": 0, "speed/wps": 12915.470608367536, "speed/FLOPS": 202855213620855.8, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07272258400917053, "optim/lr": 0.0029987561894791255, "optim/total_tokens": 2995781632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 3.159874677658081, "created_at": "2025-01-15T21:41:24.358205+00:00"} {"global_step": 5715, "acc_step": 0, "speed/wps": 12904.262955739532, "speed/FLOPS": 202679182035404.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07097161561250687, "optim/lr": 0.0029987526985668393, "optim/total_tokens": 2996305920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.0382981300354004, "created_at": "2025-01-15T21:41:34.525062+00:00"} {"global_step": 5716, "acc_step": 0, "speed/wps": 12914.867842253847, "speed/FLOPS": 202845746350752.9, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08812720328569412, "optim/lr": 0.0029987492027646087, "optim/total_tokens": 2996830208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.107800245285034, "created_at": "2025-01-15T21:41:44.680284+00:00"} {"global_step": 5717, "acc_step": 0, "speed/wps": 12910.380463563832, "speed/FLOPS": 202775265902121.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08178028464317322, "optim/lr": 0.0029987457020724455, "optim/total_tokens": 2997354496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.1504030227661133, "created_at": "2025-01-15T21:41:54.839474+00:00"} {"global_step": 5718, "acc_step": 0, "speed/wps": 12907.340781969928, "speed/FLOPS": 202727523525729.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06393489241600037, "optim/lr": 0.002998742196490361, "optim/total_tokens": 2997878784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 3.1049532890319824, "created_at": "2025-01-15T21:42:05.000261+00:00"} {"global_step": 5719, "acc_step": 0, "speed/wps": 12904.963094217566, "speed/FLOPS": 202690178672293.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07377038896083832, "optim/lr": 0.002998738686018366, "optim/total_tokens": 2998403072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.124345064163208, "created_at": "2025-01-15T21:42:15.164738+00:00"} {"global_step": 5720, "acc_step": 0, "speed/wps": 12904.844784184437, "speed/FLOPS": 202688320450647.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08069882541894913, "optim/lr": 0.0029987351706564722, "optim/total_tokens": 2998927360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 3.138291597366333, "created_at": "2025-01-15T21:42:25.330341+00:00"} {"global_step": 5721, "acc_step": 0, "speed/wps": 12910.540129353298, "speed/FLOPS": 202777773672749.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0683492124080658, "optim/lr": 0.002998731650404692, "optim/total_tokens": 2999451648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 3.1752820014953613, "created_at": "2025-01-15T21:42:35.488639+00:00"} {"global_step": 5722, "acc_step": 0, "speed/wps": 12912.775926756352, "speed/FLOPS": 202812889943271.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07255961745977402, "optim/lr": 0.0029987281252630356, "optim/total_tokens": 2999975936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.986729621887207, "created_at": "2025-01-15T21:42:45.642877+00:00"} {"global_step": 5723, "acc_step": 0, "speed/wps": 12912.463076631862, "speed/FLOPS": 202807976202165.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06101730093359947, "optim/lr": 0.002998724595231516, "optim/total_tokens": 3000500224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.1213974952697754, "created_at": "2025-01-15T21:42:55.798253+00:00"} {"global_step": 5724, "acc_step": 0, "speed/wps": 12905.663473173581, "speed/FLOPS": 202701179086218.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05826090648770332, "optim/lr": 0.002998721060310143, "optim/total_tokens": 3001024512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.18839168548584, "created_at": "2025-01-15T21:43:05.961303+00:00"} {"global_step": 5725, "acc_step": 0, "speed/wps": 12909.878242312405, "speed/FLOPS": 202767377827243.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0636272206902504, "optim/lr": 0.0029987175204989292, "optim/total_tokens": 3001548800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350938, "loss/out": 3.1282012462615967, "created_at": "2025-01-15T21:43:16.117717+00:00"} {"global_step": 5726, "acc_step": 0, "speed/wps": 12911.544385823181, "speed/FLOPS": 202793546900601.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05713082477450371, "optim/lr": 0.002998713975797886, "optim/total_tokens": 3002073088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.150259494781494, "created_at": "2025-01-15T21:43:26.276060+00:00"} {"global_step": 5727, "acc_step": 0, "speed/wps": 12907.680710048746, "speed/FLOPS": 202732862563317.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058861397206783295, "optim/lr": 0.0029987104262070245, "optim/total_tokens": 3002597376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.137396812438965, "created_at": "2025-01-15T21:43:36.434207+00:00"} {"global_step": 5728, "acc_step": 0, "speed/wps": 12909.597162979226, "speed/FLOPS": 202762963090072.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06683310121297836, "optim/lr": 0.002998706871726357, "optim/total_tokens": 3003121664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.022658348083496, "created_at": "2025-01-15T21:43:46.593949+00:00"} {"global_step": 5729, "acc_step": 0, "speed/wps": 12911.951580159312, "speed/FLOPS": 202799942447194.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0689241886138916, "optim/lr": 0.0029987033123558948, "optim/total_tokens": 3003645952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375098, "loss/out": 3.1254143714904785, "created_at": "2025-01-15T21:43:56.749097+00:00"} {"global_step": 5730, "acc_step": 0, "speed/wps": 12908.953626002047, "speed/FLOPS": 202752855457533.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06662868708372116, "optim/lr": 0.0029986997480956493, "optim/total_tokens": 3004170240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.096572160720825, "created_at": "2025-01-15T21:44:06.907282+00:00"} {"global_step": 5731, "acc_step": 0, "speed/wps": 12905.000010487494, "speed/FLOPS": 202690758493040.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05777628347277641, "optim/lr": 0.0029986961789456325, "optim/total_tokens": 3004694528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.016787052154541, "created_at": "2025-01-15T21:44:17.068987+00:00"} {"global_step": 5732, "acc_step": 0, "speed/wps": 12908.712161218995, "speed/FLOPS": 202749062921308.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0859936773777008, "optim/lr": 0.0029986926049058557, "optim/total_tokens": 3005218816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 3.094698190689087, "created_at": "2025-01-15T21:44:27.236975+00:00"} {"global_step": 5733, "acc_step": 0, "speed/wps": 12910.019422512078, "speed/FLOPS": 202769595256281.5, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08270043879747391, "optim/lr": 0.0029986890259763306, "optim/total_tokens": 3005743104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.1081948280334473, "created_at": "2025-01-15T21:44:37.393226+00:00"} {"global_step": 5734, "acc_step": 0, "speed/wps": 12910.303758119178, "speed/FLOPS": 202774061137709.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.09974496811628342, "optim/lr": 0.0029986854421570693, "optim/total_tokens": 3006267392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 3.112448215484619, "created_at": "2025-01-15T21:44:47.549249+00:00"} {"global_step": 5735, "acc_step": 0, "speed/wps": 12908.461552222621, "speed/FLOPS": 202745126762647.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10597661137580872, "optim/lr": 0.0029986818534480826, "optim/total_tokens": 3006791680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483664, "loss/out": 3.0654714107513428, "created_at": "2025-01-15T21:44:57.712995+00:00"} {"global_step": 5736, "acc_step": 0, "speed/wps": 12911.772742646555, "speed/FLOPS": 202797133558307.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07127916067838669, "optim/lr": 0.002998678259849383, "optim/total_tokens": 3007315968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 3.1322896480560303, "created_at": "2025-01-15T21:45:07.868885+00:00"} {"global_step": 5737, "acc_step": 0, "speed/wps": 12905.83400029862, "speed/FLOPS": 202703857449046.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07867906987667084, "optim/lr": 0.002998674661360982, "optim/total_tokens": 3007840256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9448370933532715, "created_at": "2025-01-15T21:45:18.029006+00:00"} {"global_step": 5738, "acc_step": 0, "speed/wps": 12905.816633879278, "speed/FLOPS": 202703584685566.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07767569273710251, "optim/lr": 0.0029986710579828916, "optim/total_tokens": 3008364544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0123701095581055, "created_at": "2025-01-15T21:45:28.193395+00:00"} {"global_step": 5739, "acc_step": 0, "speed/wps": 12902.58225101918, "speed/FLOPS": 202652784258240.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07126592844724655, "optim/lr": 0.002998667449715123, "optim/total_tokens": 3008888832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.896219253540039, "created_at": "2025-01-15T21:45:38.355573+00:00"} {"global_step": 5740, "acc_step": 0, "speed/wps": 12909.423906412863, "speed/FLOPS": 202760241857618.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0813671126961708, "optim/lr": 0.0029986638365576885, "optim/total_tokens": 3009413120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.1228575706481934, "created_at": "2025-01-15T21:45:48.512625+00:00"} {"global_step": 5741, "acc_step": 0, "speed/wps": 12908.85426196176, "speed/FLOPS": 202751294808740.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07625555992126465, "optim/lr": 0.0029986602185105996, "optim/total_tokens": 3009937408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 3.05094575881958, "created_at": "2025-01-15T21:45:58.669799+00:00"} {"global_step": 5742, "acc_step": 0, "speed/wps": 12907.1818475392, "speed/FLOPS": 202725027242095.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07115926593542099, "optim/lr": 0.002998656595573868, "optim/total_tokens": 3010461696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9999887943267822, "created_at": "2025-01-15T21:46:08.829368+00:00"} {"global_step": 5743, "acc_step": 0, "speed/wps": 12910.695534153814, "speed/FLOPS": 202780214518689.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0705304890871048, "optim/lr": 0.002998652967747506, "optim/total_tokens": 3010985984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.0152039527893066, "created_at": "2025-01-15T21:46:18.986013+00:00"} {"global_step": 5744, "acc_step": 0, "speed/wps": 12906.815201978288, "speed/FLOPS": 202719268569737.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05952852591872215, "optim/lr": 0.0029986493350315246, "optim/total_tokens": 3011510272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.146894693374634, "created_at": "2025-01-15T21:46:29.145659+00:00"} {"global_step": 5745, "acc_step": 0, "speed/wps": 12901.477095739996, "speed/FLOPS": 202635426275938.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07097126543521881, "optim/lr": 0.0029986456974259368, "optim/total_tokens": 3012034560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358144, "loss/out": 3.1916301250457764, "created_at": "2025-01-15T21:46:39.312356+00:00"} {"global_step": 5746, "acc_step": 0, "speed/wps": 12894.346718764002, "speed/FLOPS": 202523433907366.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0031, "optim/grad_norm": 0.0689348354935646, "optim/lr": 0.0029986420549307532, "optim/total_tokens": 3012558848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 3.0944113731384277, "created_at": "2025-01-15T21:46:49.484136+00:00"} {"global_step": 5747, "acc_step": 0, "speed/wps": 12890.93380959388, "speed/FLOPS": 202469829479020.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.08562653511762619, "optim/lr": 0.0029986384075459866, "optim/total_tokens": 3013083136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 2.987518787384033, "created_at": "2025-01-15T21:46:59.657900+00:00"} {"global_step": 5748, "acc_step": 0, "speed/wps": 12898.826325719503, "speed/FLOPS": 202593792290227.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06548400968313217, "optim/lr": 0.002998634755271649, "optim/total_tokens": 3013607424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441962, "loss/out": 3.00858736038208, "created_at": "2025-01-15T21:47:09.823381+00:00"} {"global_step": 5749, "acc_step": 0, "speed/wps": 12879.443497069733, "speed/FLOPS": 202289358331485.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07058904320001602, "optim/lr": 0.0029986310981077514, "optim/total_tokens": 3014131712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.024294853210449, "created_at": "2025-01-15T21:47:20.003847+00:00"} {"global_step": 5750, "acc_step": 0, "speed/wps": 12908.638099336582, "speed/FLOPS": 202747899677673.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07410438358783722, "optim/lr": 0.0029986274360543066, "optim/total_tokens": 3014656000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.1123404502868652, "created_at": "2025-01-15T21:47:30.169412+00:00"} {"global_step": 5751, "acc_step": 0, "speed/wps": 12911.1831410884, "speed/FLOPS": 202787873055639.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06568378955125809, "optim/lr": 0.0029986237691113256, "optim/total_tokens": 3015180288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 3.090601921081543, "created_at": "2025-01-15T21:47:40.324758+00:00"} {"global_step": 5752, "acc_step": 0, "speed/wps": 12905.203758443273, "speed/FLOPS": 202693958634665.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06939306110143661, "optim/lr": 0.002998620097278822, "optim/total_tokens": 3015704576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0675923824310303, "created_at": "2025-01-15T21:47:50.488100+00:00"} {"global_step": 5753, "acc_step": 0, "speed/wps": 12903.612252171273, "speed/FLOPS": 202668961841704.5, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08571279793977737, "optim/lr": 0.0029986164205568063, "optim/total_tokens": 3016228864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.205211639404297, "created_at": "2025-01-15T21:48:00.649362+00:00"} {"global_step": 5754, "acc_step": 0, "speed/wps": 12909.890423269855, "speed/FLOPS": 202767569145918.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06650792062282562, "optim/lr": 0.0029986127389452907, "optim/total_tokens": 3016753152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.167858362197876, "created_at": "2025-01-15T21:48:10.805733+00:00"} {"global_step": 5755, "acc_step": 0, "speed/wps": 12902.377788590282, "speed/FLOPS": 202649572894834.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0682467371225357, "optim/lr": 0.002998609052444288, "optim/total_tokens": 3017277440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9032669067382812, "created_at": "2025-01-15T21:48:20.968097+00:00"} {"global_step": 5756, "acc_step": 0, "speed/wps": 12904.623500948861, "speed/FLOPS": 202684844893357.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0672326385974884, "optim/lr": 0.002998605361053809, "optim/total_tokens": 3017801728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.983853340148926, "created_at": "2025-01-15T21:48:31.128921+00:00"} {"global_step": 5757, "acc_step": 0, "speed/wps": 12905.453033876262, "speed/FLOPS": 202697873847880.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19316565990447998, "optim/lr": 0.002998601664773867, "optim/total_tokens": 3018326016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.006904125213623, "created_at": "2025-01-15T21:48:41.290178+00:00"} {"global_step": 5758, "acc_step": 0, "speed/wps": 12904.171048112263, "speed/FLOPS": 202677738499823.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06779680401086807, "optim/lr": 0.0029985979636044735, "optim/total_tokens": 3018850304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.053316116333008, "created_at": "2025-01-15T21:48:51.453997+00:00"} {"global_step": 5759, "acc_step": 0, "speed/wps": 12911.354778292465, "speed/FLOPS": 202790568853783.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08747498691082001, "optim/lr": 0.00299859425754564, "optim/total_tokens": 3019374592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.054734706878662, "created_at": "2025-01-15T21:49:01.613425+00:00"} {"global_step": 5760, "acc_step": 0, "speed/wps": 12910.136823181829, "speed/FLOPS": 202771439195125.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08403220772743225, "optim/lr": 0.00299859054659738, "optim/total_tokens": 3019898880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339985, "loss/out": 3.150014877319336, "created_at": "2025-01-15T21:49:11.773140+00:00"} {"global_step": 5761, "acc_step": 0, "speed/wps": 12910.211518091963, "speed/FLOPS": 202772612381330.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07275183498859406, "optim/lr": 0.002998586830759704, "optim/total_tokens": 3020423168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.2272543907165527, "created_at": "2025-01-15T21:49:21.932639+00:00"} {"global_step": 5762, "acc_step": 0, "speed/wps": 12905.280471933487, "speed/FLOPS": 202695163525444.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07778068631887436, "optim/lr": 0.002998583110032625, "optim/total_tokens": 3020947456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 3.1257381439208984, "created_at": "2025-01-15T21:49:32.092828+00:00"} {"global_step": 5763, "acc_step": 0, "speed/wps": 12911.280619752713, "speed/FLOPS": 202789404092012.3, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.084894560277462, "optim/lr": 0.0029985793844161555, "optim/total_tokens": 3021471744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.1047961711883545, "created_at": "2025-01-15T21:49:42.249627+00:00"} {"global_step": 5764, "acc_step": 0, "speed/wps": 12912.932890064973, "speed/FLOPS": 202815355267723.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07797373831272125, "optim/lr": 0.002998575653910307, "optim/total_tokens": 3021996032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318692, "loss/out": 3.0131330490112305, "created_at": "2025-01-15T21:49:52.407319+00:00"} {"global_step": 5765, "acc_step": 0, "speed/wps": 12914.301635003558, "speed/FLOPS": 202836853287836.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06482867151498795, "optim/lr": 0.002998571918515092, "optim/total_tokens": 3022520320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.0629138946533203, "created_at": "2025-01-15T21:50:02.564158+00:00"} {"global_step": 5766, "acc_step": 0, "speed/wps": 12904.656646566304, "speed/FLOPS": 202685365490826.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07753127068281174, "optim/lr": 0.002998568178230522, "optim/total_tokens": 3023044608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.059283971786499, "created_at": "2025-01-15T21:50:12.727267+00:00"} {"global_step": 5767, "acc_step": 0, "speed/wps": 12909.795363621863, "speed/FLOPS": 202766076103521.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07641297578811646, "optim/lr": 0.0029985644330566105, "optim/total_tokens": 3023568896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.144686460494995, "created_at": "2025-01-15T21:50:22.885832+00:00"} {"global_step": 5768, "acc_step": 0, "speed/wps": 12914.04488990571, "speed/FLOPS": 202832820753270.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07028143107891083, "optim/lr": 0.002998560682993368, "optim/total_tokens": 3024093184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.0825021266937256, "created_at": "2025-01-15T21:50:33.039532+00:00"} {"global_step": 5769, "acc_step": 0, "speed/wps": 12908.237187490564, "speed/FLOPS": 202741602806221.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06953451782464981, "optim/lr": 0.0029985569280408086, "optim/total_tokens": 3024617472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.0374343395233154, "created_at": "2025-01-15T21:50:43.198973+00:00"} {"global_step": 5770, "acc_step": 0, "speed/wps": 12908.552926617653, "speed/FLOPS": 202746561923083.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0762624442577362, "optim/lr": 0.0029985531681989436, "optim/total_tokens": 3025141760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431014, "loss/out": 3.123187780380249, "created_at": "2025-01-15T21:50:53.357648+00:00"} {"global_step": 5771, "acc_step": 0, "speed/wps": 12909.337169749833, "speed/FLOPS": 202758879539136.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06916992366313934, "optim/lr": 0.0029985494034677853, "optim/total_tokens": 3025666048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.0626137256622314, "created_at": "2025-01-15T21:51:03.518768+00:00"} {"global_step": 5772, "acc_step": 0, "speed/wps": 12906.800596766452, "speed/FLOPS": 202719039174816.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07430935651063919, "optim/lr": 0.0029985456338473457, "optim/total_tokens": 3026190336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.167452573776245, "created_at": "2025-01-15T21:51:13.679793+00:00"} {"global_step": 5773, "acc_step": 0, "speed/wps": 12903.169576711052, "speed/FLOPS": 202662009015301.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06965203583240509, "optim/lr": 0.0029985418593376375, "optim/total_tokens": 3026714624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 3.059122085571289, "created_at": "2025-01-15T21:51:23.841968+00:00"} {"global_step": 5774, "acc_step": 0, "speed/wps": 12900.945783392253, "speed/FLOPS": 202627081285417.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06729133427143097, "optim/lr": 0.002998538079938673, "optim/total_tokens": 3027238912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.0967235565185547, "created_at": "2025-01-15T21:51:34.005493+00:00"} {"global_step": 5775, "acc_step": 0, "speed/wps": 12909.003152891246, "speed/FLOPS": 202753633345385.53, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07941979169845581, "optim/lr": 0.002998534295650465, "optim/total_tokens": 3027763200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 3.0179989337921143, "created_at": "2025-01-15T21:51:44.163516+00:00"} {"global_step": 5776, "acc_step": 0, "speed/wps": 12908.662387922042, "speed/FLOPS": 202748281163284.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07348700612783432, "optim/lr": 0.0029985305064730242, "optim/total_tokens": 3028287488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.0030834674835205, "created_at": "2025-01-15T21:51:54.323756+00:00"} {"global_step": 5777, "acc_step": 0, "speed/wps": 12910.60181044874, "speed/FLOPS": 202778742459111.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07522597163915634, "optim/lr": 0.002998526712406365, "optim/total_tokens": 3028811776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377259, "loss/out": 3.075681686401367, "created_at": "2025-01-15T21:52:04.480725+00:00"} {"global_step": 5778, "acc_step": 0, "speed/wps": 12904.173402638757, "speed/FLOPS": 202677775480897.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0889192745089531, "optim/lr": 0.0029985229134504984, "optim/total_tokens": 3029336064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.1484994888305664, "created_at": "2025-01-15T21:52:14.641712+00:00"} {"global_step": 5779, "acc_step": 0, "speed/wps": 12908.019502928888, "speed/FLOPS": 202738183771050.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08274021744728088, "optim/lr": 0.0029985191096054374, "optim/total_tokens": 3029860352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.072695255279541, "created_at": "2025-01-15T21:52:24.800156+00:00"} {"global_step": 5780, "acc_step": 0, "speed/wps": 12913.16528176934, "speed/FLOPS": 202819005298780.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07036877423524857, "optim/lr": 0.002998515300871194, "optim/total_tokens": 3030384640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.039055824279785, "created_at": "2025-01-15T21:52:34.954128+00:00"} {"global_step": 5781, "acc_step": 0, "speed/wps": 12908.536701340372, "speed/FLOPS": 202746307082807.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07467953115701675, "optim/lr": 0.0029985114872477813, "optim/total_tokens": 3030908928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.0753536224365234, "created_at": "2025-01-15T21:52:45.114507+00:00"} {"global_step": 5782, "acc_step": 0, "speed/wps": 12906.109007577887, "speed/FLOPS": 202708176816266.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07294894009828568, "optim/lr": 0.002998507668735211, "optim/total_tokens": 3031433216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.0511839389801025, "created_at": "2025-01-15T21:52:55.279170+00:00"} {"global_step": 5783, "acc_step": 0, "speed/wps": 12910.7398086375, "speed/FLOPS": 202780909910295.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06434913724660873, "optim/lr": 0.002998503845333496, "optim/total_tokens": 3031957504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.1434664726257324, "created_at": "2025-01-15T21:53:05.434943+00:00"} {"global_step": 5784, "acc_step": 0, "speed/wps": 12910.271728702204, "speed/FLOPS": 202773558071702.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08013620227575302, "optim/lr": 0.0029985000170426484, "optim/total_tokens": 3032481792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299229, "loss/out": 3.126227378845215, "created_at": "2025-01-15T21:53:15.596128+00:00"} {"global_step": 5785, "acc_step": 0, "speed/wps": 12895.982442528433, "speed/FLOPS": 202549125196806.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07511410117149353, "optim/lr": 0.0029984961838626814, "optim/total_tokens": 3033006080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.1176435947418213, "created_at": "2025-01-15T21:53:25.764557+00:00"} {"global_step": 5786, "acc_step": 0, "speed/wps": 12901.454702821768, "speed/FLOPS": 202635074564387.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0735420286655426, "optim/lr": 0.0029984923457936066, "optim/total_tokens": 3033530368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505844, "loss/out": 3.08660626411438, "created_at": "2025-01-15T21:53:35.929128+00:00"} {"global_step": 5787, "acc_step": 0, "speed/wps": 12900.518661100368, "speed/FLOPS": 202620372742896.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06175815314054489, "optim/lr": 0.002998488502835437, "optim/total_tokens": 3034054656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 3.154294729232788, "created_at": "2025-01-15T21:53:46.098601+00:00"} {"global_step": 5788, "acc_step": 0, "speed/wps": 12899.500602166703, "speed/FLOPS": 202604382728383.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07801926136016846, "optim/lr": 0.0029984846549881857, "optim/total_tokens": 3034578944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.0789031982421875, "created_at": "2025-01-15T21:53:56.266434+00:00"} {"global_step": 5789, "acc_step": 0, "speed/wps": 12909.174490583948, "speed/FLOPS": 202756324439292.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05548340082168579, "optim/lr": 0.0029984808022518643, "optim/total_tokens": 3035103232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 3.0607357025146484, "created_at": "2025-01-15T21:54:06.423570+00:00"} {"global_step": 5790, "acc_step": 0, "speed/wps": 12912.144639494027, "speed/FLOPS": 202802974709337.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05958276987075806, "optim/lr": 0.002998476944626486, "optim/total_tokens": 3035627520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.02103853225708, "created_at": "2025-01-15T21:54:16.580958+00:00"} {"global_step": 5791, "acc_step": 0, "speed/wps": 12910.070270925049, "speed/FLOPS": 202770393900480.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06838778406381607, "optim/lr": 0.002998473082112062, "optim/total_tokens": 3036151808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404090, "loss/out": 2.9602954387664795, "created_at": "2025-01-15T21:54:26.744284+00:00"} {"global_step": 5792, "acc_step": 0, "speed/wps": 12900.41380294499, "speed/FLOPS": 202618725801475.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0776757299900055, "optim/lr": 0.0029984692147086074, "optim/total_tokens": 3036676096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.1520609855651855, "created_at": "2025-01-15T21:54:36.908875+00:00"} {"global_step": 5793, "acc_step": 0, "speed/wps": 12893.063610010002, "speed/FLOPS": 202503280921208.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08987224847078323, "optim/lr": 0.002998465342416133, "optim/total_tokens": 3037200384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.0384373664855957, "created_at": "2025-01-15T21:54:47.079380+00:00"} {"global_step": 5794, "acc_step": 0, "speed/wps": 12911.223165474323, "speed/FLOPS": 202788501693619.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07362349331378937, "optim/lr": 0.002998461465234652, "optim/total_tokens": 3037724672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 3.069807767868042, "created_at": "2025-01-15T21:54:57.236358+00:00"} {"global_step": 5795, "acc_step": 0, "speed/wps": 12912.309586354791, "speed/FLOPS": 202805565426446.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.063698410987854, "optim/lr": 0.0029984575831641766, "optim/total_tokens": 3038248960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.0900468826293945, "created_at": "2025-01-15T21:55:07.392073+00:00"} {"global_step": 5796, "acc_step": 0, "speed/wps": 12912.189766203073, "speed/FLOPS": 202803683486313.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08132763206958771, "optim/lr": 0.00299845369620472, "optim/total_tokens": 3038773248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 3.161257028579712, "created_at": "2025-01-15T21:55:17.551725+00:00"} {"global_step": 5797, "acc_step": 0, "speed/wps": 12910.312294586658, "speed/FLOPS": 202774195214661.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06460673362016678, "optim/lr": 0.0029984498043562945, "optim/total_tokens": 3039297536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.048031806945801, "created_at": "2025-01-15T21:55:27.710652+00:00"} {"global_step": 5798, "acc_step": 0, "speed/wps": 12910.775026386102, "speed/FLOPS": 202781463053431.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06642638146877289, "optim/lr": 0.002998445907618913, "optim/total_tokens": 3039821824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451808, "loss/out": 3.0399422645568848, "created_at": "2025-01-15T21:55:37.871415+00:00"} {"global_step": 5799, "acc_step": 0, "speed/wps": 12911.758967257987, "speed/FLOPS": 202796917196901.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07138370722532272, "optim/lr": 0.0029984420059925884, "optim/total_tokens": 3040346112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.0689821243286133, "created_at": "2025-01-15T21:55:48.032796+00:00"} {"global_step": 5800, "acc_step": 0, "speed/wps": 12906.837306938725, "speed/FLOPS": 202719615758517.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0663476213812828, "optim/lr": 0.002998438099477333, "optim/total_tokens": 3040870400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.061363697052002, "created_at": "2025-01-15T21:55:58.191551+00:00"} {"global_step": 5801, "acc_step": 0, "speed/wps": 12908.948240917081, "speed/FLOPS": 202752770877374.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08291253447532654, "optim/lr": 0.0029984341880731603, "optim/total_tokens": 3041394688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 3.1576528549194336, "created_at": "2025-01-15T21:56:08.352532+00:00"} {"global_step": 5802, "acc_step": 0, "speed/wps": 12903.82871856969, "speed/FLOPS": 202672361743947.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07626424729824066, "optim/lr": 0.0029984302717800814, "optim/total_tokens": 3041918976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0605196952819824, "created_at": "2025-01-15T21:56:18.517253+00:00"} {"global_step": 5803, "acc_step": 0, "speed/wps": 12909.568721546595, "speed/FLOPS": 202762516378290.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06555383652448654, "optim/lr": 0.002998426350598111, "optim/total_tokens": 3042443264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.07700514793396, "created_at": "2025-01-15T21:56:28.674693+00:00"} {"global_step": 5804, "acc_step": 0, "speed/wps": 12908.485587958041, "speed/FLOPS": 202745504276900.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07785779982805252, "optim/lr": 0.002998422424527261, "optim/total_tokens": 3042967552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 3.0717556476593018, "created_at": "2025-01-15T21:56:38.832160+00:00"} {"global_step": 5805, "acc_step": 0, "speed/wps": 12906.716862750438, "speed/FLOPS": 202717724017032.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06288374960422516, "optim/lr": 0.0029984184935675446, "optim/total_tokens": 3043491840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464032, "loss/out": 3.0679404735565186, "created_at": "2025-01-15T21:56:48.993247+00:00"} {"global_step": 5806, "acc_step": 0, "speed/wps": 12913.476713251566, "speed/FLOPS": 202823896758160.5, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062401168048381805, "optim/lr": 0.002998414557718974, "optim/total_tokens": 3044016128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.1402831077575684, "created_at": "2025-01-15T21:56:59.151312+00:00"} {"global_step": 5807, "acc_step": 0, "speed/wps": 12905.707704866943, "speed/FLOPS": 202701873805743.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09485966712236404, "optim/lr": 0.002998410616981562, "optim/total_tokens": 3044540416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.0204033851623535, "created_at": "2025-01-15T21:57:09.311083+00:00"} {"global_step": 5808, "acc_step": 0, "speed/wps": 12904.51462994006, "speed/FLOPS": 202683134924560.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09247267246246338, "optim/lr": 0.0029984066713553224, "optim/total_tokens": 3045064704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.089735984802246, "created_at": "2025-01-15T21:57:19.472546+00:00"} {"global_step": 5809, "acc_step": 0, "speed/wps": 12909.639830099877, "speed/FLOPS": 202763633235833.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059579089283943176, "optim/lr": 0.0029984027208402675, "optim/total_tokens": 3045588992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.1359469890594482, "created_at": "2025-01-15T21:57:29.629416+00:00"} {"global_step": 5810, "acc_step": 0, "speed/wps": 12906.724542249105, "speed/FLOPS": 202717844634111.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06202138587832451, "optim/lr": 0.0029983987654364098, "optim/total_tokens": 3046113280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.0409433841705322, "created_at": "2025-01-15T21:57:39.790578+00:00"} {"global_step": 5811, "acc_step": 0, "speed/wps": 12907.32580741999, "speed/FLOPS": 202727288329845.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06670767813920975, "optim/lr": 0.0029983948051437626, "optim/total_tokens": 3046637568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.138530969619751, "created_at": "2025-01-15T21:57:49.951891+00:00"} {"global_step": 5812, "acc_step": 0, "speed/wps": 12909.587167372769, "speed/FLOPS": 202762806095337.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07846907526254654, "optim/lr": 0.0029983908399623393, "optim/total_tokens": 3047161856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.1457600593566895, "created_at": "2025-01-15T21:58:00.108626+00:00"} {"global_step": 5813, "acc_step": 0, "speed/wps": 12897.943842612809, "speed/FLOPS": 202579931680377.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06707813590765, "optim/lr": 0.002998386869892152, "optim/total_tokens": 3047686144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.075937032699585, "created_at": "2025-01-15T21:58:10.285823+00:00"} {"global_step": 5814, "acc_step": 0, "speed/wps": 12901.312502390829, "speed/FLOPS": 202632841111215.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07609739899635315, "optim/lr": 0.0029983828949332143, "optim/total_tokens": 3048210432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.132963180541992, "created_at": "2025-01-15T21:58:20.452141+00:00"} {"global_step": 5815, "acc_step": 0, "speed/wps": 12908.38258691953, "speed/FLOPS": 202743886504052.75, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10830631107091904, "optim/lr": 0.002998378915085539, "optim/total_tokens": 3048734720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 3.1191699504852295, "created_at": "2025-01-15T21:58:30.609759+00:00"} {"global_step": 5816, "acc_step": 0, "speed/wps": 12905.185351175434, "speed/FLOPS": 202693669523229.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07707808911800385, "optim/lr": 0.0029983749303491383, "optim/total_tokens": 3049259008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9778618812561035, "created_at": "2025-01-15T21:58:40.770223+00:00"} {"global_step": 5817, "acc_step": 0, "speed/wps": 12902.925876004241, "speed/FLOPS": 202658181360821.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08355046808719635, "optim/lr": 0.0029983709407240262, "optim/total_tokens": 3049783296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 3.1230244636535645, "created_at": "2025-01-15T21:58:50.932863+00:00"} {"global_step": 5818, "acc_step": 0, "speed/wps": 12910.255306063702, "speed/FLOPS": 202773300131597.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07124418765306473, "optim/lr": 0.0029983669462102154, "optim/total_tokens": 3050307584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.130225658416748, "created_at": "2025-01-15T21:59:01.091196+00:00"} {"global_step": 5819, "acc_step": 0, "speed/wps": 12908.132712901872, "speed/FLOPS": 202739961889242.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0748024731874466, "optim/lr": 0.0029983629468077187, "optim/total_tokens": 3050831872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 3.1843154430389404, "created_at": "2025-01-15T21:59:11.250563+00:00"} {"global_step": 5820, "acc_step": 0, "speed/wps": 12908.551232338761, "speed/FLOPS": 202746535312104.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07971171289682388, "optim/lr": 0.0029983589425165496, "optim/total_tokens": 3051356160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.094383716583252, "created_at": "2025-01-15T21:59:21.408094+00:00"} {"global_step": 5821, "acc_step": 0, "speed/wps": 12906.035224102181, "speed/FLOPS": 202707017945391.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08014613389968872, "optim/lr": 0.0029983549333367212, "optim/total_tokens": 3051880448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.1752219200134277, "created_at": "2025-01-15T21:59:31.569810+00:00"} {"global_step": 5822, "acc_step": 0, "speed/wps": 12904.34642955864, "speed/FLOPS": 202680493106438.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07936285436153412, "optim/lr": 0.0029983509192682465, "optim/total_tokens": 3052404736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.117349147796631, "created_at": "2025-01-15T21:59:41.730622+00:00"} {"global_step": 5823, "acc_step": 0, "speed/wps": 12890.814500357134, "speed/FLOPS": 202467955563509.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07054050266742706, "optim/lr": 0.0029983469003111375, "optim/total_tokens": 3052929024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.0976550579071045, "created_at": "2025-01-15T21:59:51.903373+00:00"} {"global_step": 5824, "acc_step": 0, "speed/wps": 12903.105412548872, "speed/FLOPS": 202661001228964.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07222002744674683, "optim/lr": 0.0029983428764654083, "optim/total_tokens": 3053453312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332882, "loss/out": 3.082585096359253, "created_at": "2025-01-15T22:00:02.066256+00:00"} {"global_step": 5825, "acc_step": 0, "speed/wps": 12893.825118932898, "speed/FLOPS": 202515241465265.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0879330039024353, "optim/lr": 0.0029983388477310725, "optim/total_tokens": 3053977600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.144697666168213, "created_at": "2025-01-15T22:00:12.235596+00:00"} {"global_step": 5826, "acc_step": 0, "speed/wps": 12901.129985859705, "speed/FLOPS": 202629974438289.97, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07284262776374817, "optim/lr": 0.002998334814108143, "optim/total_tokens": 3054501888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.19254732131958, "created_at": "2025-01-15T22:00:22.400040+00:00"} {"global_step": 5827, "acc_step": 0, "speed/wps": 12911.037526954198, "speed/FLOPS": 202785585985567.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05697164312005043, "optim/lr": 0.0029983307755966316, "optim/total_tokens": 3055026176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 3.128836154937744, "created_at": "2025-01-15T22:00:32.555716+00:00"} {"global_step": 5828, "acc_step": 0, "speed/wps": 12909.524199315756, "speed/FLOPS": 202761817095475.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06878416240215302, "optim/lr": 0.0029983267321965533, "optim/total_tokens": 3055550464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.92855167388916, "created_at": "2025-01-15T22:00:42.712539+00:00"} {"global_step": 5829, "acc_step": 0, "speed/wps": 12909.559612925415, "speed/FLOPS": 202762373314878.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061799369752407074, "optim/lr": 0.0029983226839079198, "optim/total_tokens": 3056074752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.0933074951171875, "created_at": "2025-01-15T22:00:52.869493+00:00"} {"global_step": 5830, "acc_step": 0, "speed/wps": 12908.824597144705, "speed/FLOPS": 202750828882024.8, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07846986502408981, "optim/lr": 0.0029983186307307457, "optim/total_tokens": 3056599040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.048215627670288, "created_at": "2025-01-15T22:01:03.031509+00:00"} {"global_step": 5831, "acc_step": 0, "speed/wps": 12910.910185681041, "speed/FLOPS": 202783585915886.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06868628412485123, "optim/lr": 0.0029983145726650436, "optim/total_tokens": 3057123328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.1036949157714844, "created_at": "2025-01-15T22:01:13.189121+00:00"} {"global_step": 5832, "acc_step": 0, "speed/wps": 12911.37818635752, "speed/FLOPS": 202790936509611.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0716957226395607, "optim/lr": 0.002998310509710826, "optim/total_tokens": 3057647616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.110687494277954, "created_at": "2025-01-15T22:01:23.344935+00:00"} {"global_step": 5833, "acc_step": 0, "speed/wps": 12913.814684763158, "speed/FLOPS": 202829205065171.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08577699959278107, "optim/lr": 0.0029983064418681077, "optim/total_tokens": 3058171904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312942, "loss/out": 3.150470733642578, "created_at": "2025-01-15T22:01:33.500136+00:00"} {"global_step": 5834, "acc_step": 0, "speed/wps": 12910.098909971604, "speed/FLOPS": 202770843716060.6, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08157587796449661, "optim/lr": 0.0029983023691369007, "optim/total_tokens": 3058696192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.1497702598571777, "created_at": "2025-01-15T22:01:43.660317+00:00"} {"global_step": 5835, "acc_step": 0, "speed/wps": 12898.43738096667, "speed/FLOPS": 202587683378418.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0025, "optim/grad_norm": 0.07033584266901016, "optim/lr": 0.0029982982915172187, "optim/total_tokens": 3059220480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.194599151611328, "created_at": "2025-01-15T22:01:53.826487+00:00"} {"global_step": 5836, "acc_step": 0, "speed/wps": 12908.626938978883, "speed/FLOPS": 202747724388919.78, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06504330039024353, "optim/lr": 0.002998294209009075, "optim/total_tokens": 3059744768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9825918674468994, "created_at": "2025-01-15T22:02:03.984543+00:00"} {"global_step": 5837, "acc_step": 0, "speed/wps": 12912.975288963597, "speed/FLOPS": 202816021200688.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07585037499666214, "optim/lr": 0.002998290121612483, "optim/total_tokens": 3060269056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 3.1473593711853027, "created_at": "2025-01-15T22:02:14.139059+00:00"} {"global_step": 5838, "acc_step": 0, "speed/wps": 12910.40562563053, "speed/FLOPS": 202775661106955.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07604304701089859, "optim/lr": 0.002998286029327456, "optim/total_tokens": 3060793344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 3.0587656497955322, "created_at": "2025-01-15T22:02:24.297121+00:00"} {"global_step": 5839, "acc_step": 0, "speed/wps": 12912.2758102644, "speed/FLOPS": 202805034926533.44, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.075813889503479, "optim/lr": 0.0029982819321540074, "optim/total_tokens": 3061317632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 3.12666392326355, "created_at": "2025-01-15T22:02:34.451601+00:00"} {"global_step": 5840, "acc_step": 0, "speed/wps": 12906.912189756424, "speed/FLOPS": 202720791896068.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06433757394552231, "optim/lr": 0.00299827783009215, "optim/total_tokens": 3061841920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.1324119567871094, "created_at": "2025-01-15T22:02:44.610831+00:00"} {"global_step": 5841, "acc_step": 0, "speed/wps": 12908.921663670413, "speed/FLOPS": 202752353445194.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06299620866775513, "optim/lr": 0.002998273723141898, "optim/total_tokens": 3062366208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 3.0830459594726562, "created_at": "2025-01-15T22:02:54.768305+00:00"} {"global_step": 5842, "acc_step": 0, "speed/wps": 12910.991857538684, "speed/FLOPS": 202784868684640.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06676865369081497, "optim/lr": 0.0029982696113032645, "optim/total_tokens": 3062890496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9681055545806885, "created_at": "2025-01-15T22:03:04.923853+00:00"} {"global_step": 5843, "acc_step": 0, "speed/wps": 12911.77682942584, "speed/FLOPS": 202797197746792.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07595875859260559, "optim/lr": 0.002998265494576263, "optim/total_tokens": 3063414784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313436, "loss/out": 3.169788122177124, "created_at": "2025-01-15T22:03:15.079055+00:00"} {"global_step": 5844, "acc_step": 0, "speed/wps": 12902.911542207756, "speed/FLOPS": 202657956228851.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0633213073015213, "optim/lr": 0.002998261372960907, "optim/total_tokens": 3063939072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.2106399536132812, "created_at": "2025-01-15T22:03:25.243932+00:00"} {"global_step": 5845, "acc_step": 0, "speed/wps": 12913.59777537487, "speed/FLOPS": 202825798205163.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0629383996129036, "optim/lr": 0.002998257246457209, "optim/total_tokens": 3064463360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.0664169788360596, "created_at": "2025-01-15T22:03:35.403684+00:00"} {"global_step": 5846, "acc_step": 0, "speed/wps": 12911.827126637238, "speed/FLOPS": 202797987733612.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06444882601499557, "optim/lr": 0.0029982531150651837, "optim/total_tokens": 3064987648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.077028751373291, "created_at": "2025-01-15T22:03:45.558935+00:00"} {"global_step": 5847, "acc_step": 0, "speed/wps": 12913.895777190875, "speed/FLOPS": 202830478733181.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0627412274479866, "optim/lr": 0.0029982489787848446, "optim/total_tokens": 3065511936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.153463840484619, "created_at": "2025-01-15T22:03:55.712272+00:00"} {"global_step": 5848, "acc_step": 0, "speed/wps": 12913.654958114106, "speed/FLOPS": 202826696338660.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0760410726070404, "optim/lr": 0.002998244837616204, "optim/total_tokens": 3066036224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.034039258956909, "created_at": "2025-01-15T22:04:05.868190+00:00"} {"global_step": 5849, "acc_step": 0, "speed/wps": 12910.1807859205, "speed/FLOPS": 202772129690347.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06701233983039856, "optim/lr": 0.0029982406915592765, "optim/total_tokens": 3066560512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402220, "loss/out": 3.093815565109253, "created_at": "2025-01-15T22:04:16.024544+00:00"} {"global_step": 5850, "acc_step": 0, "speed/wps": 12912.056983720748, "speed/FLOPS": 202801597954966.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06236861273646355, "optim/lr": 0.002998236540614075, "optim/total_tokens": 3067084800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.1372549533843994, "created_at": "2025-01-15T22:04:26.201344+00:00"} {"global_step": 5851, "acc_step": 0, "speed/wps": 12909.899377223779, "speed/FLOPS": 202767709780069.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05731049180030823, "optim/lr": 0.002998232384780613, "optim/total_tokens": 3067609088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.190530776977539, "created_at": "2025-01-15T22:04:36.359012+00:00"} {"global_step": 5852, "acc_step": 0, "speed/wps": 12909.882942506076, "speed/FLOPS": 202767451650244.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07671179622411728, "optim/lr": 0.002998228224058905, "optim/total_tokens": 3068133376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 3.2051525115966797, "created_at": "2025-01-15T22:04:46.516451+00:00"} {"global_step": 5853, "acc_step": 0, "speed/wps": 12909.868736539393, "speed/FLOPS": 202767228526016.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061938539147377014, "optim/lr": 0.0029982240584489634, "optim/total_tokens": 3068657664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 3.1753690242767334, "created_at": "2025-01-15T22:04:56.675641+00:00"} {"global_step": 5854, "acc_step": 0, "speed/wps": 12912.876612965134, "speed/FLOPS": 202814471358537.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0730925127863884, "optim/lr": 0.0029982198879508026, "optim/total_tokens": 3069181952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371356, "loss/out": 3.0293195247650146, "created_at": "2025-01-15T22:05:06.829689+00:00"} {"global_step": 5855, "acc_step": 0, "speed/wps": 12913.090879685064, "speed/FLOPS": 202817836711807.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06960254907608032, "optim/lr": 0.002998215712564436, "optim/total_tokens": 3069706240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.0726141929626465, "created_at": "2025-01-15T22:05:16.984497+00:00"} {"global_step": 5856, "acc_step": 0, "speed/wps": 12913.83362263036, "speed/FLOPS": 202829502510399.34, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09044540673494339, "optim/lr": 0.0029982115322898764, "optim/total_tokens": 3070230528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.0408527851104736, "created_at": "2025-01-15T22:05:27.138446+00:00"} {"global_step": 5857, "acc_step": 0, "speed/wps": 12908.543434636056, "speed/FLOPS": 202746412838468.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06418147683143616, "optim/lr": 0.0029982073471271392, "optim/total_tokens": 3070754816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.0767364501953125, "created_at": "2025-01-15T22:05:37.295946+00:00"} {"global_step": 5858, "acc_step": 0, "speed/wps": 12909.706336472023, "speed/FLOPS": 202764677809797.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10102583467960358, "optim/lr": 0.0029982031570762366, "optim/total_tokens": 3071279104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 3.054917097091675, "created_at": "2025-01-15T22:05:47.453304+00:00"} {"global_step": 5859, "acc_step": 0, "speed/wps": 12914.476691100586, "speed/FLOPS": 202839602784392.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0906718447804451, "optim/lr": 0.002998198962137183, "optim/total_tokens": 3071803392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.1473731994628906, "created_at": "2025-01-15T22:05:57.608811+00:00"} {"global_step": 5860, "acc_step": 0, "speed/wps": 12911.443106811244, "speed/FLOPS": 202791956174547.16, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06489022821187973, "optim/lr": 0.0029981947623099916, "optim/total_tokens": 3072327680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432875, "loss/out": 3.017484664916992, "created_at": "2025-01-15T22:06:07.764220+00:00"} {"global_step": 5861, "acc_step": 0, "speed/wps": 12908.057617027362, "speed/FLOPS": 202738782405340.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0966683179140091, "optim/lr": 0.002998190557594676, "optim/total_tokens": 3072851968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.0294244289398193, "created_at": "2025-01-15T22:06:17.923028+00:00"} {"global_step": 5862, "acc_step": 0, "speed/wps": 12902.471120869946, "speed/FLOPS": 202651038806538.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07262025773525238, "optim/lr": 0.0029981863479912503, "optim/total_tokens": 3073376256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342633, "loss/out": 3.090912103652954, "created_at": "2025-01-15T22:06:28.087028+00:00"} {"global_step": 5863, "acc_step": 0, "speed/wps": 12889.09843377809, "speed/FLOPS": 202441002379763.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07308003306388855, "optim/lr": 0.002998182133499728, "optim/total_tokens": 3073900544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.0116047859191895, "created_at": "2025-01-15T22:06:38.262417+00:00"} {"global_step": 5864, "acc_step": 0, "speed/wps": 12900.827123648502, "speed/FLOPS": 202625217571086.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06234003230929375, "optim/lr": 0.0029981779141201237, "optim/total_tokens": 3074424832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.1038928031921387, "created_at": "2025-01-15T22:06:48.425933+00:00"} {"global_step": 5865, "acc_step": 0, "speed/wps": 12895.944452881316, "speed/FLOPS": 202548528517195.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07423949241638184, "optim/lr": 0.00299817368985245, "optim/total_tokens": 3074949120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 2.988678455352783, "created_at": "2025-01-15T22:06:58.599840+00:00"} {"global_step": 5866, "acc_step": 0, "speed/wps": 12909.60043603747, "speed/FLOPS": 202763014497950.03, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06673716008663177, "optim/lr": 0.002998169460696721, "optim/total_tokens": 3075473408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0640921592712402, "created_at": "2025-01-15T22:07:08.764191+00:00"} {"global_step": 5867, "acc_step": 0, "speed/wps": 12912.870955662756, "speed/FLOPS": 202814382502830.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07377516478300095, "optim/lr": 0.0029981652266529513, "optim/total_tokens": 3075997696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.172567129135132, "created_at": "2025-01-15T22:07:18.919931+00:00"} {"global_step": 5868, "acc_step": 0, "speed/wps": 12910.47140560573, "speed/FLOPS": 202776694271857.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06594141572713852, "optim/lr": 0.0029981609877211537, "optim/total_tokens": 3076521984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.0523080825805664, "created_at": "2025-01-15T22:07:29.079581+00:00"} {"global_step": 5869, "acc_step": 0, "speed/wps": 12893.043171545152, "speed/FLOPS": 202502959907032.53, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07407697290182114, "optim/lr": 0.0029981567439013417, "optim/total_tokens": 3077046272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.095932960510254, "created_at": "2025-01-15T22:07:39.251996+00:00"} {"global_step": 5870, "acc_step": 0, "speed/wps": 12900.522931142683, "speed/FLOPS": 202620439809778.53, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07830868661403656, "optim/lr": 0.0029981524951935307, "optim/total_tokens": 3077570560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 3.149580478668213, "created_at": "2025-01-15T22:07:49.418711+00:00"} {"global_step": 5871, "acc_step": 0, "speed/wps": 12909.275040116767, "speed/FLOPS": 202757903707874.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06388971209526062, "optim/lr": 0.002998148241597733, "optim/total_tokens": 3078094848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.116504669189453, "created_at": "2025-01-15T22:07:59.579347+00:00"} {"global_step": 5872, "acc_step": 0, "speed/wps": 12910.437348970845, "speed/FLOPS": 202776159365607.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06872302293777466, "optim/lr": 0.002998143983113964, "optim/total_tokens": 3078619136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.0907750129699707, "created_at": "2025-01-15T22:08:09.740337+00:00"} {"global_step": 5873, "acc_step": 0, "speed/wps": 12910.6452919484, "speed/FLOPS": 202779425395813.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07731238007545471, "optim/lr": 0.0029981397197422367, "optim/total_tokens": 3079143424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.1085104942321777, "created_at": "2025-01-15T22:08:19.900424+00:00"} {"global_step": 5874, "acc_step": 0, "speed/wps": 12911.864682775724, "speed/FLOPS": 202798577604375.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0693335235118866, "optim/lr": 0.0029981354514825643, "optim/total_tokens": 3079667712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.0423390865325928, "created_at": "2025-01-15T22:08:30.057621+00:00"} {"global_step": 5875, "acc_step": 0, "speed/wps": 12909.082670463682, "speed/FLOPS": 202754882278129.25, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06165128946304321, "optim/lr": 0.0029981311783349616, "optim/total_tokens": 3080192000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0743062496185303, "created_at": "2025-01-15T22:08:40.215826+00:00"} {"global_step": 5876, "acc_step": 0, "speed/wps": 12901.013166400437, "speed/FLOPS": 202628139628155.38, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06825672835111618, "optim/lr": 0.0029981269002994433, "optim/total_tokens": 3080716288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.236599922180176, "created_at": "2025-01-15T22:08:50.383866+00:00"} {"global_step": 5877, "acc_step": 0, "speed/wps": 12913.27749314623, "speed/FLOPS": 202820767732649.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06465346366167068, "optim/lr": 0.002998122617376022, "optim/total_tokens": 3081240576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 3.0926706790924072, "created_at": "2025-01-15T22:09:00.538582+00:00"} {"global_step": 5878, "acc_step": 0, "speed/wps": 12908.613754010823, "speed/FLOPS": 202747517300878.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06775537878274918, "optim/lr": 0.0029981183295647116, "optim/total_tokens": 3081764864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.1509833335876465, "created_at": "2025-01-15T22:09:10.697748+00:00"} {"global_step": 5879, "acc_step": 0, "speed/wps": 12912.960211292844, "speed/FLOPS": 202815784385150.72, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06099041923880577, "optim/lr": 0.002998114036865528, "optim/total_tokens": 3082289152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452863, "loss/out": 3.0860595703125, "created_at": "2025-01-15T22:09:20.854796+00:00"} {"global_step": 5880, "acc_step": 0, "speed/wps": 12908.258095634457, "speed/FLOPS": 202741931197351.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07418155670166016, "optim/lr": 0.0029981097392784827, "optim/total_tokens": 3082813440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 3.1620213985443115, "created_at": "2025-01-15T22:09:31.014009+00:00"} {"global_step": 5881, "acc_step": 0, "speed/wps": 12915.864711921751, "speed/FLOPS": 202861403558729.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07282743602991104, "optim/lr": 0.0029981054368035916, "optim/total_tokens": 3083337728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432352, "loss/out": 3.0163233280181885, "created_at": "2025-01-15T22:09:41.167274+00:00"} {"global_step": 5882, "acc_step": 0, "speed/wps": 12907.237758952891, "speed/FLOPS": 202725905407678.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08877439051866531, "optim/lr": 0.0029981011294408673, "optim/total_tokens": 3083862016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9667418003082275, "created_at": "2025-01-15T22:09:51.329046+00:00"} {"global_step": 5883, "acc_step": 0, "speed/wps": 12912.218424131215, "speed/FLOPS": 202804133598454.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07227975875139236, "optim/lr": 0.0029980968171903254, "optim/total_tokens": 3084386304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0527231693267822, "created_at": "2025-01-15T22:10:01.485513+00:00"} {"global_step": 5884, "acc_step": 0, "speed/wps": 12908.863862148088, "speed/FLOPS": 202751445592858.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0715983510017395, "optim/lr": 0.0029980925000519785, "optim/total_tokens": 3084910592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.1530182361602783, "created_at": "2025-01-15T22:10:11.646282+00:00"} {"global_step": 5885, "acc_step": 0, "speed/wps": 12911.823687792805, "speed/FLOPS": 202797933721835.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05915994942188263, "optim/lr": 0.0029980881780258418, "optim/total_tokens": 3085434880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.114847183227539, "created_at": "2025-01-15T22:10:21.803301+00:00"} {"global_step": 5886, "acc_step": 0, "speed/wps": 12912.06906335813, "speed/FLOPS": 202801787682270.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09059679508209229, "optim/lr": 0.002998083851111929, "optim/total_tokens": 3085959168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 3.0119335651397705, "created_at": "2025-01-15T22:10:31.957921+00:00"} {"global_step": 5887, "acc_step": 0, "speed/wps": 12910.180121538755, "speed/FLOPS": 202772119255319.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08874912559986115, "optim/lr": 0.002998079519310254, "optim/total_tokens": 3086483456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.0255074501037598, "created_at": "2025-01-15T22:10:42.114748+00:00"} {"global_step": 5888, "acc_step": 0, "speed/wps": 12908.556694051415, "speed/FLOPS": 202746621095807.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06773137301206589, "optim/lr": 0.002998075182620831, "optim/total_tokens": 3087007744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.090852737426758, "created_at": "2025-01-15T22:10:52.275315+00:00"} {"global_step": 5889, "acc_step": 0, "speed/wps": 12910.458270164443, "speed/FLOPS": 202776487961701.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07572127878665924, "optim/lr": 0.002998070841043674, "optim/total_tokens": 3087532032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.0871288776397705, "created_at": "2025-01-15T22:11:02.433089+00:00"} {"global_step": 5890, "acc_step": 0, "speed/wps": 12910.920134775279, "speed/FLOPS": 202783742180083.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07283277064561844, "optim/lr": 0.0029980664945787982, "optim/total_tokens": 3088056320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.1111369132995605, "created_at": "2025-01-15T22:11:12.591541+00:00"} {"global_step": 5891, "acc_step": 0, "speed/wps": 12913.192158483092, "speed/FLOPS": 202819427434502.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07157525420188904, "optim/lr": 0.0029980621432262166, "optim/total_tokens": 3088580608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.1094582080841064, "created_at": "2025-01-15T22:11:22.746682+00:00"} {"global_step": 5892, "acc_step": 0, "speed/wps": 12910.875895771482, "speed/FLOPS": 202783047345738.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06057688221335411, "optim/lr": 0.002998057786985943, "optim/total_tokens": 3089104896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9820258617401123, "created_at": "2025-01-15T22:11:32.903745+00:00"} {"global_step": 5893, "acc_step": 0, "speed/wps": 12910.15669413209, "speed/FLOPS": 202771751295704.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05922113358974457, "optim/lr": 0.002998053425857994, "optim/total_tokens": 3089629184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.9605002403259277, "created_at": "2025-01-15T22:11:43.061205+00:00"} {"global_step": 5894, "acc_step": 0, "speed/wps": 12909.68676887669, "speed/FLOPS": 202764370473823.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05628439038991928, "optim/lr": 0.0029980490598423806, "optim/total_tokens": 3090153472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.1406571865081787, "created_at": "2025-01-15T22:11:53.218998+00:00"} {"global_step": 5895, "acc_step": 0, "speed/wps": 12912.211490348585, "speed/FLOPS": 202804024693870.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05835806578397751, "optim/lr": 0.0029980446889391194, "optim/total_tokens": 3090677760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.048985481262207, "created_at": "2025-01-15T22:12:03.376323+00:00"} {"global_step": 5896, "acc_step": 0, "speed/wps": 12905.137085450291, "speed/FLOPS": 202692911443691.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0660240650177002, "optim/lr": 0.0029980403131482244, "optim/total_tokens": 3091202048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.980337142944336, "created_at": "2025-01-15T22:12:13.536473+00:00"} {"global_step": 5897, "acc_step": 0, "speed/wps": 12913.831710118911, "speed/FLOPS": 202829472471778.97, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07580100744962692, "optim/lr": 0.002998035932469709, "optim/total_tokens": 3091726336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.108255624771118, "created_at": "2025-01-15T22:12:23.697960+00:00"} {"global_step": 5898, "acc_step": 0, "speed/wps": 12911.432402377515, "speed/FLOPS": 202791788046706.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06029219180345535, "optim/lr": 0.002998031546903588, "optim/total_tokens": 3092250624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.063262462615967, "created_at": "2025-01-15T22:12:33.857077+00:00"} {"global_step": 5899, "acc_step": 0, "speed/wps": 12905.972438629033, "speed/FLOPS": 202706031813259.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07303676754236221, "optim/lr": 0.0029980271564498755, "optim/total_tokens": 3092774912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 2.988101005554199, "created_at": "2025-01-15T22:12:44.017561+00:00"} {"global_step": 5900, "acc_step": 0, "speed/wps": 12911.194312973352, "speed/FLOPS": 202788048525444.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07454494386911392, "optim/lr": 0.0029980227611085854, "optim/total_tokens": 3093299200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.985738754272461, "created_at": "2025-01-15T22:12:54.176938+00:00"} {"global_step": 5901, "acc_step": 0, "speed/wps": 12912.32695366728, "speed/FLOPS": 202805838203954.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0751660093665123, "optim/lr": 0.002998018360879733, "optim/total_tokens": 3093823488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 3.174480438232422, "created_at": "2025-01-15T22:13:04.333828+00:00"} {"global_step": 5902, "acc_step": 0, "speed/wps": 12907.638021546476, "speed/FLOPS": 202732192081728.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05803076922893524, "optim/lr": 0.002998013955763333, "optim/total_tokens": 3094347776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.9931654930114746, "created_at": "2025-01-15T22:13:14.492762+00:00"} {"global_step": 5903, "acc_step": 0, "speed/wps": 12905.836833559693, "speed/FLOPS": 202703901949304.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060967110097408295, "optim/lr": 0.002998009545759398, "optim/total_tokens": 3094872064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302403, "loss/out": 2.9964234828948975, "created_at": "2025-01-15T22:13:24.657353+00:00"} {"global_step": 5904, "acc_step": 0, "speed/wps": 12908.206373556213, "speed/FLOPS": 202741118831039.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06583832949399948, "optim/lr": 0.002998005130867944, "optim/total_tokens": 3095396352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.1182992458343506, "created_at": "2025-01-15T22:13:34.815217+00:00"} {"global_step": 5905, "acc_step": 0, "speed/wps": 12914.705353128584, "speed/FLOPS": 202843194235757.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07851992547512054, "optim/lr": 0.0029980007110889844, "optim/total_tokens": 3095920640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463722, "loss/out": 3.0298569202423096, "created_at": "2025-01-15T22:13:44.971303+00:00"} {"global_step": 5906, "acc_step": 0, "speed/wps": 12909.351471645257, "speed/FLOPS": 202759104170057.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07707347720861435, "optim/lr": 0.0029979962864225343, "optim/total_tokens": 3096444928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.0623722076416016, "created_at": "2025-01-15T22:13:55.131036+00:00"} {"global_step": 5907, "acc_step": 0, "speed/wps": 12910.98366365678, "speed/FLOPS": 202784739988465.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06829976290464401, "optim/lr": 0.0029979918568686075, "optim/total_tokens": 3096969216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.1938376426696777, "created_at": "2025-01-15T22:14:05.289027+00:00"} {"global_step": 5908, "acc_step": 0, "speed/wps": 12910.595705083613, "speed/FLOPS": 202778646565962.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07014919072389603, "optim/lr": 0.002997987422427219, "optim/total_tokens": 3097493504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.044818878173828, "created_at": "2025-01-15T22:14:15.448049+00:00"} {"global_step": 5909, "acc_step": 0, "speed/wps": 12901.914595204858, "speed/FLOPS": 202642297806221.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07085975259542465, "optim/lr": 0.0029979829830983832, "optim/total_tokens": 3098017792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.1658074855804443, "created_at": "2025-01-15T22:14:25.612604+00:00"} {"global_step": 5910, "acc_step": 0, "speed/wps": 12904.472606056404, "speed/FLOPS": 202682474881720.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0726640596985817, "optim/lr": 0.002997978538882114, "optim/total_tokens": 3098542080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288153, "loss/out": 3.0794546604156494, "created_at": "2025-01-15T22:14:35.782140+00:00"} {"global_step": 5911, "acc_step": 0, "speed/wps": 12900.675992327228, "speed/FLOPS": 202622843846009.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07382804155349731, "optim/lr": 0.002997974089778427, "optim/total_tokens": 3099066368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.1213512420654297, "created_at": "2025-01-15T22:14:45.955532+00:00"} {"global_step": 5912, "acc_step": 0, "speed/wps": 12908.264375194412, "speed/FLOPS": 202742029826469.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06273075938224792, "optim/lr": 0.0029979696357873356, "optim/total_tokens": 3099590656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.102531909942627, "created_at": "2025-01-15T22:14:56.116708+00:00"} {"global_step": 5913, "acc_step": 0, "speed/wps": 12910.94955457353, "speed/FLOPS": 202784204258442.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07331059128046036, "optim/lr": 0.002997965176908855, "optim/total_tokens": 3100114944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 3.0204451084136963, "created_at": "2025-01-15T22:15:06.275979+00:00"} {"global_step": 5914, "acc_step": 0, "speed/wps": 12905.999137470284, "speed/FLOPS": 202706451155249.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06842102855443954, "optim/lr": 0.002997960713143, "optim/total_tokens": 3100639232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.1358020305633545, "created_at": "2025-01-15T22:15:16.438797+00:00"} {"global_step": 5915, "acc_step": 0, "speed/wps": 12912.664329496632, "speed/FLOPS": 202811137154957.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07485562562942505, "optim/lr": 0.0029979562444897843, "optim/total_tokens": 3101163520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.0841574668884277, "created_at": "2025-01-15T22:15:26.592943+00:00"} {"global_step": 5916, "acc_step": 0, "speed/wps": 12904.19256662329, "speed/FLOPS": 202678076477609.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07621347904205322, "optim/lr": 0.002997951770949223, "optim/total_tokens": 3101687808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.087362766265869, "created_at": "2025-01-15T22:15:36.754141+00:00"} {"global_step": 5917, "acc_step": 0, "speed/wps": 12906.236017447898, "speed/FLOPS": 202710171680805.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06853653490543365, "optim/lr": 0.0029979472925213304, "optim/total_tokens": 3102212096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 3.109086751937866, "created_at": "2025-01-15T22:15:46.918106+00:00"} {"global_step": 5918, "acc_step": 0, "speed/wps": 12907.84517233317, "speed/FLOPS": 202735445669488.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07146143168210983, "optim/lr": 0.0029979428092061213, "optim/total_tokens": 3102736384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.0952742099761963, "created_at": "2025-01-15T22:15:57.076218+00:00"} {"global_step": 5919, "acc_step": 0, "speed/wps": 12912.289419717608, "speed/FLOPS": 202805248681697.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07441066205501556, "optim/lr": 0.002997938321003611, "optim/total_tokens": 3103260672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.139314651489258, "created_at": "2025-01-15T22:16:07.232530+00:00"} {"global_step": 5920, "acc_step": 0, "speed/wps": 12903.183027381463, "speed/FLOPS": 202662220276563.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07159817963838577, "optim/lr": 0.002997933827913813, "optim/total_tokens": 3103784960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 3.223353862762451, "created_at": "2025-01-15T22:16:17.396842+00:00"} {"global_step": 5921, "acc_step": 0, "speed/wps": 12904.854187896093, "speed/FLOPS": 202688468148861.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06772476434707642, "optim/lr": 0.0029979293299367423, "optim/total_tokens": 3104309248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.14217472076416, "created_at": "2025-01-15T22:16:27.558443+00:00"} {"global_step": 5922, "acc_step": 0, "speed/wps": 12907.334946034476, "speed/FLOPS": 202727431864343.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06453736871480942, "optim/lr": 0.0029979248270724135, "optim/total_tokens": 3104833536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.0946006774902344, "created_at": "2025-01-15T22:16:37.717612+00:00"} {"global_step": 5923, "acc_step": 0, "speed/wps": 12900.630742665824, "speed/FLOPS": 202622133137898.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06422406435012817, "optim/lr": 0.0029979203193208417, "optim/total_tokens": 3105357824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.988208770751953, "created_at": "2025-01-15T22:16:47.883573+00:00"} {"global_step": 5924, "acc_step": 0, "speed/wps": 12907.953055371077, "speed/FLOPS": 202737140120846.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07086499780416489, "optim/lr": 0.0029979158066820413, "optim/total_tokens": 3105882112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 2.9931373596191406, "created_at": "2025-01-15T22:16:58.043729+00:00"} {"global_step": 5925, "acc_step": 0, "speed/wps": 12905.698380386553, "speed/FLOPS": 202701727351966.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0742158368229866, "optim/lr": 0.0029979112891560275, "optim/total_tokens": 3106406400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.0136871337890625, "created_at": "2025-01-15T22:17:08.203502+00:00"} {"global_step": 5926, "acc_step": 0, "speed/wps": 12904.962162941507, "speed/FLOPS": 202690164045322.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05776653811335564, "optim/lr": 0.002997906766742815, "optim/total_tokens": 3106930688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.03755521774292, "created_at": "2025-01-15T22:17:18.363857+00:00"} {"global_step": 5927, "acc_step": 0, "speed/wps": 12911.42441167204, "speed/FLOPS": 202791662541696.16, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07159032672643661, "optim/lr": 0.0029979022394424175, "optim/total_tokens": 3107454976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.1774914264678955, "created_at": "2025-01-15T22:17:28.520279+00:00"} {"global_step": 5928, "acc_step": 0, "speed/wps": 12907.93086963724, "speed/FLOPS": 202736791663409.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06797706335783005, "optim/lr": 0.0029978977072548513, "optim/total_tokens": 3107979264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.1107637882232666, "created_at": "2025-01-15T22:17:38.681696+00:00"} {"global_step": 5929, "acc_step": 0, "speed/wps": 12902.028665635282, "speed/FLOPS": 202644089439080.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05848481133580208, "optim/lr": 0.0029978931701801293, "optim/total_tokens": 3108503552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.2585010528564453, "created_at": "2025-01-15T22:17:48.844383+00:00"} {"global_step": 5930, "acc_step": 0, "speed/wps": 12912.419898015285, "speed/FLOPS": 202807298022658.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06866700202226639, "optim/lr": 0.002997888628218268, "optim/total_tokens": 3109027840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.1573593616485596, "created_at": "2025-01-15T22:17:58.999132+00:00"} {"global_step": 5931, "acc_step": 0, "speed/wps": 12911.03220882783, "speed/FLOPS": 202785502457084.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07379093021154404, "optim/lr": 0.0029978840813692817, "optim/total_tokens": 3109552128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 3.125896453857422, "created_at": "2025-01-15T22:18:09.155058+00:00"} {"global_step": 5932, "acc_step": 0, "speed/wps": 12907.113919486972, "speed/FLOPS": 202723960338692.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06836627423763275, "optim/lr": 0.002997879529633185, "optim/total_tokens": 3110076416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.0823934078216553, "created_at": "2025-01-15T22:18:19.320595+00:00"} {"global_step": 5933, "acc_step": 0, "speed/wps": 12913.249784316458, "speed/FLOPS": 202820332527401.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0629030168056488, "optim/lr": 0.002997874973009993, "optim/total_tokens": 3110600704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.233478546142578, "created_at": "2025-01-15T22:18:29.474686+00:00"} {"global_step": 5934, "acc_step": 0, "speed/wps": 12909.845907109577, "speed/FLOPS": 202766869958450.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058367516845464706, "optim/lr": 0.0029978704114997203, "optim/total_tokens": 3111124992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.1555137634277344, "created_at": "2025-01-15T22:18:39.632305+00:00"} {"global_step": 5935, "acc_step": 0, "speed/wps": 12914.656269552597, "speed/FLOPS": 202842423310748.4, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05940211936831474, "optim/lr": 0.0029978658451023823, "optim/total_tokens": 3111649280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299229, "loss/out": 3.1736130714416504, "created_at": "2025-01-15T22:18:49.785283+00:00"} {"global_step": 5936, "acc_step": 0, "speed/wps": 12907.14034250877, "speed/FLOPS": 202724375348559.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06512963771820068, "optim/lr": 0.0029978612738179927, "optim/total_tokens": 3112173568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.004573345184326, "created_at": "2025-01-15T22:18:59.945821+00:00"} {"global_step": 5937, "acc_step": 0, "speed/wps": 12906.801749652544, "speed/FLOPS": 202719057282476.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06648852676153183, "optim/lr": 0.0029978566976465683, "optim/total_tokens": 3112697856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.1076433658599854, "created_at": "2025-01-15T22:19:10.106387+00:00"} {"global_step": 5938, "acc_step": 0, "speed/wps": 12909.275725719433, "speed/FLOPS": 202757914476206.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06902827322483063, "optim/lr": 0.0029978521165881225, "optim/total_tokens": 3113222144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.188570022583008, "created_at": "2025-01-15T22:19:20.267456+00:00"} {"global_step": 5939, "acc_step": 0, "speed/wps": 12908.706853296191, "speed/FLOPS": 202748979553086.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06707420945167542, "optim/lr": 0.0029978475306426705, "optim/total_tokens": 3113746432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.078165054321289, "created_at": "2025-01-15T22:19:30.426028+00:00"} {"global_step": 5940, "acc_step": 0, "speed/wps": 12907.466970863754, "speed/FLOPS": 202729505495708.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07505398988723755, "optim/lr": 0.002997842939810228, "optim/total_tokens": 3114270720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 3.0368824005126953, "created_at": "2025-01-15T22:19:40.586083+00:00"} {"global_step": 5941, "acc_step": 0, "speed/wps": 12908.702843052703, "speed/FLOPS": 202748916566702.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07213106751441956, "optim/lr": 0.0029978383440908093, "optim/total_tokens": 3114795008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.1175684928894043, "created_at": "2025-01-15T22:19:50.743310+00:00"} {"global_step": 5942, "acc_step": 0, "speed/wps": 12913.440384748747, "speed/FLOPS": 202823326169103.12, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06092049926519394, "optim/lr": 0.0029978337434844295, "optim/total_tokens": 3115319296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.1139917373657227, "created_at": "2025-01-15T22:20:00.897071+00:00"} {"global_step": 5943, "acc_step": 0, "speed/wps": 12911.417308172971, "speed/FLOPS": 202791550971482.03, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0997372567653656, "optim/lr": 0.0029978291379911034, "optim/total_tokens": 3115843584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.1973609924316406, "created_at": "2025-01-15T22:20:11.052268+00:00"} {"global_step": 5944, "acc_step": 0, "speed/wps": 12908.688091487596, "speed/FLOPS": 202748684873101.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09178217500448227, "optim/lr": 0.0029978245276108466, "optim/total_tokens": 3116367872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.9855270385742188, "created_at": "2025-01-15T22:20:21.213594+00:00"} {"global_step": 5945, "acc_step": 0, "speed/wps": 12910.079659752842, "speed/FLOPS": 202770541364922.38, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08024268597364426, "optim/lr": 0.002997819912343674, "optim/total_tokens": 3116892160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.0619935989379883, "created_at": "2025-01-15T22:20:31.369809+00:00"} {"global_step": 5946, "acc_step": 0, "speed/wps": 12912.009581105947, "speed/FLOPS": 202800853431762.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08029395341873169, "optim/lr": 0.0029978152921896007, "optim/total_tokens": 3117416448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 3.0649242401123047, "created_at": "2025-01-15T22:20:41.525197+00:00"} {"global_step": 5947, "acc_step": 0, "speed/wps": 12906.00334486458, "speed/FLOPS": 202706517238158.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07702670246362686, "optim/lr": 0.002997810667148641, "optim/total_tokens": 3117940736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.2200512886047363, "created_at": "2025-01-15T22:20:51.688453+00:00"} {"global_step": 5948, "acc_step": 0, "speed/wps": 12913.751891117488, "speed/FLOPS": 202828218804678.8, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07733826339244843, "optim/lr": 0.002997806037220811, "optim/total_tokens": 3118465024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.1822612285614014, "created_at": "2025-01-15T22:21:01.843483+00:00"} {"global_step": 5949, "acc_step": 0, "speed/wps": 12901.834110902222, "speed/FLOPS": 202641033689651.9, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06404463201761246, "optim/lr": 0.0029978014024061254, "optim/total_tokens": 3118989312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.1027300357818604, "created_at": "2025-01-15T22:21:12.007654+00:00"} {"global_step": 5950, "acc_step": 0, "speed/wps": 12909.8282291005, "speed/FLOPS": 202766592301025.4, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06695402413606644, "optim/lr": 0.002997796762704599, "optim/total_tokens": 3119513600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378070, "loss/out": 3.0234549045562744, "created_at": "2025-01-15T22:21:22.164037+00:00"} {"global_step": 5951, "acc_step": 0, "speed/wps": 12913.010024847326, "speed/FLOPS": 202816566775474.84, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07076285779476166, "optim/lr": 0.0029977921181162477, "optim/total_tokens": 3120037888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0562663078308105, "created_at": "2025-01-15T22:21:32.319565+00:00"} {"global_step": 5952, "acc_step": 0, "speed/wps": 12906.25433184465, "speed/FLOPS": 202710459333572.8, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0777236595749855, "optim/lr": 0.0029977874686410856, "optim/total_tokens": 3120562176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.02816104888916, "created_at": "2025-01-15T22:21:42.478920+00:00"} {"global_step": 5953, "acc_step": 0, "speed/wps": 12898.565721122195, "speed/FLOPS": 202589699136918.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08837834000587463, "optim/lr": 0.002997782814279129, "optim/total_tokens": 3121086464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.0853939056396484, "created_at": "2025-01-15T22:21:52.646845+00:00"} {"global_step": 5954, "acc_step": 0, "speed/wps": 12905.804744642994, "speed/FLOPS": 202703397948773.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06300463527441025, "optim/lr": 0.0029977781550303917, "optim/total_tokens": 3121610752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9375662803649902, "created_at": "2025-01-15T22:22:02.809774+00:00"} {"global_step": 5955, "acc_step": 0, "speed/wps": 12913.145781885441, "speed/FLOPS": 202818699026308.16, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07733225077390671, "optim/lr": 0.0029977734908948903, "optim/total_tokens": 3122135040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.1703269481658936, "created_at": "2025-01-15T22:22:12.963833+00:00"} {"global_step": 5956, "acc_step": 0, "speed/wps": 12904.117182933287, "speed/FLOPS": 202676892473169.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06446627527475357, "optim/lr": 0.0029977688218726394, "optim/total_tokens": 3122659328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.930680990219116, "created_at": "2025-01-15T22:22:23.127479+00:00"} {"global_step": 5957, "acc_step": 0, "speed/wps": 12902.994634191264, "speed/FLOPS": 202659261302632.4, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06022121384739876, "optim/lr": 0.0029977641479636548, "optim/total_tokens": 3123183616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.0774850845336914, "created_at": "2025-01-15T22:22:33.292867+00:00"} {"global_step": 5958, "acc_step": 0, "speed/wps": 12896.563028141836, "speed/FLOPS": 202558244091673.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07671364396810532, "optim/lr": 0.0029977594691679506, "optim/total_tokens": 3123707904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.20344877243042, "created_at": "2025-01-15T22:22:43.462376+00:00"} {"global_step": 5959, "acc_step": 0, "speed/wps": 12902.004608993835, "speed/FLOPS": 202643711596469.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06442801654338837, "optim/lr": 0.002997754785485543, "optim/total_tokens": 3124232192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0818099975585938, "created_at": "2025-01-15T22:22:53.625114+00:00"} {"global_step": 5960, "acc_step": 0, "speed/wps": 12904.742301537171, "speed/FLOPS": 202686710819846.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06654125452041626, "optim/lr": 0.002997750096916447, "optim/total_tokens": 3124756480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 2.9706523418426514, "created_at": "2025-01-15T22:23:03.785931+00:00"} {"global_step": 5961, "acc_step": 0, "speed/wps": 12910.90603312315, "speed/FLOPS": 202783520694259.06, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07009531557559967, "optim/lr": 0.0029977454034606776, "optim/total_tokens": 3125280768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 3.036287784576416, "created_at": "2025-01-15T22:23:13.943944+00:00"} {"global_step": 5962, "acc_step": 0, "speed/wps": 12905.740322310688, "speed/FLOPS": 202702386107518.8, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0666891410946846, "optim/lr": 0.0029977407051182507, "optim/total_tokens": 3125805056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.100037097930908, "created_at": "2025-01-15T22:23:24.103755+00:00"} {"global_step": 5963, "acc_step": 0, "speed/wps": 12907.274678767928, "speed/FLOPS": 202726485284106.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08887793123722076, "optim/lr": 0.0029977360018891815, "optim/total_tokens": 3126329344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.032757043838501, "created_at": "2025-01-15T22:23:34.264813+00:00"} {"global_step": 5964, "acc_step": 0, "speed/wps": 12912.782952371614, "speed/FLOPS": 202813000290213.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0833500474691391, "optim/lr": 0.002997731293773485, "optim/total_tokens": 3126853632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307266, "loss/out": 2.9959070682525635, "created_at": "2025-01-15T22:23:44.422498+00:00"} {"global_step": 5965, "acc_step": 0, "speed/wps": 12905.07804690196, "speed/FLOPS": 202691984162162.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08050769567489624, "optim/lr": 0.002997726580771177, "optim/total_tokens": 3127377920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.1240389347076416, "created_at": "2025-01-15T22:23:54.586523+00:00"} {"global_step": 5966, "acc_step": 0, "speed/wps": 12904.748285405543, "speed/FLOPS": 202686804804721.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09265830367803574, "optim/lr": 0.002997721862882272, "optim/total_tokens": 3127902208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.196380376815796, "created_at": "2025-01-15T22:24:04.747467+00:00"} {"global_step": 5967, "acc_step": 0, "speed/wps": 12905.827558598983, "speed/FLOPS": 202703756273301.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07630221545696259, "optim/lr": 0.0029977171401067864, "optim/total_tokens": 3128426496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.0616302490234375, "created_at": "2025-01-15T22:24:14.911893+00:00"} {"global_step": 5968, "acc_step": 0, "speed/wps": 12904.627217575555, "speed/FLOPS": 202684903268087.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09807471930980682, "optim/lr": 0.0029977124124447355, "optim/total_tokens": 3128950784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 3.0753931999206543, "created_at": "2025-01-15T22:24:25.074556+00:00"} {"global_step": 5969, "acc_step": 0, "speed/wps": 12905.379680374663, "speed/FLOPS": 202696721730339.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06808511167764664, "optim/lr": 0.002997707679896134, "optim/total_tokens": 3129475072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.978886127471924, "created_at": "2025-01-15T22:24:35.237821+00:00"} {"global_step": 5970, "acc_step": 0, "speed/wps": 12903.998250693463, "speed/FLOPS": 202675024478913.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09805455803871155, "optim/lr": 0.002997702942460998, "optim/total_tokens": 3129999360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.1488535404205322, "created_at": "2025-01-15T22:24:45.402994+00:00"} {"global_step": 5971, "acc_step": 0, "speed/wps": 12912.638459632826, "speed/FLOPS": 202810730833197.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09506342560052872, "optim/lr": 0.0029976982001393426, "optim/total_tokens": 3130523648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 2.9178123474121094, "created_at": "2025-01-15T22:24:55.562469+00:00"} {"global_step": 5972, "acc_step": 0, "speed/wps": 12904.736790987403, "speed/FLOPS": 202686624269090.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062282487750053406, "optim/lr": 0.0029976934529311835, "optim/total_tokens": 3131047936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.0296809673309326, "created_at": "2025-01-15T22:25:05.723107+00:00"} {"global_step": 5973, "acc_step": 0, "speed/wps": 12904.962600771405, "speed/FLOPS": 202690170922043.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09195409715175629, "optim/lr": 0.002997688700836536, "optim/total_tokens": 3131572224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.2448551654815674, "created_at": "2025-01-15T22:25:15.885461+00:00"} {"global_step": 5974, "acc_step": 0, "speed/wps": 12906.308967618303, "speed/FLOPS": 202711317463475.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06213051825761795, "optim/lr": 0.0029976839438554163, "optim/total_tokens": 3132096512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.0403199195861816, "created_at": "2025-01-15T22:25:26.048350+00:00"} {"global_step": 5975, "acc_step": 0, "speed/wps": 12905.776097425238, "speed/FLOPS": 202702948004852.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08674953877925873, "optim/lr": 0.002997679181987838, "optim/total_tokens": 3132620800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 3.0261335372924805, "created_at": "2025-01-15T22:25:36.210415+00:00"} {"global_step": 5976, "acc_step": 0, "speed/wps": 12906.168057926649, "speed/FLOPS": 202709104283137.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08231551200151443, "optim/lr": 0.002997674415233819, "optim/total_tokens": 3133145088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.1649370193481445, "created_at": "2025-01-15T22:25:46.372036+00:00"} {"global_step": 5977, "acc_step": 0, "speed/wps": 12906.728641220767, "speed/FLOPS": 202717909014094.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0618709996342659, "optim/lr": 0.002997669643593374, "optim/total_tokens": 3133669376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.9301648139953613, "created_at": "2025-01-15T22:25:56.532997+00:00"} {"global_step": 5978, "acc_step": 0, "speed/wps": 12910.291272016226, "speed/FLOPS": 202773865026304.53, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06570006161928177, "optim/lr": 0.002997664867066518, "optim/total_tokens": 3134193664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 3.01357364654541, "created_at": "2025-01-15T22:26:06.689378+00:00"} {"global_step": 5979, "acc_step": 0, "speed/wps": 12903.27760046385, "speed/FLOPS": 202663705676779.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0881357416510582, "optim/lr": 0.0029976600856532667, "optim/total_tokens": 3134717952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 3.164280891418457, "created_at": "2025-01-15T22:26:16.850892+00:00"} {"global_step": 5980, "acc_step": 0, "speed/wps": 12908.102084154694, "speed/FLOPS": 202739480822679.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06195603311061859, "optim/lr": 0.0029976552993536366, "optim/total_tokens": 3135242240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377784, "loss/out": 3.187387466430664, "created_at": "2025-01-15T22:26:27.009179+00:00"} {"global_step": 5981, "acc_step": 0, "speed/wps": 12903.877871317436, "speed/FLOPS": 202673133755393.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08977905660867691, "optim/lr": 0.0029976505081676422, "optim/total_tokens": 3135766528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0609850883483887, "created_at": "2025-01-15T22:26:37.170385+00:00"} {"global_step": 5982, "acc_step": 0, "speed/wps": 12899.459737679486, "speed/FLOPS": 202603740895458.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09677069634199142, "optim/lr": 0.0029976457120953, "optim/total_tokens": 3136290816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404614, "loss/out": 3.0337395668029785, "created_at": "2025-01-15T22:26:47.336084+00:00"} {"global_step": 5983, "acc_step": 0, "speed/wps": 12905.678093563844, "speed/FLOPS": 202701408719538.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06479501724243164, "optim/lr": 0.002997640911136625, "optim/total_tokens": 3136815104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.1055099964141846, "created_at": "2025-01-15T22:26:57.498490+00:00"} {"global_step": 5984, "acc_step": 0, "speed/wps": 12908.697034301715, "speed/FLOPS": 202748825332285.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06700751930475235, "optim/lr": 0.0029976361052916336, "optim/total_tokens": 3137339392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.012528657913208, "created_at": "2025-01-15T22:27:07.657658+00:00"} {"global_step": 5985, "acc_step": 0, "speed/wps": 12904.183393412995, "speed/FLOPS": 202677932399736.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055523816496133804, "optim/lr": 0.0029976312945603403, "optim/total_tokens": 3137863680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 3.1145520210266113, "created_at": "2025-01-15T22:27:17.818844+00:00"} {"global_step": 5986, "acc_step": 0, "speed/wps": 12913.129274920748, "speed/FLOPS": 202818439761744.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08174606412649155, "optim/lr": 0.002997626478942762, "optim/total_tokens": 3138387968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.1637089252471924, "created_at": "2025-01-15T22:27:27.972681+00:00"} {"global_step": 5987, "acc_step": 0, "speed/wps": 12908.617694343284, "speed/FLOPS": 202747579189214.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07386202365159988, "optim/lr": 0.002997621658438914, "optim/total_tokens": 3138912256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 3.0066757202148438, "created_at": "2025-01-15T22:27:38.130020+00:00"} {"global_step": 5988, "acc_step": 0, "speed/wps": 12904.33948174687, "speed/FLOPS": 202680383981507.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.07159174233675003, "optim/lr": 0.002997616833048811, "optim/total_tokens": 3139436544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.1152849197387695, "created_at": "2025-01-15T22:27:48.293636+00:00"} {"global_step": 5989, "acc_step": 0, "speed/wps": 12906.968617291, "speed/FLOPS": 202721678168038.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07926671206951141, "optim/lr": 0.0029976120027724704, "optim/total_tokens": 3139960832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.9439587593078613, "created_at": "2025-01-15T22:27:58.455659+00:00"} {"global_step": 5990, "acc_step": 0, "speed/wps": 12908.011314003206, "speed/FLOPS": 202738055152719.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0894419476389885, "optim/lr": 0.0029976071676099075, "optim/total_tokens": 3140485120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.050168514251709, "created_at": "2025-01-15T22:28:08.617203+00:00"} {"global_step": 5991, "acc_step": 0, "speed/wps": 12906.565618259572, "speed/FLOPS": 202715348514471.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07690704613924026, "optim/lr": 0.0029976023275611373, "optim/total_tokens": 3141009408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.084836483001709, "created_at": "2025-01-15T22:28:18.779156+00:00"} {"global_step": 5992, "acc_step": 0, "speed/wps": 12909.247034671802, "speed/FLOPS": 202757463843877.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07465247809886932, "optim/lr": 0.0029975974826261767, "optim/total_tokens": 3141533696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.165365695953369, "created_at": "2025-01-15T22:28:28.940873+00:00"} {"global_step": 5993, "acc_step": 0, "speed/wps": 12911.188176251711, "speed/FLOPS": 202787952139798.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07732413709163666, "optim/lr": 0.0029975926328050405, "optim/total_tokens": 3142057984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.1583142280578613, "created_at": "2025-01-15T22:28:39.101260+00:00"} {"global_step": 5994, "acc_step": 0, "speed/wps": 12912.889848110119, "speed/FLOPS": 202814679234676.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06424331665039062, "optim/lr": 0.002997587778097745, "optim/total_tokens": 3142582272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.0747628211975098, "created_at": "2025-01-15T22:28:49.255193+00:00"} {"global_step": 5995, "acc_step": 0, "speed/wps": 12903.620307743902, "speed/FLOPS": 202669088365542.16, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07386474311351776, "optim/lr": 0.002997582918504306, "optim/total_tokens": 3143106560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.114680051803589, "created_at": "2025-01-15T22:28:59.418684+00:00"} {"global_step": 5996, "acc_step": 0, "speed/wps": 12904.011766981686, "speed/FLOPS": 202675236770792.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0665731132030487, "optim/lr": 0.0029975780540247394, "optim/total_tokens": 3143630848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 3.1078197956085205, "created_at": "2025-01-15T22:29:09.589032+00:00"} {"global_step": 5997, "acc_step": 0, "speed/wps": 12902.730004834539, "speed/FLOPS": 202655104934946.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06390728801488876, "optim/lr": 0.002997573184659061, "optim/total_tokens": 3144155136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0812952518463135, "created_at": "2025-01-15T22:29:19.753876+00:00"} {"global_step": 5998, "acc_step": 0, "speed/wps": 12911.104217104585, "speed/FLOPS": 202786633446020.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06289643049240112, "optim/lr": 0.002997568310407287, "optim/total_tokens": 3144679424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.1014187335968018, "created_at": "2025-01-15T22:29:29.909440+00:00"} {"global_step": 5999, "acc_step": 0, "speed/wps": 12909.022757488548, "speed/FLOPS": 202753941262525.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07004806399345398, "optim/lr": 0.0029975634312694325, "optim/total_tokens": 3145203712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 2.8900463581085205, "created_at": "2025-01-15T22:29:40.069978+00:00"} {"global_step": 6000, "acc_step": 0, "speed/wps": 12906.57930800083, "speed/FLOPS": 202715563530669.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07223547250032425, "optim/lr": 0.002997558547245514, "optim/total_tokens": 3145728000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 2.958296060562134, "created_at": "2025-01-15T22:29:50.228900+00:00"} {"global_step": 6001, "acc_step": 0, "speed/wps": 12608.717016203824, "speed/FLOPS": 198037226932316.56, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.2448, "optim/grad_norm": 0.06780937314033508, "optim/lr": 0.0029975536583355476, "optim/total_tokens": 3146252288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 2.997556209564209, "created_at": "2025-01-15T22:30:00.631141+00:00"} {"global_step": 6002, "acc_step": 0, "speed/wps": 12893.95779013088, "speed/FLOPS": 202517325248739.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06409307569265366, "optim/lr": 0.002997548764539549, "optim/total_tokens": 3146776576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9846017360687256, "created_at": "2025-01-15T22:30:10.800053+00:00"} {"global_step": 6003, "acc_step": 0, "speed/wps": 12889.892129805618, "speed/FLOPS": 202453468466528.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05731421709060669, "optim/lr": 0.002997543865857534, "optim/total_tokens": 3147300864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.088007688522339, "created_at": "2025-01-15T22:30:20.974171+00:00"} {"global_step": 6004, "acc_step": 0, "speed/wps": 12891.728822835186, "speed/FLOPS": 202482316254438.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05906693637371063, "optim/lr": 0.002997538962289519, "optim/total_tokens": 3147825152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.1226141452789307, "created_at": "2025-01-15T22:30:31.149018+00:00"} {"global_step": 6005, "acc_step": 0, "speed/wps": 12890.232945792215, "speed/FLOPS": 202458821449927.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.070818230509758, "optim/lr": 0.00299753405383552, "optim/total_tokens": 3148349440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 3.029912233352661, "created_at": "2025-01-15T22:30:41.324715+00:00"} {"global_step": 6006, "acc_step": 0, "speed/wps": 12885.650102742407, "speed/FLOPS": 202386841602344.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08375848084688187, "optim/lr": 0.002997529140495552, "optim/total_tokens": 3148873728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.0991365909576416, "created_at": "2025-01-15T22:30:51.506674+00:00"} {"global_step": 6007, "acc_step": 0, "speed/wps": 12890.639458200334, "speed/FLOPS": 202465206285903.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059316281229257584, "optim/lr": 0.0029975242222696325, "optim/total_tokens": 3149398016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 3.2621331214904785, "created_at": "2025-01-15T22:31:01.686329+00:00"} {"global_step": 6008, "acc_step": 0, "speed/wps": 12891.421617080552, "speed/FLOPS": 202477491165916.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0842650830745697, "optim/lr": 0.0029975192991577764, "optim/total_tokens": 3149922304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 2.9192631244659424, "created_at": "2025-01-15T22:31:11.860356+00:00"} {"global_step": 6009, "acc_step": 0, "speed/wps": 12891.778328647191, "speed/FLOPS": 202483093811243.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06147393211722374, "optim/lr": 0.0029975143711600005, "optim/total_tokens": 3150446592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.107118606567383, "created_at": "2025-01-15T22:31:22.033252+00:00"} {"global_step": 6010, "acc_step": 0, "speed/wps": 12891.909611465293, "speed/FLOPS": 202485155788303.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07456660270690918, "optim/lr": 0.0029975094382763204, "optim/total_tokens": 3150970880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.143918514251709, "created_at": "2025-01-15T22:31:32.204625+00:00"} {"global_step": 6011, "acc_step": 0, "speed/wps": 12892.160938664148, "speed/FLOPS": 202489103227321.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07588068395853043, "optim/lr": 0.0029975045005067526, "optim/total_tokens": 3151495168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.0099987983703613, "created_at": "2025-01-15T22:31:42.375045+00:00"} {"global_step": 6012, "acc_step": 0, "speed/wps": 12885.827638056357, "speed/FLOPS": 202389630038408.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09268879145383835, "optim/lr": 0.0029974995578513123, "optim/total_tokens": 3152019456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.0430431365966797, "created_at": "2025-01-15T22:31:52.554186+00:00"} {"global_step": 6013, "acc_step": 0, "speed/wps": 12891.814940424834, "speed/FLOPS": 202483668849521.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09368821233510971, "optim/lr": 0.0029974946103100172, "optim/total_tokens": 3152543744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 3.0106077194213867, "created_at": "2025-01-15T22:32:02.728133+00:00"} {"global_step": 6014, "acc_step": 0, "speed/wps": 12886.881861753232, "speed/FLOPS": 202406188070223.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06882539391517639, "optim/lr": 0.002997489657882882, "optim/total_tokens": 3153068032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472720, "loss/out": 3.034548759460449, "created_at": "2025-01-15T22:32:12.904461+00:00"} {"global_step": 6015, "acc_step": 0, "speed/wps": 12890.643942495992, "speed/FLOPS": 202465276717929.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10397401452064514, "optim/lr": 0.002997484700569924, "optim/total_tokens": 3153592320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0608811378479004, "created_at": "2025-01-15T22:32:23.076314+00:00"} {"global_step": 6016, "acc_step": 0, "speed/wps": 12890.346313881013, "speed/FLOPS": 202460602051546.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09506956487894058, "optim/lr": 0.0029974797383711584, "optim/total_tokens": 3154116608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 3.0213139057159424, "created_at": "2025-01-15T22:32:33.250656+00:00"} {"global_step": 6017, "acc_step": 0, "speed/wps": 12886.748893633932, "speed/FLOPS": 202404099623192.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06251682341098785, "optim/lr": 0.002997474771286602, "optim/total_tokens": 3154640896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.095355987548828, "created_at": "2025-01-15T22:32:43.426527+00:00"} {"global_step": 6018, "acc_step": 0, "speed/wps": 12885.110334780498, "speed/FLOPS": 202378363804783.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06716958433389664, "optim/lr": 0.0029974697993162703, "optim/total_tokens": 3155165184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.0642824172973633, "created_at": "2025-01-15T22:32:53.608147+00:00"} {"global_step": 6019, "acc_step": 0, "speed/wps": 12894.686905543678, "speed/FLOPS": 202528777008206.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0734662413597107, "optim/lr": 0.0029974648224601807, "optim/total_tokens": 3155689472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.015223741531372, "created_at": "2025-01-15T22:33:03.783837+00:00"} {"global_step": 6020, "acc_step": 0, "speed/wps": 12889.085167090301, "speed/FLOPS": 202440794008201.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061307333409786224, "optim/lr": 0.002997459840718348, "optim/total_tokens": 3156213760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 3.1175975799560547, "created_at": "2025-01-15T22:33:13.960286+00:00"} {"global_step": 6021, "acc_step": 0, "speed/wps": 12889.998010551475, "speed/FLOPS": 202455131469137.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06731599569320679, "optim/lr": 0.0029974548540907895, "optim/total_tokens": 3156738048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9571025371551514, "created_at": "2025-01-15T22:33:24.135716+00:00"} {"global_step": 6022, "acc_step": 0, "speed/wps": 12889.134298473487, "speed/FLOPS": 202441565684087.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06213586777448654, "optim/lr": 0.002997449862577522, "optim/total_tokens": 3157262336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 3.081967353820801, "created_at": "2025-01-15T22:33:34.310497+00:00"} {"global_step": 6023, "acc_step": 0, "speed/wps": 12890.670969948023, "speed/FLOPS": 202465701221203.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05978269502520561, "optim/lr": 0.00299744486617856, "optim/total_tokens": 3157786624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.1030168533325195, "created_at": "2025-01-15T22:33:44.482824+00:00"} {"global_step": 6024, "acc_step": 0, "speed/wps": 12888.7618349893, "speed/FLOPS": 202435715633249.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058295611292123795, "optim/lr": 0.0029974398648939214, "optim/total_tokens": 3158310912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 3.03956937789917, "created_at": "2025-01-15T22:33:54.658425+00:00"} {"global_step": 6025, "acc_step": 0, "speed/wps": 12888.553022970294, "speed/FLOPS": 202432435953553.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05815030261874199, "optim/lr": 0.002997434858723621, "optim/total_tokens": 3158835200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1888651847839355, "created_at": "2025-01-15T22:34:04.836385+00:00"} {"global_step": 6026, "acc_step": 0, "speed/wps": 12890.60878371813, "speed/FLOPS": 202464724501009.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055349916219711304, "optim/lr": 0.0029974298476676763, "optim/total_tokens": 3159359488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.0061662197113037, "created_at": "2025-01-15T22:34:15.010032+00:00"} {"global_step": 6027, "acc_step": 0, "speed/wps": 12891.462952590806, "speed/FLOPS": 202478140396906.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.057404883205890656, "optim/lr": 0.0029974248317261036, "optim/total_tokens": 3159883776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 3.245633602142334, "created_at": "2025-01-15T22:34:25.181927+00:00"} {"global_step": 6028, "acc_step": 0, "speed/wps": 12890.264797935202, "speed/FLOPS": 202459321731602.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05964280292391777, "optim/lr": 0.0029974198108989195, "optim/total_tokens": 3160408064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.144392967224121, "created_at": "2025-01-15T22:34:35.353887+00:00"} {"global_step": 6029, "acc_step": 0, "speed/wps": 12888.52071896521, "speed/FLOPS": 202431928574763.47, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061867713928222656, "optim/lr": 0.002997414785186139, "optim/total_tokens": 3160932352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 2.9610016345977783, "created_at": "2025-01-15T22:34:45.530022+00:00"} {"global_step": 6030, "acc_step": 0, "speed/wps": 12888.468698160468, "speed/FLOPS": 202431111516540.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059217724949121475, "optim/lr": 0.00299740975458778, "optim/total_tokens": 3161456640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0243592262268066, "created_at": "2025-01-15T22:34:55.703239+00:00"} {"global_step": 6031, "acc_step": 0, "speed/wps": 12887.954403618942, "speed/FLOPS": 202423033814051.5, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09424396604299545, "optim/lr": 0.0029974047191038584, "optim/total_tokens": 3161980928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 2.9927382469177246, "created_at": "2025-01-15T22:35:05.882946+00:00"} {"global_step": 6032, "acc_step": 0, "speed/wps": 12887.173795901857, "speed/FLOPS": 202410773297187.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08960005640983582, "optim/lr": 0.0029973996787343903, "optim/total_tokens": 3162505216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9630026817321777, "created_at": "2025-01-15T22:35:16.062237+00:00"} {"global_step": 6033, "acc_step": 0, "speed/wps": 12884.166916816323, "speed/FLOPS": 202363546129264.1, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09366489946842194, "optim/lr": 0.0029973946334793927, "optim/total_tokens": 3163029504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.0097498893737793, "created_at": "2025-01-15T22:35:26.239571+00:00"} {"global_step": 6034, "acc_step": 0, "speed/wps": 12891.134380097325, "speed/FLOPS": 202472979714390.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06468874216079712, "optim/lr": 0.002997389583338881, "optim/total_tokens": 3163553792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402407, "loss/out": 3.1387932300567627, "created_at": "2025-01-15T22:35:36.415131+00:00"} {"global_step": 6035, "acc_step": 0, "speed/wps": 12891.039289460405, "speed/FLOPS": 202471486185269.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06275064498186111, "optim/lr": 0.002997384528312873, "optim/total_tokens": 3164078080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.0646157264709473, "created_at": "2025-01-15T22:35:46.587206+00:00"} {"global_step": 6036, "acc_step": 0, "speed/wps": 12891.022243857495, "speed/FLOPS": 202471218460652.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07349596172571182, "optim/lr": 0.0029973794684013846, "optim/total_tokens": 3164602368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 3.044185161590576, "created_at": "2025-01-15T22:35:56.759564+00:00"} {"global_step": 6037, "acc_step": 0, "speed/wps": 12890.991719871985, "speed/FLOPS": 202470739039515.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05826419219374657, "optim/lr": 0.0029973744036044326, "optim/total_tokens": 3165126656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0647690296173096, "created_at": "2025-01-15T22:36:06.935858+00:00"} {"global_step": 6038, "acc_step": 0, "speed/wps": 12897.521345508007, "speed/FLOPS": 202573295782775.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06753747910261154, "optim/lr": 0.002997369333922033, "optim/total_tokens": 3165650944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383767, "loss/out": 3.1523776054382324, "created_at": "2025-01-15T22:36:17.102034+00:00"} {"global_step": 6039, "acc_step": 0, "speed/wps": 12886.548801757344, "speed/FLOPS": 202400956905315.75, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06979364901781082, "optim/lr": 0.0029973642593542024, "optim/total_tokens": 3166175232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9460129737854004, "created_at": "2025-01-15T22:36:27.278713+00:00"} {"global_step": 6040, "acc_step": 0, "speed/wps": 12887.44912370414, "speed/FLOPS": 202415097698662.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05459311977028847, "optim/lr": 0.002997359179900958, "optim/total_tokens": 3166699520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.093385696411133, "created_at": "2025-01-15T22:36:37.452962+00:00"} {"global_step": 6041, "acc_step": 0, "speed/wps": 12893.627358744661, "speed/FLOPS": 202512135369758.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07208188623189926, "optim/lr": 0.0029973540955623157, "optim/total_tokens": 3167223808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433677, "loss/out": 3.06969952583313, "created_at": "2025-01-15T22:36:47.626112+00:00"} {"global_step": 6042, "acc_step": 0, "speed/wps": 12886.100176947433, "speed/FLOPS": 202393910636201.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0834411084651947, "optim/lr": 0.002997349006338292, "optim/total_tokens": 3167748096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 3.0428411960601807, "created_at": "2025-01-15T22:36:57.806262+00:00"} {"global_step": 6043, "acc_step": 0, "speed/wps": 12890.569698169105, "speed/FLOPS": 202464110608753.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06774666905403137, "optim/lr": 0.002997343912228904, "optim/total_tokens": 3168272384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.0312252044677734, "created_at": "2025-01-15T22:37:07.980160+00:00"} {"global_step": 6044, "acc_step": 0, "speed/wps": 12892.579945400557, "speed/FLOPS": 202495684303892.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07341807335615158, "optim/lr": 0.0029973388132341684, "optim/total_tokens": 3168796672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.0004398822784424, "created_at": "2025-01-15T22:37:18.150531+00:00"} {"global_step": 6045, "acc_step": 0, "speed/wps": 12894.899193788671, "speed/FLOPS": 202532111286808.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08673606067895889, "optim/lr": 0.002997333709354101, "optim/total_tokens": 3169320960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 3.0535178184509277, "created_at": "2025-01-15T22:37:28.318678+00:00"} {"global_step": 6046, "acc_step": 0, "speed/wps": 12889.66989167582, "speed/FLOPS": 202449977911312.38, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07851545512676239, "optim/lr": 0.00299732860058872, "optim/total_tokens": 3169845248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.163301944732666, "created_at": "2025-01-15T22:37:38.493022+00:00"} {"global_step": 6047, "acc_step": 0, "speed/wps": 12887.927217608078, "speed/FLOPS": 202422606820393.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06502630561590195, "optim/lr": 0.0029973234869380403, "optim/total_tokens": 3170369536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.0905117988586426, "created_at": "2025-01-15T22:37:48.670064+00:00"} {"global_step": 6048, "acc_step": 0, "speed/wps": 12891.397133546949, "speed/FLOPS": 202477106618377.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0677972361445427, "optim/lr": 0.00299731836840208, "optim/total_tokens": 3170893824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432352, "loss/out": 3.0529351234436035, "created_at": "2025-01-15T22:37:58.842749+00:00"} {"global_step": 6049, "acc_step": 0, "speed/wps": 12887.211457755613, "speed/FLOPS": 202411364828353.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0772399753332138, "optim/lr": 0.0029973132449808548, "optim/total_tokens": 3171418112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 3.070490598678589, "created_at": "2025-01-15T22:38:09.016828+00:00"} {"global_step": 6050, "acc_step": 0, "speed/wps": 12891.96185962946, "speed/FLOPS": 202485976417518.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06415603309869766, "optim/lr": 0.0029973081166743817, "optim/total_tokens": 3171942400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.1159520149230957, "created_at": "2025-01-15T22:38:19.191946+00:00"} {"global_step": 6051, "acc_step": 0, "speed/wps": 12891.113290050936, "speed/FLOPS": 202472648466230.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058071885257959366, "optim/lr": 0.002997302983482678, "optim/total_tokens": 3172466688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.081634521484375, "created_at": "2025-01-15T22:38:29.367279+00:00"} {"global_step": 6052, "acc_step": 0, "speed/wps": 12889.213510072312, "speed/FLOPS": 202442809811095.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057381562888622284, "optim/lr": 0.002997297845405759, "optim/total_tokens": 3172990976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.9986414909362793, "created_at": "2025-01-15T22:38:39.542034+00:00"} {"global_step": 6053, "acc_step": 0, "speed/wps": 12889.997677627553, "speed/FLOPS": 202455126240109.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06221339479088783, "optim/lr": 0.0029972927024436435, "optim/total_tokens": 3173515264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 3.083120346069336, "created_at": "2025-01-15T22:38:49.717401+00:00"} {"global_step": 6054, "acc_step": 0, "speed/wps": 12891.28654314886, "speed/FLOPS": 202475369644208.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06892654299736023, "optim/lr": 0.0029972875545963467, "optim/total_tokens": 3174039552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9692349433898926, "created_at": "2025-01-15T22:38:59.890099+00:00"} {"global_step": 6055, "acc_step": 0, "speed/wps": 12889.790108813313, "speed/FLOPS": 202451866086652.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05720856785774231, "optim/lr": 0.002997282401863886, "optim/total_tokens": 3174563840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0286154747009277, "created_at": "2025-01-15T22:39:10.065100+00:00"} {"global_step": 6056, "acc_step": 0, "speed/wps": 12891.103357289989, "speed/FLOPS": 202472492458571.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05817912146449089, "optim/lr": 0.002997277244246278, "optim/total_tokens": 3175088128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 3.0968708992004395, "created_at": "2025-01-15T22:39:20.238833+00:00"} {"global_step": 6057, "acc_step": 0, "speed/wps": 12887.842268406857, "speed/FLOPS": 202421272576455.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0738009363412857, "optim/lr": 0.00299727208174354, "optim/total_tokens": 3175612416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.1609394550323486, "created_at": "2025-01-15T22:39:30.416132+00:00"} {"global_step": 6058, "acc_step": 0, "speed/wps": 12878.65520728207, "speed/FLOPS": 202276977157146.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061892032623291016, "optim/lr": 0.0029972669143556886, "optim/total_tokens": 3176136704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.1606335639953613, "created_at": "2025-01-15T22:39:40.600401+00:00"} {"global_step": 6059, "acc_step": 0, "speed/wps": 12889.959334823723, "speed/FLOPS": 202454524013686.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06730977445840836, "optim/lr": 0.0029972617420827405, "optim/total_tokens": 3176660992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9428796768188477, "created_at": "2025-01-15T22:39:50.778090+00:00"} {"global_step": 6060, "acc_step": 0, "speed/wps": 12891.282642910319, "speed/FLOPS": 202475308385602.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05529182031750679, "optim/lr": 0.0029972565649247123, "optim/total_tokens": 3177185280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 3.0833933353424072, "created_at": "2025-01-15T22:40:00.949140+00:00"} {"global_step": 6061, "acc_step": 0, "speed/wps": 12887.719218620528, "speed/FLOPS": 202419339910473.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07078394293785095, "optim/lr": 0.0029972513828816216, "optim/total_tokens": 3177709568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.1269640922546387, "created_at": "2025-01-15T22:40:11.130436+00:00"} {"global_step": 6062, "acc_step": 0, "speed/wps": 12885.255450790039, "speed/FLOPS": 202380643051123.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06147594377398491, "optim/lr": 0.0029972461959534846, "optim/total_tokens": 3178233856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.1012704372406006, "created_at": "2025-01-15T22:40:21.306263+00:00"} {"global_step": 6063, "acc_step": 0, "speed/wps": 12884.031057174407, "speed/FLOPS": 202361412266897.1, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08867739886045456, "optim/lr": 0.0029972410041403186, "optim/total_tokens": 3178758144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.1476945877075195, "created_at": "2025-01-15T22:40:31.483358+00:00"} {"global_step": 6064, "acc_step": 0, "speed/wps": 12883.926414979502, "speed/FLOPS": 202359768717432.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.09702972322702408, "optim/lr": 0.002997235807442141, "optim/total_tokens": 3179282432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312376, "loss/out": 3.062809944152832, "created_at": "2025-01-15T22:40:41.664762+00:00"} {"global_step": 6065, "acc_step": 0, "speed/wps": 12889.7560844667, "speed/FLOPS": 202451331687534.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061048708856105804, "optim/lr": 0.002997230605858968, "optim/total_tokens": 3179806720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.048063278198242, "created_at": "2025-01-15T22:40:51.836999+00:00"} {"global_step": 6066, "acc_step": 0, "speed/wps": 12888.409335185395, "speed/FLOPS": 202430179139444.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08392041176557541, "optim/lr": 0.0029972253993908165, "optim/total_tokens": 3180331008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.1472208499908447, "created_at": "2025-01-15T22:41:02.015172+00:00"} {"global_step": 6067, "acc_step": 0, "speed/wps": 12886.896141076984, "speed/FLOPS": 202406412346624.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08495955914258957, "optim/lr": 0.0029972201880377036, "optim/total_tokens": 3180855296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 3.1515233516693115, "created_at": "2025-01-15T22:41:12.192040+00:00"} {"global_step": 6068, "acc_step": 0, "speed/wps": 12885.783871694435, "speed/FLOPS": 202388942627553.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0795615017414093, "optim/lr": 0.002997214971799647, "optim/total_tokens": 3181379584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.085765838623047, "created_at": "2025-01-15T22:41:22.368835+00:00"} {"global_step": 6069, "acc_step": 0, "speed/wps": 12886.151860067339, "speed/FLOPS": 202394722390619.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06615903228521347, "optim/lr": 0.002997209750676663, "optim/total_tokens": 3181903872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 3.097554922103882, "created_at": "2025-01-15T22:41:32.545525+00:00"} {"global_step": 6070, "acc_step": 0, "speed/wps": 12885.97779754105, "speed/FLOPS": 202391988499456.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09224256128072739, "optim/lr": 0.002997204524668769, "optim/total_tokens": 3182428160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.187743663787842, "created_at": "2025-01-15T22:41:42.720765+00:00"} {"global_step": 6071, "acc_step": 0, "speed/wps": 12888.32486118056, "speed/FLOPS": 202428852359118.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09432020038366318, "optim/lr": 0.0029971992937759826, "optim/total_tokens": 3182952448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 3.092062473297119, "created_at": "2025-01-15T22:41:52.897757+00:00"} {"global_step": 6072, "acc_step": 0, "speed/wps": 12889.406978438064, "speed/FLOPS": 202445848497632.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05973803624510765, "optim/lr": 0.002997194057998319, "optim/total_tokens": 3183476736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.0881872177124023, "created_at": "2025-01-15T22:42:03.074190+00:00"} {"global_step": 6073, "acc_step": 0, "speed/wps": 12889.402467852115, "speed/FLOPS": 202445777652682.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08571594953536987, "optim/lr": 0.0029971888173357976, "optim/total_tokens": 3184001024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.123786687850952, "created_at": "2025-01-15T22:42:13.247906+00:00"} {"global_step": 6074, "acc_step": 0, "speed/wps": 12886.299237815168, "speed/FLOPS": 202397037160667.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0659787580370903, "optim/lr": 0.0029971835717884335, "optim/total_tokens": 3184525312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0939741134643555, "created_at": "2025-01-15T22:42:23.424788+00:00"} {"global_step": 6075, "acc_step": 0, "speed/wps": 12884.199081048248, "speed/FLOPS": 202364051312724.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055795274674892426, "optim/lr": 0.002997178321356245, "optim/total_tokens": 3185049600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363731, "loss/out": 3.162935733795166, "created_at": "2025-01-15T22:42:33.603098+00:00"} {"global_step": 6076, "acc_step": 0, "speed/wps": 12888.346345759439, "speed/FLOPS": 202429189803952.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07151748985052109, "optim/lr": 0.002997173066039249, "optim/total_tokens": 3185573888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.1213538646698, "created_at": "2025-01-15T22:42:43.777280+00:00"} {"global_step": 6077, "acc_step": 0, "speed/wps": 12888.542138099456, "speed/FLOPS": 202432264991699.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059169452637434006, "optim/lr": 0.0029971678058374625, "optim/total_tokens": 3186098176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.1299498081207275, "created_at": "2025-01-15T22:42:53.956198+00:00"} {"global_step": 6078, "acc_step": 0, "speed/wps": 12891.774145874362, "speed/FLOPS": 202483028115048.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06322433054447174, "optim/lr": 0.0029971625407509027, "optim/total_tokens": 3186622464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.226668357849121, "created_at": "2025-01-15T22:43:04.127005+00:00"} {"global_step": 6079, "acc_step": 0, "speed/wps": 12887.197063266452, "speed/FLOPS": 202411138743120.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.069393590092659, "optim/lr": 0.0029971572707795865, "optim/total_tokens": 3187146752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 3.0703933238983154, "created_at": "2025-01-15T22:43:14.303885+00:00"} {"global_step": 6080, "acc_step": 0, "speed/wps": 12888.580407554784, "speed/FLOPS": 202432866066083.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061059024184942245, "optim/lr": 0.002997151995923532, "optim/total_tokens": 3187671040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.030644655227661, "created_at": "2025-01-15T22:43:24.478203+00:00"} {"global_step": 6081, "acc_step": 0, "speed/wps": 12887.060201578968, "speed/FLOPS": 202408989142251.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06500278413295746, "optim/lr": 0.002997146716182755, "optim/total_tokens": 3188195328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506128, "loss/out": 3.0800704956054688, "created_at": "2025-01-15T22:43:34.652903+00:00"} {"global_step": 6082, "acc_step": 0, "speed/wps": 12886.175520182012, "speed/FLOPS": 202395094005232.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06860918551683426, "optim/lr": 0.0029971414315572748, "optim/total_tokens": 3188719616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 3.0802931785583496, "created_at": "2025-01-15T22:43:44.828677+00:00"} {"global_step": 6083, "acc_step": 0, "speed/wps": 12887.182019773369, "speed/FLOPS": 202410902464390.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07769256830215454, "optim/lr": 0.0029971361420471064, "optim/total_tokens": 3189243904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 3.157824993133545, "created_at": "2025-01-15T22:43:55.010606+00:00"} {"global_step": 6084, "acc_step": 0, "speed/wps": 12888.672401300724, "speed/FLOPS": 202434310954276.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07285192608833313, "optim/lr": 0.002997130847652268, "optim/total_tokens": 3189768192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.0345141887664795, "created_at": "2025-01-15T22:44:05.188696+00:00"} {"global_step": 6085, "acc_step": 0, "speed/wps": 12892.465915615177, "speed/FLOPS": 202493893309419.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08573459833860397, "optim/lr": 0.002997125548372777, "optim/total_tokens": 3190292480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.0651071071624756, "created_at": "2025-01-15T22:44:15.360513+00:00"} {"global_step": 6086, "acc_step": 0, "speed/wps": 12882.659509907746, "speed/FLOPS": 202339870232370.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0895778089761734, "optim/lr": 0.002997120244208651, "optim/total_tokens": 3190816768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.1380248069763184, "created_at": "2025-01-15T22:44:25.540512+00:00"} {"global_step": 6087, "acc_step": 0, "speed/wps": 12888.197537928576, "speed/FLOPS": 202426852572485.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08106748014688492, "optim/lr": 0.002997114935159907, "optim/total_tokens": 3191341056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0756168365478516, "created_at": "2025-01-15T22:44:35.717580+00:00"} {"global_step": 6088, "acc_step": 0, "speed/wps": 12887.71417932254, "speed/FLOPS": 202419260761373.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07634840905666351, "optim/lr": 0.0029971096212265615, "optim/total_tokens": 3191865344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 3.0646605491638184, "created_at": "2025-01-15T22:44:45.897372+00:00"} {"global_step": 6089, "acc_step": 0, "speed/wps": 12886.883420547387, "speed/FLOPS": 202406212553227.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06876068562269211, "optim/lr": 0.002997104302408633, "optim/total_tokens": 3192389632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.958892345428467, "created_at": "2025-01-15T22:44:56.072300+00:00"} {"global_step": 6090, "acc_step": 0, "speed/wps": 12886.940494385293, "speed/FLOPS": 202407108976279.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07388999313116074, "optim/lr": 0.0029970989787061384, "optim/total_tokens": 3192913920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428074, "loss/out": 3.041274070739746, "created_at": "2025-01-15T22:45:06.247143+00:00"} {"global_step": 6091, "acc_step": 0, "speed/wps": 12886.434939989007, "speed/FLOPS": 202399168549781.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06793937087059021, "optim/lr": 0.002997093650119095, "optim/total_tokens": 3193438208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.1501598358154297, "created_at": "2025-01-15T22:45:16.423208+00:00"} {"global_step": 6092, "acc_step": 0, "speed/wps": 12886.855558191262, "speed/FLOPS": 202405774936637.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0738406628370285, "optim/lr": 0.0029970883166475205, "optim/total_tokens": 3193962496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.1647250652313232, "created_at": "2025-01-15T22:45:26.599700+00:00"} {"global_step": 6093, "acc_step": 0, "speed/wps": 12886.253936744997, "speed/FLOPS": 202396325645109.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07630418986082077, "optim/lr": 0.0029970829782914317, "optim/total_tokens": 3194486784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.01332688331604, "created_at": "2025-01-15T22:45:36.775070+00:00"} {"global_step": 6094, "acc_step": 0, "speed/wps": 12886.897397791152, "speed/FLOPS": 202406432085047.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06955428421497345, "optim/lr": 0.002997077635050847, "optim/total_tokens": 3195011072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 3.0490357875823975, "created_at": "2025-01-15T22:45:46.952417+00:00"} {"global_step": 6095, "acc_step": 0, "speed/wps": 12893.566605175218, "speed/FLOPS": 202511181151465.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07981625944375992, "optim/lr": 0.002997072286925782, "optim/total_tokens": 3195535360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.0602097511291504, "created_at": "2025-01-15T22:45:57.127075+00:00"} {"global_step": 6096, "acc_step": 0, "speed/wps": 12885.654237890181, "speed/FLOPS": 202386906550522.38, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06001981347799301, "optim/lr": 0.0029970669339162564, "optim/total_tokens": 3196059648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.034898281097412, "created_at": "2025-01-15T22:46:07.306752+00:00"} {"global_step": 6097, "acc_step": 0, "speed/wps": 12886.13548693657, "speed/FLOPS": 202394465228101.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06935764104127884, "optim/lr": 0.0029970615760222864, "optim/total_tokens": 3196583936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.0963728427886963, "created_at": "2025-01-15T22:46:17.509187+00:00"} {"global_step": 6098, "acc_step": 0, "speed/wps": 12888.257391122059, "speed/FLOPS": 202427792649136.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06178215891122818, "optim/lr": 0.0029970562132438897, "optim/total_tokens": 3197108224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.0835986137390137, "created_at": "2025-01-15T22:46:27.690363+00:00"} {"global_step": 6099, "acc_step": 0, "speed/wps": 12891.83404292055, "speed/FLOPS": 202483968880466.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07726194709539413, "optim/lr": 0.0029970508455810837, "optim/total_tokens": 3197632512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352528, "loss/out": 3.0181498527526855, "created_at": "2025-01-15T22:46:37.863888+00:00"} {"global_step": 6100, "acc_step": 0, "speed/wps": 12887.986040325404, "speed/FLOPS": 202423530711999.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06975840777158737, "optim/lr": 0.0029970454730338866, "optim/total_tokens": 3198156800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.104076385498047, "created_at": "2025-01-15T22:46:48.039814+00:00"} {"global_step": 6101, "acc_step": 0, "speed/wps": 12888.02435722785, "speed/FLOPS": 202424132531605.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0642552375793457, "optim/lr": 0.0029970400956023146, "optim/total_tokens": 3198681088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.1826634407043457, "created_at": "2025-01-15T22:46:58.213520+00:00"} {"global_step": 6102, "acc_step": 0, "speed/wps": 12886.784445536867, "speed/FLOPS": 202404658014681.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0681716576218605, "optim/lr": 0.0029970347132863865, "optim/total_tokens": 3199205376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351344, "loss/out": 3.154313564300537, "created_at": "2025-01-15T22:47:08.391742+00:00"} {"global_step": 6103, "acc_step": 0, "speed/wps": 12885.568810169414, "speed/FLOPS": 202385564790777.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05949195474386215, "optim/lr": 0.002997029326086119, "optim/total_tokens": 3199729664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 3.1228818893432617, "created_at": "2025-01-15T22:47:18.568187+00:00"} {"global_step": 6104, "acc_step": 0, "speed/wps": 12889.973674094248, "speed/FLOPS": 202454749231634.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061030879616737366, "optim/lr": 0.0029970239340015308, "optim/total_tokens": 3200253952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.04648494720459, "created_at": "2025-01-15T22:47:28.743471+00:00"} {"global_step": 6105, "acc_step": 0, "speed/wps": 12887.38326067268, "speed/FLOPS": 202414063229247.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.060748644173145294, "optim/lr": 0.002997018537032638, "optim/total_tokens": 3200778240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 3.0750255584716797, "created_at": "2025-01-15T22:47:38.921024+00:00"} {"global_step": 6106, "acc_step": 0, "speed/wps": 12887.021928170447, "speed/FLOPS": 202408388005777.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06757582724094391, "optim/lr": 0.002997013135179459, "optim/total_tokens": 3201302528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0007925033569336, "created_at": "2025-01-15T22:47:49.096512+00:00"} {"global_step": 6107, "acc_step": 0, "speed/wps": 12889.360536434631, "speed/FLOPS": 202445119062150.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10166410356760025, "optim/lr": 0.0029970077284420115, "optim/total_tokens": 3201826816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.0097551345825195, "created_at": "2025-01-15T22:47:59.269407+00:00"} {"global_step": 6108, "acc_step": 0, "speed/wps": 12887.196896876825, "speed/FLOPS": 202411136129743.2, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07815255969762802, "optim/lr": 0.0029970023168203125, "optim/total_tokens": 3202351104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9695372581481934, "created_at": "2025-01-15T22:48:09.447581+00:00"} {"global_step": 6109, "acc_step": 0, "speed/wps": 12884.430504535661, "speed/FLOPS": 202367686136603.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06098543480038643, "optim/lr": 0.0029969969003143812, "optim/total_tokens": 3202875392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.158682346343994, "created_at": "2025-01-15T22:48:19.629460+00:00"} {"global_step": 6110, "acc_step": 0, "speed/wps": 12886.11434245129, "speed/FLOPS": 202394133124904.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07102515548467636, "optim/lr": 0.0029969914789242332, "optim/total_tokens": 3203399680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0798261165618896, "created_at": "2025-01-15T22:48:29.804593+00:00"} {"global_step": 6111, "acc_step": 0, "speed/wps": 12888.045864409349, "speed/FLOPS": 202424470331445.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06916987150907516, "optim/lr": 0.002996986052649888, "optim/total_tokens": 3203923968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 3.0853452682495117, "created_at": "2025-01-15T22:48:39.981313+00:00"} {"global_step": 6112, "acc_step": 0, "speed/wps": 12885.800053981882, "speed/FLOPS": 202389196792614.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0651242733001709, "optim/lr": 0.002996980621491362, "optim/total_tokens": 3204448256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.1572463512420654, "created_at": "2025-01-15T22:48:50.159254+00:00"} {"global_step": 6113, "acc_step": 0, "speed/wps": 12888.529057865848, "speed/FLOPS": 202432059548656.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07599213719367981, "optim/lr": 0.002996975185448674, "optim/total_tokens": 3204972544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.1840639114379883, "created_at": "2025-01-15T22:49:00.333500+00:00"} {"global_step": 6114, "acc_step": 0, "speed/wps": 12889.263377852236, "speed/FLOPS": 202443593053104.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06788328289985657, "optim/lr": 0.0029969697445218404, "optim/total_tokens": 3205496832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 2.9327144622802734, "created_at": "2025-01-15T22:49:10.508360+00:00"} {"global_step": 6115, "acc_step": 0, "speed/wps": 12888.064389227948, "speed/FLOPS": 202424761289177.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07204222679138184, "optim/lr": 0.0029969642987108806, "optim/total_tokens": 3206021120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.98252010345459, "created_at": "2025-01-15T22:49:20.683170+00:00"} {"global_step": 6116, "acc_step": 0, "speed/wps": 12887.472558459678, "speed/FLOPS": 202415465773700.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06528186798095703, "optim/lr": 0.0029969588480158113, "optim/total_tokens": 3206545408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.960465908050537, "created_at": "2025-01-15T22:49:30.857208+00:00"} {"global_step": 6117, "acc_step": 0, "speed/wps": 12885.875674505593, "speed/FLOPS": 202390384516853.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07435403764247894, "optim/lr": 0.00299695339243665, "optim/total_tokens": 3207069696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.999950885772705, "created_at": "2025-01-15T22:49:41.035896+00:00"} {"global_step": 6118, "acc_step": 0, "speed/wps": 12893.337038585285, "speed/FLOPS": 202507575492713.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06631164252758026, "optim/lr": 0.002996947931973415, "optim/total_tokens": 3207593984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 3.107114315032959, "created_at": "2025-01-15T22:49:51.208177+00:00"} {"global_step": 6119, "acc_step": 0, "speed/wps": 12886.683516829024, "speed/FLOPS": 202403072790634.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06065238267183304, "optim/lr": 0.002996942466626125, "optim/total_tokens": 3208118272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0470285415649414, "created_at": "2025-01-15T22:50:01.383025+00:00"} {"global_step": 6120, "acc_step": 0, "speed/wps": 12888.177793531, "speed/FLOPS": 202426542459590.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05815810710191727, "optim/lr": 0.0029969369963947956, "optim/total_tokens": 3208642560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.0699071884155273, "created_at": "2025-01-15T22:50:11.557883+00:00"} {"global_step": 6121, "acc_step": 0, "speed/wps": 12885.86267637222, "speed/FLOPS": 202390180363307.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06576719135046005, "optim/lr": 0.002996931521279447, "optim/total_tokens": 3209166848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286610, "loss/out": 2.8391056060791016, "created_at": "2025-01-15T22:50:21.733461+00:00"} {"global_step": 6122, "acc_step": 0, "speed/wps": 12884.039265240466, "speed/FLOPS": 202361541185853.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06398172676563263, "optim/lr": 0.0029969260412800957, "optim/total_tokens": 3209691136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.0965147018432617, "created_at": "2025-01-15T22:50:31.913570+00:00"} {"global_step": 6123, "acc_step": 0, "speed/wps": 12890.63579685435, "speed/FLOPS": 202465148779434.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06162186339497566, "optim/lr": 0.00299692055639676, "optim/total_tokens": 3210215424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 3.001629114151001, "created_at": "2025-01-15T22:50:42.087478+00:00"} {"global_step": 6124, "acc_step": 0, "speed/wps": 12887.351140725701, "speed/FLOPS": 202413558741343.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07198876887559891, "optim/lr": 0.0029969150666294578, "optim/total_tokens": 3210739712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.010899782180786, "created_at": "2025-01-15T22:50:52.292874+00:00"} {"global_step": 6125, "acc_step": 0, "speed/wps": 12893.242094988862, "speed/FLOPS": 202506084273065.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06316839158535004, "optim/lr": 0.0029969095719782073, "optim/total_tokens": 3211264000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.124958038330078, "created_at": "2025-01-15T22:51:02.464428+00:00"} {"global_step": 6126, "acc_step": 0, "speed/wps": 12889.21192120348, "speed/FLOPS": 202442784855726.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06914002448320389, "optim/lr": 0.002996904072443025, "optim/total_tokens": 3211788288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 3.1789116859436035, "created_at": "2025-01-15T22:51:12.637210+00:00"} {"global_step": 6127, "acc_step": 0, "speed/wps": 12888.393884023339, "speed/FLOPS": 202429936457711.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06537614017724991, "optim/lr": 0.002996898568023931, "optim/total_tokens": 3212312576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 3.1208956241607666, "created_at": "2025-01-15T22:51:22.812806+00:00"} {"global_step": 6128, "acc_step": 0, "speed/wps": 12889.837936338643, "speed/FLOPS": 202452617283659.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07030440121889114, "optim/lr": 0.0029968930587209413, "optim/total_tokens": 3212836864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.103963851928711, "created_at": "2025-01-15T22:51:32.988116+00:00"} {"global_step": 6129, "acc_step": 0, "speed/wps": 12887.82072295884, "speed/FLOPS": 202420934175588.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07050839066505432, "optim/lr": 0.0029968875445340747, "optim/total_tokens": 3213361152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 3.092254877090454, "created_at": "2025-01-15T22:51:43.171612+00:00"} {"global_step": 6130, "acc_step": 0, "speed/wps": 12889.80433201384, "speed/FLOPS": 202452089481561.38, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05702376365661621, "optim/lr": 0.0029968820254633495, "optim/total_tokens": 3213885440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.08913516998291, "created_at": "2025-01-15T22:51:53.343762+00:00"} {"global_step": 6131, "acc_step": 0, "speed/wps": 12891.14592350561, "speed/FLOPS": 202473161019479.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07415731996297836, "optim/lr": 0.0029968765015087836, "optim/total_tokens": 3214409728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.1209731101989746, "created_at": "2025-01-15T22:52:03.514885+00:00"} {"global_step": 6132, "acc_step": 0, "speed/wps": 12890.114278789602, "speed/FLOPS": 202456957621586.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06272179633378983, "optim/lr": 0.002996870972670395, "optim/total_tokens": 3214934016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.0797014236450195, "created_at": "2025-01-15T22:52:13.692222+00:00"} {"global_step": 6133, "acc_step": 0, "speed/wps": 12885.136396460035, "speed/FLOPS": 202378773139273.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07375071197748184, "optim/lr": 0.0029968654389482016, "optim/total_tokens": 3215458304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 3.024528741836548, "created_at": "2025-01-15T22:52:23.868809+00:00"} {"global_step": 6134, "acc_step": 0, "speed/wps": 12893.19307166184, "speed/FLOPS": 202505314294349.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08486083894968033, "optim/lr": 0.002996859900342221, "optim/total_tokens": 3215982592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 3.042048931121826, "created_at": "2025-01-15T22:52:34.039821+00:00"} {"global_step": 6135, "acc_step": 0, "speed/wps": 12888.656427799566, "speed/FLOPS": 202434060068490.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07365180552005768, "optim/lr": 0.0029968543568524722, "optim/total_tokens": 3216506880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.881786346435547, "created_at": "2025-01-15T22:52:44.216045+00:00"} {"global_step": 6136, "acc_step": 0, "speed/wps": 12886.263696848631, "speed/FLOPS": 202396478940949.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07662727683782578, "optim/lr": 0.0029968488084789724, "optim/total_tokens": 3217031168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 3.1975698471069336, "created_at": "2025-01-15T22:52:54.396535+00:00"} {"global_step": 6137, "acc_step": 0, "speed/wps": 12886.967813116542, "speed/FLOPS": 202407538054493.88, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06990549713373184, "optim/lr": 0.0029968432552217403, "optim/total_tokens": 3217555456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 3.0925726890563965, "created_at": "2025-01-15T22:53:04.573712+00:00"} {"global_step": 6138, "acc_step": 0, "speed/wps": 12886.280643736813, "speed/FLOPS": 202396745115115.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07809333503246307, "optim/lr": 0.0029968376970807937, "optim/total_tokens": 3218079744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 3.10408353805542, "created_at": "2025-01-15T22:53:14.750223+00:00"} {"global_step": 6139, "acc_step": 0, "speed/wps": 12886.96330659763, "speed/FLOPS": 202407467273421.78, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06697912514209747, "optim/lr": 0.002996832134056151, "optim/total_tokens": 3218604032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.152489423751831, "created_at": "2025-01-15T22:53:24.927959+00:00"} {"global_step": 6140, "acc_step": 0, "speed/wps": 12892.862150333074, "speed/FLOPS": 202500116720147.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07103037089109421, "optim/lr": 0.00299682656614783, "optim/total_tokens": 3219128320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 2.9696145057678223, "created_at": "2025-01-15T22:53:35.097708+00:00"} {"global_step": 6141, "acc_step": 0, "speed/wps": 12895.216416961994, "speed/FLOPS": 202537093712656.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07317808270454407, "optim/lr": 0.0029968209933558493, "optim/total_tokens": 3219652608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.07277250289917, "created_at": "2025-01-15T22:53:45.267994+00:00"} {"global_step": 6142, "acc_step": 0, "speed/wps": 12893.85005467253, "speed/FLOPS": 202515633115321.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08161604404449463, "optim/lr": 0.002996815415680227, "optim/total_tokens": 3220176896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.1617958545684814, "created_at": "2025-01-15T22:53:55.439195+00:00"} {"global_step": 6143, "acc_step": 0, "speed/wps": 12889.314153788671, "speed/FLOPS": 202444390558959.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07998392730951309, "optim/lr": 0.002996809833120981, "optim/total_tokens": 3220701184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422232, "loss/out": 3.076361656188965, "created_at": "2025-01-15T22:54:05.613821+00:00"} {"global_step": 6144, "acc_step": 0, "speed/wps": 12886.541163948252, "speed/FLOPS": 202400836943028.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0651121586561203, "optim/lr": 0.00299680424567813, "optim/total_tokens": 3221225472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 3.010176181793213, "created_at": "2025-01-15T22:54:15.792122+00:00"} {"global_step": 6145, "acc_step": 0, "speed/wps": 12890.31973289171, "speed/FLOPS": 202460184560583.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06787434965372086, "optim/lr": 0.002996798653351691, "optim/total_tokens": 3221749760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 3.077826499938965, "created_at": "2025-01-15T22:54:25.968915+00:00"} {"global_step": 6146, "acc_step": 0, "speed/wps": 12891.608560085144, "speed/FLOPS": 202480427362690.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06137577444314957, "optim/lr": 0.002996793056141684, "optim/total_tokens": 3222274048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.109571933746338, "created_at": "2025-01-15T22:54:36.139801+00:00"} {"global_step": 6147, "acc_step": 0, "speed/wps": 12884.52576900112, "speed/FLOPS": 202369182395940.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05796222761273384, "optim/lr": 0.0029967874540481262, "optim/total_tokens": 3222798336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 3.1581380367279053, "created_at": "2025-01-15T22:54:46.321334+00:00"} {"global_step": 6148, "acc_step": 0, "speed/wps": 12893.49579406414, "speed/FLOPS": 202510068965661.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05465112626552582, "optim/lr": 0.002996781847071036, "optim/total_tokens": 3223322624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.1086442470550537, "created_at": "2025-01-15T22:54:56.492814+00:00"} {"global_step": 6149, "acc_step": 0, "speed/wps": 12885.163118874632, "speed/FLOPS": 202379192851514.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05548741668462753, "optim/lr": 0.0029967762352104315, "optim/total_tokens": 3223846912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.142974376678467, "created_at": "2025-01-15T22:55:06.676724+00:00"} {"global_step": 6150, "acc_step": 0, "speed/wps": 12886.081640077009, "speed/FLOPS": 202393619489178.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06274890154600143, "optim/lr": 0.0029967706184663317, "optim/total_tokens": 3224371200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.0891294479370117, "created_at": "2025-01-15T22:55:16.851971+00:00"} {"global_step": 6151, "acc_step": 0, "speed/wps": 12889.819654336447, "speed/FLOPS": 202452330139692.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06137322634458542, "optim/lr": 0.002996764996838754, "optim/total_tokens": 3224895488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.026637554168701, "created_at": "2025-01-15T22:55:27.030570+00:00"} {"global_step": 6152, "acc_step": 0, "speed/wps": 12891.78110141848, "speed/FLOPS": 202483137361427.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06161738187074661, "optim/lr": 0.002996759370327718, "optim/total_tokens": 3225419776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 3.128575325012207, "created_at": "2025-01-15T22:55:37.201449+00:00"} {"global_step": 6153, "acc_step": 0, "speed/wps": 12883.986209637322, "speed/FLOPS": 202360707874700.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.059546131640672684, "optim/lr": 0.0029967537389332414, "optim/total_tokens": 3225944064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.1202337741851807, "created_at": "2025-01-15T22:55:47.383563+00:00"} {"global_step": 6154, "acc_step": 0, "speed/wps": 12890.832057849233, "speed/FLOPS": 202468231328049.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08600182831287384, "optim/lr": 0.002996748102655342, "optim/total_tokens": 3226468352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 3.0467467308044434, "created_at": "2025-01-15T22:55:57.554916+00:00"} {"global_step": 6155, "acc_step": 0, "speed/wps": 12893.301442250477, "speed/FLOPS": 202507016403360.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08508920669555664, "optim/lr": 0.002996742461494039, "optim/total_tokens": 3226992640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 3.00315260887146, "created_at": "2025-01-15T22:56:07.735925+00:00"} {"global_step": 6156, "acc_step": 0, "speed/wps": 12892.917566884515, "speed/FLOPS": 202500987113238.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07448381185531616, "optim/lr": 0.0029967368154493502, "optim/total_tokens": 3227516928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.0980048179626465, "created_at": "2025-01-15T22:56:17.915456+00:00"} {"global_step": 6157, "acc_step": 0, "speed/wps": 12885.982485066994, "speed/FLOPS": 202392062123493.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07722191512584686, "optim/lr": 0.0029967311645212945, "optim/total_tokens": 3228041216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.1454784870147705, "created_at": "2025-01-15T22:56:28.090726+00:00"} {"global_step": 6158, "acc_step": 0, "speed/wps": 12891.295566961404, "speed/FLOPS": 202475511375584.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07094407081604004, "optim/lr": 0.0029967255087098897, "optim/total_tokens": 3228565504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.00673770904541, "created_at": "2025-01-15T22:56:38.265705+00:00"} {"global_step": 6159, "acc_step": 0, "speed/wps": 12893.870388198195, "speed/FLOPS": 202515952481283.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06746964901685715, "optim/lr": 0.0029967198480151547, "optim/total_tokens": 3229089792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379944, "loss/out": 3.0345346927642822, "created_at": "2025-01-15T22:56:48.436325+00:00"} {"global_step": 6160, "acc_step": 0, "speed/wps": 12888.402765692823, "speed/FLOPS": 202430075956535.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06440149992704391, "optim/lr": 0.002996714182437109, "optim/total_tokens": 3229614080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.094675064086914, "created_at": "2025-01-15T22:56:58.613062+00:00"} {"global_step": 6161, "acc_step": 0, "speed/wps": 12888.58257226506, "speed/FLOPS": 202432900065832.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07498887181282043, "optim/lr": 0.002996708511975769, "optim/total_tokens": 3230138368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.1239144802093506, "created_at": "2025-01-15T22:57:08.786187+00:00"} {"global_step": 6162, "acc_step": 0, "speed/wps": 12886.804064125892, "speed/FLOPS": 202404966151581.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07144927978515625, "optim/lr": 0.0029967028366311545, "optim/total_tokens": 3230662656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.003297805786133, "created_at": "2025-01-15T22:57:18.964444+00:00"} {"global_step": 6163, "acc_step": 0, "speed/wps": 12897.856717598846, "speed/FLOPS": 202578563262311.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07334110885858536, "optim/lr": 0.002996697156403284, "optim/total_tokens": 3231186944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 2.9998779296875, "created_at": "2025-01-15T22:57:29.130643+00:00"} {"global_step": 6164, "acc_step": 0, "speed/wps": 12890.688603795152, "speed/FLOPS": 202465978185003.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06753326207399368, "optim/lr": 0.002996691471292176, "optim/total_tokens": 3231711232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.0454673767089844, "created_at": "2025-01-15T22:57:39.303810+00:00"} {"global_step": 6165, "acc_step": 0, "speed/wps": 12889.484790969434, "speed/FLOPS": 202447070650363.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05757037550210953, "optim/lr": 0.0029966857812978483, "optim/total_tokens": 3232235520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.0173044204711914, "created_at": "2025-01-15T22:57:49.488821+00:00"} {"global_step": 6166, "acc_step": 0, "speed/wps": 12887.147901808758, "speed/FLOPS": 202410366594872.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08848312497138977, "optim/lr": 0.0029966800864203204, "optim/total_tokens": 3232759808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 3.1591134071350098, "created_at": "2025-01-15T22:57:59.665935+00:00"} {"global_step": 6167, "acc_step": 0, "speed/wps": 12894.163730628932, "speed/FLOPS": 202520559827252.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06267333775758743, "optim/lr": 0.00299667438665961, "optim/total_tokens": 3233284096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.103595495223999, "created_at": "2025-01-15T22:58:09.837890+00:00"} {"global_step": 6168, "acc_step": 0, "speed/wps": 12893.659960041583, "speed/FLOPS": 202512647417925.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0769386813044548, "optim/lr": 0.0029966686820157365, "optim/total_tokens": 3233808384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.0982537269592285, "created_at": "2025-01-15T22:58:20.011901+00:00"} {"global_step": 6169, "acc_step": 0, "speed/wps": 12890.562885614283, "speed/FLOPS": 202464003608218.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08925511687994003, "optim/lr": 0.0029966629724887182, "optim/total_tokens": 3234332672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470959, "loss/out": 3.1452555656433105, "created_at": "2025-01-15T22:58:30.186413+00:00"} {"global_step": 6170, "acc_step": 0, "speed/wps": 12892.116347468187, "speed/FLOPS": 202488402861314.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06850642710924149, "optim/lr": 0.002996657258078574, "optim/total_tokens": 3234856960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329658, "loss/out": 3.1467761993408203, "created_at": "2025-01-15T22:58:40.357152+00:00"} {"global_step": 6171, "acc_step": 0, "speed/wps": 12892.370520708493, "speed/FLOPS": 202492395001323.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057739656418561935, "optim/lr": 0.0029966515387853214, "optim/total_tokens": 3235381248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.0577759742736816, "created_at": "2025-01-15T22:58:50.528332+00:00"} {"global_step": 6172, "acc_step": 0, "speed/wps": 12893.420595540374, "speed/FLOPS": 202508887869512.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07127156108617783, "optim/lr": 0.0029966458146089806, "optim/total_tokens": 3235905536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 3.18670916557312, "created_at": "2025-01-15T22:59:00.699916+00:00"} {"global_step": 6173, "acc_step": 0, "speed/wps": 12889.24480472987, "speed/FLOPS": 202443301336695.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07395799458026886, "optim/lr": 0.00299664008554957, "optim/total_tokens": 3236429824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 2.995553731918335, "created_at": "2025-01-15T22:59:10.872811+00:00"} {"global_step": 6174, "acc_step": 0, "speed/wps": 12889.959837748365, "speed/FLOPS": 202454531912809.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07190896570682526, "optim/lr": 0.0029966343516071073, "optim/total_tokens": 3236954112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 3.0294971466064453, "created_at": "2025-01-15T22:59:21.049204+00:00"} {"global_step": 6175, "acc_step": 0, "speed/wps": 12885.836296742926, "speed/FLOPS": 202389766034979.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06612490117549896, "optim/lr": 0.0029966286127816122, "optim/total_tokens": 3237478400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.0759479999542236, "created_at": "2025-01-15T22:59:31.224974+00:00"} {"global_step": 6176, "acc_step": 0, "speed/wps": 12888.903457885053, "speed/FLOPS": 202437940015438.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07455245405435562, "optim/lr": 0.0029966228690731025, "optim/total_tokens": 3238002688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460308, "loss/out": 3.1277289390563965, "created_at": "2025-01-15T22:59:41.397990+00:00"} {"global_step": 6177, "acc_step": 0, "speed/wps": 12884.847749068951, "speed/FLOPS": 202374239535353.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07184641063213348, "optim/lr": 0.0029966171204815987, "optim/total_tokens": 3238526976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 3.068925142288208, "created_at": "2025-01-15T22:59:51.575204+00:00"} {"global_step": 6178, "acc_step": 0, "speed/wps": 12888.245933143982, "speed/FLOPS": 202427612685845.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05956176668405533, "optim/lr": 0.002996611367007117, "optim/total_tokens": 3239051264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.110395908355713, "created_at": "2025-01-15T23:00:01.748895+00:00"} {"global_step": 6179, "acc_step": 0, "speed/wps": 12889.074224660832, "speed/FLOPS": 202440622142309.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07232324779033661, "optim/lr": 0.002996605608649678, "optim/total_tokens": 3239575552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.993130683898926, "created_at": "2025-01-15T23:00:11.926551+00:00"} {"global_step": 6180, "acc_step": 0, "speed/wps": 12889.964626159586, "speed/FLOPS": 202454607121387.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05904987081885338, "optim/lr": 0.0029965998454093003, "optim/total_tokens": 3240099840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 3.1163742542266846, "created_at": "2025-01-15T23:00:22.104871+00:00"} {"global_step": 6181, "acc_step": 0, "speed/wps": 12895.785373591149, "speed/FLOPS": 202546029958345.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08492092788219452, "optim/lr": 0.002996594077286002, "optim/total_tokens": 3240624128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0479519367218018, "created_at": "2025-01-15T23:00:32.272882+00:00"} {"global_step": 6182, "acc_step": 0, "speed/wps": 12886.676604543583, "speed/FLOPS": 202402964223693.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08045874536037445, "optim/lr": 0.0029965883042798024, "optim/total_tokens": 3241148416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.020183563232422, "created_at": "2025-01-15T23:00:42.451743+00:00"} {"global_step": 6183, "acc_step": 0, "speed/wps": 12882.30272381198, "speed/FLOPS": 202334266416461.25, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06122402846813202, "optim/lr": 0.00299658252639072, "optim/total_tokens": 3241672704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0317702293395996, "created_at": "2025-01-15T23:00:52.633651+00:00"} {"global_step": 6184, "acc_step": 0, "speed/wps": 12881.52772300423, "speed/FLOPS": 202322093963812.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06575611233711243, "optim/lr": 0.002996576743618774, "optim/total_tokens": 3242196992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8313090801239014, "created_at": "2025-01-15T23:01:02.815205+00:00"} {"global_step": 6185, "acc_step": 0, "speed/wps": 12886.14601731964, "speed/FLOPS": 202394630622237.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06731735169887543, "optim/lr": 0.0029965709559639834, "optim/total_tokens": 3242721280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.169236898422241, "created_at": "2025-01-15T23:01:12.990776+00:00"} {"global_step": 6186, "acc_step": 0, "speed/wps": 12884.895932746433, "speed/FLOPS": 202374996326219.34, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06618396937847137, "optim/lr": 0.002996565163426367, "optim/total_tokens": 3243245568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.998429775238037, "created_at": "2025-01-15T23:01:23.171500+00:00"} {"global_step": 6187, "acc_step": 0, "speed/wps": 12879.132315388015, "speed/FLOPS": 202284470795565.16, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055527474731206894, "optim/lr": 0.002996559366005943, "optim/total_tokens": 3243769856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.193558692932129, "created_at": "2025-01-15T23:01:33.359089+00:00"} {"global_step": 6188, "acc_step": 0, "speed/wps": 12887.597550769306, "speed/FLOPS": 202417428949680.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06055012345314026, "optim/lr": 0.0029965535637027315, "optim/total_tokens": 3244294144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.0086817741394043, "created_at": "2025-01-15T23:01:43.538811+00:00"} {"global_step": 6189, "acc_step": 0, "speed/wps": 12884.46791916151, "speed/FLOPS": 202368273784715.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07408657670021057, "optim/lr": 0.00299654775651675, "optim/total_tokens": 3244818432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365596, "loss/out": 3.0329880714416504, "created_at": "2025-01-15T23:01:53.716266+00:00"} {"global_step": 6190, "acc_step": 0, "speed/wps": 12882.862975212054, "speed/FLOPS": 202343065934565.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06056203320622444, "optim/lr": 0.0029965419444480195, "optim/total_tokens": 3245342720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 2.998760223388672, "created_at": "2025-01-15T23:02:03.896223+00:00"} {"global_step": 6191, "acc_step": 0, "speed/wps": 12881.057166818575, "speed/FLOPS": 202314703232305.16, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059687234461307526, "optim/lr": 0.002996536127496556, "optim/total_tokens": 3245867008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.1035988330841064, "created_at": "2025-01-15T23:02:14.077083+00:00"} {"global_step": 6192, "acc_step": 0, "speed/wps": 12881.747951935084, "speed/FLOPS": 202325552961798.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06335945427417755, "optim/lr": 0.0029965303056623815, "optim/total_tokens": 3246391296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.895859718322754, "created_at": "2025-01-15T23:02:24.259855+00:00"} {"global_step": 6193, "acc_step": 0, "speed/wps": 12877.006317073477, "speed/FLOPS": 202251079070607.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06793902069330215, "optim/lr": 0.0029965244789455134, "optim/total_tokens": 3246915584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 2.9296610355377197, "created_at": "2025-01-15T23:02:34.444529+00:00"} {"global_step": 6194, "acc_step": 0, "speed/wps": 12886.84290618198, "speed/FLOPS": 202405576219446.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08186429738998413, "optim/lr": 0.0029965186473459708, "optim/total_tokens": 3247439872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312376, "loss/out": 3.130387306213379, "created_at": "2025-01-15T23:02:44.620235+00:00"} {"global_step": 6195, "acc_step": 0, "speed/wps": 12890.96992655977, "speed/FLOPS": 202470396745599.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0644088089466095, "optim/lr": 0.002996512810863773, "optim/total_tokens": 3247964160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.9393198490142822, "created_at": "2025-01-15T23:02:54.793123+00:00"} {"global_step": 6196, "acc_step": 0, "speed/wps": 12892.214658013634, "speed/FLOPS": 202489946963523.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06443312019109726, "optim/lr": 0.002996506969498939, "optim/total_tokens": 3248488448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0099782943725586, "created_at": "2025-01-15T23:03:04.969462+00:00"} {"global_step": 6197, "acc_step": 0, "speed/wps": 12890.990579254696, "speed/FLOPS": 202470721124553.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07886093854904175, "optim/lr": 0.002996501123251488, "optim/total_tokens": 3249012736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459256, "loss/out": 3.0729832649230957, "created_at": "2025-01-15T23:03:15.140857+00:00"} {"global_step": 6198, "acc_step": 0, "speed/wps": 12892.067083518885, "speed/FLOPS": 202487629103294.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06896587461233139, "optim/lr": 0.0029964952721214387, "optim/total_tokens": 3249537024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9347167015075684, "created_at": "2025-01-15T23:03:25.313579+00:00"} {"global_step": 6199, "acc_step": 0, "speed/wps": 12890.832997713644, "speed/FLOPS": 202468246089911.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06094122678041458, "optim/lr": 0.002996489416108811, "optim/total_tokens": 3250061312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.0296263694763184, "created_at": "2025-01-15T23:03:35.494382+00:00"} {"global_step": 6200, "acc_step": 0, "speed/wps": 12892.386427831867, "speed/FLOPS": 202492644844554.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07084112614393234, "optim/lr": 0.0029964835552136223, "optim/total_tokens": 3250585600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 2.981945514678955, "created_at": "2025-01-15T23:03:45.666087+00:00"} {"global_step": 6201, "acc_step": 0, "speed/wps": 12889.371120488426, "speed/FLOPS": 202445285299259.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07601160556077957, "optim/lr": 0.0029964776894358934, "optim/total_tokens": 3251109888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 3.0767812728881836, "created_at": "2025-01-15T23:03:55.841182+00:00"} {"global_step": 6202, "acc_step": 0, "speed/wps": 12876.283419538007, "speed/FLOPS": 202239724971447.66, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06254124641418457, "optim/lr": 0.002996471818775643, "optim/total_tokens": 3251634176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.110821485519409, "created_at": "2025-01-15T23:04:06.028127+00:00"} {"global_step": 6203, "acc_step": 0, "speed/wps": 12891.405015642848, "speed/FLOPS": 202477230417524.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.061158206313848495, "optim/lr": 0.002996465943232889, "optim/total_tokens": 3252158464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.0440945625305176, "created_at": "2025-01-15T23:04:16.199385+00:00"} {"global_step": 6204, "acc_step": 0, "speed/wps": 12892.674481224134, "speed/FLOPS": 202497169118907.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06423542648553848, "optim/lr": 0.002996460062807653, "optim/total_tokens": 3252682752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.013828992843628, "created_at": "2025-01-15T23:04:26.371184+00:00"} {"global_step": 6205, "acc_step": 0, "speed/wps": 12892.667824707843, "speed/FLOPS": 202497064569172.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06450849771499634, "optim/lr": 0.0029964541774999526, "optim/total_tokens": 3253207040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 3.0972237586975098, "created_at": "2025-01-15T23:04:36.543855+00:00"} {"global_step": 6206, "acc_step": 0, "speed/wps": 12892.063741401957, "speed/FLOPS": 202487576610755.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06290686130523682, "optim/lr": 0.0029964482873098067, "optim/total_tokens": 3253731328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.1598401069641113, "created_at": "2025-01-15T23:04:46.716561+00:00"} {"global_step": 6207, "acc_step": 0, "speed/wps": 12892.71359016777, "speed/FLOPS": 202497783378608.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06349830329418182, "optim/lr": 0.0029964423922372355, "optim/total_tokens": 3254255616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.063901424407959, "created_at": "2025-01-15T23:04:56.890018+00:00"} {"global_step": 6208, "acc_step": 0, "speed/wps": 12882.266206133902, "speed/FLOPS": 202333692856146.5, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06889620423316956, "optim/lr": 0.002996436492282258, "optim/total_tokens": 3254779904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 3.1040725708007812, "created_at": "2025-01-15T23:05:07.069736+00:00"} {"global_step": 6209, "acc_step": 0, "speed/wps": 12887.21654386334, "speed/FLOPS": 202411444712664.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07545419782400131, "optim/lr": 0.0029964305874448927, "optim/total_tokens": 3255304192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 2.943276882171631, "created_at": "2025-01-15T23:05:17.246322+00:00"} {"global_step": 6210, "acc_step": 0, "speed/wps": 12881.510741453885, "speed/FLOPS": 202321827245229.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05493579059839249, "optim/lr": 0.00299642467772516, "optim/total_tokens": 3255828480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.09108829498291, "created_at": "2025-01-15T23:05:27.426193+00:00"} {"global_step": 6211, "acc_step": 0, "speed/wps": 12881.547536638105, "speed/FLOPS": 202322405164159.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06855014711618423, "optim/lr": 0.002996418763123078, "optim/total_tokens": 3256352768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.064453601837158, "created_at": "2025-01-15T23:05:37.607528+00:00"} {"global_step": 6212, "acc_step": 0, "speed/wps": 12886.982582342795, "speed/FLOPS": 202407770025487.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06849819421768188, "optim/lr": 0.002996412843638667, "optim/total_tokens": 3256877056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9749112129211426, "created_at": "2025-01-15T23:05:47.782426+00:00"} {"global_step": 6213, "acc_step": 0, "speed/wps": 12891.588043572126, "speed/FLOPS": 202480105122660.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062407758086919785, "optim/lr": 0.0029964069192719464, "optim/total_tokens": 3257401344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.0231564044952393, "created_at": "2025-01-15T23:05:57.954488+00:00"} {"global_step": 6214, "acc_step": 0, "speed/wps": 12886.538067764974, "speed/FLOPS": 202400788313215.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07688068598508835, "optim/lr": 0.0029964009900229345, "optim/total_tokens": 3257925632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0763397216796875, "created_at": "2025-01-15T23:06:08.129742+00:00"} {"global_step": 6215, "acc_step": 0, "speed/wps": 12884.630406776922, "speed/FLOPS": 202370825875997.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06579428166151047, "optim/lr": 0.002996395055891651, "optim/total_tokens": 3258449920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0702247619628906, "created_at": "2025-01-15T23:06:18.307745+00:00"} {"global_step": 6216, "acc_step": 0, "speed/wps": 12877.852623864457, "speed/FLOPS": 202264371481707.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05902353674173355, "optim/lr": 0.0029963891168781156, "optim/total_tokens": 3258974208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.0302093029022217, "created_at": "2025-01-15T23:06:28.490604+00:00"} {"global_step": 6217, "acc_step": 0, "speed/wps": 12884.081018144207, "speed/FLOPS": 202362196972581.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07115383446216583, "optim/lr": 0.0029963831729823477, "optim/total_tokens": 3259498496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 2.9919142723083496, "created_at": "2025-01-15T23:06:38.668489+00:00"} {"global_step": 6218, "acc_step": 0, "speed/wps": 12888.211574675295, "speed/FLOPS": 202427073038881.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06961045414209366, "optim/lr": 0.002996377224204367, "optim/total_tokens": 3260022784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.079998254776001, "created_at": "2025-01-15T23:06:48.842068+00:00"} {"global_step": 6219, "acc_step": 0, "speed/wps": 12887.006979488915, "speed/FLOPS": 202408153216192.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06237403303384781, "optim/lr": 0.002996371270544192, "optim/total_tokens": 3260547072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.935662269592285, "created_at": "2025-01-15T23:06:59.021632+00:00"} {"global_step": 6220, "acc_step": 0, "speed/wps": 12892.69680587881, "speed/FLOPS": 202497519758286.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05986211448907852, "optim/lr": 0.0029963653120018423, "optim/total_tokens": 3261071360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370920, "loss/out": 2.9132351875305176, "created_at": "2025-01-15T23:07:09.191683+00:00"} {"global_step": 6221, "acc_step": 0, "speed/wps": 12882.230394978991, "speed/FLOPS": 202333130392748.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0812634825706482, "optim/lr": 0.0029963593485773377, "optim/total_tokens": 3261595648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 3.0352790355682373, "created_at": "2025-01-15T23:07:19.380617+00:00"} {"global_step": 6222, "acc_step": 0, "speed/wps": 12882.618085788012, "speed/FLOPS": 202339219609647.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05828886106610298, "optim/lr": 0.002996353380270698, "optim/total_tokens": 3262119936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.048790693283081, "created_at": "2025-01-15T23:07:29.562480+00:00"} {"global_step": 6223, "acc_step": 0, "speed/wps": 12886.056304906526, "speed/FLOPS": 202393221565512.03, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08391033858060837, "optim/lr": 0.0029963474070819414, "optim/total_tokens": 3262644224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480757, "loss/out": 3.08168625831604, "created_at": "2025-01-15T23:07:39.743006+00:00"} {"global_step": 6224, "acc_step": 0, "speed/wps": 12887.690723625561, "speed/FLOPS": 202418892357421.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07141049206256866, "optim/lr": 0.002996341429011089, "optim/total_tokens": 3263168512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.0098960399627686, "created_at": "2025-01-15T23:07:49.917202+00:00"} {"global_step": 6225, "acc_step": 0, "speed/wps": 12889.105741741005, "speed/FLOPS": 202441117161362.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06570687144994736, "optim/lr": 0.0029963354460581593, "optim/total_tokens": 3263692800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 2.901939630508423, "created_at": "2025-01-15T23:08:00.091003+00:00"} {"global_step": 6226, "acc_step": 0, "speed/wps": 12879.675400386894, "speed/FLOPS": 202293000691749.56, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.06279958784580231, "optim/lr": 0.002996329458223172, "optim/total_tokens": 3264217088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.0938773155212402, "created_at": "2025-01-15T23:08:10.275535+00:00"} {"global_step": 6227, "acc_step": 0, "speed/wps": 12879.572615798827, "speed/FLOPS": 202291386318553.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06585917621850967, "optim/lr": 0.002996323465506146, "optim/total_tokens": 3264741376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.2334418296813965, "created_at": "2025-01-15T23:08:20.458960+00:00"} {"global_step": 6228, "acc_step": 0, "speed/wps": 12888.347406830151, "speed/FLOPS": 202429206469526.53, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06643272191286087, "optim/lr": 0.0029963174679071025, "optim/total_tokens": 3265265664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.0118398666381836, "created_at": "2025-01-15T23:08:30.636334+00:00"} {"global_step": 6229, "acc_step": 0, "speed/wps": 12880.456519264548, "speed/FLOPS": 202305269237092.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.065131276845932, "optim/lr": 0.00299631146542606, "optim/total_tokens": 3265789952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422503, "loss/out": 3.0580484867095947, "created_at": "2025-01-15T23:08:40.816542+00:00"} {"global_step": 6230, "acc_step": 0, "speed/wps": 12886.357834467923, "speed/FLOPS": 202397957501618.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07465602457523346, "optim/lr": 0.0029963054580630383, "optim/total_tokens": 3266314240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.0784354209899902, "created_at": "2025-01-15T23:08:50.994629+00:00"} {"global_step": 6231, "acc_step": 0, "speed/wps": 12882.899398393214, "speed/FLOPS": 202343638010676.34, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059403859078884125, "optim/lr": 0.0029962994458180565, "optim/total_tokens": 3266838528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.0728702545166016, "created_at": "2025-01-15T23:09:01.173421+00:00"} {"global_step": 6232, "acc_step": 0, "speed/wps": 12877.098310620908, "speed/FLOPS": 202252523955682.3, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07443142682313919, "optim/lr": 0.002996293428691135, "optim/total_tokens": 3267362816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.139110565185547, "created_at": "2025-01-15T23:09:11.367544+00:00"} {"global_step": 6233, "acc_step": 0, "speed/wps": 12883.057533709907, "speed/FLOPS": 202346121743125.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08634557574987411, "optim/lr": 0.0029962874066822927, "optim/total_tokens": 3267887104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 2.986435890197754, "created_at": "2025-01-15T23:09:21.547745+00:00"} {"global_step": 6234, "acc_step": 0, "speed/wps": 12881.967123659308, "speed/FLOPS": 202328995354900.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05761700123548508, "optim/lr": 0.0029962813797915493, "optim/total_tokens": 3268411392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.9880547523498535, "created_at": "2025-01-15T23:09:31.732791+00:00"} {"global_step": 6235, "acc_step": 0, "speed/wps": 12880.462922694938, "speed/FLOPS": 202305369811766.44, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07148576527833939, "optim/lr": 0.002996275348018925, "optim/total_tokens": 3268935680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.1416513919830322, "created_at": "2025-01-15T23:09:41.916267+00:00"} {"global_step": 6236, "acc_step": 0, "speed/wps": 12890.572803374127, "speed/FLOPS": 202464159380265.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0697258859872818, "optim/lr": 0.0029962693113644396, "optim/total_tokens": 3269459968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.9693353176116943, "created_at": "2025-01-15T23:09:52.090572+00:00"} {"global_step": 6237, "acc_step": 0, "speed/wps": 12889.29864851713, "speed/FLOPS": 202444147027363.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07231413573026657, "optim/lr": 0.002996263269828112, "optim/total_tokens": 3269984256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.006960868835449, "created_at": "2025-01-15T23:10:02.263260+00:00"} {"global_step": 6238, "acc_step": 0, "speed/wps": 12890.412743220731, "speed/FLOPS": 202461645415610.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0762261301279068, "optim/lr": 0.002996257223409963, "optim/total_tokens": 3270508544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431014, "loss/out": 3.0332489013671875, "created_at": "2025-01-15T23:10:12.435231+00:00"} {"global_step": 6239, "acc_step": 0, "speed/wps": 12881.228270429809, "speed/FLOPS": 202317390649640.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058049362152814865, "optim/lr": 0.0029962511721100104, "optim/total_tokens": 3271032832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9299840927124023, "created_at": "2025-01-15T23:10:22.621024+00:00"} {"global_step": 6240, "acc_step": 0, "speed/wps": 12887.978263908004, "speed/FLOPS": 202423408572678.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05411284789443016, "optim/lr": 0.002996245115928276, "optim/total_tokens": 3271557120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0140132904052734, "created_at": "2025-01-15T23:10:32.794686+00:00"} {"global_step": 6241, "acc_step": 0, "speed/wps": 12889.729677088708, "speed/FLOPS": 202450916923375.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05911632254719734, "optim/lr": 0.0029962390548647785, "optim/total_tokens": 3272081408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 3.034898519515991, "created_at": "2025-01-15T23:10:42.967841+00:00"} {"global_step": 6242, "acc_step": 0, "speed/wps": 12888.717902330294, "speed/FLOPS": 202435025610470.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06708239763975143, "optim/lr": 0.0029962329889195384, "optim/total_tokens": 3272605696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.154604911804199, "created_at": "2025-01-15T23:10:53.144498+00:00"} {"global_step": 6243, "acc_step": 0, "speed/wps": 12887.757808916265, "speed/FLOPS": 202419946024096.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05214090645313263, "optim/lr": 0.002996226918092575, "optim/total_tokens": 3273129984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.914909601211548, "created_at": "2025-01-15T23:11:03.318543+00:00"} {"global_step": 6244, "acc_step": 0, "speed/wps": 12883.286628246482, "speed/FLOPS": 202349719987631.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06354016065597534, "optim/lr": 0.0029962208423839074, "optim/total_tokens": 3273654272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.125727653503418, "created_at": "2025-01-15T23:11:13.497070+00:00"} {"global_step": 6245, "acc_step": 0, "speed/wps": 12888.516678762879, "speed/FLOPS": 202431865117834.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08777052909135818, "optim/lr": 0.002996214761793557, "optim/total_tokens": 3274178560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 3.0196123123168945, "created_at": "2025-01-15T23:11:23.677157+00:00"} {"global_step": 6246, "acc_step": 0, "speed/wps": 12890.747851985736, "speed/FLOPS": 202466908759252.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0862949937582016, "optim/lr": 0.002996208676321543, "optim/total_tokens": 3274702848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.9941606521606445, "created_at": "2025-01-15T23:11:33.857646+00:00"} {"global_step": 6247, "acc_step": 0, "speed/wps": 12885.69113225621, "speed/FLOPS": 202387486027239.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07905778288841248, "optim/lr": 0.002996202585967884, "optim/total_tokens": 3275227136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.1186699867248535, "created_at": "2025-01-15T23:11:44.034140+00:00"} {"global_step": 6248, "acc_step": 0, "speed/wps": 12887.455025459252, "speed/FLOPS": 202415190393836.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07211043685674667, "optim/lr": 0.0029961964907326013, "optim/total_tokens": 3275751424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.970458507537842, "created_at": "2025-01-15T23:11:54.213659+00:00"} {"global_step": 6249, "acc_step": 0, "speed/wps": 12886.394804838365, "speed/FLOPS": 202398538172089.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07200770080089569, "optim/lr": 0.002996190390615715, "optim/total_tokens": 3276275712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0962977409362793, "created_at": "2025-01-15T23:12:04.389470+00:00"} {"global_step": 6250, "acc_step": 0, "speed/wps": 12885.519162094739, "speed/FLOPS": 202384784999541.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08100161701440811, "optim/lr": 0.0029961842856172437, "optim/total_tokens": 3276800000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405668, "loss/out": 3.0251643657684326, "created_at": "2025-01-15T23:12:14.567946+00:00"} {"global_step": 6251, "acc_step": 0, "speed/wps": 12887.579797923036, "speed/FLOPS": 202417150116835.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11066003143787384, "optim/lr": 0.0029961781757372086, "optim/total_tokens": 3277324288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 3.0785810947418213, "created_at": "2025-01-15T23:12:24.743282+00:00"} {"global_step": 6252, "acc_step": 0, "speed/wps": 12886.019756505879, "speed/FLOPS": 202392647522657.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07555235922336578, "optim/lr": 0.0029961720609756287, "optim/total_tokens": 3277848576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.1852164268493652, "created_at": "2025-01-15T23:12:34.926837+00:00"} {"global_step": 6253, "acc_step": 0, "speed/wps": 12881.634704381726, "speed/FLOPS": 202323774253354.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10111948847770691, "optim/lr": 0.0029961659413325244, "optim/total_tokens": 3278372864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 3.0474343299865723, "created_at": "2025-01-15T23:12:45.105640+00:00"} {"global_step": 6254, "acc_step": 0, "speed/wps": 12883.484599716907, "speed/FLOPS": 202352829401615.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0664878860116005, "optim/lr": 0.0029961598168079156, "optim/total_tokens": 3278897152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348781, "loss/out": 3.0032384395599365, "created_at": "2025-01-15T23:12:55.283265+00:00"} {"global_step": 6255, "acc_step": 0, "speed/wps": 12887.760925748611, "speed/FLOPS": 202419994978231.5, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08286647498607635, "optim/lr": 0.002996153687401822, "optim/total_tokens": 3279421440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 3.0750203132629395, "created_at": "2025-01-15T23:13:05.458260+00:00"} {"global_step": 6256, "acc_step": 0, "speed/wps": 12891.888266163747, "speed/FLOPS": 202484820531011.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06006980314850807, "optim/lr": 0.0029961475531142642, "optim/total_tokens": 3279945728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.987774133682251, "created_at": "2025-01-15T23:13:15.635566+00:00"} {"global_step": 6257, "acc_step": 0, "speed/wps": 12893.009897759059, "speed/FLOPS": 202502437296498.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07516723871231079, "optim/lr": 0.002996141413945262, "optim/total_tokens": 3280470016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.023895263671875, "created_at": "2025-01-15T23:13:25.805304+00:00"} {"global_step": 6258, "acc_step": 0, "speed/wps": 12893.965463904211, "speed/FLOPS": 202517445775894.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06653943657875061, "optim/lr": 0.0029961352698948352, "optim/total_tokens": 3280994304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 3.0562045574188232, "created_at": "2025-01-15T23:13:35.974287+00:00"} {"global_step": 6259, "acc_step": 0, "speed/wps": 12888.738049701942, "speed/FLOPS": 202435342052627.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06356891244649887, "optim/lr": 0.002996129120963004, "optim/total_tokens": 3281518592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0210800170898438, "created_at": "2025-01-15T23:13:46.151038+00:00"} {"global_step": 6260, "acc_step": 0, "speed/wps": 12888.847822055493, "speed/FLOPS": 202437066178283.12, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05503339692950249, "optim/lr": 0.002996122967149788, "optim/total_tokens": 3282042880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.04239559173584, "created_at": "2025-01-15T23:13:56.327798+00:00"} {"global_step": 6261, "acc_step": 0, "speed/wps": 12887.425102519102, "speed/FLOPS": 202414720412942.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06319370865821838, "optim/lr": 0.0029961168084552087, "optim/total_tokens": 3282567168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0029826164245605, "created_at": "2025-01-15T23:14:06.509777+00:00"} {"global_step": 6262, "acc_step": 0, "speed/wps": 12888.69422815728, "speed/FLOPS": 202434653775051.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05584203451871872, "optim/lr": 0.0029961106448792843, "optim/total_tokens": 3283091456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 3.1088907718658447, "created_at": "2025-01-15T23:14:16.682785+00:00"} {"global_step": 6263, "acc_step": 0, "speed/wps": 12890.17820165433, "speed/FLOPS": 202457961618015.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06425987184047699, "optim/lr": 0.002996104476422036, "optim/total_tokens": 3283615744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.0748114585876465, "created_at": "2025-01-15T23:14:26.858411+00:00"} {"global_step": 6264, "acc_step": 0, "speed/wps": 12891.401264118249, "speed/FLOPS": 202477171494675.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0556703545153141, "optim/lr": 0.0029960983030834836, "optim/total_tokens": 3284140032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.0203356742858887, "created_at": "2025-01-15T23:14:37.031170+00:00"} {"global_step": 6265, "acc_step": 0, "speed/wps": 12889.280205142595, "speed/FLOPS": 202443857348822.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06172710657119751, "optim/lr": 0.0029960921248636473, "optim/total_tokens": 3284664320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385060, "loss/out": 3.1765880584716797, "created_at": "2025-01-15T23:14:47.203776+00:00"} {"global_step": 6266, "acc_step": 0, "speed/wps": 12890.006875542647, "speed/FLOPS": 202455270706005.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060866717249155045, "optim/lr": 0.0029960859417625476, "optim/total_tokens": 3285188608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9985666275024414, "created_at": "2025-01-15T23:14:57.375829+00:00"} {"global_step": 6267, "acc_step": 0, "speed/wps": 12888.424675436945, "speed/FLOPS": 202430420079174.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07394952327013016, "optim/lr": 0.0029960797537802045, "optim/total_tokens": 3285712896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.033458948135376, "created_at": "2025-01-15T23:15:07.550724+00:00"} {"global_step": 6268, "acc_step": 0, "speed/wps": 12885.594360501571, "speed/FLOPS": 202385966093854.1, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07970145344734192, "optim/lr": 0.002996073560916638, "optim/total_tokens": 3286237184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.0869011878967285, "created_at": "2025-01-15T23:15:17.734259+00:00"} {"global_step": 6269, "acc_step": 0, "speed/wps": 12893.222677671893, "speed/FLOPS": 202505779297420.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10010094195604324, "optim/lr": 0.0029960673631718683, "optim/total_tokens": 3286761472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 3.131546974182129, "created_at": "2025-01-15T23:15:27.906909+00:00"} {"global_step": 6270, "acc_step": 0, "speed/wps": 12883.749417977791, "speed/FLOPS": 202356988736301.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08969953656196594, "optim/lr": 0.0029960611605459155, "optim/total_tokens": 3287285760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.0820834636688232, "created_at": "2025-01-15T23:15:38.083937+00:00"} {"global_step": 6271, "acc_step": 0, "speed/wps": 12889.730355894737, "speed/FLOPS": 202450927584957.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08662997931241989, "optim/lr": 0.0029960549530388004, "optim/total_tokens": 3287810048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.098632335662842, "created_at": "2025-01-15T23:15:48.260160+00:00"} {"global_step": 6272, "acc_step": 0, "speed/wps": 12890.696918344964, "speed/FLOPS": 202466108776433.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06970388442277908, "optim/lr": 0.002996048740650543, "optim/total_tokens": 3288334336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407007, "loss/out": 3.128476142883301, "created_at": "2025-01-15T23:15:58.431612+00:00"} {"global_step": 6273, "acc_step": 0, "speed/wps": 12890.153958965482, "speed/FLOPS": 202457580853274.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06299344450235367, "optim/lr": 0.002996042523381163, "optim/total_tokens": 3288858624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.024749517440796, "created_at": "2025-01-15T23:16:08.604843+00:00"} {"global_step": 6274, "acc_step": 0, "speed/wps": 12890.830465039135, "speed/FLOPS": 202468206310778.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06394966691732407, "optim/lr": 0.002996036301230681, "optim/total_tokens": 3289382912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 3.107628107070923, "created_at": "2025-01-15T23:16:18.779046+00:00"} {"global_step": 6275, "acc_step": 0, "speed/wps": 12885.383404884156, "speed/FLOPS": 202382652745999.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08658632636070251, "optim/lr": 0.0029960300741991184, "optim/total_tokens": 3289907200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9792277812957764, "created_at": "2025-01-15T23:16:28.957612+00:00"} {"global_step": 6276, "acc_step": 0, "speed/wps": 12889.673169984051, "speed/FLOPS": 202450029401648.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06558509916067123, "optim/lr": 0.002996023842286494, "optim/total_tokens": 3290431488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385871, "loss/out": 3.0100252628326416, "created_at": "2025-01-15T23:16:39.134774+00:00"} {"global_step": 6277, "acc_step": 0, "speed/wps": 12895.336800205623, "speed/FLOPS": 202538984496921.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0793602392077446, "optim/lr": 0.0029960176054928287, "optim/total_tokens": 3290955776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.0732839107513428, "created_at": "2025-01-15T23:16:49.304489+00:00"} {"global_step": 6278, "acc_step": 0, "speed/wps": 12891.918629064545, "speed/FLOPS": 202485297422091.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07499514520168304, "optim/lr": 0.002996011363818143, "optim/total_tokens": 3291480064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.148195743560791, "created_at": "2025-01-15T23:16:59.475393+00:00"} {"global_step": 6279, "acc_step": 0, "speed/wps": 12891.476885451151, "speed/FLOPS": 202478359231623.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06521774083375931, "optim/lr": 0.0029960051172624575, "optim/total_tokens": 3292004352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 3.132800817489624, "created_at": "2025-01-15T23:17:09.646281+00:00"} {"global_step": 6280, "acc_step": 0, "speed/wps": 12887.902029785806, "speed/FLOPS": 202422211211032.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09321743249893188, "optim/lr": 0.0029959988658257914, "optim/total_tokens": 3292528640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.988222360610962, "created_at": "2025-01-15T23:17:19.822967+00:00"} {"global_step": 6281, "acc_step": 0, "speed/wps": 12886.87326538195, "speed/FLOPS": 202406053052399.9, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0721481516957283, "optim/lr": 0.0029959926095081664, "optim/total_tokens": 3293052928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0264620780944824, "created_at": "2025-01-15T23:17:30.002800+00:00"} {"global_step": 6282, "acc_step": 0, "speed/wps": 12883.70870158284, "speed/FLOPS": 202356349229368.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05749469995498657, "optim/lr": 0.002995986348309602, "optim/total_tokens": 3293577216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.048705577850342, "created_at": "2025-01-15T23:17:40.181981+00:00"} {"global_step": 6283, "acc_step": 0, "speed/wps": 12889.276304937535, "speed/FLOPS": 202443796090742.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06885872036218643, "optim/lr": 0.00299598008223012, "optim/total_tokens": 3294101504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 3.0427591800689697, "created_at": "2025-01-15T23:17:50.357404+00:00"} {"global_step": 6284, "acc_step": 0, "speed/wps": 12892.026575525155, "speed/FLOPS": 202486992869589.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08304455876350403, "optim/lr": 0.0029959738112697388, "optim/total_tokens": 3294625792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.134361743927002, "created_at": "2025-01-15T23:18:00.529974+00:00"} {"global_step": 6285, "acc_step": 0, "speed/wps": 12886.323368074527, "speed/FLOPS": 202397416159548.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06020444631576538, "optim/lr": 0.0029959675354284806, "optim/total_tokens": 3295150080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.127331256866455, "created_at": "2025-01-15T23:18:10.715352+00:00"} {"global_step": 6286, "acc_step": 0, "speed/wps": 12889.204405363427, "speed/FLOPS": 202442666809131.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08832484483718872, "optim/lr": 0.0029959612547063646, "optim/total_tokens": 3295674368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443538, "loss/out": 3.023189067840576, "created_at": "2025-01-15T23:18:20.889525+00:00"} {"global_step": 6287, "acc_step": 0, "speed/wps": 12890.419621733736, "speed/FLOPS": 202461753452109.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06610589474439621, "optim/lr": 0.0029959549691034124, "optim/total_tokens": 3296198656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 2.746190071105957, "created_at": "2025-01-15T23:18:31.066596+00:00"} {"global_step": 6288, "acc_step": 0, "speed/wps": 12887.604185529033, "speed/FLOPS": 202417533157699.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059781379997730255, "optim/lr": 0.0029959486786196436, "optim/total_tokens": 3296722944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 3.094630479812622, "created_at": "2025-01-15T23:18:41.240931+00:00"} {"global_step": 6289, "acc_step": 0, "speed/wps": 12886.259940056509, "speed/FLOPS": 202396419935366.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08367779850959778, "optim/lr": 0.00299594238325508, "optim/total_tokens": 3297247232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9625468254089355, "created_at": "2025-01-15T23:18:51.425692+00:00"} {"global_step": 6290, "acc_step": 0, "speed/wps": 12887.159284670754, "speed/FLOPS": 202410545378362.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08078815788030624, "optim/lr": 0.002995936083009741, "optim/total_tokens": 3297771520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.000532865524292, "created_at": "2025-01-15T23:19:01.602318+00:00"} {"global_step": 6291, "acc_step": 0, "speed/wps": 12886.726246198192, "speed/FLOPS": 202403743914093.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06512576341629028, "optim/lr": 0.002995929777883647, "optim/total_tokens": 3298295808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438471, "loss/out": 3.2033324241638184, "created_at": "2025-01-15T23:19:11.779217+00:00"} {"global_step": 6292, "acc_step": 0, "speed/wps": 12889.222436547727, "speed/FLOPS": 202442950013657.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08407890051603317, "optim/lr": 0.0029959234678768186, "optim/total_tokens": 3298820096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0474352836608887, "created_at": "2025-01-15T23:19:21.952075+00:00"} {"global_step": 6293, "acc_step": 0, "speed/wps": 12888.2261838771, "speed/FLOPS": 202427302496471.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06273456662893295, "optim/lr": 0.0029959171529892778, "optim/total_tokens": 3299344384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.1160595417022705, "created_at": "2025-01-15T23:19:32.125625+00:00"} {"global_step": 6294, "acc_step": 0, "speed/wps": 12881.80072381267, "speed/FLOPS": 202326381816651.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06540790945291519, "optim/lr": 0.002995910833221044, "optim/total_tokens": 3299868672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.0190582275390625, "created_at": "2025-01-15T23:19:42.309014+00:00"} {"global_step": 6295, "acc_step": 0, "speed/wps": 12888.388258775614, "speed/FLOPS": 202429848105466.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07342468202114105, "optim/lr": 0.0029959045085721377, "optim/total_tokens": 3300392960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 3.150808572769165, "created_at": "2025-01-15T23:19:52.486346+00:00"} {"global_step": 6296, "acc_step": 0, "speed/wps": 12893.569822860407, "speed/FLOPS": 202511231689633.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05624474957585335, "optim/lr": 0.00299589817904258, "optim/total_tokens": 3300917248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.063986301422119, "created_at": "2025-01-15T23:20:02.655739+00:00"} {"global_step": 6297, "acc_step": 0, "speed/wps": 12886.684989437055, "speed/FLOPS": 202403095919967.06, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06010308489203453, "optim/lr": 0.0029958918446323915, "optim/total_tokens": 3301441536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9860639572143555, "created_at": "2025-01-15T23:20:12.834270+00:00"} {"global_step": 6298, "acc_step": 0, "speed/wps": 12887.476188505978, "speed/FLOPS": 202415522788565.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06339728832244873, "optim/lr": 0.002995885505341592, "optim/total_tokens": 3301965824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 3.047333240509033, "created_at": "2025-01-15T23:20:23.009027+00:00"} {"global_step": 6299, "acc_step": 0, "speed/wps": 12891.018354405789, "speed/FLOPS": 202471157371468.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07751882821321487, "optim/lr": 0.0029958791611702043, "optim/total_tokens": 3302490112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9580702781677246, "created_at": "2025-01-15T23:20:33.181228+00:00"} {"global_step": 6300, "acc_step": 0, "speed/wps": 12889.942395966962, "speed/FLOPS": 202454257965664.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.074492909014225, "optim/lr": 0.002995872812118247, "optim/total_tokens": 3303014400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.137727737426758, "created_at": "2025-01-15T23:20:43.353568+00:00"} {"global_step": 6301, "acc_step": 0, "speed/wps": 12885.413633334954, "speed/FLOPS": 202383127525357.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0662742629647255, "optim/lr": 0.0029958664581857413, "optim/total_tokens": 3303538688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454731, "loss/out": 3.0662424564361572, "created_at": "2025-01-15T23:20:53.529233+00:00"} {"global_step": 6302, "acc_step": 0, "speed/wps": 12892.868594450727, "speed/FLOPS": 202500217933870.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07156431674957275, "optim/lr": 0.0029958600993727088, "optim/total_tokens": 3304062976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.0098044872283936, "created_at": "2025-01-15T23:21:03.700667+00:00"} {"global_step": 6303, "acc_step": 0, "speed/wps": 12888.71266513351, "speed/FLOPS": 202434943353098.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06208239123225212, "optim/lr": 0.002995853735679169, "optim/total_tokens": 3304587264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.1843016147613525, "created_at": "2025-01-15T23:21:13.877296+00:00"} {"global_step": 6304, "acc_step": 0, "speed/wps": 12888.639020233988, "speed/FLOPS": 202433786658752.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05421365797519684, "optim/lr": 0.002995847367105144, "optim/total_tokens": 3305111552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.140834331512451, "created_at": "2025-01-15T23:21:24.059713+00:00"} {"global_step": 6305, "acc_step": 0, "speed/wps": 12888.473559822916, "speed/FLOPS": 202431187875630.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.10756371170282364, "optim/lr": 0.002995840993650654, "optim/total_tokens": 3305635840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 2.974430561065674, "created_at": "2025-01-15T23:21:34.234097+00:00"} {"global_step": 6306, "acc_step": 0, "speed/wps": 12889.454878671058, "speed/FLOPS": 202446600836613.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0881769061088562, "optim/lr": 0.002995834615315719, "optim/total_tokens": 3306160128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.069855213165283, "created_at": "2025-01-15T23:21:44.409038+00:00"} {"global_step": 6307, "acc_step": 0, "speed/wps": 12887.090809613019, "speed/FLOPS": 202409469883485.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07547684013843536, "optim/lr": 0.002995828232100361, "optim/total_tokens": 3306684416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.1197781562805176, "created_at": "2025-01-15T23:21:54.585858+00:00"} {"global_step": 6308, "acc_step": 0, "speed/wps": 12887.683799638151, "speed/FLOPS": 202418783606684.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06258638203144073, "optim/lr": 0.0029958218440046, "optim/total_tokens": 3307208704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 3.0861763954162598, "created_at": "2025-01-15T23:22:04.760871+00:00"} {"global_step": 6309, "acc_step": 0, "speed/wps": 12886.792658291513, "speed/FLOPS": 202404787007278.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058857791125774384, "optim/lr": 0.0029958154510284574, "optim/total_tokens": 3307732992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 3.120107650756836, "created_at": "2025-01-15T23:22:14.939039+00:00"} {"global_step": 6310, "acc_step": 0, "speed/wps": 12887.044842154564, "speed/FLOPS": 202408747901385.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06406623870134354, "optim/lr": 0.002995809053171954, "optim/total_tokens": 3308257280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.083696126937866, "created_at": "2025-01-15T23:22:25.115789+00:00"} {"global_step": 6311, "acc_step": 0, "speed/wps": 12879.71321880495, "speed/FLOPS": 202293594681973.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07565031945705414, "optim/lr": 0.00299580265043511, "optim/total_tokens": 3308781568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.975645065307617, "created_at": "2025-01-15T23:22:35.301229+00:00"} {"global_step": 6312, "acc_step": 0, "speed/wps": 12887.684688296764, "speed/FLOPS": 202418797564289.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06468303501605988, "optim/lr": 0.002995796242817947, "optim/total_tokens": 3309305856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.053354263305664, "created_at": "2025-01-15T23:22:45.476436+00:00"} {"global_step": 6313, "acc_step": 0, "speed/wps": 12888.594920806176, "speed/FLOPS": 202433094016639.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06262742727994919, "optim/lr": 0.0029957898303204855, "optim/total_tokens": 3309830144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 3.0832111835479736, "created_at": "2025-01-15T23:22:55.650664+00:00"} {"global_step": 6314, "acc_step": 0, "speed/wps": 12889.096759947659, "speed/FLOPS": 202440976089956.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05982507765293121, "optim/lr": 0.002995783412942747, "optim/total_tokens": 3310354432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.165459156036377, "created_at": "2025-01-15T23:23:05.823492+00:00"} {"global_step": 6315, "acc_step": 0, "speed/wps": 12882.212744173814, "speed/FLOPS": 202332853162598.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05524052679538727, "optim/lr": 0.002995776990684752, "optim/total_tokens": 3310878720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.0780892372131348, "created_at": "2025-01-15T23:23:16.004238+00:00"} {"global_step": 6316, "acc_step": 0, "speed/wps": 12892.534486780329, "speed/FLOPS": 202494970313794.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07691758126020432, "optim/lr": 0.002995770563546522, "optim/total_tokens": 3311403008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.987407922744751, "created_at": "2025-01-15T23:23:26.176640+00:00"} {"global_step": 6317, "acc_step": 0, "speed/wps": 12883.169425285554, "speed/FLOPS": 202347879154076.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06593179702758789, "optim/lr": 0.0029957641315280763, "optim/total_tokens": 3311927296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 3.0738234519958496, "created_at": "2025-01-15T23:23:36.354969+00:00"} {"global_step": 6318, "acc_step": 0, "speed/wps": 12890.325174464899, "speed/FLOPS": 202460270027967.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06828456372022629, "optim/lr": 0.0029957576946294377, "optim/total_tokens": 3312451584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.203036308288574, "created_at": "2025-01-15T23:23:46.526978+00:00"} {"global_step": 6319, "acc_step": 0, "speed/wps": 12884.693953229853, "speed/FLOPS": 202371823960362.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09014267474412918, "optim/lr": 0.0029957512528506263, "optim/total_tokens": 3312975872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9843173027038574, "created_at": "2025-01-15T23:23:56.708636+00:00"} {"global_step": 6320, "acc_step": 0, "speed/wps": 12890.141700704768, "speed/FLOPS": 202457388320445.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08207263052463531, "optim/lr": 0.002995744806191664, "optim/total_tokens": 3313500160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 2.9610934257507324, "created_at": "2025-01-15T23:24:06.882499+00:00"} {"global_step": 6321, "acc_step": 0, "speed/wps": 12885.512347798638, "speed/FLOPS": 202384677971657.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060674332082271576, "optim/lr": 0.0029957383546525703, "optim/total_tokens": 3314024448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 3.1091041564941406, "created_at": "2025-01-15T23:24:17.059009+00:00"} {"global_step": 6322, "acc_step": 0, "speed/wps": 12889.019933015332, "speed/FLOPS": 202439769417412.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06112821400165558, "optim/lr": 0.002995731898233368, "optim/total_tokens": 3314548736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.02418851852417, "created_at": "2025-01-15T23:24:27.234545+00:00"} {"global_step": 6323, "acc_step": 0, "speed/wps": 12889.77096987489, "speed/FLOPS": 202451565483324.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05662094056606293, "optim/lr": 0.002995725436934077, "optim/total_tokens": 3315073024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.211458444595337, "created_at": "2025-01-15T23:24:37.409479+00:00"} {"global_step": 6324, "acc_step": 0, "speed/wps": 12888.857255584495, "speed/FLOPS": 202437214344819.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08525662124156952, "optim/lr": 0.002995718970754719, "optim/total_tokens": 3315597312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.0823516845703125, "created_at": "2025-01-15T23:24:47.585405+00:00"} {"global_step": 6325, "acc_step": 0, "speed/wps": 12884.739509926283, "speed/FLOPS": 202372539490881.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.070799320936203, "optim/lr": 0.0029957124996953144, "optim/total_tokens": 3316121600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.1295297145843506, "created_at": "2025-01-15T23:24:57.764256+00:00"} {"global_step": 6326, "acc_step": 0, "speed/wps": 12888.773462675745, "speed/FLOPS": 202435898262043.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08635543286800385, "optim/lr": 0.0029957060237558847, "optim/total_tokens": 3316645888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.9979119300842285, "created_at": "2025-01-15T23:25:07.938617+00:00"} {"global_step": 6327, "acc_step": 0, "speed/wps": 12889.061812639538, "speed/FLOPS": 202440427194459.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06729398667812347, "optim/lr": 0.0029956995429364507, "optim/total_tokens": 3317170176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0411949157714844, "created_at": "2025-01-15T23:25:18.112646+00:00"} {"global_step": 6328, "acc_step": 0, "speed/wps": 12887.524241967638, "speed/FLOPS": 202416277534213.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07445094734430313, "optim/lr": 0.0029956930572370352, "optim/total_tokens": 3317694464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 3.1194896697998047, "created_at": "2025-01-15T23:25:28.293603+00:00"} {"global_step": 6329, "acc_step": 0, "speed/wps": 12887.361780484656, "speed/FLOPS": 202413725853378.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07718565315008163, "optim/lr": 0.0029956865666576566, "optim/total_tokens": 3318218752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 3.020810842514038, "created_at": "2025-01-15T23:25:38.468014+00:00"} {"global_step": 6330, "acc_step": 0, "speed/wps": 12886.27059099264, "speed/FLOPS": 202396587222954.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07663867622613907, "optim/lr": 0.002995680071198338, "optim/total_tokens": 3318743040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.025716543197632, "created_at": "2025-01-15T23:25:48.643103+00:00"} {"global_step": 6331, "acc_step": 0, "speed/wps": 12888.461763894691, "speed/FLOPS": 202431002604368.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057977981865406036, "optim/lr": 0.002995673570859101, "optim/total_tokens": 3319267328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.0839319229125977, "created_at": "2025-01-15T23:25:58.816411+00:00"} {"global_step": 6332, "acc_step": 0, "speed/wps": 12886.563880350648, "speed/FLOPS": 202401193735343.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06328170001506805, "optim/lr": 0.002995667065639965, "optim/total_tokens": 3319791616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 2.9818787574768066, "created_at": "2025-01-15T23:26:08.991183+00:00"} {"global_step": 6333, "acc_step": 0, "speed/wps": 12893.13635228885, "speed/FLOPS": 202504423438656.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06640321761369705, "optim/lr": 0.002995660555540952, "optim/total_tokens": 3320315904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.096713066101074, "created_at": "2025-01-15T23:26:19.161300+00:00"} {"global_step": 6334, "acc_step": 0, "speed/wps": 12883.437795854275, "speed/FLOPS": 202352094282637.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05835821479558945, "optim/lr": 0.002995654040562084, "optim/total_tokens": 3320840192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.0553884506225586, "created_at": "2025-01-15T23:26:29.339999+00:00"} {"global_step": 6335, "acc_step": 0, "speed/wps": 12888.645685574433, "speed/FLOPS": 202433891347083.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054106563329696655, "optim/lr": 0.0029956475207033816, "optim/total_tokens": 3321364480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 2.9306206703186035, "created_at": "2025-01-15T23:26:39.523412+00:00"} {"global_step": 6336, "acc_step": 0, "speed/wps": 12884.658942502372, "speed/FLOPS": 202371274068777.66, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05098464712500572, "optim/lr": 0.002995640995964866, "optim/total_tokens": 3321888768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460071, "loss/out": 3.050013303756714, "created_at": "2025-01-15T23:26:49.699745+00:00"} {"global_step": 6337, "acc_step": 0, "speed/wps": 12888.68821077402, "speed/FLOPS": 202434559263778.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055021703243255615, "optim/lr": 0.0029956344663465586, "optim/total_tokens": 3322413056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 3.1197996139526367, "created_at": "2025-01-15T23:26:59.878024+00:00"} {"global_step": 6338, "acc_step": 0, "speed/wps": 12888.352352200965, "speed/FLOPS": 202429284143370.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06389198452234268, "optim/lr": 0.00299562793184848, "optim/total_tokens": 3322937344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1059560775756836, "created_at": "2025-01-15T23:27:10.056772+00:00"} {"global_step": 6339, "acc_step": 0, "speed/wps": 12884.850026950498, "speed/FLOPS": 202374275312613.0, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0604332871735096, "optim/lr": 0.0029956213924706536, "optim/total_tokens": 3323461632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.0061068534851074, "created_at": "2025-01-15T23:27:20.232873+00:00"} {"global_step": 6340, "acc_step": 0, "speed/wps": 12887.794506437778, "speed/FLOPS": 202420522409099.56, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09540204703807831, "optim/lr": 0.002995614848213099, "optim/total_tokens": 3323985920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335820, "loss/out": 3.056147575378418, "created_at": "2025-01-15T23:27:30.406911+00:00"} {"global_step": 6341, "acc_step": 0, "speed/wps": 12887.016115314635, "speed/FLOPS": 202408296706889.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06286854296922684, "optim/lr": 0.0029956082990758373, "optim/total_tokens": 3324510208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0604279041290283, "created_at": "2025-01-15T23:27:40.582944+00:00"} {"global_step": 6342, "acc_step": 0, "speed/wps": 12884.74440888281, "speed/FLOPS": 202372616435725.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07336167246103287, "optim/lr": 0.002995601745058891, "optim/total_tokens": 3325034496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0326175689697266, "created_at": "2025-01-15T23:27:50.764814+00:00"} {"global_step": 6343, "acc_step": 0, "speed/wps": 12889.49976900975, "speed/FLOPS": 202447305901068.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09543963521718979, "optim/lr": 0.0029955951861622807, "optim/total_tokens": 3325558784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 3.1433606147766113, "created_at": "2025-01-15T23:28:00.939254+00:00"} {"global_step": 6344, "acc_step": 0, "speed/wps": 12890.98490923261, "speed/FLOPS": 202470632069065.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0850735530257225, "optim/lr": 0.002995588622386028, "optim/total_tokens": 3326083072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392005, "loss/out": 3.0757627487182617, "created_at": "2025-01-15T23:28:11.113273+00:00"} {"global_step": 6345, "acc_step": 0, "speed/wps": 12889.499364101677, "speed/FLOPS": 202447299541431.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07135313749313354, "optim/lr": 0.002995582053730155, "optim/total_tokens": 3326607360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.0571017265319824, "created_at": "2025-01-15T23:28:21.286188+00:00"} {"global_step": 6346, "acc_step": 0, "speed/wps": 12883.365371719168, "speed/FLOPS": 202350956762074.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08059155195951462, "optim/lr": 0.0029955754801946816, "optim/total_tokens": 3327131648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.0550484657287598, "created_at": "2025-01-15T23:28:31.465891+00:00"} {"global_step": 6347, "acc_step": 0, "speed/wps": 12886.959814903159, "speed/FLOPS": 202407412431562.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04910682141780853, "optim/lr": 0.0029955689017796315, "optim/total_tokens": 3327655936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473347, "loss/out": 3.020381212234497, "created_at": "2025-01-15T23:28:41.643688+00:00"} {"global_step": 6348, "acc_step": 0, "speed/wps": 12884.41772753612, "speed/FLOPS": 202367485456268.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06509966403245926, "optim/lr": 0.0029955623184850237, "optim/total_tokens": 3328180224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 3.10030460357666, "created_at": "2025-01-15T23:28:51.820345+00:00"} {"global_step": 6349, "acc_step": 0, "speed/wps": 12888.988938090028, "speed/FLOPS": 202439282599519.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05545985326170921, "optim/lr": 0.0029955557303108812, "optim/total_tokens": 3328704512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.130072593688965, "created_at": "2025-01-15T23:29:01.998173+00:00"} {"global_step": 6350, "acc_step": 0, "speed/wps": 12886.627068288215, "speed/FLOPS": 202402186188732.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05567227676510811, "optim/lr": 0.0029955491372572252, "optim/total_tokens": 3329228800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 3.0966339111328125, "created_at": "2025-01-15T23:29:12.172906+00:00"} {"global_step": 6351, "acc_step": 0, "speed/wps": 12885.640913403548, "speed/FLOPS": 202386697271150.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06658992916345596, "optim/lr": 0.0029955425393240766, "optim/total_tokens": 3329753088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.1118319034576416, "created_at": "2025-01-15T23:29:22.348409+00:00"} {"global_step": 6352, "acc_step": 0, "speed/wps": 12884.737569454524, "speed/FLOPS": 202372509013105.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07339229434728622, "optim/lr": 0.0029955359365114584, "optim/total_tokens": 3330277376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.0514817237854004, "created_at": "2025-01-15T23:29:32.524606+00:00"} {"global_step": 6353, "acc_step": 0, "speed/wps": 12886.247134650566, "speed/FLOPS": 202396218808869.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0815105140209198, "optim/lr": 0.00299552932881939, "optim/total_tokens": 3330801664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.086176872253418, "created_at": "2025-01-15T23:29:42.701836+00:00"} {"global_step": 6354, "acc_step": 0, "speed/wps": 12887.613059029287, "speed/FLOPS": 202417672528214.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08389939367771149, "optim/lr": 0.002995522716247895, "optim/total_tokens": 3331325952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.103080987930298, "created_at": "2025-01-15T23:29:52.881965+00:00"} {"global_step": 6355, "acc_step": 0, "speed/wps": 12889.68181849263, "speed/FLOPS": 202450165238359.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08549313247203827, "optim/lr": 0.0029955160987969937, "optim/total_tokens": 3331850240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.1001744270324707, "created_at": "2025-01-15T23:30:03.055348+00:00"} {"global_step": 6356, "acc_step": 0, "speed/wps": 12890.036644010133, "speed/FLOPS": 202455738260693.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.061932485550642014, "optim/lr": 0.0029955094764667085, "optim/total_tokens": 3332374528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286083, "loss/out": 3.0051684379577637, "created_at": "2025-01-15T23:30:13.230620+00:00"} {"global_step": 6357, "acc_step": 0, "speed/wps": 12888.983933210895, "speed/FLOPS": 202439203991015.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0840090960264206, "optim/lr": 0.0029955028492570607, "optim/total_tokens": 3332898816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.0910863876342773, "created_at": "2025-01-15T23:30:23.406814+00:00"} {"global_step": 6358, "acc_step": 0, "speed/wps": 12887.302087840606, "speed/FLOPS": 202412788298376.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06828786432743073, "optim/lr": 0.0029954962171680716, "optim/total_tokens": 3333423104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.040243148803711, "created_at": "2025-01-15T23:30:33.583170+00:00"} {"global_step": 6359, "acc_step": 0, "speed/wps": 12885.577124014817, "speed/FLOPS": 202385695371144.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05873776227235794, "optim/lr": 0.002995489580199763, "optim/total_tokens": 3333947392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388029, "loss/out": 3.0805282592773438, "created_at": "2025-01-15T23:30:43.763759+00:00"} {"global_step": 6360, "acc_step": 0, "speed/wps": 12888.59125235926, "speed/FLOPS": 202433036398640.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06186441704630852, "optim/lr": 0.0029954829383521562, "optim/total_tokens": 3334471680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 3.044041395187378, "created_at": "2025-01-15T23:30:53.937174+00:00"} {"global_step": 6361, "acc_step": 0, "speed/wps": 12886.214019821262, "speed/FLOPS": 202395698694970.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06146452575922012, "optim/lr": 0.002995476291625274, "optim/total_tokens": 3334995968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.042600393295288, "created_at": "2025-01-15T23:31:04.115195+00:00"} {"global_step": 6362, "acc_step": 0, "speed/wps": 12888.471795271544, "speed/FLOPS": 202431160160926.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07321064919233322, "optim/lr": 0.0029954696400191373, "optim/total_tokens": 3335520256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.0891201496124268, "created_at": "2025-01-15T23:31:14.291912+00:00"} {"global_step": 6363, "acc_step": 0, "speed/wps": 12885.182517822119, "speed/FLOPS": 202379497538642.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0660729631781578, "optim/lr": 0.002995462983533768, "optim/total_tokens": 3336044544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 3.0409133434295654, "created_at": "2025-01-15T23:31:24.470311+00:00"} {"global_step": 6364, "acc_step": 0, "speed/wps": 12888.258191340077, "speed/FLOPS": 202427805217660.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07678474485874176, "optim/lr": 0.0029954563221691872, "optim/total_tokens": 3336568832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 2.9990482330322266, "created_at": "2025-01-15T23:31:34.644705+00:00"} {"global_step": 6365, "acc_step": 0, "speed/wps": 12891.347083938961, "speed/FLOPS": 202476320520508.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0856134369969368, "optim/lr": 0.0029954496559254176, "optim/total_tokens": 3337093120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9922707080841064, "created_at": "2025-01-15T23:31:44.817589+00:00"} {"global_step": 6366, "acc_step": 0, "speed/wps": 12890.614408542966, "speed/FLOPS": 202464812846613.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07776085287332535, "optim/lr": 0.0029954429848024805, "optim/total_tokens": 3337617408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.9486782550811768, "created_at": "2025-01-15T23:31:54.989917+00:00"} {"global_step": 6367, "acc_step": 0, "speed/wps": 12888.651666327607, "speed/FLOPS": 202433985283030.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06287914514541626, "optim/lr": 0.0029954363088003973, "optim/total_tokens": 3338141696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 3.005683660507202, "created_at": "2025-01-15T23:32:05.167197+00:00"} {"global_step": 6368, "acc_step": 0, "speed/wps": 12890.072851577148, "speed/FLOPS": 202456306950287.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09406300634145737, "optim/lr": 0.0029954296279191906, "optim/total_tokens": 3338665984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0479910373687744, "created_at": "2025-01-15T23:32:15.340196+00:00"} {"global_step": 6369, "acc_step": 0, "speed/wps": 12890.356992824922, "speed/FLOPS": 202460769779034.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07453219592571259, "optim/lr": 0.002995422942158881, "optim/total_tokens": 3339190272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.020714282989502, "created_at": "2025-01-15T23:32:25.513158+00:00"} {"global_step": 6370, "acc_step": 0, "speed/wps": 12888.539331311844, "speed/FLOPS": 202432220907243.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07382509112358093, "optim/lr": 0.0029954162515194917, "optim/total_tokens": 3339714560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 3.1036739349365234, "created_at": "2025-01-15T23:32:35.687616+00:00"} {"global_step": 6371, "acc_step": 0, "speed/wps": 12890.53550795358, "speed/FLOPS": 202463573604436.03, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0865664929151535, "optim/lr": 0.002995409556001044, "optim/total_tokens": 3340238848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.012718439102173, "created_at": "2025-01-15T23:32:45.859359+00:00"} {"global_step": 6372, "acc_step": 0, "speed/wps": 12887.518363751675, "speed/FLOPS": 202416185208754.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06762399524450302, "optim/lr": 0.0029954028556035592, "optim/total_tokens": 3340763136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.104501485824585, "created_at": "2025-01-15T23:32:56.038005+00:00"} {"global_step": 6373, "acc_step": 0, "speed/wps": 12890.722296561487, "speed/FLOPS": 202466507376197.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09590628743171692, "optim/lr": 0.00299539615032706, "optim/total_tokens": 3341287424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0666189193725586, "created_at": "2025-01-15T23:33:06.210143+00:00"} {"global_step": 6374, "acc_step": 0, "speed/wps": 12886.634494985352, "speed/FLOPS": 202402302835216.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09133584052324295, "optim/lr": 0.0029953894401715674, "optim/total_tokens": 3341811712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 3.1032443046569824, "created_at": "2025-01-15T23:33:16.386988+00:00"} {"global_step": 6375, "acc_step": 0, "speed/wps": 12892.073505584474, "speed/FLOPS": 202487729970659.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0750894770026207, "optim/lr": 0.002995382725137104, "optim/total_tokens": 3342336000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.0941944122314453, "created_at": "2025-01-15T23:33:26.558537+00:00"} {"global_step": 6376, "acc_step": 0, "speed/wps": 12889.61049756643, "speed/FLOPS": 202449045045208.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08959721028804779, "optim/lr": 0.002995376005223692, "optim/total_tokens": 3342860288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.970182418823242, "created_at": "2025-01-15T23:33:36.733646+00:00"} {"global_step": 6377, "acc_step": 0, "speed/wps": 12888.929281128057, "speed/FLOPS": 202438345604954.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06973114609718323, "optim/lr": 0.002995369280431351, "optim/total_tokens": 3343384576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.9357244968414307, "created_at": "2025-01-15T23:33:46.910300+00:00"} {"global_step": 6378, "acc_step": 0, "speed/wps": 12887.303230165437, "speed/FLOPS": 202412806240158.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07479538023471832, "optim/lr": 0.002995362550760107, "optim/total_tokens": 3343908864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.0152809619903564, "created_at": "2025-01-15T23:33:57.087549+00:00"} {"global_step": 6379, "acc_step": 0, "speed/wps": 12884.409389236476, "speed/FLOPS": 202367354491814.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07828827947378159, "optim/lr": 0.0029953558162099784, "optim/total_tokens": 3344433152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.1009161472320557, "created_at": "2025-01-15T23:34:07.265328+00:00"} {"global_step": 6380, "acc_step": 0, "speed/wps": 12887.983565419652, "speed/FLOPS": 202423491840204.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06864194571971893, "optim/lr": 0.0029953490767809887, "optim/total_tokens": 3344957440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.1389002799987793, "created_at": "2025-01-15T23:34:17.440844+00:00"} {"global_step": 6381, "acc_step": 0, "speed/wps": 12885.376575395088, "speed/FLOPS": 202382545479488.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07025110721588135, "optim/lr": 0.00299534233247316, "optim/total_tokens": 3345481728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.9841907024383545, "created_at": "2025-01-15T23:34:27.619512+00:00"} {"global_step": 6382, "acc_step": 0, "speed/wps": 12889.247492599929, "speed/FLOPS": 202443343553388.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09502577781677246, "optim/lr": 0.0029953355832865137, "optim/total_tokens": 3346006016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 3.0167412757873535, "created_at": "2025-01-15T23:34:37.796917+00:00"} {"global_step": 6383, "acc_step": 0, "speed/wps": 12889.284864379757, "speed/FLOPS": 202443930528544.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09491783380508423, "optim/lr": 0.002995328829221072, "optim/total_tokens": 3346530304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.9927361011505127, "created_at": "2025-01-15T23:34:47.969722+00:00"} {"global_step": 6384, "acc_step": 0, "speed/wps": 12889.550466534729, "speed/FLOPS": 202448102175364.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08710868656635284, "optim/lr": 0.0029953220702768567, "optim/total_tokens": 3347054592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 3.036365032196045, "created_at": "2025-01-15T23:34:58.146163+00:00"} {"global_step": 6385, "acc_step": 0, "speed/wps": 12888.69833220946, "speed/FLOPS": 202434718234830.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0913587212562561, "optim/lr": 0.0029953153064538906, "optim/total_tokens": 3347578880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351344, "loss/out": 3.0942041873931885, "created_at": "2025-01-15T23:35:08.319480+00:00"} {"global_step": 6386, "acc_step": 0, "speed/wps": 12888.566617871664, "speed/FLOPS": 202432649480161.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06502122431993484, "optim/lr": 0.002995308537752195, "optim/total_tokens": 3348103168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.069390296936035, "created_at": "2025-01-15T23:35:18.494430+00:00"} {"global_step": 6387, "acc_step": 0, "speed/wps": 12888.59468002002, "speed/FLOPS": 202433090234762.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10303878039121628, "optim/lr": 0.0029953017641717924, "optim/total_tokens": 3348627456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 3.0884852409362793, "created_at": "2025-01-15T23:35:28.669982+00:00"} {"global_step": 6388, "acc_step": 0, "speed/wps": 12884.974049438375, "speed/FLOPS": 202376223256209.88, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06851154565811157, "optim/lr": 0.0029952949857127054, "optim/total_tokens": 3349151744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.1368346214294434, "created_at": "2025-01-15T23:35:38.851687+00:00"} {"global_step": 6389, "acc_step": 0, "speed/wps": 12891.83562062673, "speed/FLOPS": 202483993660509.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09122882783412933, "optim/lr": 0.002995288202374955, "optim/total_tokens": 3349676032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 3.0497989654541016, "created_at": "2025-01-15T23:35:49.022424+00:00"} {"global_step": 6390, "acc_step": 0, "speed/wps": 12888.968264659237, "speed/FLOPS": 202438957894880.4, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06010870262980461, "optim/lr": 0.002995281414158564, "optim/total_tokens": 3350200320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.976270914077759, "created_at": "2025-01-15T23:35:59.198433+00:00"} {"global_step": 6391, "acc_step": 0, "speed/wps": 12885.05160083329, "speed/FLOPS": 202377441307433.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08289286494255066, "optim/lr": 0.0029952746210635543, "optim/total_tokens": 3350724608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.979118824005127, "created_at": "2025-01-15T23:36:09.381126+00:00"} {"global_step": 6392, "acc_step": 0, "speed/wps": 12893.64889880551, "speed/FLOPS": 202512473686012.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06564696133136749, "optim/lr": 0.0029952678230899484, "optim/total_tokens": 3351248896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.048140287399292, "created_at": "2025-01-15T23:36:19.552733+00:00"} {"global_step": 6393, "acc_step": 0, "speed/wps": 12889.657447813768, "speed/FLOPS": 202449782463359.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07055135071277618, "optim/lr": 0.0029952610202377684, "optim/total_tokens": 3351773184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 3.066256523132324, "created_at": "2025-01-15T23:36:29.725471+00:00"} {"global_step": 6394, "acc_step": 0, "speed/wps": 12886.226088909687, "speed/FLOPS": 202395888256589.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05862346291542053, "optim/lr": 0.002995254212507036, "optim/total_tokens": 3352297472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.067120313644409, "created_at": "2025-01-15T23:36:39.901059+00:00"} {"global_step": 6395, "acc_step": 0, "speed/wps": 12887.308042565037, "speed/FLOPS": 202412881825506.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05914563685655594, "optim/lr": 0.002995247399897774, "optim/total_tokens": 3352821760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.054464817047119, "created_at": "2025-01-15T23:36:50.079488+00:00"} {"global_step": 6396, "acc_step": 0, "speed/wps": 12892.532506166517, "speed/FLOPS": 202494939205533.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05724499374628067, "optim/lr": 0.002995240582410004, "optim/total_tokens": 3353346048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.027935266494751, "created_at": "2025-01-15T23:37:00.258590+00:00"} {"global_step": 6397, "acc_step": 0, "speed/wps": 12888.435607321186, "speed/FLOPS": 202430591779438.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059719618409872055, "optim/lr": 0.002995233760043749, "optim/total_tokens": 3353870336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289212, "loss/out": 3.0274288654327393, "created_at": "2025-01-15T23:37:10.431898+00:00"} {"global_step": 6398, "acc_step": 0, "speed/wps": 12887.538103768575, "speed/FLOPS": 202416495252845.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06011670082807541, "optim/lr": 0.002995226932799031, "optim/total_tokens": 3354394624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.922830581665039, "created_at": "2025-01-15T23:37:20.609129+00:00"} {"global_step": 6399, "acc_step": 0, "speed/wps": 12884.796348903637, "speed/FLOPS": 202373432225125.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0582316517829895, "optim/lr": 0.002995220100675872, "optim/total_tokens": 3354918912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.005434036254883, "created_at": "2025-01-15T23:37:30.786018+00:00"} {"global_step": 6400, "acc_step": 0, "speed/wps": 12891.566505672954, "speed/FLOPS": 202479766840358.12, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06197719648480415, "optim/lr": 0.0029952132636742942, "optim/total_tokens": 3355443200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 3.063253402709961, "created_at": "2025-01-15T23:37:40.960255+00:00"} {"global_step": 6401, "acc_step": 0, "speed/wps": 12888.349622214173, "speed/FLOPS": 202429241265176.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0628570094704628, "optim/lr": 0.00299520642179432, "optim/total_tokens": 3355967488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.050612449645996, "created_at": "2025-01-15T23:37:51.138431+00:00"} {"global_step": 6402, "acc_step": 0, "speed/wps": 12883.244848648963, "speed/FLOPS": 202349063781641.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060364238917827606, "optim/lr": 0.002995199575035973, "optim/total_tokens": 3356491776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.011078357696533, "created_at": "2025-01-15T23:38:01.322768+00:00"} {"global_step": 6403, "acc_step": 0, "speed/wps": 12888.292548877045, "speed/FLOPS": 202428344849990.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05632530525326729, "optim/lr": 0.0029951927233992735, "optim/total_tokens": 3357016064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.0759501457214355, "created_at": "2025-01-15T23:38:11.504427+00:00"} {"global_step": 6404, "acc_step": 0, "speed/wps": 12888.437880570013, "speed/FLOPS": 202430627483935.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055684760212898254, "optim/lr": 0.0029951858668842454, "optim/total_tokens": 3357540352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 2.941565752029419, "created_at": "2025-01-15T23:38:21.683053+00:00"} {"global_step": 6405, "acc_step": 0, "speed/wps": 12892.689759584891, "speed/FLOPS": 202497409086557.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06751306354999542, "optim/lr": 0.0029951790054909095, "optim/total_tokens": 3358064640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.942521095275879, "created_at": "2025-01-15T23:38:31.855926+00:00"} {"global_step": 6406, "acc_step": 0, "speed/wps": 12892.702402999093, "speed/FLOPS": 202497607668751.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06377467513084412, "optim/lr": 0.00299517213921929, "optim/total_tokens": 3358588928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.014958381652832, "created_at": "2025-01-15T23:38:42.028694+00:00"} {"global_step": 6407, "acc_step": 0, "speed/wps": 12885.973622070902, "speed/FLOPS": 202391922917960.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06309977173805237, "optim/lr": 0.0029951652680694075, "optim/total_tokens": 3359113216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.012918710708618, "created_at": "2025-01-15T23:38:52.203870+00:00"} {"global_step": 6408, "acc_step": 0, "speed/wps": 12886.504831086348, "speed/FLOPS": 202400266285505.7, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06028105318546295, "optim/lr": 0.0029951583920412865, "optim/total_tokens": 3359637504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 3.082639217376709, "created_at": "2025-01-15T23:39:02.382200+00:00"} {"global_step": 6409, "acc_step": 0, "speed/wps": 12893.690198987464, "speed/FLOPS": 202513122362122.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06223383545875549, "optim/lr": 0.002995151511134948, "optim/total_tokens": 3360161792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.974818229675293, "created_at": "2025-01-15T23:39:12.552346+00:00"} {"global_step": 6410, "acc_step": 0, "speed/wps": 12890.325900559305, "speed/FLOPS": 202460281432278.2, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06994955986738205, "optim/lr": 0.002995144625350414, "optim/total_tokens": 3360686080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.056175470352173, "created_at": "2025-01-15T23:39:22.726178+00:00"} {"global_step": 6411, "acc_step": 0, "speed/wps": 12884.615999179507, "speed/FLOPS": 202370599584881.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07327018678188324, "optim/lr": 0.002995137734687709, "optim/total_tokens": 3361210368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394754, "loss/out": 3.006424903869629, "created_at": "2025-01-15T23:39:32.902810+00:00"} {"global_step": 6412, "acc_step": 0, "speed/wps": 12885.77318616864, "speed/FLOPS": 202388774796687.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06579633057117462, "optim/lr": 0.002995130839146853, "optim/total_tokens": 3361734656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.001056671142578, "created_at": "2025-01-15T23:39:43.078726+00:00"} {"global_step": 6413, "acc_step": 0, "speed/wps": 12888.422808213485, "speed/FLOPS": 202430390751864.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05129990354180336, "optim/lr": 0.0029951239387278708, "optim/total_tokens": 3362258944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.0478549003601074, "created_at": "2025-01-15T23:39:53.254639+00:00"} {"global_step": 6414, "acc_step": 0, "speed/wps": 12889.877764744215, "speed/FLOPS": 202453242843499.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07056298851966858, "optim/lr": 0.0029951170334307823, "optim/total_tokens": 3362783232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.0123355388641357, "created_at": "2025-01-15T23:40:03.429902+00:00"} {"global_step": 6415, "acc_step": 0, "speed/wps": 12887.89669765613, "speed/FLOPS": 202422127462609.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06279108673334122, "optim/lr": 0.002995110123255613, "optim/total_tokens": 3363307520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418530, "loss/out": 3.1179094314575195, "created_at": "2025-01-15T23:40:13.608685+00:00"} {"global_step": 6416, "acc_step": 0, "speed/wps": 12890.814689274077, "speed/FLOPS": 202467958530709.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06089768558740616, "optim/lr": 0.0029951032082023835, "optim/total_tokens": 3363831808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329658, "loss/out": 3.036147117614746, "created_at": "2025-01-15T23:40:23.783825+00:00"} {"global_step": 6417, "acc_step": 0, "speed/wps": 12889.147045875203, "speed/FLOPS": 202441765899548.28, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06273434311151505, "optim/lr": 0.0029950962882711167, "optim/total_tokens": 3364356096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.926454782485962, "created_at": "2025-01-15T23:40:33.958252+00:00"} {"global_step": 6418, "acc_step": 0, "speed/wps": 12885.968887357672, "speed/FLOPS": 202391848552783.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05659458041191101, "optim/lr": 0.0029950893634618355, "optim/total_tokens": 3364880384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.978170156478882, "created_at": "2025-01-15T23:40:44.135581+00:00"} {"global_step": 6419, "acc_step": 0, "speed/wps": 12886.126565916267, "speed/FLOPS": 202394325111218.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05810333043336868, "optim/lr": 0.002995082433774562, "optim/total_tokens": 3365404672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 3.1359620094299316, "created_at": "2025-01-15T23:40:54.310717+00:00"} {"global_step": 6420, "acc_step": 0, "speed/wps": 12890.29565968206, "speed/FLOPS": 202459806457745.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06728188693523407, "optim/lr": 0.0029950754992093195, "optim/total_tokens": 3365928960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.062574625015259, "created_at": "2025-01-15T23:41:04.483240+00:00"} {"global_step": 6421, "acc_step": 0, "speed/wps": 12886.91780969557, "speed/FLOPS": 202406752682055.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06061677634716034, "optim/lr": 0.0029950685597661298, "optim/total_tokens": 3366453248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.0975353717803955, "created_at": "2025-01-15T23:41:14.659605+00:00"} {"global_step": 6422, "acc_step": 0, "speed/wps": 12889.247948251266, "speed/FLOPS": 202443350710018.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08529041707515717, "optim/lr": 0.002995061615445016, "optim/total_tokens": 3366977536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.134157419204712, "created_at": "2025-01-15T23:41:24.833927+00:00"} {"global_step": 6423, "acc_step": 0, "speed/wps": 12891.203937238495, "speed/FLOPS": 202474072204873.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06937454640865326, "optim/lr": 0.002995054666246001, "optim/total_tokens": 3367501824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379498, "loss/out": 3.035083770751953, "created_at": "2025-01-15T23:41:35.006903+00:00"} {"global_step": 6424, "acc_step": 0, "speed/wps": 12888.310439418325, "speed/FLOPS": 202428625845525.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06731638312339783, "optim/lr": 0.0029950477121691074, "optim/total_tokens": 3368026112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.9831910133361816, "created_at": "2025-01-15T23:41:45.180204+00:00"} {"global_step": 6425, "acc_step": 0, "speed/wps": 12888.784358560544, "speed/FLOPS": 202436069396886.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06420617550611496, "optim/lr": 0.0029950407532143572, "optim/total_tokens": 3368550400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.0508968830108643, "created_at": "2025-01-15T23:41:55.355620+00:00"} {"global_step": 6426, "acc_step": 0, "speed/wps": 12889.38505945173, "speed/FLOPS": 202445504229832.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06878044456243515, "optim/lr": 0.0029950337893817737, "optim/total_tokens": 3369074688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 3.022301197052002, "created_at": "2025-01-15T23:42:05.531612+00:00"} {"global_step": 6427, "acc_step": 0, "speed/wps": 12883.300964468455, "speed/FLOPS": 202349945157697.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0760699212551117, "optim/lr": 0.0029950268206713803, "optim/total_tokens": 3369598976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.159115791320801, "created_at": "2025-01-15T23:42:15.708992+00:00"} {"global_step": 6428, "acc_step": 0, "speed/wps": 12887.953105391374, "speed/FLOPS": 202423013423603.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06094597280025482, "optim/lr": 0.0029950198470831985, "optim/total_tokens": 3370123264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.022958517074585, "created_at": "2025-01-15T23:42:25.884371+00:00"} {"global_step": 6429, "acc_step": 0, "speed/wps": 12889.73932086314, "speed/FLOPS": 202451068392104.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05962303280830383, "optim/lr": 0.0029950128686172515, "optim/total_tokens": 3370647552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9479517936706543, "created_at": "2025-01-15T23:42:36.056812+00:00"} {"global_step": 6430, "acc_step": 0, "speed/wps": 12889.195782284613, "speed/FLOPS": 202442531371828.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06330269575119019, "optim/lr": 0.0029950058852735617, "optim/total_tokens": 3371171840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392297, "loss/out": 3.0121777057647705, "created_at": "2025-01-15T23:42:46.229533+00:00"} {"global_step": 6431, "acc_step": 0, "speed/wps": 12883.6839477634, "speed/FLOPS": 202355960436618.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07165282964706421, "optim/lr": 0.0029949988970521526, "optim/total_tokens": 3371696128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.983189582824707, "created_at": "2025-01-15T23:42:56.409862+00:00"} {"global_step": 6432, "acc_step": 0, "speed/wps": 12887.635403923661, "speed/FLOPS": 202418023485485.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06690244376659393, "optim/lr": 0.0029949919039530468, "optim/total_tokens": 3372220416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0430614948272705, "created_at": "2025-01-15T23:43:06.586067+00:00"} {"global_step": 6433, "acc_step": 0, "speed/wps": 12886.122644022082, "speed/FLOPS": 202394263512481.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0636545717716217, "optim/lr": 0.002994984905976267, "optim/total_tokens": 3372744704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.9975249767303467, "created_at": "2025-01-15T23:43:16.763175+00:00"} {"global_step": 6434, "acc_step": 0, "speed/wps": 12890.03532293186, "speed/FLOPS": 202455717511343.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06981796771287918, "optim/lr": 0.002994977903121836, "optim/total_tokens": 3373268992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.0630784034729004, "created_at": "2025-01-15T23:43:26.936558+00:00"} {"global_step": 6435, "acc_step": 0, "speed/wps": 12892.466641950772, "speed/FLOPS": 202493904717518.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07194513082504272, "optim/lr": 0.0029949708953897767, "optim/total_tokens": 3373793280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.056161642074585, "created_at": "2025-01-15T23:43:37.107445+00:00"} {"global_step": 6436, "acc_step": 0, "speed/wps": 12884.001494541184, "speed/FLOPS": 202360947945119.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06157829985022545, "optim/lr": 0.002994963882780112, "optim/total_tokens": 3374317568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.0971462726593018, "created_at": "2025-01-15T23:43:47.284619+00:00"} {"global_step": 6437, "acc_step": 0, "speed/wps": 12887.34468088062, "speed/FLOPS": 202413457280599.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056111328303813934, "optim/lr": 0.0029949568652928643, "optim/total_tokens": 3374841856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.9692840576171875, "created_at": "2025-01-15T23:43:57.459611+00:00"} {"global_step": 6438, "acc_step": 0, "speed/wps": 12888.010532147822, "speed/FLOPS": 202423915389726.25, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06519044935703278, "optim/lr": 0.0029949498429280574, "optim/total_tokens": 3375366144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.109776496887207, "created_at": "2025-01-15T23:44:07.637332+00:00"} {"global_step": 6439, "acc_step": 0, "speed/wps": 12891.250046549381, "speed/FLOPS": 202474796414962.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051139943301677704, "optim/lr": 0.0029949428156857134, "optim/total_tokens": 3375890432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 3.020699977874756, "created_at": "2025-01-15T23:44:17.815114+00:00"} {"global_step": 6440, "acc_step": 0, "speed/wps": 12888.49449841589, "speed/FLOPS": 202431516745005.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05671446770429611, "optim/lr": 0.002994935783565855, "optim/total_tokens": 3376414720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 3.047621726989746, "created_at": "2025-01-15T23:44:27.991594+00:00"} {"global_step": 6441, "acc_step": 0, "speed/wps": 12885.192277482527, "speed/FLOPS": 202379650827520.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.060489121824502945, "optim/lr": 0.002994928746568507, "optim/total_tokens": 3376939008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.045207977294922, "created_at": "2025-01-15T23:44:38.169223+00:00"} {"global_step": 6442, "acc_step": 0, "speed/wps": 12888.000310294081, "speed/FLOPS": 202423754841466.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07573024928569794, "optim/lr": 0.0029949217046936903, "optim/total_tokens": 3377463296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.995558023452759, "created_at": "2025-01-15T23:44:48.343312+00:00"} {"global_step": 6443, "acc_step": 0, "speed/wps": 12892.522175568743, "speed/FLOPS": 202494776949299.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054683323949575424, "optim/lr": 0.0029949146579414288, "optim/total_tokens": 3377987584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.1367404460906982, "created_at": "2025-01-15T23:44:58.516033+00:00"} {"global_step": 6444, "acc_step": 0, "speed/wps": 12888.299354311292, "speed/FLOPS": 202428451738686.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06334982812404633, "optim/lr": 0.0029949076063117454, "optim/total_tokens": 3378511872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 3.0534214973449707, "created_at": "2025-01-15T23:45:08.690678+00:00"} {"global_step": 6445, "acc_step": 0, "speed/wps": 12889.565287720907, "speed/FLOPS": 202448334962459.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05192846059799194, "optim/lr": 0.0029949005498046627, "optim/total_tokens": 3379036160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.004303455352783, "created_at": "2025-01-15T23:45:18.864329+00:00"} {"global_step": 6446, "acc_step": 0, "speed/wps": 12886.954992143412, "speed/FLOPS": 202407336683493.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060759253799915314, "optim/lr": 0.0029948934884202042, "optim/total_tokens": 3379560448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.0272457599639893, "created_at": "2025-01-15T23:45:29.042610+00:00"} {"global_step": 6447, "acc_step": 0, "speed/wps": 12886.893044723127, "speed/FLOPS": 202406363714132.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05513059347867966, "optim/lr": 0.002994886422158393, "optim/total_tokens": 3380084736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.0678703784942627, "created_at": "2025-01-15T23:45:39.220405+00:00"} {"global_step": 6448, "acc_step": 0, "speed/wps": 12889.989119599742, "speed/FLOPS": 202454991824522.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05359647050499916, "optim/lr": 0.002994879351019252, "optim/total_tokens": 3380609024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 3.101550340652466, "created_at": "2025-01-15T23:45:49.398251+00:00"} {"global_step": 6449, "acc_step": 0, "speed/wps": 12888.99127173429, "speed/FLOPS": 202439319252609.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0664990022778511, "optim/lr": 0.002994872275002804, "optim/total_tokens": 3381133312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9856748580932617, "created_at": "2025-01-15T23:45:59.573562+00:00"} {"global_step": 6450, "acc_step": 0, "speed/wps": 12886.220020735802, "speed/FLOPS": 202395792947579.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054617591202259064, "optim/lr": 0.0029948651941090723, "optim/total_tokens": 3381657600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.0735726356506348, "created_at": "2025-01-15T23:46:09.754598+00:00"} {"global_step": 6451, "acc_step": 0, "speed/wps": 12883.964659388072, "speed/FLOPS": 202360369398422.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07986176013946533, "optim/lr": 0.00299485810833808, "optim/total_tokens": 3382181888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.04353666305542, "created_at": "2025-01-15T23:46:19.932451+00:00"} {"global_step": 6452, "acc_step": 0, "speed/wps": 12888.539406851949, "speed/FLOPS": 202432222093704.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06924688816070557, "optim/lr": 0.00299485101768985, "optim/total_tokens": 3382706176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 3.081587076187134, "created_at": "2025-01-15T23:46:30.109266+00:00"} {"global_step": 6453, "acc_step": 0, "speed/wps": 12886.358515278685, "speed/FLOPS": 202397968194687.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0565638467669487, "optim/lr": 0.002994843922164406, "optim/total_tokens": 3383230464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.8683252334594727, "created_at": "2025-01-15T23:46:40.287879+00:00"} {"global_step": 6454, "acc_step": 0, "speed/wps": 12887.329915544755, "speed/FLOPS": 202413225370709.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06936739385128021, "optim/lr": 0.002994836821761771, "optim/total_tokens": 3383754752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 3.0397791862487793, "created_at": "2025-01-15T23:46:50.462138+00:00"} {"global_step": 6455, "acc_step": 0, "speed/wps": 12886.86821494542, "speed/FLOPS": 202405973728354.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06466825306415558, "optim/lr": 0.0029948297164819676, "optim/total_tokens": 3384279040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.045034170150757, "created_at": "2025-01-15T23:47:00.638614+00:00"} {"global_step": 6456, "acc_step": 0, "speed/wps": 12887.58198117943, "speed/FLOPS": 202417184407876.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06676886230707169, "optim/lr": 0.0029948226063250196, "optim/total_tokens": 3384803328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371737, "loss/out": 3.053762435913086, "created_at": "2025-01-15T23:47:10.820770+00:00"} {"global_step": 6457, "acc_step": 0, "speed/wps": 12887.560283184153, "speed/FLOPS": 202416843611044.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0640694871544838, "optim/lr": 0.00299481549129095, "optim/total_tokens": 3385327616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.0325632095336914, "created_at": "2025-01-15T23:47:20.994712+00:00"} {"global_step": 6458, "acc_step": 0, "speed/wps": 12887.68624964546, "speed/FLOPS": 202418822087415.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05685357376933098, "optim/lr": 0.0029948083713797814, "optim/total_tokens": 3385851904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.0379889011383057, "created_at": "2025-01-15T23:47:31.171691+00:00"} {"global_step": 6459, "acc_step": 0, "speed/wps": 12884.610821936267, "speed/FLOPS": 202370518269162.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0929885134100914, "optim/lr": 0.0029948012465915385, "optim/total_tokens": 3386376192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471956, "loss/out": 2.9764397144317627, "created_at": "2025-01-15T23:47:41.350037+00:00"} {"global_step": 6460, "acc_step": 0, "speed/wps": 12887.454581735825, "speed/FLOPS": 202415183424549.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11120975762605667, "optim/lr": 0.002994794116926243, "optim/total_tokens": 3386900480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286322, "loss/out": 2.9828991889953613, "created_at": "2025-01-15T23:47:51.530151+00:00"} {"global_step": 6461, "acc_step": 0, "speed/wps": 12894.531924877158, "speed/FLOPS": 202526342823873.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05733528733253479, "optim/lr": 0.0029947869823839187, "optim/total_tokens": 3387424768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0251715183258057, "created_at": "2025-01-15T23:48:01.708272+00:00"} {"global_step": 6462, "acc_step": 0, "speed/wps": 12885.030369070186, "speed/FLOPS": 202377107833418.75, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09907374531030655, "optim/lr": 0.002994779842964589, "optim/total_tokens": 3387949056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.028028964996338, "created_at": "2025-01-15T23:48:11.884904+00:00"} {"global_step": 6463, "acc_step": 0, "speed/wps": 12886.889819749444, "speed/FLOPS": 202406313061489.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07966217398643494, "optim/lr": 0.002994772698668278, "optim/total_tokens": 3388473344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.0997567176818848, "created_at": "2025-01-15T23:48:22.060351+00:00"} {"global_step": 6464, "acc_step": 0, "speed/wps": 12887.058882290768, "speed/FLOPS": 202408968421017.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06931594759225845, "optim/lr": 0.0029947655494950076, "optim/total_tokens": 3388997632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.804392099380493, "created_at": "2025-01-15T23:48:32.239887+00:00"} {"global_step": 6465, "acc_step": 0, "speed/wps": 12891.765367020054, "speed/FLOPS": 202482890231078.44, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10446734726428986, "optim/lr": 0.0029947583954448016, "optim/total_tokens": 3389521920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.154435157775879, "created_at": "2025-01-15T23:48:42.411659+00:00"} {"global_step": 6466, "acc_step": 0, "speed/wps": 12886.56697654633, "speed/FLOPS": 202401242365351.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0940803736448288, "optim/lr": 0.0029947512365176834, "optim/total_tokens": 3390046208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.0436580181121826, "created_at": "2025-01-15T23:48:52.587690+00:00"} {"global_step": 6467, "acc_step": 0, "speed/wps": 12885.7982134747, "speed/FLOPS": 202389167884920.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06430070847272873, "optim/lr": 0.002994744072713677, "optim/total_tokens": 3390570496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.135327100753784, "created_at": "2025-01-15T23:49:02.763895+00:00"} {"global_step": 6468, "acc_step": 0, "speed/wps": 12888.189439047675, "speed/FLOPS": 202426725368432.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07796341925859451, "optim/lr": 0.0029947369040328044, "optim/total_tokens": 3391094784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307266, "loss/out": 3.0410008430480957, "created_at": "2025-01-15T23:49:12.944505+00:00"} {"global_step": 6469, "acc_step": 0, "speed/wps": 12890.232079214002, "speed/FLOPS": 202458807839126.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056402336806058884, "optim/lr": 0.0029947297304750904, "optim/total_tokens": 3391619072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.07883358001709, "created_at": "2025-01-15T23:49:23.120574+00:00"} {"global_step": 6470, "acc_step": 0, "speed/wps": 12886.745881129815, "speed/FLOPS": 202404052307675.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.002, "optim/grad_norm": 0.08058034628629684, "optim/lr": 0.0029947225520405574, "optim/total_tokens": 3392143360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.992766857147217, "created_at": "2025-01-15T23:49:33.295084+00:00"} {"global_step": 6471, "acc_step": 0, "speed/wps": 12887.693701166665, "speed/FLOPS": 202418939123795.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058383144438266754, "optim/lr": 0.0029947153687292292, "optim/total_tokens": 3392667648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413664, "loss/out": 3.091031312942505, "created_at": "2025-01-15T23:49:43.469016+00:00"} {"global_step": 6472, "acc_step": 0, "speed/wps": 12888.919972614309, "speed/FLOPS": 202438199401954.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10592354834079742, "optim/lr": 0.0029947081805411294, "optim/total_tokens": 3393191936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.0228755474090576, "created_at": "2025-01-15T23:49:53.641941+00:00"} {"global_step": 6473, "acc_step": 0, "speed/wps": 12885.332117238042, "speed/FLOPS": 202381847203040.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.06772714853286743, "optim/lr": 0.002994700987476281, "optim/total_tokens": 3393716224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0075502395629883, "created_at": "2025-01-15T23:50:03.818534+00:00"} {"global_step": 6474, "acc_step": 0, "speed/wps": 12883.658013439303, "speed/FLOPS": 202355553102421.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09722080081701279, "optim/lr": 0.0029946937895347085, "optim/total_tokens": 3394240512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.0681309700012207, "created_at": "2025-01-15T23:50:13.998227+00:00"} {"global_step": 6475, "acc_step": 0, "speed/wps": 12891.474852023941, "speed/FLOPS": 202478327293855.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08480823040008545, "optim/lr": 0.0029946865867164344, "optim/total_tokens": 3394764800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.906386375427246, "created_at": "2025-01-15T23:50:24.170311+00:00"} {"global_step": 6476, "acc_step": 0, "speed/wps": 12888.641952180307, "speed/FLOPS": 202433832708998.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08627011626958847, "optim/lr": 0.0029946793790214823, "optim/total_tokens": 3395289088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 2.999624013900757, "created_at": "2025-01-15T23:50:34.344180+00:00"} {"global_step": 6477, "acc_step": 0, "speed/wps": 12884.535355447855, "speed/FLOPS": 202369332964260.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09373855590820312, "optim/lr": 0.0029946721664498755, "optim/total_tokens": 3395813376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.128246307373047, "created_at": "2025-01-15T23:50:44.521756+00:00"} {"global_step": 6478, "acc_step": 0, "speed/wps": 12887.251930775501, "speed/FLOPS": 202412000512746.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06868430972099304, "optim/lr": 0.0029946649490016383, "optim/total_tokens": 3396337664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.036871910095215, "created_at": "2025-01-15T23:50:54.698968+00:00"} {"global_step": 6479, "acc_step": 0, "speed/wps": 12886.888524096365, "speed/FLOPS": 202406292711477.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07185252010822296, "optim/lr": 0.002994657726676794, "optim/total_tokens": 3396861952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 3.1064388751983643, "created_at": "2025-01-15T23:51:04.878268+00:00"} {"global_step": 6480, "acc_step": 0, "speed/wps": 12889.540509030541, "speed/FLOPS": 202447945779077.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06887419521808624, "optim/lr": 0.0029946504994753658, "optim/total_tokens": 3397386240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.981962203979492, "created_at": "2025-01-15T23:51:15.053504+00:00"} {"global_step": 6481, "acc_step": 0, "speed/wps": 12887.952471620374, "speed/FLOPS": 202423003469359.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05811670795083046, "optim/lr": 0.002994643267397377, "optim/total_tokens": 3397910528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9740307331085205, "created_at": "2025-01-15T23:51:25.232938+00:00"} {"global_step": 6482, "acc_step": 0, "speed/wps": 12882.641445194671, "speed/FLOPS": 202339586501228.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05971609801054001, "optim/lr": 0.0029946360304428524, "optim/total_tokens": 3398434816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 2.9563515186309814, "created_at": "2025-01-15T23:51:35.413021+00:00"} {"global_step": 6483, "acc_step": 0, "speed/wps": 12888.346561750459, "speed/FLOPS": 202429193196388.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059458281844854355, "optim/lr": 0.0029946287886118144, "optim/total_tokens": 3398959104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.925881862640381, "created_at": "2025-01-15T23:51:45.591040+00:00"} {"global_step": 6484, "acc_step": 0, "speed/wps": 12888.981303290118, "speed/FLOPS": 202439162684495.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0601736381649971, "optim/lr": 0.0029946215419042873, "optim/total_tokens": 3399483392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.0048255920410156, "created_at": "2025-01-15T23:51:55.766980+00:00"} {"global_step": 6485, "acc_step": 0, "speed/wps": 12891.036136799778, "speed/FLOPS": 202471436668401.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06486860662698746, "optim/lr": 0.002994614290320295, "optim/total_tokens": 3400007680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 3.042656421661377, "created_at": "2025-01-15T23:52:05.939050+00:00"} {"global_step": 6486, "acc_step": 0, "speed/wps": 12889.406202867316, "speed/FLOPS": 202445836316228.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061332471668720245, "optim/lr": 0.00299460703385986, "optim/total_tokens": 3400531968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.033506155014038, "created_at": "2025-01-15T23:52:16.112539+00:00"} {"global_step": 6487, "acc_step": 0, "speed/wps": 12887.469472449884, "speed/FLOPS": 202415417303676.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061923664063215256, "optim/lr": 0.0029945997725230075, "optim/total_tokens": 3401056256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 3.030766487121582, "created_at": "2025-01-15T23:52:26.287177+00:00"} {"global_step": 6488, "acc_step": 0, "speed/wps": 12887.47384006877, "speed/FLOPS": 202415485903132.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07262937724590302, "optim/lr": 0.0029945925063097597, "optim/total_tokens": 3401580544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.941246271133423, "created_at": "2025-01-15T23:52:36.463827+00:00"} {"global_step": 6489, "acc_step": 0, "speed/wps": 12886.471345612796, "speed/FLOPS": 202399740350129.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06155454367399216, "optim/lr": 0.0029945852352201417, "optim/total_tokens": 3402104832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.052292823791504, "created_at": "2025-01-15T23:52:46.641387+00:00"} {"global_step": 6490, "acc_step": 0, "speed/wps": 12888.03316875498, "speed/FLOPS": 202424270928747.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05781581252813339, "optim/lr": 0.0029945779592541757, "optim/total_tokens": 3402629120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.042682647705078, "created_at": "2025-01-15T23:52:56.818234+00:00"} {"global_step": 6491, "acc_step": 0, "speed/wps": 12889.94335931307, "speed/FLOPS": 202454273096339.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06618734449148178, "optim/lr": 0.0029945706784118863, "optim/total_tokens": 3403153408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 3.058518648147583, "created_at": "2025-01-15T23:53:06.993308+00:00"} {"global_step": 6492, "acc_step": 0, "speed/wps": 12888.066165470042, "speed/FLOPS": 202424789187499.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061972688883543015, "optim/lr": 0.0029945633926932976, "optim/total_tokens": 3403677696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.0592269897460938, "created_at": "2025-01-15T23:53:17.167336+00:00"} {"global_step": 6493, "acc_step": 0, "speed/wps": 12883.21019959761, "speed/FLOPS": 202348519570676.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06338558346033096, "optim/lr": 0.002994556102098433, "optim/total_tokens": 3404201984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475100, "loss/out": 2.98762845993042, "created_at": "2025-01-15T23:53:27.346687+00:00"} {"global_step": 6494, "acc_step": 0, "speed/wps": 12885.941989344408, "speed/FLOPS": 202391426082522.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07550635933876038, "optim/lr": 0.0029945488066273162, "optim/total_tokens": 3404726272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.065703868865967, "created_at": "2025-01-15T23:53:37.524647+00:00"} {"global_step": 6495, "acc_step": 0, "speed/wps": 12885.64708013393, "speed/FLOPS": 202386794128124.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05533114820718765, "optim/lr": 0.0029945415062799714, "optim/total_tokens": 3405250560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 2.9976248741149902, "created_at": "2025-01-15T23:53:47.703439+00:00"} {"global_step": 6496, "acc_step": 0, "speed/wps": 12890.23704374063, "speed/FLOPS": 202458885813838.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0778261050581932, "optim/lr": 0.0029945342010564216, "optim/total_tokens": 3405774848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.0101675987243652, "created_at": "2025-01-15T23:53:57.877135+00:00"} {"global_step": 6497, "acc_step": 0, "speed/wps": 12888.620341501008, "speed/FLOPS": 202433493283583.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06266740709543228, "optim/lr": 0.002994526890956691, "optim/total_tokens": 3406299136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.04486346244812, "created_at": "2025-01-15T23:54:08.050268+00:00"} {"global_step": 6498, "acc_step": 0, "speed/wps": 12888.37189176406, "speed/FLOPS": 202429591039059.4, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07511454075574875, "optim/lr": 0.0029945195759808045, "optim/total_tokens": 3406823424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.010200023651123, "created_at": "2025-01-15T23:54:18.223677+00:00"} {"global_step": 6499, "acc_step": 0, "speed/wps": 12887.588975865838, "speed/FLOPS": 202417294269038.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.086789570748806, "optim/lr": 0.0029945122561287844, "optim/total_tokens": 3407347712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.0719075202941895, "created_at": "2025-01-15T23:54:28.401401+00:00"} {"global_step": 6500, "acc_step": 0, "speed/wps": 12888.899821132763, "speed/FLOPS": 202437882895245.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058173585683107376, "optim/lr": 0.002994504931400655, "optim/total_tokens": 3407872000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0753235816955566, "created_at": "2025-01-15T23:54:38.579225+00:00"} {"global_step": 6501, "acc_step": 0, "speed/wps": 12887.489569893927, "speed/FLOPS": 202415732961651.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07739429175853729, "optim/lr": 0.002994497601796441, "optim/total_tokens": 3408396288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.0506951808929443, "created_at": "2025-01-15T23:54:48.759477+00:00"} {"global_step": 6502, "acc_step": 0, "speed/wps": 12892.607022308608, "speed/FLOPS": 202496109583940.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056996189057826996, "optim/lr": 0.0029944902673161656, "optim/total_tokens": 3408920576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 3.067412853240967, "created_at": "2025-01-15T23:54:58.934595+00:00"} {"global_step": 6503, "acc_step": 0, "speed/wps": 12890.296244096826, "speed/FLOPS": 202459815636782.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06527918577194214, "optim/lr": 0.0029944829279598523, "optim/total_tokens": 3409444864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.127479076385498, "created_at": "2025-01-15T23:55:09.107265+00:00"} {"global_step": 6504, "acc_step": 0, "speed/wps": 12886.81435252206, "speed/FLOPS": 202405127744980.6, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.062411755323410034, "optim/lr": 0.0029944755837275263, "optim/total_tokens": 3409969152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.032925844192505, "created_at": "2025-01-15T23:55:19.285395+00:00"} {"global_step": 6505, "acc_step": 0, "speed/wps": 12893.416331379685, "speed/FLOPS": 202508820895009.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05910055339336395, "optim/lr": 0.002994468234619211, "optim/total_tokens": 3410493440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 2.8865602016448975, "created_at": "2025-01-15T23:55:29.454722+00:00"} {"global_step": 6506, "acc_step": 0, "speed/wps": 12886.207773495784, "speed/FLOPS": 202395600587846.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06199663132429123, "optim/lr": 0.0029944608806349296, "optim/total_tokens": 3411017728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 2.984516143798828, "created_at": "2025-01-15T23:55:39.629772+00:00"} {"global_step": 6507, "acc_step": 0, "speed/wps": 12884.246157015661, "speed/FLOPS": 202364790705482.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06586121022701263, "optim/lr": 0.0029944535217747066, "optim/total_tokens": 3411542016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.0867552757263184, "created_at": "2025-01-15T23:55:49.813379+00:00"} {"global_step": 6508, "acc_step": 0, "speed/wps": 12886.717700773683, "speed/FLOPS": 202403609696458.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06012019142508507, "optim/lr": 0.0029944461580385666, "optim/total_tokens": 3412066304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.989932060241699, "created_at": "2025-01-15T23:55:59.990921+00:00"} {"global_step": 6509, "acc_step": 0, "speed/wps": 12888.709630464973, "speed/FLOPS": 202434895689458.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07128427177667618, "optim/lr": 0.0029944387894265336, "optim/total_tokens": 3412590592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.993908166885376, "created_at": "2025-01-15T23:56:10.165143+00:00"} {"global_step": 6510, "acc_step": 0, "speed/wps": 12890.808400704293, "speed/FLOPS": 202467859760079.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0539788156747818, "optim/lr": 0.0029944314159386304, "optim/total_tokens": 3413114880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 3.152250289916992, "created_at": "2025-01-15T23:56:20.336581+00:00"} {"global_step": 6511, "acc_step": 0, "speed/wps": 12885.014911852726, "speed/FLOPS": 202376865056578.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0916653648018837, "optim/lr": 0.0029944240375748827, "optim/total_tokens": 3413639168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.0346240997314453, "created_at": "2025-01-15T23:56:30.515374+00:00"} {"global_step": 6512, "acc_step": 0, "speed/wps": 12889.188061068045, "speed/FLOPS": 202442410099512.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07806742936372757, "optim/lr": 0.002994416654335313, "optim/total_tokens": 3414163456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.1866183280944824, "created_at": "2025-01-15T23:56:40.689141+00:00"} {"global_step": 6513, "acc_step": 0, "speed/wps": 12887.954489774052, "speed/FLOPS": 202423035167235.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0660221204161644, "optim/lr": 0.0029944092662199467, "optim/total_tokens": 3414687744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374328, "loss/out": 3.148975133895874, "created_at": "2025-01-15T23:56:50.863434+00:00"} {"global_step": 6514, "acc_step": 0, "speed/wps": 12883.486870040171, "speed/FLOPS": 202352865060162.44, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06296860426664352, "optim/lr": 0.0029944018732288072, "optim/total_tokens": 3415212032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.1271305084228516, "created_at": "2025-01-15T23:57:01.042184+00:00"} {"global_step": 6515, "acc_step": 0, "speed/wps": 12888.212457505422, "speed/FLOPS": 202427086904941.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06271197646856308, "optim/lr": 0.002994394475361919, "optim/total_tokens": 3415736320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.9489176273345947, "created_at": "2025-01-15T23:57:11.215695+00:00"} {"global_step": 6516, "acc_step": 0, "speed/wps": 12889.14178590159, "speed/FLOPS": 202441683284434.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07266507297754288, "optim/lr": 0.002994387072619306, "optim/total_tokens": 3416260608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.15303373336792, "created_at": "2025-01-15T23:57:21.390867+00:00"} {"global_step": 6517, "acc_step": 0, "speed/wps": 12888.090696544648, "speed/FLOPS": 202425174481735.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057665519416332245, "optim/lr": 0.0029943796650009915, "optim/total_tokens": 3416784896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 3.0044307708740234, "created_at": "2025-01-15T23:57:31.565392+00:00"} {"global_step": 6518, "acc_step": 0, "speed/wps": 12885.7177590486, "speed/FLOPS": 202387904237602.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09081890434026718, "optim/lr": 0.0029943722525070018, "optim/total_tokens": 3417309184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.134936809539795, "created_at": "2025-01-15T23:57:41.741904+00:00"} {"global_step": 6519, "acc_step": 0, "speed/wps": 12889.781242939756, "speed/FLOPS": 202451726835925.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0893864780664444, "optim/lr": 0.0029943648351373592, "optim/total_tokens": 3417833472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.033522605895996, "created_at": "2025-01-15T23:57:51.914791+00:00"} {"global_step": 6520, "acc_step": 0, "speed/wps": 12890.496158378253, "speed/FLOPS": 202462955565283.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0689588189125061, "optim/lr": 0.002994357412892088, "optim/total_tokens": 3418357760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378395, "loss/out": 3.08660626411438, "created_at": "2025-01-15T23:58:02.091942+00:00"} {"global_step": 6521, "acc_step": 0, "speed/wps": 12888.530442372263, "speed/FLOPS": 202432081294232.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05602893978357315, "optim/lr": 0.0029943499857712137, "optim/total_tokens": 3418882048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313676, "loss/out": 3.0915188789367676, "created_at": "2025-01-15T23:58:12.266674+00:00"} {"global_step": 6522, "acc_step": 0, "speed/wps": 12889.497841269791, "speed/FLOPS": 202447275623263.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06103404983878136, "optim/lr": 0.0029943425537747592, "optim/total_tokens": 3419406336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.023475408554077, "created_at": "2025-01-15T23:58:22.440929+00:00"} {"global_step": 6523, "acc_step": 0, "speed/wps": 12884.461314760645, "speed/FLOPS": 202368170053524.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057523805648088455, "optim/lr": 0.00299433511690275, "optim/total_tokens": 3419930624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0356245040893555, "created_at": "2025-01-15T23:58:32.617364+00:00"} {"global_step": 6524, "acc_step": 0, "speed/wps": 12890.261024642243, "speed/FLOPS": 202459262466851.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07536007463932037, "optim/lr": 0.0029943276751552093, "optim/total_tokens": 3420454912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333415, "loss/out": 2.922635078430176, "created_at": "2025-01-15T23:58:42.794087+00:00"} {"global_step": 6525, "acc_step": 0, "speed/wps": 12887.655067589216, "speed/FLOPS": 202418332330373.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08750429004430771, "optim/lr": 0.0029943202285321615, "optim/total_tokens": 3420979200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.095733165740967, "created_at": "2025-01-15T23:58:52.968001+00:00"} {"global_step": 6526, "acc_step": 0, "speed/wps": 12895.247010582541, "speed/FLOPS": 202537574227506.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08514774590730667, "optim/lr": 0.0029943127770336317, "optim/total_tokens": 3421503488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.047762870788574, "created_at": "2025-01-15T23:59:03.136120+00:00"} {"global_step": 6527, "acc_step": 0, "speed/wps": 12885.978506627944, "speed/FLOPS": 202391999636640.88, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07130473107099533, "optim/lr": 0.0029943053206596434, "optim/total_tokens": 3422027776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.123530864715576, "created_at": "2025-01-15T23:59:13.315595+00:00"} {"global_step": 6528, "acc_step": 0, "speed/wps": 12893.405593030735, "speed/FLOPS": 202508652234482.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0521363765001297, "optim/lr": 0.0029942978594102213, "optim/total_tokens": 3422552064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 3.0948171615600586, "created_at": "2025-01-15T23:59:23.485142+00:00"} {"global_step": 6529, "acc_step": 0, "speed/wps": 12887.8189467844, "speed/FLOPS": 202420906278327.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.065409354865551, "optim/lr": 0.002994290393285389, "optim/total_tokens": 3423076352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.023524522781372, "created_at": "2025-01-15T23:59:33.664591+00:00"} {"global_step": 6530, "acc_step": 0, "speed/wps": 12890.737383742502, "speed/FLOPS": 202466744341107.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07111600041389465, "optim/lr": 0.0029942829222851723, "optim/total_tokens": 3423600640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 3.004335641860962, "created_at": "2025-01-15T23:59:43.837214+00:00"} {"global_step": 6531, "acc_step": 0, "speed/wps": 12885.050926058988, "speed/FLOPS": 202377430709176.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06474599242210388, "optim/lr": 0.002994275446409594, "optim/total_tokens": 3424124928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 3.0537798404693604, "created_at": "2025-01-15T23:59:54.014555+00:00"} {"global_step": 6532, "acc_step": 0, "speed/wps": 12889.598063263122, "speed/FLOPS": 202448849747388.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05876307561993599, "optim/lr": 0.0029942679656586803, "optim/total_tokens": 3424649216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9783613681793213, "created_at": "2025-01-16T00:00:04.186925+00:00"} {"global_step": 6533, "acc_step": 0, "speed/wps": 12886.272393878, "speed/FLOPS": 202396615539746.4, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07084064930677414, "optim/lr": 0.002994260480032454, "optim/total_tokens": 3425173504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9690537452697754, "created_at": "2025-01-16T00:00:14.370625+00:00"} {"global_step": 6534, "acc_step": 0, "speed/wps": 12889.681042888807, "speed/FLOPS": 202450153056435.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06310772895812988, "optim/lr": 0.00299425298953094, "optim/total_tokens": 3425697792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.974015235900879, "created_at": "2025-01-16T00:00:24.548021+00:00"} {"global_step": 6535, "acc_step": 0, "speed/wps": 12890.317415295325, "speed/FLOPS": 202460148159547.4, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057275183498859406, "optim/lr": 0.0029942454941541633, "optim/total_tokens": 3426222080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372985, "loss/out": 3.055563449859619, "created_at": "2025-01-16T00:00:34.725229+00:00"} {"global_step": 6536, "acc_step": 0, "speed/wps": 12891.279605850634, "speed/FLOPS": 202475260684406.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06475840508937836, "optim/lr": 0.0029942379939021475, "optim/total_tokens": 3426746368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 3.038599729537964, "created_at": "2025-01-16T00:00:44.898908+00:00"} {"global_step": 6537, "acc_step": 0, "speed/wps": 12885.540585410881, "speed/FLOPS": 202385121482160.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06647957861423492, "optim/lr": 0.0029942304887749173, "optim/total_tokens": 3427270656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.0225353240966797, "created_at": "2025-01-16T00:00:55.074561+00:00"} {"global_step": 6538, "acc_step": 0, "speed/wps": 12891.775963290003, "speed/FLOPS": 202483056660058.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05897616222500801, "optim/lr": 0.0029942229787724975, "optim/total_tokens": 3427794944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 2.957566261291504, "created_at": "2025-01-16T00:01:05.245351+00:00"} {"global_step": 6539, "acc_step": 0, "speed/wps": 12890.845922043905, "speed/FLOPS": 202468449084278.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0635102316737175, "optim/lr": 0.0029942154638949127, "optim/total_tokens": 3428319232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.027366876602173, "created_at": "2025-01-16T00:01:15.417074+00:00"} {"global_step": 6540, "acc_step": 0, "speed/wps": 12891.56651748164, "speed/FLOPS": 202479767025829.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06022133305668831, "optim/lr": 0.0029942079441421868, "optim/total_tokens": 3428843520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.1129322052001953, "created_at": "2025-01-16T00:01:25.587878+00:00"} {"global_step": 6541, "acc_step": 0, "speed/wps": 12891.487034884018, "speed/FLOPS": 202478518642413.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061853326857089996, "optim/lr": 0.002994200419514345, "optim/total_tokens": 3429367808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9222021102905273, "created_at": "2025-01-16T00:01:35.759905+00:00"} {"global_step": 6542, "acc_step": 0, "speed/wps": 12889.961392564932, "speed/FLOPS": 202454556333340.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07637949287891388, "optim/lr": 0.0029941928900114115, "optim/total_tokens": 3429892096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375182, "loss/out": 3.038778781890869, "created_at": "2025-01-16T00:01:45.937832+00:00"} {"global_step": 6543, "acc_step": 0, "speed/wps": 12888.190566186568, "speed/FLOPS": 202426743071697.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08250108361244202, "optim/lr": 0.0029941853556334106, "optim/total_tokens": 3430416384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.955197334289551, "created_at": "2025-01-16T00:01:56.112884+00:00"} {"global_step": 6544, "acc_step": 0, "speed/wps": 12889.388628010613, "speed/FLOPS": 202445560278953.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05800821632146835, "optim/lr": 0.0029941778163803676, "optim/total_tokens": 3430940672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.9955015182495117, "created_at": "2025-01-16T00:02:06.285898+00:00"} {"global_step": 6545, "acc_step": 0, "speed/wps": 12887.35553422652, "speed/FLOPS": 202413627747310.72, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05312773212790489, "optim/lr": 0.002994170272252307, "optim/total_tokens": 3431464960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418245, "loss/out": 3.0525312423706055, "created_at": "2025-01-16T00:02:16.461339+00:00"} {"global_step": 6546, "acc_step": 0, "speed/wps": 12888.382495439408, "speed/FLOPS": 202429757584351.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07379256188869476, "optim/lr": 0.0029941627232492525, "optim/total_tokens": 3431989248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 3.075197696685791, "created_at": "2025-01-16T00:02:26.635828+00:00"} {"global_step": 6547, "acc_step": 0, "speed/wps": 12891.282654718485, "speed/FLOPS": 202475308571066.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07383275032043457, "optim/lr": 0.0029941551693712294, "optim/total_tokens": 3432513536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.11959171295166, "created_at": "2025-01-16T00:02:36.812732+00:00"} {"global_step": 6548, "acc_step": 0, "speed/wps": 12885.76880790433, "speed/FLOPS": 202388706030030.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06270827353000641, "optim/lr": 0.0029941476106182625, "optim/total_tokens": 3433037824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.080078363418579, "created_at": "2025-01-16T00:02:46.988647+00:00"} {"global_step": 6549, "acc_step": 0, "speed/wps": 12893.44979625782, "speed/FLOPS": 202509346506905.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06290852278470993, "optim/lr": 0.002994140046990376, "optim/total_tokens": 3433562112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0225796699523926, "created_at": "2025-01-16T00:02:57.159020+00:00"} {"global_step": 6550, "acc_step": 0, "speed/wps": 12888.434025725903, "speed/FLOPS": 202430566938311.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06022590771317482, "optim/lr": 0.002994132478487595, "optim/total_tokens": 3434086400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.0637454986572266, "created_at": "2025-01-16T00:03:07.332366+00:00"} {"global_step": 6551, "acc_step": 0, "speed/wps": 12889.09795925066, "speed/FLOPS": 202440994926657.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053775377571582794, "optim/lr": 0.002994124905109944, "optim/total_tokens": 3434610688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.0385289192199707, "created_at": "2025-01-16T00:03:17.506897+00:00"} {"global_step": 6552, "acc_step": 0, "speed/wps": 12885.564210247358, "speed/FLOPS": 202385492542680.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07529919594526291, "optim/lr": 0.002994117326857448, "optim/total_tokens": 3435134976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475100, "loss/out": 3.093924045562744, "created_at": "2025-01-16T00:03:27.686834+00:00"} {"global_step": 6553, "acc_step": 0, "speed/wps": 12887.852305900527, "speed/FLOPS": 202421430229086.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05859414115548134, "optim/lr": 0.0029941097437301313, "optim/total_tokens": 3435659264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 2.9482648372650146, "created_at": "2025-01-16T00:03:37.860615+00:00"} {"global_step": 6554, "acc_step": 0, "speed/wps": 12890.215614250093, "speed/FLOPS": 202458549234242.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08169076591730118, "optim/lr": 0.002994102155728019, "optim/total_tokens": 3436183552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.026988983154297, "created_at": "2025-01-16T00:03:48.039160+00:00"} {"global_step": 6555, "acc_step": 0, "speed/wps": 12885.534906024906, "speed/FLOPS": 202385032279599.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0959589034318924, "optim/lr": 0.002994094562851135, "optim/total_tokens": 3436707840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.9704365730285645, "created_at": "2025-01-16T00:03:58.217108+00:00"} {"global_step": 6556, "acc_step": 0, "speed/wps": 12886.70968635105, "speed/FLOPS": 202403483818938.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05802427977323532, "optim/lr": 0.002994086965099505, "optim/total_tokens": 3437232128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.968182325363159, "created_at": "2025-01-16T00:04:08.396775+00:00"} {"global_step": 6557, "acc_step": 0, "speed/wps": 12886.137387713687, "speed/FLOPS": 202394495082417.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07736675441265106, "optim/lr": 0.0029940793624731537, "optim/total_tokens": 3437756416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293649, "loss/out": 3.095350742340088, "created_at": "2025-01-16T00:04:18.574079+00:00"} {"global_step": 6558, "acc_step": 0, "speed/wps": 12897.069726200278, "speed/FLOPS": 202566202480957.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08634187281131744, "optim/lr": 0.0029940717549721053, "optim/total_tokens": 3438280704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.000980854034424, "created_at": "2025-01-16T00:04:28.741347+00:00"} {"global_step": 6559, "acc_step": 0, "speed/wps": 12882.303894730007, "speed/FLOPS": 202334284807337.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057185154408216476, "optim/lr": 0.002994064142596386, "optim/total_tokens": 3438804992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.0348873138427734, "created_at": "2025-01-16T00:04:38.921951+00:00"} {"global_step": 6560, "acc_step": 0, "speed/wps": 12880.275026155587, "speed/FLOPS": 202302418638419.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057799331843853, "optim/lr": 0.002994056525346019, "optim/total_tokens": 3439329280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.0490245819091797, "created_at": "2025-01-16T00:04:49.103047+00:00"} {"global_step": 6561, "acc_step": 0, "speed/wps": 12887.689379424935, "speed/FLOPS": 202418871244903.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0805746540427208, "optim/lr": 0.0029940489032210297, "optim/total_tokens": 3439853568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382950, "loss/out": 3.012465238571167, "created_at": "2025-01-16T00:04:59.279392+00:00"} {"global_step": 6562, "acc_step": 0, "speed/wps": 12887.513291567224, "speed/FLOPS": 202416105543127.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06771545112133026, "optim/lr": 0.0029940412762214435, "optim/total_tokens": 3440377856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 3.0441603660583496, "created_at": "2025-01-16T00:05:09.456031+00:00"} {"global_step": 6563, "acc_step": 0, "speed/wps": 12895.420000252021, "speed/FLOPS": 202540291267978.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05794991925358772, "optim/lr": 0.0029940336443472844, "optim/total_tokens": 3440902144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.06201171875, "created_at": "2025-01-16T00:05:19.623925+00:00"} {"global_step": 6564, "acc_step": 0, "speed/wps": 12889.901431459679, "speed/FLOPS": 202453614561786.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07823571562767029, "optim/lr": 0.0029940260075985783, "optim/total_tokens": 3441426432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 3.0513672828674316, "created_at": "2025-01-16T00:05:29.798201+00:00"} {"global_step": 6565, "acc_step": 0, "speed/wps": 12889.4709001903, "speed/FLOPS": 202446852476589.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07627756148576736, "optim/lr": 0.002994018365975349, "optim/total_tokens": 3441950720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.01621675491333, "created_at": "2025-01-16T00:05:39.972800+00:00"} {"global_step": 6566, "acc_step": 0, "speed/wps": 12886.91649279656, "speed/FLOPS": 202406731998347.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07296845316886902, "optim/lr": 0.002994010719477623, "optim/total_tokens": 3442475008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.139702320098877, "created_at": "2025-01-16T00:05:50.147913+00:00"} {"global_step": 6567, "acc_step": 0, "speed/wps": 12884.676207202563, "speed/FLOPS": 202371545234618.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06507251411676407, "optim/lr": 0.0029940030681054235, "optim/total_tokens": 3442999296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.937222480773926, "created_at": "2025-01-16T00:06:00.326556+00:00"} {"global_step": 6568, "acc_step": 0, "speed/wps": 12885.603061346588, "speed/FLOPS": 202386102752581.3, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06767193973064423, "optim/lr": 0.0029939954118587764, "optim/total_tokens": 3443523584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367755, "loss/out": 2.976325035095215, "created_at": "2025-01-16T00:06:10.503928+00:00"} {"global_step": 6569, "acc_step": 0, "speed/wps": 12884.005522477362, "speed/FLOPS": 202361011209391.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06608026474714279, "optim/lr": 0.002993987750737707, "optim/total_tokens": 3444047872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.0217204093933105, "created_at": "2025-01-16T00:06:20.687566+00:00"} {"global_step": 6570, "acc_step": 0, "speed/wps": 12885.522811096635, "speed/FLOPS": 202384842312130.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06218843162059784, "optim/lr": 0.002993980084742239, "optim/total_tokens": 3444572160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.0553441047668457, "created_at": "2025-01-16T00:06:30.870604+00:00"} {"global_step": 6571, "acc_step": 0, "speed/wps": 12883.908178049573, "speed/FLOPS": 202359482281387.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06753849983215332, "optim/lr": 0.0029939724138723987, "optim/total_tokens": 3445096448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.854210376739502, "created_at": "2025-01-16T00:06:41.051289+00:00"} {"global_step": 6572, "acc_step": 0, "speed/wps": 12887.880085232599, "speed/FLOPS": 202421866541669.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07254806905984879, "optim/lr": 0.0029939647381282105, "optim/total_tokens": 3445620736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374944, "loss/out": 3.0274157524108887, "created_at": "2025-01-16T00:06:51.225146+00:00"} {"global_step": 6573, "acc_step": 0, "speed/wps": 12889.328034786293, "speed/FLOPS": 202444608579101.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06880630552768707, "optim/lr": 0.0029939570575097, "optim/total_tokens": 3446145024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.171626091003418, "created_at": "2025-01-16T00:07:01.400917+00:00"} {"global_step": 6574, "acc_step": 0, "speed/wps": 12889.680099650797, "speed/FLOPS": 202450138241586.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05990084633231163, "optim/lr": 0.0029939493720168913, "optim/total_tokens": 3446669312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9633164405822754, "created_at": "2025-01-16T00:07:11.576612+00:00"} {"global_step": 6575, "acc_step": 0, "speed/wps": 12882.81158498446, "speed/FLOPS": 202342258780423.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06808799505233765, "optim/lr": 0.0029939416816498108, "optim/total_tokens": 3447193600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.9877920150756836, "created_at": "2025-01-16T00:07:21.756208+00:00"} {"global_step": 6576, "acc_step": 0, "speed/wps": 12889.438937464212, "speed/FLOPS": 202446350458054.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0748305693268776, "optim/lr": 0.002993933986408482, "optim/total_tokens": 3447717888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397090, "loss/out": 2.964717388153076, "created_at": "2025-01-16T00:07:31.929432+00:00"} {"global_step": 6577, "acc_step": 0, "speed/wps": 12891.558306907176, "speed/FLOPS": 202479638067475.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06313889473676682, "optim/lr": 0.0029939262862929313, "optim/total_tokens": 3448242176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.0617284774780273, "created_at": "2025-01-16T00:07:42.100253+00:00"} {"global_step": 6578, "acc_step": 0, "speed/wps": 12886.473626428316, "speed/FLOPS": 202399776173471.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06469593942165375, "optim/lr": 0.002993918581303183, "optim/total_tokens": 3448766464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0409698486328125, "created_at": "2025-01-16T00:07:52.275774+00:00"} {"global_step": 6579, "acc_step": 0, "speed/wps": 12888.867259248014, "speed/FLOPS": 202437371466101.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06098059564828873, "optim/lr": 0.002993910871439263, "optim/total_tokens": 3449290752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487960, "loss/out": 2.97269606590271, "created_at": "2025-01-16T00:08:02.452260+00:00"} {"global_step": 6580, "acc_step": 0, "speed/wps": 12887.596715229076, "speed/FLOPS": 202417415826373.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07293946295976639, "optim/lr": 0.0029939031567011963, "optim/total_tokens": 3449815040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 3.000995635986328, "created_at": "2025-01-16T00:08:12.631329+00:00"} {"global_step": 6581, "acc_step": 0, "speed/wps": 12887.882014849074, "speed/FLOPS": 202421896848948.12, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0606832392513752, "optim/lr": 0.0029938954370890072, "optim/total_tokens": 3450339328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.111541509628296, "created_at": "2025-01-16T00:08:22.806202+00:00"} {"global_step": 6582, "acc_step": 0, "speed/wps": 12886.917741254576, "speed/FLOPS": 202406751607095.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06046074628829956, "optim/lr": 0.0029938877126027223, "optim/total_tokens": 3450863616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478609, "loss/out": 3.1103949546813965, "created_at": "2025-01-16T00:08:32.982787+00:00"} {"global_step": 6583, "acc_step": 0, "speed/wps": 12885.596853366204, "speed/FLOPS": 202386005247718.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06934262067079544, "optim/lr": 0.002993879983242365, "optim/total_tokens": 3451387904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.117375373840332, "created_at": "2025-01-16T00:08:43.160018+00:00"} {"global_step": 6584, "acc_step": 0, "speed/wps": 12888.51662328831, "speed/FLOPS": 202431864246529.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056901462376117706, "optim/lr": 0.0029938722490079617, "optim/total_tokens": 3451912192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.0392603874206543, "created_at": "2025-01-16T00:08:53.335777+00:00"} {"global_step": 6585, "acc_step": 0, "speed/wps": 12892.440933263626, "speed/FLOPS": 202493500927259.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057628024369478226, "optim/lr": 0.002993864509899538, "optim/total_tokens": 3452436480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.965506076812744, "created_at": "2025-01-16T00:09:03.506277+00:00"} {"global_step": 6586, "acc_step": 0, "speed/wps": 12890.577900400212, "speed/FLOPS": 202464239436063.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05614529922604561, "optim/lr": 0.002993856765917119, "optim/total_tokens": 3452960768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9524765014648438, "created_at": "2025-01-16T00:09:13.683239+00:00"} {"global_step": 6587, "acc_step": 0, "speed/wps": 12883.182500539382, "speed/FLOPS": 202348084518905.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05625336244702339, "optim/lr": 0.0029938490170607285, "optim/total_tokens": 3453485056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.0730385780334473, "created_at": "2025-01-16T00:09:23.865877+00:00"} {"global_step": 6588, "acc_step": 0, "speed/wps": 12886.887314584392, "speed/FLOPS": 202406273714429.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06295674294233322, "optim/lr": 0.002993841263330393, "optim/total_tokens": 3454009344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 278573, "loss/out": 3.0837857723236084, "created_at": "2025-01-16T00:09:34.043375+00:00"} {"global_step": 6589, "acc_step": 0, "speed/wps": 12888.762792257181, "speed/FLOPS": 202435730668457.44, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06426522880792618, "optim/lr": 0.002993833504726138, "optim/total_tokens": 3454533632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467434, "loss/out": 3.1366143226623535, "created_at": "2025-01-16T00:09:44.221906+00:00"} {"global_step": 6590, "acc_step": 0, "speed/wps": 12891.382584463117, "speed/FLOPS": 202476878105023.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06412655115127563, "optim/lr": 0.002993825741247989, "optim/total_tokens": 3455057920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.096062421798706, "created_at": "2025-01-16T00:09:54.393504+00:00"} {"global_step": 6591, "acc_step": 0, "speed/wps": 12887.9792930524, "speed/FLOPS": 202423424736805.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07979739457368851, "optim/lr": 0.00299381797289597, "optim/total_tokens": 3455582208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 3.080348253250122, "created_at": "2025-01-16T00:10:04.568432+00:00"} {"global_step": 6592, "acc_step": 0, "speed/wps": 12890.13725217674, "speed/FLOPS": 202457318450199.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06648728996515274, "optim/lr": 0.002993810199670108, "optim/total_tokens": 3456106496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.989832639694214, "created_at": "2025-01-16T00:10:14.743752+00:00"} {"global_step": 6593, "acc_step": 0, "speed/wps": 12890.013701676799, "speed/FLOPS": 202455377919822.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05962754040956497, "optim/lr": 0.002993802421570427, "optim/total_tokens": 3456630784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.0350003242492676, "created_at": "2025-01-16T00:10:24.916661+00:00"} {"global_step": 6594, "acc_step": 0, "speed/wps": 12888.073081606713, "speed/FLOPS": 202424897814930.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06635081768035889, "optim/lr": 0.0029937946385969524, "optim/total_tokens": 3457155072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.9989748001098633, "created_at": "2025-01-16T00:10:35.090332+00:00"} {"global_step": 6595, "acc_step": 0, "speed/wps": 12886.6196239002, "speed/FLOPS": 202402069264389.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059329330921173096, "optim/lr": 0.0029937868507497107, "optim/total_tokens": 3457679360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414236, "loss/out": 3.0083765983581543, "created_at": "2025-01-16T00:10:45.269952+00:00"} {"global_step": 6596, "acc_step": 0, "speed/wps": 12886.261180128038, "speed/FLOPS": 202396439412393.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07172535359859467, "optim/lr": 0.0029937790580287266, "optim/total_tokens": 3458203648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.0235230922698975, "created_at": "2025-01-16T00:10:55.449432+00:00"} {"global_step": 6597, "acc_step": 0, "speed/wps": 12883.03850674542, "speed/FLOPS": 202345822898502.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07863155752420425, "optim/lr": 0.002993771260434026, "optim/total_tokens": 3458727936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.073521614074707, "created_at": "2025-01-16T00:11:05.628202+00:00"} {"global_step": 6598, "acc_step": 0, "speed/wps": 12887.820719418292, "speed/FLOPS": 202420934119978.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05463534966111183, "optim/lr": 0.0029937634579656333, "optim/total_tokens": 3459252224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.956002950668335, "created_at": "2025-01-16T00:11:15.801949+00:00"} {"global_step": 6599, "acc_step": 0, "speed/wps": 12884.393153823064, "speed/FLOPS": 202367099492337.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06331629306077957, "optim/lr": 0.002993755650623575, "optim/total_tokens": 3459776512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 2.9460809230804443, "created_at": "2025-01-16T00:11:25.986622+00:00"} {"global_step": 6600, "acc_step": 0, "speed/wps": 12883.254067592348, "speed/FLOPS": 202349208577815.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05681494623422623, "optim/lr": 0.0029937478384078763, "optim/total_tokens": 3460300800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 3.0402417182922363, "created_at": "2025-01-16T00:11:36.164152+00:00"} {"global_step": 6601, "acc_step": 0, "speed/wps": 12887.491938398947, "speed/FLOPS": 202415770162277.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07174640148878098, "optim/lr": 0.002993740021318562, "optim/total_tokens": 3460825088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.079780101776123, "created_at": "2025-01-16T00:11:46.341695+00:00"} {"global_step": 6602, "acc_step": 0, "speed/wps": 12887.61882993839, "speed/FLOPS": 202417763168271.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09794910997152328, "optim/lr": 0.0029937321993556585, "optim/total_tokens": 3461349376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.0431551933288574, "created_at": "2025-01-16T00:11:56.517991+00:00"} {"global_step": 6603, "acc_step": 0, "speed/wps": 12889.767993734933, "speed/FLOPS": 202451518738957.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08408382534980774, "optim/lr": 0.0029937243725191914, "optim/total_tokens": 3461873664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375182, "loss/out": 2.9865546226501465, "created_at": "2025-01-16T00:12:06.692180+00:00"} {"global_step": 6604, "acc_step": 0, "speed/wps": 12890.00604323173, "speed/FLOPS": 202455257633418.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11050839722156525, "optim/lr": 0.002993716540809185, "optim/total_tokens": 3462397952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.0757672786712646, "created_at": "2025-01-16T00:12:16.866631+00:00"} {"global_step": 6605, "acc_step": 0, "speed/wps": 12889.1660873461, "speed/FLOPS": 202442064972014.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18402230739593506, "optim/lr": 0.0029937087042256664, "optim/total_tokens": 3462922240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 3.0657410621643066, "created_at": "2025-01-16T00:12:27.042414+00:00"} {"global_step": 6606, "acc_step": 0, "speed/wps": 12883.878677312196, "speed/FLOPS": 202359018931769.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09496892988681793, "optim/lr": 0.00299370086276866, "optim/total_tokens": 3463446528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9846057891845703, "created_at": "2025-01-16T00:12:37.221291+00:00"} {"global_step": 6607, "acc_step": 0, "speed/wps": 12888.151306461785, "speed/FLOPS": 202426126443771.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09233743697404861, "optim/lr": 0.0029936930164381923, "optim/total_tokens": 3463970816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431829, "loss/out": 3.060758113861084, "created_at": "2025-01-16T00:12:47.400304+00:00"} {"global_step": 6608, "acc_step": 0, "speed/wps": 12884.644787275734, "speed/FLOPS": 202371051741491.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07024163752794266, "optim/lr": 0.0029936851652342883, "optim/total_tokens": 3464495104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 3.090599536895752, "created_at": "2025-01-16T00:12:57.579567+00:00"} {"global_step": 6609, "acc_step": 0, "speed/wps": 12886.999505161155, "speed/FLOPS": 202408035821603.97, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09386538714170456, "optim/lr": 0.0029936773091569736, "optim/total_tokens": 3465019392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.0178604125976562, "created_at": "2025-01-16T00:13:07.755894+00:00"} {"global_step": 6610, "acc_step": 0, "speed/wps": 12880.015498316574, "speed/FLOPS": 202298342397078.72, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07103583216667175, "optim/lr": 0.002993669448206274, "optim/total_tokens": 3465543680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.9932985305786133, "created_at": "2025-01-16T00:13:17.936008+00:00"} {"global_step": 6611, "acc_step": 0, "speed/wps": 12883.831311526008, "speed/FLOPS": 202358274987009.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08813074231147766, "optim/lr": 0.0029936615823822153, "optim/total_tokens": 3466067968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 3.046959400177002, "created_at": "2025-01-16T00:13:28.115137+00:00"} {"global_step": 6612, "acc_step": 0, "speed/wps": 12886.446485601944, "speed/FLOPS": 202399349889498.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0846271887421608, "optim/lr": 0.0029936537116848233, "optim/total_tokens": 3466592256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 2.9899213314056396, "created_at": "2025-01-16T00:13:38.290117+00:00"} {"global_step": 6613, "acc_step": 0, "speed/wps": 12888.162643886604, "speed/FLOPS": 202426304513607.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057777371257543564, "optim/lr": 0.0029936458361141225, "optim/total_tokens": 3467116544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.188045024871826, "created_at": "2025-01-16T00:13:48.466964+00:00"} {"global_step": 6614, "acc_step": 0, "speed/wps": 12888.394792844769, "speed/FLOPS": 202429950732001.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07955913245677948, "optim/lr": 0.0029936379556701398, "optim/total_tokens": 3467640832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.082000732421875, "created_at": "2025-01-16T00:13:58.642319+00:00"} {"global_step": 6615, "acc_step": 0, "speed/wps": 12887.650277353314, "speed/FLOPS": 202418257093136.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050030503422021866, "optim/lr": 0.002993630070352901, "optim/total_tokens": 3468165120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 3.045931816101074, "created_at": "2025-01-16T00:14:08.819757+00:00"} {"global_step": 6616, "acc_step": 0, "speed/wps": 12889.152106361049, "speed/FLOPS": 202441845381432.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06390172243118286, "optim/lr": 0.0029936221801624315, "optim/total_tokens": 3468689408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 2.8433072566986084, "created_at": "2025-01-16T00:14:18.993801+00:00"} {"global_step": 6617, "acc_step": 0, "speed/wps": 12888.239318968137, "speed/FLOPS": 202427508801125.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07117360830307007, "optim/lr": 0.0029936142850987566, "optim/total_tokens": 3469213696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.081822633743286, "created_at": "2025-01-16T00:14:29.170512+00:00"} {"global_step": 6618, "acc_step": 0, "speed/wps": 12884.662286679208, "speed/FLOPS": 202371326593670.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06352603435516357, "optim/lr": 0.002993606385161903, "optim/total_tokens": 3469737984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8846242427825928, "created_at": "2025-01-16T00:14:39.351500+00:00"} {"global_step": 6619, "acc_step": 0, "speed/wps": 12887.091413798053, "speed/FLOPS": 202409479373042.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06439603120088577, "optim/lr": 0.002993598480351895, "optim/total_tokens": 3470262272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423524, "loss/out": 3.0824105739593506, "created_at": "2025-01-16T00:14:49.527191+00:00"} {"global_step": 6620, "acc_step": 0, "speed/wps": 12891.42874818994, "speed/FLOPS": 202477603169788.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09420941025018692, "optim/lr": 0.0029935905706687596, "optim/total_tokens": 3470786560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 3.035787582397461, "created_at": "2025-01-16T00:14:59.700136+00:00"} {"global_step": 6621, "acc_step": 0, "speed/wps": 12887.70156811304, "speed/FLOPS": 202419062684998.9, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.078011654317379, "optim/lr": 0.0029935826561125227, "optim/total_tokens": 3471310848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.015033721923828, "created_at": "2025-01-16T00:15:09.874463+00:00"} {"global_step": 6622, "acc_step": 0, "speed/wps": 12892.061862495286, "speed/FLOPS": 202487547099944.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05627300217747688, "optim/lr": 0.0029935747366832094, "optim/total_tokens": 3471835136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.079561233520508, "created_at": "2025-01-16T00:15:20.045464+00:00"} {"global_step": 6623, "acc_step": 0, "speed/wps": 12891.07704593836, "speed/FLOPS": 202472079202638.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0799485594034195, "optim/lr": 0.0029935668123808458, "optim/total_tokens": 3472359424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.0170164108276367, "created_at": "2025-01-16T00:15:30.217226+00:00"} {"global_step": 6624, "acc_step": 0, "speed/wps": 12885.733327673837, "speed/FLOPS": 202388148764255.44, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06623083353042603, "optim/lr": 0.002993558883205458, "optim/total_tokens": 3472883712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.9248955249786377, "created_at": "2025-01-16T00:15:40.392746+00:00"} {"global_step": 6625, "acc_step": 0, "speed/wps": 12887.22890984827, "speed/FLOPS": 202411638937450.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051203157752752304, "optim/lr": 0.0029935509491570713, "optim/total_tokens": 3473408000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9793930053710938, "created_at": "2025-01-16T00:15:50.575462+00:00"} {"global_step": 6626, "acc_step": 0, "speed/wps": 12889.02419544893, "speed/FLOPS": 202439836364789.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07000115513801575, "optim/lr": 0.0029935430102357124, "optim/total_tokens": 3473932288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.961803913116455, "created_at": "2025-01-16T00:16:00.751131+00:00"} {"global_step": 6627, "acc_step": 0, "speed/wps": 12888.542271475011, "speed/FLOPS": 202432267086546.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061854857951402664, "optim/lr": 0.002993535066441407, "optim/total_tokens": 3474456576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 3.0308940410614014, "created_at": "2025-01-16T00:16:10.924419+00:00"} {"global_step": 6628, "acc_step": 0, "speed/wps": 12888.639547842587, "speed/FLOPS": 202433794945570.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05338495597243309, "optim/lr": 0.00299352711777418, "optim/total_tokens": 3474980864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.895986557006836, "created_at": "2025-01-16T00:16:21.098425+00:00"} {"global_step": 6629, "acc_step": 0, "speed/wps": 12890.75199985176, "speed/FLOPS": 202466973907187.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0649477168917656, "optim/lr": 0.0029935191642340583, "optim/total_tokens": 3475505152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.0104713439941406, "created_at": "2025-01-16T00:16:31.270033+00:00"} {"global_step": 6630, "acc_step": 0, "speed/wps": 12886.791460597477, "speed/FLOPS": 202404768195848.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06319234520196915, "optim/lr": 0.0029935112058210684, "optim/total_tokens": 3476029440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 2.9692349433898926, "created_at": "2025-01-16T00:16:41.448589+00:00"} {"global_step": 6631, "acc_step": 0, "speed/wps": 12888.278679551771, "speed/FLOPS": 202428127013178.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6789439916610718, "optim/lr": 0.0029935032425352347, "optim/total_tokens": 3476553728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.9879069328308105, "created_at": "2025-01-16T00:16:51.625028+00:00"} {"global_step": 6632, "acc_step": 0, "speed/wps": 12879.986321877397, "speed/FLOPS": 202297884141008.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2350482940673828, "optim/lr": 0.002993495274376584, "optim/total_tokens": 3477078016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.5790154933929443, "created_at": "2025-01-16T00:17:01.812063+00:00"} {"global_step": 6633, "acc_step": 0, "speed/wps": 12883.22548383963, "speed/FLOPS": 202348759630700.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8524719476699829, "optim/lr": 0.0029934873013451433, "optim/total_tokens": 3477602304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.0273470878601074, "created_at": "2025-01-16T00:17:11.989595+00:00"} {"global_step": 6634, "acc_step": 0, "speed/wps": 12885.928443633335, "speed/FLOPS": 202391213328516.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1655902862548828, "optim/lr": 0.002993479323440937, "optim/total_tokens": 3478126592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9194366931915283, "created_at": "2025-01-16T00:17:22.165064+00:00"} {"global_step": 6635, "acc_step": 0, "speed/wps": 12891.156856464253, "speed/FLOPS": 202473332736618.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15317387878894806, "optim/lr": 0.0029934713406639915, "optim/total_tokens": 3478650880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 3.0745630264282227, "created_at": "2025-01-16T00:17:32.339330+00:00"} {"global_step": 6636, "acc_step": 0, "speed/wps": 12887.106380376244, "speed/FLOPS": 202409714443718.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1151270642876625, "optim/lr": 0.0029934633530143335, "optim/total_tokens": 3479175168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.048557758331299, "created_at": "2025-01-16T00:17:42.518876+00:00"} {"global_step": 6637, "acc_step": 0, "speed/wps": 12890.142864783998, "speed/FLOPS": 202457406603909.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12455811351537704, "optim/lr": 0.0029934553604919886, "optim/total_tokens": 3479699456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.0397090911865234, "created_at": "2025-01-16T00:17:52.692794+00:00"} {"global_step": 6638, "acc_step": 0, "speed/wps": 12880.735747515442, "speed/FLOPS": 202309654900473.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09492365270853043, "optim/lr": 0.002993447363096983, "optim/total_tokens": 3480223744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 3.102931022644043, "created_at": "2025-01-16T00:18:02.873958+00:00"} {"global_step": 6639, "acc_step": 0, "speed/wps": 12888.324629846693, "speed/FLOPS": 202428848725702.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09335660189390182, "optim/lr": 0.002993439360829342, "optim/total_tokens": 3480748032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.043618679046631, "created_at": "2025-01-16T00:18:13.050526+00:00"} {"global_step": 6640, "acc_step": 0, "speed/wps": 12885.129973046069, "speed/FLOPS": 202378672250730.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08578545600175858, "optim/lr": 0.002993431353689093, "optim/total_tokens": 3481272320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9544143676757812, "created_at": "2025-01-16T00:18:23.228599+00:00"} {"global_step": 6641, "acc_step": 0, "speed/wps": 12884.774019751138, "speed/FLOPS": 202373081515102.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0644218847155571, "optim/lr": 0.0029934233416762618, "optim/total_tokens": 3481796608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9556241035461426, "created_at": "2025-01-16T00:18:33.408449+00:00"} {"global_step": 6642, "acc_step": 0, "speed/wps": 12882.666824721764, "speed/FLOPS": 202339985121576.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0676354467868805, "optim/lr": 0.002993415324790874, "optim/total_tokens": 3482320896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.091269016265869, "created_at": "2025-01-16T00:18:43.590174+00:00"} {"global_step": 6643, "acc_step": 0, "speed/wps": 12886.659851308956, "speed/FLOPS": 202402701091121.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.070152148604393, "optim/lr": 0.0029934073030329564, "optim/total_tokens": 3482845184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.0313196182250977, "created_at": "2025-01-16T00:18:53.767608+00:00"} {"global_step": 6644, "acc_step": 0, "speed/wps": 12887.436168420634, "speed/FLOPS": 202414894218132.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05673995986580849, "optim/lr": 0.0029933992764025344, "optim/total_tokens": 3483369472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.045070171356201, "created_at": "2025-01-16T00:19:03.941616+00:00"} {"global_step": 6645, "acc_step": 0, "speed/wps": 12886.988346787122, "speed/FLOPS": 202407860564006.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057338595390319824, "optim/lr": 0.002993391244899635, "optim/total_tokens": 3483893760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.978886127471924, "created_at": "2025-01-16T00:19:14.117541+00:00"} {"global_step": 6646, "acc_step": 0, "speed/wps": 12890.594856276246, "speed/FLOPS": 202464505751396.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06741355359554291, "optim/lr": 0.002993383208524284, "optim/total_tokens": 3484418048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376285, "loss/out": 3.097221851348877, "created_at": "2025-01-16T00:19:24.290514+00:00"} {"global_step": 6647, "acc_step": 0, "speed/wps": 12891.936142327002, "speed/FLOPS": 202485572491943.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06639599800109863, "optim/lr": 0.0029933751672765077, "optim/total_tokens": 3484942336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.1946444511413574, "created_at": "2025-01-16T00:19:34.460972+00:00"} {"global_step": 6648, "acc_step": 0, "speed/wps": 12891.978192065199, "speed/FLOPS": 202486232940864.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06698500365018845, "optim/lr": 0.002993367121156332, "optim/total_tokens": 3485466624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.079829692840576, "created_at": "2025-01-16T00:19:44.631714+00:00"} {"global_step": 6649, "acc_step": 0, "speed/wps": 12890.626621655607, "speed/FLOPS": 202465004670329.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06995117664337158, "optim/lr": 0.0029933590701637833, "optim/total_tokens": 3485990912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8922038078308105, "created_at": "2025-01-16T00:19:54.808345+00:00"} {"global_step": 6650, "acc_step": 0, "speed/wps": 12889.293989270001, "speed/FLOPS": 202444073847484.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05400524288415909, "optim/lr": 0.0029933510142988885, "optim/total_tokens": 3486515200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.9915173053741455, "created_at": "2025-01-16T00:20:04.982602+00:00"} {"global_step": 6651, "acc_step": 0, "speed/wps": 12890.260803864532, "speed/FLOPS": 202459258999234.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06444542855024338, "optim/lr": 0.002993342953561673, "optim/total_tokens": 3487039488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.084712505340576, "created_at": "2025-01-16T00:20:15.155329+00:00"} {"global_step": 6652, "acc_step": 0, "speed/wps": 12889.769674822943, "speed/FLOPS": 202451545142754.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06138550490140915, "optim/lr": 0.002993334887952164, "optim/total_tokens": 3487563776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.958583116531372, "created_at": "2025-01-16T00:20:25.328123+00:00"} {"global_step": 6653, "acc_step": 0, "speed/wps": 12889.46986254428, "speed/FLOPS": 202446836178932.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04935305193066597, "optim/lr": 0.002993326817470387, "optim/total_tokens": 3488088064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 3.0783209800720215, "created_at": "2025-01-16T00:20:35.503280+00:00"} {"global_step": 6654, "acc_step": 0, "speed/wps": 12891.687023891644, "speed/FLOPS": 202481659744592.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06525501608848572, "optim/lr": 0.0029933187421163684, "optim/total_tokens": 3488612352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.997128486633301, "created_at": "2025-01-16T00:20:45.676271+00:00"} {"global_step": 6655, "acc_step": 0, "speed/wps": 12886.799348862305, "speed/FLOPS": 202404892091886.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053520578891038895, "optim/lr": 0.002993310661890135, "optim/total_tokens": 3489136640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 3.0405020713806152, "created_at": "2025-01-16T00:20:55.855992+00:00"} {"global_step": 6656, "acc_step": 0, "speed/wps": 12893.125691181287, "speed/FLOPS": 202504255991311.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06797412782907486, "optim/lr": 0.0029933025767917127, "optim/total_tokens": 3489660928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 3.111104965209961, "created_at": "2025-01-16T00:21:06.025825+00:00"} {"global_step": 6657, "acc_step": 0, "speed/wps": 12883.43892452097, "speed/FLOPS": 202352112009899.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08638196438550949, "optim/lr": 0.002993294486821128, "optim/total_tokens": 3490185216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.0589163303375244, "created_at": "2025-01-16T00:21:16.203513+00:00"} {"global_step": 6658, "acc_step": 0, "speed/wps": 12891.773677054425, "speed/FLOPS": 202483020751587.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06564714014530182, "optim/lr": 0.0029932863919784084, "optim/total_tokens": 3490709504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 3.074075698852539, "created_at": "2025-01-16T00:21:26.376531+00:00"} {"global_step": 6659, "acc_step": 0, "speed/wps": 12891.91646913817, "speed/FLOPS": 202485263497479.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0876065343618393, "optim/lr": 0.0029932782922635786, "optim/total_tokens": 3491233792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.1158885955810547, "created_at": "2025-01-16T00:21:36.548187+00:00"} {"global_step": 6660, "acc_step": 0, "speed/wps": 12892.253960212738, "speed/FLOPS": 202490564258566.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06831065565347672, "optim/lr": 0.002993270187676666, "optim/total_tokens": 3491758080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 3.0601236820220947, "created_at": "2025-01-16T00:21:46.720820+00:00"} {"global_step": 6661, "acc_step": 0, "speed/wps": 12890.599295674596, "speed/FLOPS": 202464575478248.1, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.059861984103918076, "optim/lr": 0.0029932620782176966, "optim/total_tokens": 3492282368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483664, "loss/out": 3.0636157989501953, "created_at": "2025-01-16T00:21:56.893797+00:00"} {"global_step": 6662, "acc_step": 0, "speed/wps": 12895.687487138443, "speed/FLOPS": 202544492517095.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05734686553478241, "optim/lr": 0.002993253963886697, "optim/total_tokens": 3492806656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 3.0815000534057617, "created_at": "2025-01-16T00:22:07.064402+00:00"} {"global_step": 6663, "acc_step": 0, "speed/wps": 12888.98748974371, "speed/FLOPS": 202439259851250.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05828635022044182, "optim/lr": 0.0029932458446836944, "optim/total_tokens": 3493330944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 3.023257255554199, "created_at": "2025-01-16T00:22:17.239750+00:00"} {"global_step": 6664, "acc_step": 0, "speed/wps": 12891.05576361323, "speed/FLOPS": 202471744934477.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06260996311903, "optim/lr": 0.0029932377206087135, "optim/total_tokens": 3493855232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.9637279510498047, "created_at": "2025-01-16T00:22:27.411000+00:00"} {"global_step": 6665, "acc_step": 0, "speed/wps": 12888.965631203677, "speed/FLOPS": 202438916532842.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051780398935079575, "optim/lr": 0.002993229591661783, "optim/total_tokens": 3494379520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 2.9450461864471436, "created_at": "2025-01-16T00:22:37.586078+00:00"} {"global_step": 6666, "acc_step": 0, "speed/wps": 12892.807350790084, "speed/FLOPS": 202499256018022.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07837047427892685, "optim/lr": 0.002993221457842928, "optim/total_tokens": 3494903808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.1017308235168457, "created_at": "2025-01-16T00:22:47.759942+00:00"} {"global_step": 6667, "acc_step": 0, "speed/wps": 12887.25980072279, "speed/FLOPS": 202412124121083.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06015734747052193, "optim/lr": 0.002993213319152175, "optim/total_tokens": 3495428096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 3.0766496658325195, "created_at": "2025-01-16T00:22:57.940018+00:00"} {"global_step": 6668, "acc_step": 0, "speed/wps": 12894.929456269754, "speed/FLOPS": 202532586600659.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0631842389702797, "optim/lr": 0.0029932051755895512, "optim/total_tokens": 3495952384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.995471954345703, "created_at": "2025-01-16T00:23:08.108261+00:00"} {"global_step": 6669, "acc_step": 0, "speed/wps": 12888.675258900874, "speed/FLOPS": 202434355836813.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059046611189842224, "optim/lr": 0.0029931970271550824, "optim/total_tokens": 3496476672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.0298891067504883, "created_at": "2025-01-16T00:23:18.283875+00:00"} {"global_step": 6670, "acc_step": 0, "speed/wps": 12895.613202427403, "speed/FLOPS": 202543325773630.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.052246566861867905, "optim/lr": 0.0029931888738487966, "optim/total_tokens": 3497000960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.1696462631225586, "created_at": "2025-01-16T00:23:28.455390+00:00"} {"global_step": 6671, "acc_step": 0, "speed/wps": 12889.07503206276, "speed/FLOPS": 202440634823666.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06012491509318352, "optim/lr": 0.0029931807156707195, "optim/total_tokens": 3497525248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9796390533447266, "created_at": "2025-01-16T00:23:38.629302+00:00"} {"global_step": 6672, "acc_step": 0, "speed/wps": 12894.148335365742, "speed/FLOPS": 202520318023489.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06268147379159927, "optim/lr": 0.002993172552620877, "optim/total_tokens": 3498049536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9652702808380127, "created_at": "2025-01-16T00:23:48.802111+00:00"} {"global_step": 6673, "acc_step": 0, "speed/wps": 12890.395107767014, "speed/FLOPS": 202461368426575.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0685693621635437, "optim/lr": 0.0029931643846992967, "optim/total_tokens": 3498573824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.0025041103363037, "created_at": "2025-01-16T00:23:58.974538+00:00"} {"global_step": 6674, "acc_step": 0, "speed/wps": 12891.13202678432, "speed/FLOPS": 202472942752375.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06462783366441727, "optim/lr": 0.002993156211906005, "optim/total_tokens": 3499098112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 3.0167369842529297, "created_at": "2025-01-16T00:24:09.145614+00:00"} {"global_step": 6675, "acc_step": 0, "speed/wps": 12889.557959168356, "speed/FLOPS": 202448219857471.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06168049946427345, "optim/lr": 0.002993148034241029, "optim/total_tokens": 3499622400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.048168659210205, "created_at": "2025-01-16T00:24:19.318357+00:00"} {"global_step": 6676, "acc_step": 0, "speed/wps": 12894.404632538097, "speed/FLOPS": 202524343522770.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0528857558965683, "optim/lr": 0.0029931398517043946, "optim/total_tokens": 3500146688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 3.097506523132324, "created_at": "2025-01-16T00:24:29.490787+00:00"} {"global_step": 6677, "acc_step": 0, "speed/wps": 12891.05468202335, "speed/FLOPS": 202471727946621.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0530097521841526, "optim/lr": 0.002993131664296129, "optim/total_tokens": 3500670976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.04587984085083, "created_at": "2025-01-16T00:24:39.662052+00:00"} {"global_step": 6678, "acc_step": 0, "speed/wps": 12891.50227854742, "speed/FLOPS": 202478758065094.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05847804248332977, "optim/lr": 0.002993123472016259, "optim/total_tokens": 3501195264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 3.1247365474700928, "created_at": "2025-01-16T00:24:49.836662+00:00"} {"global_step": 6679, "acc_step": 0, "speed/wps": 12885.891591562035, "speed/FLOPS": 202390634516096.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06761021167039871, "optim/lr": 0.0029931152748648104, "optim/total_tokens": 3501719552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.027050733566284, "created_at": "2025-01-16T00:25:00.014342+00:00"} {"global_step": 6680, "acc_step": 0, "speed/wps": 12890.869333732446, "speed/FLOPS": 202468816797017.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04912446439266205, "optim/lr": 0.0029931070728418108, "optim/total_tokens": 3502243840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394230, "loss/out": 3.0035293102264404, "created_at": "2025-01-16T00:25:10.188110+00:00"} {"global_step": 6681, "acc_step": 0, "speed/wps": 12887.952901215607, "speed/FLOPS": 202423010216742.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06414882093667984, "optim/lr": 0.002993098865947287, "optim/total_tokens": 3502768128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454173, "loss/out": 3.131892204284668, "created_at": "2025-01-16T00:25:20.363506+00:00"} {"global_step": 6682, "acc_step": 0, "speed/wps": 12888.074050574576, "speed/FLOPS": 202424913033901.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056547392159700394, "optim/lr": 0.0029930906541812655, "optim/total_tokens": 3503292416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.1349775791168213, "created_at": "2025-01-16T00:25:30.542837+00:00"} {"global_step": 6683, "acc_step": 0, "speed/wps": 12887.527287879284, "speed/FLOPS": 202416325374441.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08269548416137695, "optim/lr": 0.0029930824375437736, "optim/total_tokens": 3503816704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 3.113224506378174, "created_at": "2025-01-16T00:25:40.719863+00:00"} {"global_step": 6684, "acc_step": 0, "speed/wps": 12894.106473477788, "speed/FLOPS": 202519660525014.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06549186259508133, "optim/lr": 0.002993074216034837, "optim/total_tokens": 3504340992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 3.088583469390869, "created_at": "2025-01-16T00:25:50.894737+00:00"} {"global_step": 6685, "acc_step": 0, "speed/wps": 12891.003137863645, "speed/FLOPS": 202470918374764.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05401315167546272, "optim/lr": 0.0029930659896544834, "optim/total_tokens": 3504865280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0491514205932617, "created_at": "2025-01-16T00:26:01.068350+00:00"} {"global_step": 6686, "acc_step": 0, "speed/wps": 12894.705854693628, "speed/FLOPS": 202529074630645.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06040564924478531, "optim/lr": 0.0029930577584027394, "optim/total_tokens": 3505389568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.0262253284454346, "created_at": "2025-01-16T00:26:11.238039+00:00"} {"global_step": 6687, "acc_step": 0, "speed/wps": 12895.46889045415, "speed/FLOPS": 202541059155785.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05895071104168892, "optim/lr": 0.002993049522279632, "optim/total_tokens": 3505913856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.1180405616760254, "created_at": "2025-01-16T00:26:21.409266+00:00"} {"global_step": 6688, "acc_step": 0, "speed/wps": 12893.274534878032, "speed/FLOPS": 202506593786101.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057019446045160294, "optim/lr": 0.002993041281285188, "optim/total_tokens": 3506438144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 3.034364700317383, "created_at": "2025-01-16T00:26:31.579761+00:00"} {"global_step": 6689, "acc_step": 0, "speed/wps": 12897.045458764478, "speed/FLOPS": 202565821327531.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057637669146060944, "optim/lr": 0.0029930330354194343, "optim/total_tokens": 3506962432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 2.9577841758728027, "created_at": "2025-01-16T00:26:41.752576+00:00"} {"global_step": 6690, "acc_step": 0, "speed/wps": 12893.434858646828, "speed/FLOPS": 202509111891198.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05284985899925232, "optim/lr": 0.002993024784682398, "optim/total_tokens": 3507486720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.022005319595337, "created_at": "2025-01-16T00:26:51.924562+00:00"} {"global_step": 6691, "acc_step": 0, "speed/wps": 12889.563103792487, "speed/FLOPS": 202448300660862.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04867705702781677, "optim/lr": 0.002993016529074105, "optim/total_tokens": 3508011008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.0525171756744385, "created_at": "2025-01-16T00:27:02.097068+00:00"} {"global_step": 6692, "acc_step": 0, "speed/wps": 12891.830652506847, "speed/FLOPS": 202483915629360.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061069149523973465, "optim/lr": 0.0029930082685945838, "optim/total_tokens": 3508535296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.9939398765563965, "created_at": "2025-01-16T00:27:12.274762+00:00"} {"global_step": 6693, "acc_step": 0, "speed/wps": 12890.545007710027, "speed/FLOPS": 202463722811165.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054887909442186356, "optim/lr": 0.00299300000324386, "optim/total_tokens": 3509059584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.930225372314453, "created_at": "2025-01-16T00:27:22.451965+00:00"} {"global_step": 6694, "acc_step": 0, "speed/wps": 12895.713635342956, "speed/FLOPS": 202544903210578.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06485560536384583, "optim/lr": 0.0029929917330219612, "optim/total_tokens": 3509583872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487147, "loss/out": 3.199094772338867, "created_at": "2025-01-16T00:27:32.620814+00:00"} {"global_step": 6695, "acc_step": 0, "speed/wps": 12897.67364519355, "speed/FLOPS": 202575687858618.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06865384429693222, "optim/lr": 0.0029929834579289142, "optim/total_tokens": 3510108160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.0662648677825928, "created_at": "2025-01-16T00:27:42.789129+00:00"} {"global_step": 6696, "acc_step": 0, "speed/wps": 12897.87675289088, "speed/FLOPS": 202578877944103.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07572473585605621, "optim/lr": 0.002992975177964747, "optim/total_tokens": 3510632448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9576406478881836, "created_at": "2025-01-16T00:27:52.955374+00:00"} {"global_step": 6697, "acc_step": 0, "speed/wps": 12893.94041549435, "speed/FLOPS": 202517052356197.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06190812960267067, "optim/lr": 0.002992966893129485, "optim/total_tokens": 3511156736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.97322416305542, "created_at": "2025-01-16T00:28:03.126588+00:00"} {"global_step": 6698, "acc_step": 0, "speed/wps": 12893.842501527703, "speed/FLOPS": 202515514482803.0, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06717235594987869, "optim/lr": 0.0029929586034231556, "optim/total_tokens": 3511681024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9925308227539062, "created_at": "2025-01-16T00:28:13.296898+00:00"} {"global_step": 6699, "acc_step": 0, "speed/wps": 12890.741678010763, "speed/FLOPS": 202466811788491.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07815922051668167, "optim/lr": 0.002992950308845787, "optim/total_tokens": 3512205312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9322032928466797, "created_at": "2025-01-16T00:28:23.470517+00:00"} {"global_step": 6700, "acc_step": 0, "speed/wps": 12889.170715809587, "speed/FLOPS": 202442137668393.66, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056221459060907364, "optim/lr": 0.002992942009397405, "optim/total_tokens": 3512729600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9384765625, "created_at": "2025-01-16T00:28:33.644952+00:00"} {"global_step": 6701, "acc_step": 0, "speed/wps": 12893.61357243474, "speed/FLOPS": 202511918836816.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06844547390937805, "optim/lr": 0.002992933705078037, "optim/total_tokens": 3513253888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.0039167404174805, "created_at": "2025-01-16T00:28:43.815237+00:00"} {"global_step": 6702, "acc_step": 0, "speed/wps": 12894.575874733702, "speed/FLOPS": 202527033116763.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05907578393816948, "optim/lr": 0.0029929253958877106, "optim/total_tokens": 3513778176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 3.075230360031128, "created_at": "2025-01-16T00:28:53.990790+00:00"} {"global_step": 6703, "acc_step": 0, "speed/wps": 12892.572309888945, "speed/FLOPS": 202495564377690.1, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05660587176680565, "optim/lr": 0.0029929170818264526, "optim/total_tokens": 3514302464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 3.0358433723449707, "created_at": "2025-01-16T00:29:04.162022+00:00"} {"global_step": 6704, "acc_step": 0, "speed/wps": 12892.954791049238, "speed/FLOPS": 202501571769896.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06625054031610489, "optim/lr": 0.0029929087628942897, "optim/total_tokens": 3514826752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.1394660472869873, "created_at": "2025-01-16T00:29:14.336741+00:00"} {"global_step": 6705, "acc_step": 0, "speed/wps": 12893.80713393256, "speed/FLOPS": 202514958986121.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05646994337439537, "optim/lr": 0.0029929004390912496, "optim/total_tokens": 3515351040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.9838173389434814, "created_at": "2025-01-16T00:29:24.507633+00:00"} {"global_step": 6706, "acc_step": 0, "speed/wps": 12887.33928430693, "speed/FLOPS": 202413372519993.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06410437822341919, "optim/lr": 0.0029928921104173593, "optim/total_tokens": 3515875328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.0000405311584473, "created_at": "2025-01-16T00:29:34.682463+00:00"} {"global_step": 6707, "acc_step": 0, "speed/wps": 12892.949251584789, "speed/FLOPS": 202501484764995.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05640111863613129, "optim/lr": 0.002992883776872646, "optim/total_tokens": 3516399616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374042, "loss/out": 3.0991673469543457, "created_at": "2025-01-16T00:29:44.853005+00:00"} {"global_step": 6708, "acc_step": 0, "speed/wps": 12896.498175697252, "speed/FLOPS": 202557225494914.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06230387091636658, "optim/lr": 0.002992875438457137, "optim/total_tokens": 3516923904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.012302875518799, "created_at": "2025-01-16T00:29:55.020015+00:00"} {"global_step": 6709, "acc_step": 0, "speed/wps": 12896.185492237133, "speed/FLOPS": 202552314371501.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06393307447433472, "optim/lr": 0.002992867095170859, "optim/total_tokens": 3517448192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 3.051485538482666, "created_at": "2025-01-16T00:30:05.196806+00:00"} {"global_step": 6710, "acc_step": 0, "speed/wps": 12888.86341595079, "speed/FLOPS": 202437311101837.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06273096054792404, "optim/lr": 0.00299285874701384, "optim/total_tokens": 3517972480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 268894, "loss/out": 3.1258926391601562, "created_at": "2025-01-16T00:30:15.369841+00:00"} {"global_step": 6711, "acc_step": 0, "speed/wps": 12884.949631677197, "speed/FLOPS": 202375839741717.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08826126158237457, "optim/lr": 0.002992850393986106, "optim/total_tokens": 3518496768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9280033111572266, "created_at": "2025-01-16T00:30:25.546177+00:00"} {"global_step": 6712, "acc_step": 0, "speed/wps": 12891.293479273736, "speed/FLOPS": 202475478585580.8, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06303395330905914, "optim/lr": 0.0029928420360876857, "optim/total_tokens": 3519021056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.0958364009857178, "created_at": "2025-01-16T00:30:35.722374+00:00"} {"global_step": 6713, "acc_step": 0, "speed/wps": 12892.783020258863, "speed/FLOPS": 202498873873596.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06043185666203499, "optim/lr": 0.002992833673318605, "optim/total_tokens": 3519545344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.028719425201416, "created_at": "2025-01-16T00:30:45.892263+00:00"} {"global_step": 6714, "acc_step": 0, "speed/wps": 12889.32795333465, "speed/FLOPS": 202444607299791.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06963484734296799, "optim/lr": 0.0029928253056788925, "optim/total_tokens": 3520069632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368810, "loss/out": 3.0651700496673584, "created_at": "2025-01-16T00:30:56.064811+00:00"} {"global_step": 6715, "acc_step": 0, "speed/wps": 12889.936438807817, "speed/FLOPS": 202454164400294.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07237377762794495, "optim/lr": 0.002992816933168575, "optim/total_tokens": 3520593920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.942620277404785, "created_at": "2025-01-16T00:31:06.238687+00:00"} {"global_step": 6716, "acc_step": 0, "speed/wps": 12890.426429414492, "speed/FLOPS": 202461860376090.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08259721100330353, "optim/lr": 0.0029928085557876798, "optim/total_tokens": 3521118208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.0781362056732178, "created_at": "2025-01-16T00:31:16.411541+00:00"} {"global_step": 6717, "acc_step": 0, "speed/wps": 12892.642888816423, "speed/FLOPS": 202496672916731.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07188282907009125, "optim/lr": 0.002992800173536234, "optim/total_tokens": 3521642496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311555, "loss/out": 2.982966423034668, "created_at": "2025-01-16T00:31:26.584655+00:00"} {"global_step": 6718, "acc_step": 0, "speed/wps": 12883.871983876474, "speed/FLOPS": 202358913802163.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06056928634643555, "optim/lr": 0.0029927917864142647, "optim/total_tokens": 3522166784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.9704670906066895, "created_at": "2025-01-16T00:31:36.761587+00:00"} {"global_step": 6719, "acc_step": 0, "speed/wps": 12889.638892443942, "speed/FLOPS": 202449491025778.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05769440531730652, "optim/lr": 0.0029927833944218, "optim/total_tokens": 3522691072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.05503511428833, "created_at": "2025-01-16T00:31:46.936714+00:00"} {"global_step": 6720, "acc_step": 0, "speed/wps": 12889.671150111551, "speed/FLOPS": 202449997676774.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05625345557928085, "optim/lr": 0.0029927749975588672, "optim/total_tokens": 3523215360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.0744476318359375, "created_at": "2025-01-16T00:31:57.110337+00:00"} {"global_step": 6721, "acc_step": 0, "speed/wps": 12887.564561174286, "speed/FLOPS": 202416910802758.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054004162549972534, "optim/lr": 0.0029927665958254935, "optim/total_tokens": 3523739648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9756736755371094, "created_at": "2025-01-16T00:32:07.287089+00:00"} {"global_step": 6722, "acc_step": 0, "speed/wps": 12890.357717742265, "speed/FLOPS": 202460781164857.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06107553467154503, "optim/lr": 0.002992758189221706, "optim/total_tokens": 3524263936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.948140859603882, "created_at": "2025-01-16T00:32:17.461362+00:00"} {"global_step": 6723, "acc_step": 0, "speed/wps": 12890.3398592974, "speed/FLOPS": 202460500673441.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07319169491529465, "optim/lr": 0.002992749777747532, "optim/total_tokens": 3524788224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.1345951557159424, "created_at": "2025-01-16T00:32:27.633230+00:00"} {"global_step": 6724, "acc_step": 0, "speed/wps": 12889.775529120918, "speed/FLOPS": 202451637092548.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07987241446971893, "optim/lr": 0.0029927413614029997, "optim/total_tokens": 3525312512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444883, "loss/out": 3.001063346862793, "created_at": "2025-01-16T00:32:37.809777+00:00"} {"global_step": 6725, "acc_step": 0, "speed/wps": 12895.573657602934, "speed/FLOPS": 202542704667822.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0721455067396164, "optim/lr": 0.0029927329401881362, "optim/total_tokens": 3525836800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.0464727878570557, "created_at": "2025-01-16T00:32:47.978114+00:00"} {"global_step": 6726, "acc_step": 0, "speed/wps": 12894.671589368907, "speed/FLOPS": 202528536446635.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08569791913032532, "optim/lr": 0.0029927245141029694, "optim/total_tokens": 3526361088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.9736571311950684, "created_at": "2025-01-16T00:32:58.148374+00:00"} {"global_step": 6727, "acc_step": 0, "speed/wps": 12891.267336002384, "speed/FLOPS": 202475067969579.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06031284108757973, "optim/lr": 0.0029927160831475256, "optim/total_tokens": 3526885376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0412211418151855, "created_at": "2025-01-16T00:33:08.319979+00:00"} {"global_step": 6728, "acc_step": 0, "speed/wps": 12888.691041233316, "speed/FLOPS": 202434603720031.4, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056288328021764755, "optim/lr": 0.0029927076473218336, "optim/total_tokens": 3527409664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 3.043492078781128, "created_at": "2025-01-16T00:33:18.496275+00:00"} {"global_step": 6729, "acc_step": 0, "speed/wps": 12877.53638873256, "speed/FLOPS": 202259404574408.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06642412394285202, "optim/lr": 0.00299269920662592, "optim/total_tokens": 3527933952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0525083541870117, "created_at": "2025-01-16T00:33:28.678325+00:00"} {"global_step": 6730, "acc_step": 0, "speed/wps": 12884.334433491149, "speed/FLOPS": 202366177208834.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.058909617364406586, "optim/lr": 0.002992690761059813, "optim/total_tokens": 3528458240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0357232093811035, "created_at": "2025-01-16T00:33:38.856283+00:00"} {"global_step": 6731, "acc_step": 0, "speed/wps": 12885.335392160412, "speed/FLOPS": 202381898640196.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05454149469733238, "optim/lr": 0.00299268231062354, "optim/total_tokens": 3528982528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.9841346740722656, "created_at": "2025-01-16T00:33:49.032270+00:00"} {"global_step": 6732, "acc_step": 0, "speed/wps": 12889.994732078005, "speed/FLOPS": 202455079976206.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05838928371667862, "optim/lr": 0.002992673855317128, "optim/total_tokens": 3529506816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 2.93035626411438, "created_at": "2025-01-16T00:33:59.211144+00:00"} {"global_step": 6733, "acc_step": 0, "speed/wps": 12883.000438833455, "speed/FLOPS": 202345224989634.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05345704033970833, "optim/lr": 0.0029926653951406054, "optim/total_tokens": 3530031104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0787742137908936, "created_at": "2025-01-16T00:34:09.393217+00:00"} {"global_step": 6734, "acc_step": 0, "speed/wps": 12887.496032244506, "speed/FLOPS": 202415834461747.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04742690920829773, "optim/lr": 0.002992656930093999, "optim/total_tokens": 3530555392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 3.0869081020355225, "created_at": "2025-01-16T00:34:19.570101+00:00"} {"global_step": 6735, "acc_step": 0, "speed/wps": 12897.764981578803, "speed/FLOPS": 202577122422057.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05556159093976021, "optim/lr": 0.0029926484601773373, "optim/total_tokens": 3531079680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.996201992034912, "created_at": "2025-01-16T00:34:29.738600+00:00"} {"global_step": 6736, "acc_step": 0, "speed/wps": 12887.932475410218, "speed/FLOPS": 202422689401400.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05406085029244423, "optim/lr": 0.0029926399853906476, "optim/total_tokens": 3531603968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.9990720748901367, "created_at": "2025-01-16T00:34:39.913284+00:00"} {"global_step": 6737, "acc_step": 0, "speed/wps": 12890.955719678896, "speed/FLOPS": 202470173607013.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050272297114133835, "optim/lr": 0.0029926315057339573, "optim/total_tokens": 3532128256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.097881317138672, "created_at": "2025-01-16T00:34:50.084564+00:00"} {"global_step": 6738, "acc_step": 0, "speed/wps": 12890.57203592788, "speed/FLOPS": 202464147326467.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05506764352321625, "optim/lr": 0.0029926230212072944, "optim/total_tokens": 3532652544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 3.078366279602051, "created_at": "2025-01-16T00:35:00.257543+00:00"} {"global_step": 6739, "acc_step": 0, "speed/wps": 12896.23630145521, "speed/FLOPS": 202553112400090.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05555584281682968, "optim/lr": 0.002992614531810686, "optim/total_tokens": 3533176832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303224, "loss/out": 2.9116766452789307, "created_at": "2025-01-16T00:35:10.428024+00:00"} {"global_step": 6740, "acc_step": 0, "speed/wps": 12890.03096656307, "speed/FLOPS": 202455649088585.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0698428750038147, "optim/lr": 0.0029926060375441606, "optim/total_tokens": 3533701120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477322, "loss/out": 3.0570266246795654, "created_at": "2025-01-16T00:35:20.601481+00:00"} {"global_step": 6741, "acc_step": 0, "speed/wps": 12885.252533361092, "speed/FLOPS": 202380597228893.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07913952320814133, "optim/lr": 0.002992597538407745, "optim/total_tokens": 3534225408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.089642286300659, "created_at": "2025-01-16T00:35:30.778067+00:00"} {"global_step": 6742, "acc_step": 0, "speed/wps": 12891.0815990237, "speed/FLOPS": 202472150715100.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07317254692316055, "optim/lr": 0.002992589034401468, "optim/total_tokens": 3534749696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419023, "loss/out": 2.938607931137085, "created_at": "2025-01-16T00:35:40.949861+00:00"} {"global_step": 6743, "acc_step": 0, "speed/wps": 12895.591580180153, "speed/FLOPS": 202542986166525.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050997164100408554, "optim/lr": 0.0029925805255253566, "optim/total_tokens": 3535273984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.965877056121826, "created_at": "2025-01-16T00:35:51.123393+00:00"} {"global_step": 6744, "acc_step": 0, "speed/wps": 12897.514272664303, "speed/FLOPS": 202573184694045.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06004185602068901, "optim/lr": 0.0029925720117794383, "optim/total_tokens": 3535798272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.0135486125946045, "created_at": "2025-01-16T00:36:01.289503+00:00"} {"global_step": 6745, "acc_step": 0, "speed/wps": 12889.618028046223, "speed/FLOPS": 202449163321741.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06324394047260284, "optim/lr": 0.0029925634931637418, "optim/total_tokens": 3536322560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491917, "loss/out": 3.0296080112457275, "created_at": "2025-01-16T00:36:11.464357+00:00"} {"global_step": 6746, "acc_step": 0, "speed/wps": 12890.400341608556, "speed/FLOPS": 202461450631249.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06735625863075256, "optim/lr": 0.002992554969678295, "optim/total_tokens": 3536846848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 2.791675090789795, "created_at": "2025-01-16T00:36:21.638190+00:00"} {"global_step": 6747, "acc_step": 0, "speed/wps": 12884.165265497737, "speed/FLOPS": 202363520193036.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06495184451341629, "optim/lr": 0.002992546441323124, "optim/total_tokens": 3537371136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.091181755065918, "created_at": "2025-01-16T00:36:31.814816+00:00"} {"global_step": 6748, "acc_step": 0, "speed/wps": 12889.965483257236, "speed/FLOPS": 202454620583283.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0644250437617302, "optim/lr": 0.002992537908098258, "optim/total_tokens": 3537895424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.9625322818756104, "created_at": "2025-01-16T00:36:41.991546+00:00"} {"global_step": 6749, "acc_step": 0, "speed/wps": 12890.834769996023, "speed/FLOPS": 202468273926041.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05692675709724426, "optim/lr": 0.002992529370003725, "optim/total_tokens": 3538419712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294366, "loss/out": 2.9900364875793457, "created_at": "2025-01-16T00:36:52.162854+00:00"} {"global_step": 6750, "acc_step": 0, "speed/wps": 12890.386700334986, "speed/FLOPS": 202461236376302.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09042168408632278, "optim/lr": 0.002992520827039552, "optim/total_tokens": 3538944000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.0326552391052246, "created_at": "2025-01-16T00:37:02.337953+00:00"} {"global_step": 6751, "acc_step": 0, "speed/wps": 12893.989964282295, "speed/FLOPS": 202517830587998.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09990978986024857, "optim/lr": 0.0029925122792057676, "optim/total_tokens": 3539468288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.007983922958374, "created_at": "2025-01-16T00:37:12.506894+00:00"} {"global_step": 6752, "acc_step": 0, "speed/wps": 12896.289876330908, "speed/FLOPS": 202553953867132.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06027267500758171, "optim/lr": 0.0029925037265023996, "optim/total_tokens": 3539992576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410217, "loss/out": 2.9560675621032715, "created_at": "2025-01-16T00:37:22.674641+00:00"} {"global_step": 6753, "acc_step": 0, "speed/wps": 12892.69410594027, "speed/FLOPS": 202497477352041.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08760162442922592, "optim/lr": 0.0029924951689294754, "optim/total_tokens": 3540516864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0565237998962402, "created_at": "2025-01-16T00:37:32.844796+00:00"} {"global_step": 6754, "acc_step": 0, "speed/wps": 12888.04457796559, "speed/FLOPS": 202424450126078.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08480618894100189, "optim/lr": 0.0029924866064870233, "optim/total_tokens": 3541041152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 3.002387523651123, "created_at": "2025-01-16T00:37:43.018912+00:00"} {"global_step": 6755, "acc_step": 0, "speed/wps": 12891.021542480783, "speed/FLOPS": 202471207444567.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06461834162473679, "optim/lr": 0.0029924780391750714, "optim/total_tokens": 3541565440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.9476871490478516, "created_at": "2025-01-16T00:37:53.191107+00:00"} {"global_step": 6756, "acc_step": 0, "speed/wps": 12887.561731209737, "speed/FLOPS": 202416866354276.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07704456150531769, "optim/lr": 0.0029924694669936475, "optim/total_tokens": 3542089728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8680026531219482, "created_at": "2025-01-16T00:38:03.369270+00:00"} {"global_step": 6757, "acc_step": 0, "speed/wps": 12886.227524836562, "speed/FLOPS": 202395910809794.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07604724168777466, "optim/lr": 0.002992460889942779, "optim/total_tokens": 3542614016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.9855098724365234, "created_at": "2025-01-16T00:38:13.545677+00:00"} {"global_step": 6758, "acc_step": 0, "speed/wps": 12893.920859043174, "speed/FLOPS": 202516745195257.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08020824193954468, "optim/lr": 0.0029924523080224947, "optim/total_tokens": 3543138304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.008057117462158, "created_at": "2025-01-16T00:38:23.719613+00:00"} {"global_step": 6759, "acc_step": 0, "speed/wps": 12889.17199893688, "speed/FLOPS": 202442157821671.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08072153478860855, "optim/lr": 0.002992443721232822, "optim/total_tokens": 3543662592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 3.0838847160339355, "created_at": "2025-01-16T00:38:33.892412+00:00"} {"global_step": 6760, "acc_step": 0, "speed/wps": 12888.27261652618, "speed/FLOPS": 202428031785030.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05622350424528122, "optim/lr": 0.0029924351295737895, "optim/total_tokens": 3544186880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.956228256225586, "created_at": "2025-01-16T00:38:44.067258+00:00"} {"global_step": 6761, "acc_step": 0, "speed/wps": 12890.500935381217, "speed/FLOPS": 202463030594679.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07147089391946793, "optim/lr": 0.002992426533045425, "optim/total_tokens": 3544711168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.0454206466674805, "created_at": "2025-01-16T00:38:54.245660+00:00"} {"global_step": 6762, "acc_step": 0, "speed/wps": 12890.967787024178, "speed/FLOPS": 202470363141253.44, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059307314455509186, "optim/lr": 0.002992417931647756, "optim/total_tokens": 3545235456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 3.125439167022705, "created_at": "2025-01-16T00:39:04.418942+00:00"} {"global_step": 6763, "acc_step": 0, "speed/wps": 12889.482677897508, "speed/FLOPS": 202447037461665.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05758051946759224, "optim/lr": 0.0029924093253808114, "optim/total_tokens": 3545759744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.015522003173828, "created_at": "2025-01-16T00:39:14.593918+00:00"} {"global_step": 6764, "acc_step": 0, "speed/wps": 12891.744852369731, "speed/FLOPS": 202482568020305.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06820569932460785, "optim/lr": 0.002992400714244619, "optim/total_tokens": 3546284032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.022918701171875, "created_at": "2025-01-16T00:39:24.767439+00:00"} {"global_step": 6765, "acc_step": 0, "speed/wps": 12890.240767433266, "speed/FLOPS": 202458944299548.56, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06609661132097244, "optim/lr": 0.0029923920982392067, "optim/total_tokens": 3546808320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.009944200515747, "created_at": "2025-01-16T00:39:34.939302+00:00"} {"global_step": 6766, "acc_step": 0, "speed/wps": 12894.123046433942, "speed/FLOPS": 202519920826064.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08184940367937088, "optim/lr": 0.0029923834773646026, "optim/total_tokens": 3547332608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 3.051448345184326, "created_at": "2025-01-16T00:39:45.107946+00:00"} {"global_step": 6767, "acc_step": 0, "speed/wps": 12886.666278608287, "speed/FLOPS": 202402802040689.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06341954320669174, "optim/lr": 0.002992374851620835, "optim/total_tokens": 3547856896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0135421752929688, "created_at": "2025-01-16T00:39:55.289063+00:00"} {"global_step": 6768, "acc_step": 0, "speed/wps": 12893.909752470132, "speed/FLOPS": 202516570751266.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07801944762468338, "optim/lr": 0.0029923662210079322, "optim/total_tokens": 3548381184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470624, "loss/out": 2.9966671466827393, "created_at": "2025-01-16T00:40:05.458138+00:00"} {"global_step": 6769, "acc_step": 0, "speed/wps": 12889.48454542808, "speed/FLOPS": 202447066793799.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07501830905675888, "optim/lr": 0.0029923575855259218, "optim/total_tokens": 3548905472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 2.914689540863037, "created_at": "2025-01-16T00:40:15.635106+00:00"} {"global_step": 6770, "acc_step": 0, "speed/wps": 12892.113194280722, "speed/FLOPS": 202488353336172.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06085038185119629, "optim/lr": 0.0029923489451748326, "optim/total_tokens": 3549429760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.0716843605041504, "created_at": "2025-01-16T00:40:25.807297+00:00"} {"global_step": 6771, "acc_step": 0, "speed/wps": 12884.772846024782, "speed/FLOPS": 202373063080116.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08163507282733917, "optim/lr": 0.002992340299954693, "optim/total_tokens": 3549954048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.039496421813965, "created_at": "2025-01-16T00:40:35.986099+00:00"} {"global_step": 6772, "acc_step": 0, "speed/wps": 12884.656619863468, "speed/FLOPS": 202371237588542.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05710151046514511, "optim/lr": 0.00299233164986553, "optim/total_tokens": 3550478336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.995983600616455, "created_at": "2025-01-16T00:40:46.162558+00:00"} {"global_step": 6773, "acc_step": 0, "speed/wps": 12883.339597891476, "speed/FLOPS": 202350551948694.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07952658832073212, "optim/lr": 0.002992322994907373, "optim/total_tokens": 3551002624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414801, "loss/out": 2.935271739959717, "created_at": "2025-01-16T00:40:56.339809+00:00"} {"global_step": 6774, "acc_step": 0, "speed/wps": 12888.02532382795, "speed/FLOPS": 202424147713388.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0726415142416954, "optim/lr": 0.0029923143350802494, "optim/total_tokens": 3551526912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.092971086502075, "created_at": "2025-01-16T00:41:06.515671+00:00"} {"global_step": 6775, "acc_step": 0, "speed/wps": 12888.888050392587, "speed/FLOPS": 202437698019596.7, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06608317047357559, "optim/lr": 0.0029923056703841876, "optim/total_tokens": 3552051200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.054677724838257, "created_at": "2025-01-16T00:41:16.697970+00:00"} {"global_step": 6776, "acc_step": 0, "speed/wps": 12893.074000532722, "speed/FLOPS": 202503444118645.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08068902045488358, "optim/lr": 0.0029922970008192163, "optim/total_tokens": 3552575488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 3.0871076583862305, "created_at": "2025-01-16T00:41:26.869109+00:00"} {"global_step": 6777, "acc_step": 0, "speed/wps": 12893.557457697167, "speed/FLOPS": 202511037477753.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0639975443482399, "optim/lr": 0.002992288326385364, "optim/total_tokens": 3553099776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.0633668899536133, "created_at": "2025-01-16T00:41:37.042030+00:00"} {"global_step": 6778, "acc_step": 0, "speed/wps": 12889.528053613352, "speed/FLOPS": 202447750149635.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06514748185873032, "optim/lr": 0.0029922796470826583, "optim/total_tokens": 3553624064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.9674293994903564, "created_at": "2025-01-16T00:41:47.218209+00:00"} {"global_step": 6779, "acc_step": 0, "speed/wps": 12892.027375030339, "speed/FLOPS": 202487005426916.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06204395741224289, "optim/lr": 0.002992270962911128, "optim/total_tokens": 3554148352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.0160152912139893, "created_at": "2025-01-16T00:41:57.389457+00:00"} {"global_step": 6780, "acc_step": 0, "speed/wps": 12893.130850465352, "speed/FLOPS": 202504337024957.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0719103142619133, "optim/lr": 0.0029922622738708013, "optim/total_tokens": 3554672640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 2.995671272277832, "created_at": "2025-01-16T00:42:07.560748+00:00"} {"global_step": 6781, "acc_step": 0, "speed/wps": 12892.473058509268, "speed/FLOPS": 202494005498386.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05679188296198845, "optim/lr": 0.0029922535799617065, "optim/total_tokens": 3555196928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9025959968566895, "created_at": "2025-01-16T00:42:17.734178+00:00"} {"global_step": 6782, "acc_step": 0, "speed/wps": 12888.031901197295, "speed/FLOPS": 202424251020012.34, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07164805382490158, "optim/lr": 0.0029922448811838718, "optim/total_tokens": 3555721216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.9845516681671143, "created_at": "2025-01-16T00:42:27.908803+00:00"} {"global_step": 6783, "acc_step": 0, "speed/wps": 12895.668904950462, "speed/FLOPS": 202544200658298.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05001988261938095, "optim/lr": 0.0029922361775373257, "optim/total_tokens": 3556245504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373603, "loss/out": 2.950875759124756, "created_at": "2025-01-16T00:42:38.084835+00:00"} {"global_step": 6784, "acc_step": 0, "speed/wps": 12888.933408937657, "speed/FLOPS": 202438410437876.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05888485908508301, "optim/lr": 0.0029922274690220964, "optim/total_tokens": 3556769792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 2.9530258178710938, "created_at": "2025-01-16T00:42:48.258374+00:00"} {"global_step": 6785, "acc_step": 0, "speed/wps": 12887.920833898515, "speed/FLOPS": 202422506555462.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06283807754516602, "optim/lr": 0.0029922187556382135, "optim/total_tokens": 3557294080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.002751350402832, "created_at": "2025-01-16T00:42:58.435977+00:00"} {"global_step": 6786, "acc_step": 0, "speed/wps": 12891.293606802148, "speed/FLOPS": 202475480588589.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0728151798248291, "optim/lr": 0.002992210037385704, "optim/total_tokens": 3557818368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.110574722290039, "created_at": "2025-01-16T00:43:08.606896+00:00"} {"global_step": 6787, "acc_step": 0, "speed/wps": 12893.930109810119, "speed/FLOPS": 202516890491264.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07194051891565323, "optim/lr": 0.002992201314264597, "optim/total_tokens": 3558342656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 2.9420907497406006, "created_at": "2025-01-16T00:43:18.778784+00:00"} {"global_step": 6788, "acc_step": 0, "speed/wps": 12890.292395224802, "speed/FLOPS": 202459755184958.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08327076584100723, "optim/lr": 0.00299219258627492, "optim/total_tokens": 3558866944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.0061333179473877, "created_at": "2025-01-16T00:43:28.955954+00:00"} {"global_step": 6789, "acc_step": 0, "speed/wps": 12888.621091009945, "speed/FLOPS": 202433505055651.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09876259416341782, "optim/lr": 0.0029921838534167026, "optim/total_tokens": 3559391232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.046110153198242, "created_at": "2025-01-16T00:43:39.129879+00:00"} {"global_step": 6790, "acc_step": 0, "speed/wps": 12887.484059907289, "speed/FLOPS": 202415646419740.0, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06768352538347244, "optim/lr": 0.002992175115689973, "optim/total_tokens": 3559915520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410217, "loss/out": 3.075712203979492, "created_at": "2025-01-16T00:43:49.307420+00:00"} {"global_step": 6791, "acc_step": 0, "speed/wps": 12887.42090250877, "speed/FLOPS": 202414654446008.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06113305687904358, "optim/lr": 0.00299216637309476, "optim/total_tokens": 3560439808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.9860472679138184, "created_at": "2025-01-16T00:43:59.483663+00:00"} {"global_step": 6792, "acc_step": 0, "speed/wps": 12889.046065991672, "speed/FLOPS": 202440179871716.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06023016944527626, "optim/lr": 0.0029921576256310917, "optim/total_tokens": 3560964096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.168400526046753, "created_at": "2025-01-16T00:44:09.658452+00:00"} {"global_step": 6793, "acc_step": 0, "speed/wps": 12888.318857125203, "speed/FLOPS": 202428758057178.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06907204538583755, "optim/lr": 0.0029921488732989965, "optim/total_tokens": 3561488384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9853618144989014, "created_at": "2025-01-16T00:44:19.832273+00:00"} {"global_step": 6794, "acc_step": 0, "speed/wps": 12890.643432433768, "speed/FLOPS": 202465268706701.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07226040214300156, "optim/lr": 0.0029921401160985035, "optim/total_tokens": 3562012672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.9639084339141846, "created_at": "2025-01-16T00:44:30.005685+00:00"} {"global_step": 6795, "acc_step": 0, "speed/wps": 12893.96155023142, "speed/FLOPS": 202517384306284.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05908527970314026, "optim/lr": 0.0029921313540296405, "optim/total_tokens": 3562536960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.025142192840576, "created_at": "2025-01-16T00:44:40.175925+00:00"} {"global_step": 6796, "acc_step": 0, "speed/wps": 12886.316324025856, "speed/FLOPS": 202397305523084.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08011127263307571, "optim/lr": 0.002992122587092437, "optim/total_tokens": 3563061248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9037063121795654, "created_at": "2025-01-16T00:44:50.354387+00:00"} {"global_step": 6797, "acc_step": 0, "speed/wps": 12885.081841883488, "speed/FLOPS": 202377916284682.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06019468232989311, "optim/lr": 0.002992113815286921, "optim/total_tokens": 3563585536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458442, "loss/out": 3.0222525596618652, "created_at": "2025-01-16T00:45:00.533210+00:00"} {"global_step": 6798, "acc_step": 0, "speed/wps": 12891.975950633752, "speed/FLOPS": 202486197736103.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054076217114925385, "optim/lr": 0.0029921050386131215, "optim/total_tokens": 3564109824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 2.943392038345337, "created_at": "2025-01-16T00:45:10.710521+00:00"} {"global_step": 6799, "acc_step": 0, "speed/wps": 12893.997463259442, "speed/FLOPS": 202517948369739.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0638391301035881, "optim/lr": 0.0029920962570710668, "optim/total_tokens": 3564634112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9990270137786865, "created_at": "2025-01-16T00:45:20.879374+00:00"} {"global_step": 6800, "acc_step": 0, "speed/wps": 12893.102679974209, "speed/FLOPS": 202503894568684.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07417742908000946, "optim/lr": 0.0029920874706607856, "optim/total_tokens": 3565158400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.86842679977417, "created_at": "2025-01-16T00:45:31.049940+00:00"} {"global_step": 6801, "acc_step": 0, "speed/wps": 12888.345825256516, "speed/FLOPS": 202429181628739.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05257813259959221, "optim/lr": 0.002992078679382306, "optim/total_tokens": 3565682688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384291, "loss/out": 3.054645538330078, "created_at": "2025-01-16T00:45:41.225423+00:00"} {"global_step": 6802, "acc_step": 0, "speed/wps": 12890.105327466827, "speed/FLOPS": 202456817028761.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0590582937002182, "optim/lr": 0.0029920698832356578, "optim/total_tokens": 3566206976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.949705123901367, "created_at": "2025-01-16T00:45:51.397537+00:00"} {"global_step": 6803, "acc_step": 0, "speed/wps": 12891.134339950482, "speed/FLOPS": 202472979083828.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05693496763706207, "optim/lr": 0.0029920610822208697, "optim/total_tokens": 3566731264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.040348529815674, "created_at": "2025-01-16T00:46:01.595396+00:00"} {"global_step": 6804, "acc_step": 0, "speed/wps": 12891.598595858222, "speed/FLOPS": 202480270860814.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06265299767255783, "optim/lr": 0.0029920522763379693, "optim/total_tokens": 3567255552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.001255512237549, "created_at": "2025-01-16T00:46:11.766664+00:00"} {"global_step": 6805, "acc_step": 0, "speed/wps": 12890.825700779993, "speed/FLOPS": 202468131481541.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060422834008932114, "optim/lr": 0.0029920434655869863, "optim/total_tokens": 3567779840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 3.0271475315093994, "created_at": "2025-01-16T00:46:21.940729+00:00"} {"global_step": 6806, "acc_step": 0, "speed/wps": 12894.14305004522, "speed/FLOPS": 202520235010267.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057074952870607376, "optim/lr": 0.002992034649967949, "optim/total_tokens": 3568304128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.0542664527893066, "created_at": "2025-01-16T00:46:32.109927+00:00"} {"global_step": 6807, "acc_step": 0, "speed/wps": 12889.199250401714, "speed/FLOPS": 202442585843373.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06505844742059708, "optim/lr": 0.0029920258294808865, "optim/total_tokens": 3568828416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.876147985458374, "created_at": "2025-01-16T00:46:42.285022+00:00"} {"global_step": 6808, "acc_step": 0, "speed/wps": 12889.1181066301, "speed/FLOPS": 202441311368937.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059007398784160614, "optim/lr": 0.002992017004125827, "optim/total_tokens": 3569352704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 2.944227695465088, "created_at": "2025-01-16T00:46:52.460648+00:00"} {"global_step": 6809, "acc_step": 0, "speed/wps": 12890.983701313096, "speed/FLOPS": 202470613097029.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05300162360072136, "optim/lr": 0.0029920081739027997, "optim/total_tokens": 3569876992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9980828762054443, "created_at": "2025-01-16T00:47:02.634354+00:00"} {"global_step": 6810, "acc_step": 0, "speed/wps": 12896.391971877101, "speed/FLOPS": 202555557417979.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06131719797849655, "optim/lr": 0.002991999338811833, "optim/total_tokens": 3570401280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.081897735595703, "created_at": "2025-01-16T00:47:12.801775+00:00"} {"global_step": 6811, "acc_step": 0, "speed/wps": 12894.326765319884, "speed/FLOPS": 202523120511107.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05221269652247429, "optim/lr": 0.0029919904988529567, "optim/total_tokens": 3570925568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490877, "loss/out": 3.1750988960266113, "created_at": "2025-01-16T00:47:22.971531+00:00"} {"global_step": 6812, "acc_step": 0, "speed/wps": 12892.517056927254, "speed/FLOPS": 202494696554001.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060213714838027954, "optim/lr": 0.002991981654026199, "optim/total_tokens": 3571449856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 2.931326389312744, "created_at": "2025-01-16T00:47:33.144021+00:00"} {"global_step": 6813, "acc_step": 0, "speed/wps": 12891.92350750125, "speed/FLOPS": 202485374044643.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05991300195455551, "optim/lr": 0.002991972804331588, "optim/total_tokens": 3571974144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.1016640663146973, "created_at": "2025-01-16T00:47:43.315662+00:00"} {"global_step": 6814, "acc_step": 0, "speed/wps": 12892.393296641207, "speed/FLOPS": 202492752728644.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061914823949337006, "optim/lr": 0.002991963949769154, "optim/total_tokens": 3572498432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.038705587387085, "created_at": "2025-01-16T00:47:53.485866+00:00"} {"global_step": 6815, "acc_step": 0, "speed/wps": 12892.782358847646, "speed/FLOPS": 202498863485224.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0629267618060112, "optim/lr": 0.0029919550903389253, "optim/total_tokens": 3573022720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0498476028442383, "created_at": "2025-01-16T00:48:03.656987+00:00"} {"global_step": 6816, "acc_step": 0, "speed/wps": 12890.526933862779, "speed/FLOPS": 202463438936557.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061213310807943344, "optim/lr": 0.0029919462260409304, "optim/total_tokens": 3573547008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 2.891719102859497, "created_at": "2025-01-16T00:48:13.831023+00:00"} {"global_step": 6817, "acc_step": 0, "speed/wps": 12886.892676559199, "speed/FLOPS": 202406357931612.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06319627165794373, "optim/lr": 0.0029919373568751987, "optim/total_tokens": 3574071296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.0177812576293945, "created_at": "2025-01-16T00:48:24.005508+00:00"} {"global_step": 6818, "acc_step": 0, "speed/wps": 12893.154114541116, "speed/FLOPS": 202504702419235.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05610910803079605, "optim/lr": 0.0029919284828417592, "optim/total_tokens": 3574595584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 3.039356231689453, "created_at": "2025-01-16T00:48:34.175309+00:00"} {"global_step": 6819, "acc_step": 0, "speed/wps": 12887.598497242972, "speed/FLOPS": 202417443815350.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05409662798047066, "optim/lr": 0.0029919196039406406, "optim/total_tokens": 3575119872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0425167083740234, "created_at": "2025-01-16T00:48:44.352708+00:00"} {"global_step": 6820, "acc_step": 0, "speed/wps": 12891.157013509803, "speed/FLOPS": 202473335203234.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061330024152994156, "optim/lr": 0.0029919107201718718, "optim/total_tokens": 3575644160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.053077220916748, "created_at": "2025-01-16T00:48:54.523852+00:00"} {"global_step": 6821, "acc_step": 0, "speed/wps": 12891.246733195496, "speed/FLOPS": 202474744374186.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06617993116378784, "optim/lr": 0.002991901831535482, "optim/total_tokens": 3576168448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 3.2207422256469727, "created_at": "2025-01-16T00:49:04.696499+00:00"} {"global_step": 6822, "acc_step": 0, "speed/wps": 12891.227310080052, "speed/FLOPS": 202474439307468.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05276791378855705, "optim/lr": 0.0029918929380315, "optim/total_tokens": 3576692736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.0218586921691895, "created_at": "2025-01-16T00:49:14.867613+00:00"} {"global_step": 6823, "acc_step": 0, "speed/wps": 12891.696806397813, "speed/FLOPS": 202481813392293.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06540136784315109, "optim/lr": 0.0029918840396599552, "optim/total_tokens": 3577217024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.006350517272949, "created_at": "2025-01-16T00:49:25.050041+00:00"} {"global_step": 6824, "acc_step": 0, "speed/wps": 12894.61315336894, "speed/FLOPS": 202527618628956.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07962922006845474, "optim/lr": 0.0029918751364208756, "optim/total_tokens": 3577741312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.957383155822754, "created_at": "2025-01-16T00:49:35.218839+00:00"} {"global_step": 6825, "acc_step": 0, "speed/wps": 12890.308107137698, "speed/FLOPS": 202460001962141.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05685844644904137, "optim/lr": 0.002991866228314292, "optim/total_tokens": 3578265600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 3.04207706451416, "created_at": "2025-01-16T00:49:45.390659+00:00"} {"global_step": 6826, "acc_step": 0, "speed/wps": 12892.8926477437, "speed/FLOPS": 202500595723889.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06466472148895264, "optim/lr": 0.002991857315340232, "optim/total_tokens": 3578789888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.859189510345459, "created_at": "2025-01-16T00:49:55.562009+00:00"} {"global_step": 6827, "acc_step": 0, "speed/wps": 12888.748715358128, "speed/FLOPS": 202435509571414.12, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06197332218289375, "optim/lr": 0.002991848397498725, "optim/total_tokens": 3579314176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.000730037689209, "created_at": "2025-01-16T00:50:05.735194+00:00"} {"global_step": 6828, "acc_step": 0, "speed/wps": 12888.975642117794, "speed/FLOPS": 202439073768004.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06125505268573761, "optim/lr": 0.0029918394747898004, "optim/total_tokens": 3579838464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369775, "loss/out": 3.0349833965301514, "created_at": "2025-01-16T00:50:15.908938+00:00"} {"global_step": 6829, "acc_step": 0, "speed/wps": 12891.746845731463, "speed/FLOPS": 202482599328790.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06367573142051697, "optim/lr": 0.002991830547213487, "optim/total_tokens": 3580362752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.0218656063079834, "created_at": "2025-01-16T00:50:26.080114+00:00"} {"global_step": 6830, "acc_step": 0, "speed/wps": 12890.514842548659, "speed/FLOPS": 202463249025854.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05760713294148445, "optim/lr": 0.0029918216147698145, "optim/total_tokens": 3580887040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.075087785720825, "created_at": "2025-01-16T00:50:36.254051+00:00"} {"global_step": 6831, "acc_step": 0, "speed/wps": 12887.053500116937, "speed/FLOPS": 202408883886581.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058522965759038925, "optim/lr": 0.002991812677458811, "optim/total_tokens": 3581411328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 3.0569348335266113, "created_at": "2025-01-16T00:50:46.431830+00:00"} {"global_step": 6832, "acc_step": 0, "speed/wps": 12881.006534904665, "speed/FLOPS": 202313907988521.66, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058854229748249054, "optim/lr": 0.002991803735280507, "optim/total_tokens": 3581935616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299645, "loss/out": 3.019901752471924, "created_at": "2025-01-16T00:50:56.612175+00:00"} {"global_step": 6833, "acc_step": 0, "speed/wps": 12894.147930165558, "speed/FLOPS": 202520311659263.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06575459241867065, "optim/lr": 0.0029917947882349303, "optim/total_tokens": 3582459904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.148676872253418, "created_at": "2025-01-16T00:51:06.781160+00:00"} {"global_step": 6834, "acc_step": 0, "speed/wps": 12893.008678829827, "speed/FLOPS": 202502418151540.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050942692905664444, "optim/lr": 0.0029917858363221112, "optim/total_tokens": 3582984192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0885210037231445, "created_at": "2025-01-16T00:51:16.952783+00:00"} {"global_step": 6835, "acc_step": 0, "speed/wps": 12886.807829495265, "speed/FLOPS": 202405025291881.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07776306569576263, "optim/lr": 0.0029917768795420787, "optim/total_tokens": 3583508480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.9503493309020996, "created_at": "2025-01-16T00:51:27.128444+00:00"} {"global_step": 6836, "acc_step": 0, "speed/wps": 12891.21068316512, "speed/FLOPS": 202474178158921.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07161612808704376, "optim/lr": 0.002991767917894861, "optim/total_tokens": 3584032768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.0703859329223633, "created_at": "2025-01-16T00:51:37.302920+00:00"} {"global_step": 6837, "acc_step": 0, "speed/wps": 12890.625445682037, "speed/FLOPS": 202464986200048.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05607084184885025, "optim/lr": 0.0029917589513804885, "optim/total_tokens": 3584557056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.92073130607605, "created_at": "2025-01-16T00:51:47.478777+00:00"} {"global_step": 6838, "acc_step": 0, "speed/wps": 12890.619844461473, "speed/FLOPS": 202464898225182.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05756387114524841, "optim/lr": 0.0029917499799989905, "optim/total_tokens": 3585081344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 3.066073417663574, "created_at": "2025-01-16T00:51:57.652258+00:00"} {"global_step": 6839, "acc_step": 0, "speed/wps": 12890.881933431172, "speed/FLOPS": 202469014692600.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07405362278223038, "optim/lr": 0.0029917410037503954, "optim/total_tokens": 3585605632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 2.8771257400512695, "created_at": "2025-01-16T00:52:07.825402+00:00"} {"global_step": 6840, "acc_step": 0, "speed/wps": 12892.17068055529, "speed/FLOPS": 202489256237108.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07712417840957642, "optim/lr": 0.002991732022634733, "optim/total_tokens": 3586129920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0549545288085938, "created_at": "2025-01-16T00:52:17.996899+00:00"} {"global_step": 6841, "acc_step": 0, "speed/wps": 12891.38377120235, "speed/FLOPS": 202476896744393.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.076125368475914, "optim/lr": 0.002991723036652033, "optim/total_tokens": 3586654208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.066311836242676, "created_at": "2025-01-16T00:52:28.167919+00:00"} {"global_step": 6842, "acc_step": 0, "speed/wps": 12888.41812245282, "speed/FLOPS": 202430317155554.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06586487591266632, "optim/lr": 0.0029917140458023237, "optim/total_tokens": 3587178496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 3.0496296882629395, "created_at": "2025-01-16T00:52:38.342710+00:00"} {"global_step": 6843, "acc_step": 0, "speed/wps": 12897.284935419573, "speed/FLOPS": 202569582637474.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0626109391450882, "optim/lr": 0.0029917050500856347, "optim/total_tokens": 3587702784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.1429574489593506, "created_at": "2025-01-16T00:52:48.509318+00:00"} {"global_step": 6844, "acc_step": 0, "speed/wps": 12888.398938016353, "speed/FLOPS": 202430015837617.1, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06241539865732193, "optim/lr": 0.002991696049501996, "optim/total_tokens": 3588227072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.0229341983795166, "created_at": "2025-01-16T00:52:58.683770+00:00"} {"global_step": 6845, "acc_step": 0, "speed/wps": 12893.78962385531, "speed/FLOPS": 202514683966296.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05780673399567604, "optim/lr": 0.002991687044051437, "optim/total_tokens": 3588751360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 2.9507768154144287, "created_at": "2025-01-16T00:53:08.854575+00:00"} {"global_step": 6846, "acc_step": 0, "speed/wps": 12893.868768648776, "speed/FLOPS": 202515927044034.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06662920862436295, "optim/lr": 0.0029916780337339863, "optim/total_tokens": 3589275648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9517300128936768, "created_at": "2025-01-16T00:53:19.024467+00:00"} {"global_step": 6847, "acc_step": 0, "speed/wps": 12890.757481934408, "speed/FLOPS": 202467060010829.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058251187205314636, "optim/lr": 0.0029916690185496737, "optim/total_tokens": 3589799936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.1123464107513428, "created_at": "2025-01-16T00:53:29.195859+00:00"} {"global_step": 6848, "acc_step": 0, "speed/wps": 12893.895315843856, "speed/FLOPS": 202516344004212.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056172437965869904, "optim/lr": 0.002991659998498529, "optim/total_tokens": 3590324224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 2.8187789916992188, "created_at": "2025-01-16T00:53:39.366900+00:00"} {"global_step": 6849, "acc_step": 0, "speed/wps": 12889.58737259107, "speed/FLOPS": 202448681835693.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06880529969930649, "optim/lr": 0.002991650973580581, "optim/total_tokens": 3590848512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 3.047844171524048, "created_at": "2025-01-16T00:53:49.539294+00:00"} {"global_step": 6850, "acc_step": 0, "speed/wps": 12892.243073813812, "speed/FLOPS": 202490393272712.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05221409723162651, "optim/lr": 0.002991641943795859, "optim/total_tokens": 3591372800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 3.039174795150757, "created_at": "2025-01-16T00:53:59.712662+00:00"} {"global_step": 6851, "acc_step": 0, "speed/wps": 12890.665304929904, "speed/FLOPS": 202465612244308.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052768200635910034, "optim/lr": 0.0029916329091443927, "optim/total_tokens": 3591897088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.0089988708496094, "created_at": "2025-01-16T00:54:09.887520+00:00"} {"global_step": 6852, "acc_step": 0, "speed/wps": 12893.002718836147, "speed/FLOPS": 202502324541649.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06852871924638748, "optim/lr": 0.0029916238696262123, "optim/total_tokens": 3592421376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 2.894813060760498, "created_at": "2025-01-16T00:54:20.058254+00:00"} {"global_step": 6853, "acc_step": 0, "speed/wps": 12887.022081575156, "speed/FLOPS": 202408390415209.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07878720760345459, "optim/lr": 0.0029916148252413467, "optim/total_tokens": 3592945664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9575719833374023, "created_at": "2025-01-16T00:54:30.236106+00:00"} {"global_step": 6854, "acc_step": 0, "speed/wps": 12888.734646752271, "speed/FLOPS": 202435288604626.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0665208175778389, "optim/lr": 0.002991605775989825, "optim/total_tokens": 3593469952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9324610233306885, "created_at": "2025-01-16T00:54:40.409657+00:00"} {"global_step": 6855, "acc_step": 0, "speed/wps": 12890.75295977673, "speed/FLOPS": 202466988984128.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05653400346636772, "optim/lr": 0.0029915967218716775, "optim/total_tokens": 3593994240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376285, "loss/out": 2.9046833515167236, "created_at": "2025-01-16T00:54:50.584299+00:00"} {"global_step": 6856, "acc_step": 0, "speed/wps": 12890.541027633879, "speed/FLOPS": 202463660298599.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06667956709861755, "optim/lr": 0.0029915876628869337, "optim/total_tokens": 3594518528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0451912879943848, "created_at": "2025-01-16T00:55:00.756760+00:00"} {"global_step": 6857, "acc_step": 0, "speed/wps": 12890.317817893516, "speed/FLOPS": 202460154482905.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056357137858867645, "optim/lr": 0.002991578599035622, "optim/total_tokens": 3595042816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.988185405731201, "created_at": "2025-01-16T00:55:10.929191+00:00"} {"global_step": 6858, "acc_step": 0, "speed/wps": 12887.188503056715, "speed/FLOPS": 202411004293264.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07235205173492432, "optim/lr": 0.0029915695303177735, "optim/total_tokens": 3595567104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.9186882972717285, "created_at": "2025-01-16T00:55:21.105446+00:00"} {"global_step": 6859, "acc_step": 0, "speed/wps": 12887.480350788503, "speed/FLOPS": 202415588162932.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0818769708275795, "optim/lr": 0.002991560456733417, "optim/total_tokens": 3596091392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 3.0872673988342285, "created_at": "2025-01-16T00:55:31.279465+00:00"} {"global_step": 6860, "acc_step": 0, "speed/wps": 12896.273192630122, "speed/FLOPS": 202553691826686.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06454236060380936, "optim/lr": 0.0029915513782825818, "optim/total_tokens": 3596615680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9763195514678955, "created_at": "2025-01-16T00:55:41.449207+00:00"} {"global_step": 6861, "acc_step": 0, "speed/wps": 12890.047382651068, "speed/FLOPS": 202455906925805.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06210075691342354, "optim/lr": 0.002991542294965298, "optim/total_tokens": 3597139968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.093338966369629, "created_at": "2025-01-16T00:55:51.623074+00:00"} {"global_step": 6862, "acc_step": 0, "speed/wps": 12878.721383005435, "speed/FLOPS": 202278016537816.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.003, "optim/grad_norm": 0.06971771270036697, "optim/lr": 0.0029915332067815955, "optim/total_tokens": 3597664256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371206, "loss/out": 2.932863712310791, "created_at": "2025-01-16T00:56:01.809610+00:00"} {"global_step": 6863, "acc_step": 0, "speed/wps": 12887.795110688796, "speed/FLOPS": 202420531899692.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05237189680337906, "optim/lr": 0.002991524113731503, "optim/total_tokens": 3598188544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.0042033195495605, "created_at": "2025-01-16T00:56:11.983279+00:00"} {"global_step": 6864, "acc_step": 0, "speed/wps": 12889.107180670555, "speed/FLOPS": 202441139761728.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08762810379266739, "optim/lr": 0.0029915150158150517, "optim/total_tokens": 3598712832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9666152000427246, "created_at": "2025-01-16T00:56:22.161464+00:00"} {"global_step": 6865, "acc_step": 0, "speed/wps": 12891.822479385224, "speed/FLOPS": 202483787259254.06, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09655903279781342, "optim/lr": 0.0029915059130322702, "optim/total_tokens": 3599237120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.0445127487182617, "created_at": "2025-01-16T00:56:32.332945+00:00"} {"global_step": 6866, "acc_step": 0, "speed/wps": 12892.802087828557, "speed/FLOPS": 202499173355979.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05921761691570282, "optim/lr": 0.002991496805383188, "optim/total_tokens": 3599761408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.0252556800842285, "created_at": "2025-01-16T00:56:42.504701+00:00"} {"global_step": 6867, "acc_step": 0, "speed/wps": 12889.321407695172, "speed/FLOPS": 202444504491529.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08084987103939056, "optim/lr": 0.002991487692867835, "optim/total_tokens": 3600285696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 3.0326733589172363, "created_at": "2025-01-16T00:56:52.677609+00:00"} {"global_step": 6868, "acc_step": 0, "speed/wps": 12894.00347024412, "speed/FLOPS": 202518042717688.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07776976376771927, "optim/lr": 0.0029914785754862412, "optim/total_tokens": 3600809984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.9621493816375732, "created_at": "2025-01-16T00:57:02.846802+00:00"} {"global_step": 6869, "acc_step": 0, "speed/wps": 12887.707828868553, "speed/FLOPS": 202419161018767.44, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05635900795459747, "optim/lr": 0.002991469453238437, "optim/total_tokens": 3601334272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419023, "loss/out": 3.0429506301879883, "created_at": "2025-01-16T00:57:13.022842+00:00"} {"global_step": 6870, "acc_step": 0, "speed/wps": 12891.283197894163, "speed/FLOPS": 202475317102386.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07040572911500931, "optim/lr": 0.0029914603261244504, "optim/total_tokens": 3601858560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 3.043820858001709, "created_at": "2025-01-16T00:57:23.197329+00:00"} {"global_step": 6871, "acc_step": 0, "speed/wps": 12890.932013669219, "speed/FLOPS": 202469801271556.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07384970784187317, "optim/lr": 0.0029914511941443126, "optim/total_tokens": 3602382848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.066959857940674, "created_at": "2025-01-16T00:57:33.368973+00:00"} {"global_step": 6872, "acc_step": 0, "speed/wps": 12887.430178174469, "speed/FLOPS": 202414800133084.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06202051043510437, "optim/lr": 0.0029914420572980534, "optim/total_tokens": 3602907136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9495954513549805, "created_at": "2025-01-16T00:57:43.544158+00:00"} {"global_step": 6873, "acc_step": 0, "speed/wps": 12894.518501688313, "speed/FLOPS": 202526131994246.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0574534609913826, "optim/lr": 0.002991432915585702, "optim/total_tokens": 3603431424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 2.9898576736450195, "created_at": "2025-01-16T00:57:53.716228+00:00"} {"global_step": 6874, "acc_step": 0, "speed/wps": 12889.667957982369, "speed/FLOPS": 202449947539999.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07911205291748047, "optim/lr": 0.002991423769007288, "optim/total_tokens": 3603955712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.0636496543884277, "created_at": "2025-01-16T00:58:03.889805+00:00"} {"global_step": 6875, "acc_step": 0, "speed/wps": 12892.599102129981, "speed/FLOPS": 202495985186651.97, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05568474903702736, "optim/lr": 0.002991414617562842, "optim/total_tokens": 3604480000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8777318000793457, "created_at": "2025-01-16T00:58:14.061491+00:00"} {"global_step": 6876, "acc_step": 0, "speed/wps": 12890.442415567883, "speed/FLOPS": 202462111460596.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06135149672627449, "optim/lr": 0.002991405461252394, "optim/total_tokens": 3605004288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 2.9972524642944336, "created_at": "2025-01-16T00:58:24.234216+00:00"} {"global_step": 6877, "acc_step": 0, "speed/wps": 12893.030020754886, "speed/FLOPS": 202502753355800.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07041479647159576, "optim/lr": 0.002991396300075972, "optim/total_tokens": 3605528576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.003891944885254, "created_at": "2025-01-16T00:58:34.413209+00:00"} {"global_step": 6878, "acc_step": 0, "speed/wps": 12889.22031175743, "speed/FLOPS": 202442916640906.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057772617787122726, "optim/lr": 0.002991387134033608, "optim/total_tokens": 3606052864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9129388332366943, "created_at": "2025-01-16T00:58:44.587138+00:00"} {"global_step": 6879, "acc_step": 0, "speed/wps": 12890.087034080749, "speed/FLOPS": 202456529705995.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08396067470312119, "optim/lr": 0.0029913779631253316, "optim/total_tokens": 3606577152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.111708879470825, "created_at": "2025-01-16T00:58:54.759368+00:00"} {"global_step": 6880, "acc_step": 0, "speed/wps": 12893.77295961696, "speed/FLOPS": 202514422231534.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06981191039085388, "optim/lr": 0.0029913687873511724, "optim/total_tokens": 3607101440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 3.148646593093872, "created_at": "2025-01-16T00:59:04.931264+00:00"} {"global_step": 6881, "acc_step": 0, "speed/wps": 12888.285055353965, "speed/FLOPS": 202428227153913.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0567425936460495, "optim/lr": 0.0029913596067111603, "optim/total_tokens": 3607625728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.013152599334717, "created_at": "2025-01-16T00:59:15.106814+00:00"} {"global_step": 6882, "acc_step": 0, "speed/wps": 12884.873258793084, "speed/FLOPS": 202374640200624.72, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07203186303377151, "optim/lr": 0.0029913504212053244, "optim/total_tokens": 3608150016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.1620631217956543, "created_at": "2025-01-16T00:59:25.283159+00:00"} {"global_step": 6883, "acc_step": 0, "speed/wps": 12887.346059232115, "speed/FLOPS": 202413478929503.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0590028278529644, "optim/lr": 0.0029913412308336957, "optim/total_tokens": 3608674304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 3.007519483566284, "created_at": "2025-01-16T00:59:35.457834+00:00"} {"global_step": 6884, "acc_step": 0, "speed/wps": 12891.306495450588, "speed/FLOPS": 202475683022524.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07612462341785431, "optim/lr": 0.0029913320355963047, "optim/total_tokens": 3609198592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334467, "loss/out": 2.993396043777466, "created_at": "2025-01-16T00:59:45.634277+00:00"} {"global_step": 6885, "acc_step": 0, "speed/wps": 12890.268437819082, "speed/FLOPS": 202459378900980.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07075052708387375, "optim/lr": 0.0029913228354931805, "optim/total_tokens": 3609722880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.8940515518188477, "created_at": "2025-01-16T00:59:55.806215+00:00"} {"global_step": 6886, "acc_step": 0, "speed/wps": 12888.332934265578, "speed/FLOPS": 202428979158012.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06847196072340012, "optim/lr": 0.002991313630524353, "optim/total_tokens": 3610247168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.024190902709961, "created_at": "2025-01-16T01:00:05.982008+00:00"} {"global_step": 6887, "acc_step": 0, "speed/wps": 12888.32146671099, "speed/FLOPS": 202428799044309.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048762086778879166, "optim/lr": 0.002991304420689852, "optim/total_tokens": 3610771456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463004, "loss/out": 3.0571582317352295, "created_at": "2025-01-16T01:00:16.156469+00:00"} {"global_step": 6888, "acc_step": 0, "speed/wps": 12888.550996366946, "speed/FLOPS": 202432404122963.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07052149623632431, "optim/lr": 0.002991295205989709, "optim/total_tokens": 3611295744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9882521629333496, "created_at": "2025-01-16T01:00:26.333015+00:00"} {"global_step": 6889, "acc_step": 0, "speed/wps": 12888.008177614782, "speed/FLOPS": 202423878408549.38, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06146056950092316, "optim/lr": 0.0029912859864239523, "optim/total_tokens": 3611820032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.0567474365234375, "created_at": "2025-01-16T01:00:36.511784+00:00"} {"global_step": 6890, "acc_step": 0, "speed/wps": 12894.219880177696, "speed/FLOPS": 202521441733073.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0610387958586216, "optim/lr": 0.0029912767619926133, "optim/total_tokens": 3612344320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448320, "loss/out": 2.941875457763672, "created_at": "2025-01-16T01:00:46.686181+00:00"} {"global_step": 6891, "acc_step": 0, "speed/wps": 12888.106285067239, "speed/FLOPS": 202425419320903.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06726086139678955, "optim/lr": 0.0029912675326957214, "optim/total_tokens": 3612868608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 3.0704636573791504, "created_at": "2025-01-16T01:00:56.863087+00:00"} {"global_step": 6892, "acc_step": 0, "speed/wps": 12890.809438565839, "speed/FLOPS": 202467876061121.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05582961440086365, "optim/lr": 0.0029912582985333076, "optim/total_tokens": 3613392896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.0266950130462646, "created_at": "2025-01-16T01:01:07.041336+00:00"} {"global_step": 6893, "acc_step": 0, "speed/wps": 12890.393858635147, "speed/FLOPS": 202461348807243.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05610241740942001, "optim/lr": 0.0029912490595054007, "optim/total_tokens": 3613917184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.049290180206299, "created_at": "2025-01-16T01:01:17.214230+00:00"} {"global_step": 6894, "acc_step": 0, "speed/wps": 12894.628016917348, "speed/FLOPS": 202527852081408.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056539978832006454, "optim/lr": 0.0029912398156120316, "optim/total_tokens": 3614441472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392360, "loss/out": 3.020134449005127, "created_at": "2025-01-16T01:01:27.382777+00:00"} {"global_step": 6895, "acc_step": 0, "speed/wps": 12888.226318426372, "speed/FLOPS": 202427304609752.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05931773781776428, "optim/lr": 0.0029912305668532303, "optim/total_tokens": 3614965760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9661121368408203, "created_at": "2025-01-16T01:01:37.556101+00:00"} {"global_step": 6896, "acc_step": 0, "speed/wps": 12891.444716755106, "speed/FLOPS": 202477853978047.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061759866774082184, "optim/lr": 0.002991221313229027, "optim/total_tokens": 3615490048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 3.0645980834960938, "created_at": "2025-01-16T01:01:47.727423+00:00"} {"global_step": 6897, "acc_step": 0, "speed/wps": 12889.288694943363, "speed/FLOPS": 202443990692809.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05765213817358017, "optim/lr": 0.002991212054739452, "optim/total_tokens": 3616014336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345973, "loss/out": 2.971792697906494, "created_at": "2025-01-16T01:01:57.901927+00:00"} {"global_step": 6898, "acc_step": 0, "speed/wps": 12891.56499179953, "speed/FLOPS": 202479743062895.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057434532791376114, "optim/lr": 0.0029912027913845357, "optim/total_tokens": 3616538624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.954073905944824, "created_at": "2025-01-16T01:02:08.075595+00:00"} {"global_step": 6899, "acc_step": 0, "speed/wps": 12893.226601526718, "speed/FLOPS": 202505840926952.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06574856489896774, "optim/lr": 0.002991193523164307, "optim/total_tokens": 3617062912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8738369941711426, "created_at": "2025-01-16T01:02:18.251041+00:00"} {"global_step": 6900, "acc_step": 0, "speed/wps": 12887.984397469405, "speed/FLOPS": 202423504908688.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07045523822307587, "optim/lr": 0.0029911842500787984, "optim/total_tokens": 3617587200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9936418533325195, "created_at": "2025-01-16T01:02:28.428113+00:00"} {"global_step": 6901, "acc_step": 0, "speed/wps": 12890.069013463875, "speed/FLOPS": 202456246667444.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058889515697956085, "optim/lr": 0.0029911749721280387, "optim/total_tokens": 3618111488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 3.1195178031921387, "created_at": "2025-01-16T01:02:38.600119+00:00"} {"global_step": 6902, "acc_step": 0, "speed/wps": 12888.224984736327, "speed/FLOPS": 202427283662317.44, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08710929751396179, "optim/lr": 0.0029911656893120584, "optim/total_tokens": 3618635776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.047454595565796, "created_at": "2025-01-16T01:02:48.773594+00:00"} {"global_step": 6903, "acc_step": 0, "speed/wps": 12887.247406356715, "speed/FLOPS": 202411929450532.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05489465594291687, "optim/lr": 0.0029911564016308874, "optim/total_tokens": 3619160064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8697104454040527, "created_at": "2025-01-16T01:02:58.950293+00:00"} {"global_step": 6904, "acc_step": 0, "speed/wps": 12891.948021376815, "speed/FLOPS": 202485759068744.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0662323534488678, "optim/lr": 0.002991147109084557, "optim/total_tokens": 3619684352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480757, "loss/out": 3.0093436241149902, "created_at": "2025-01-16T01:03:09.125820+00:00"} {"global_step": 6905, "acc_step": 0, "speed/wps": 12889.921701768933, "speed/FLOPS": 202453932934848.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06318400055170059, "optim/lr": 0.0029911378116730964, "optim/total_tokens": 3620208640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 2.9682984352111816, "created_at": "2025-01-16T01:03:19.301649+00:00"} {"global_step": 6906, "acc_step": 0, "speed/wps": 12894.372475219347, "speed/FLOPS": 202523838447890.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05509178712964058, "optim/lr": 0.0029911285093965367, "optim/total_tokens": 3620732928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.0060911178588867, "created_at": "2025-01-16T01:03:29.470270+00:00"} {"global_step": 6907, "acc_step": 0, "speed/wps": 12888.035248305467, "speed/FLOPS": 202424303590945.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061459362506866455, "optim/lr": 0.0029911192022549077, "optim/total_tokens": 3621257216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.0306644439697266, "created_at": "2025-01-16T01:03:39.644159+00:00"} {"global_step": 6908, "acc_step": 0, "speed/wps": 12889.548621414535, "speed/FLOPS": 202448073195216.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06379420310258865, "optim/lr": 0.0029911098902482406, "optim/total_tokens": 3621781504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 3.047208070755005, "created_at": "2025-01-16T01:03:49.816684+00:00"} {"global_step": 6909, "acc_step": 0, "speed/wps": 12886.665806619867, "speed/FLOPS": 202402794627463.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05929247662425041, "optim/lr": 0.002991100573376565, "optim/total_tokens": 3622305792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0519556999206543, "created_at": "2025-01-16T01:03:59.994733+00:00"} {"global_step": 6910, "acc_step": 0, "speed/wps": 12893.607621331745, "speed/FLOPS": 202511825366566.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0628855898976326, "optim/lr": 0.0029910912516399116, "optim/total_tokens": 3622830080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8980212211608887, "created_at": "2025-01-16T01:04:10.167590+00:00"} {"global_step": 6911, "acc_step": 0, "speed/wps": 12884.975351784486, "speed/FLOPS": 202376243711345.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05891751870512962, "optim/lr": 0.0029910819250383107, "optim/total_tokens": 3623354368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.029448986053467, "created_at": "2025-01-16T01:04:20.343686+00:00"} {"global_step": 6912, "acc_step": 0, "speed/wps": 12891.488453089929, "speed/FLOPS": 202478540917286.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06083866208791733, "optim/lr": 0.002991072593571793, "optim/total_tokens": 3623878656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 3.034015655517578, "created_at": "2025-01-16T01:04:30.517479+00:00"} {"global_step": 6913, "acc_step": 0, "speed/wps": 12891.913503816697, "speed/FLOPS": 202485216923031.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0637085810303688, "optim/lr": 0.0029910632572403887, "optim/total_tokens": 3624402944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.9936256408691406, "created_at": "2025-01-16T01:04:40.689149+00:00"} {"global_step": 6914, "acc_step": 0, "speed/wps": 12890.389414655685, "speed/FLOPS": 202461279008439.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08534061908721924, "optim/lr": 0.002991053916044128, "optim/total_tokens": 3624927232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.0357062816619873, "created_at": "2025-01-16T01:04:50.861093+00:00"} {"global_step": 6915, "acc_step": 0, "speed/wps": 12881.423569492865, "speed/FLOPS": 202320458089794.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06223228573799133, "optim/lr": 0.002991044569983042, "optim/total_tokens": 3625451520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 3.0996036529541016, "created_at": "2025-01-16T01:05:01.040465+00:00"} {"global_step": 6916, "acc_step": 0, "speed/wps": 12888.455793940448, "speed/FLOPS": 202430908838033.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06826650351285934, "optim/lr": 0.002991035219057161, "optim/total_tokens": 3625975808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 2.9827122688293457, "created_at": "2025-01-16T01:05:11.218616+00:00"} {"global_step": 6917, "acc_step": 0, "speed/wps": 12893.345792412756, "speed/FLOPS": 202507712983602.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06748572736978531, "optim/lr": 0.0029910258632665147, "optim/total_tokens": 3626500096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.03963041305542, "created_at": "2025-01-16T01:05:21.389110+00:00"} {"global_step": 6918, "acc_step": 0, "speed/wps": 12890.513935789564, "speed/FLOPS": 202463234783956.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0557023249566555, "optim/lr": 0.0029910165026111346, "optim/total_tokens": 3627024384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.074181079864502, "created_at": "2025-01-16T01:05:31.560716+00:00"} {"global_step": 6919, "acc_step": 0, "speed/wps": 12891.109401725076, "speed/FLOPS": 202472587394730.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05995842069387436, "optim/lr": 0.002991007137091051, "optim/total_tokens": 3627548672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 3.002185344696045, "created_at": "2025-01-16T01:05:41.732320+00:00"} {"global_step": 6920, "acc_step": 0, "speed/wps": 12889.329243576003, "speed/FLOPS": 202444627564805.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0664425790309906, "optim/lr": 0.0029909977667062953, "optim/total_tokens": 3628072960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9244794845581055, "created_at": "2025-01-16T01:05:51.904933+00:00"} {"global_step": 6921, "acc_step": 0, "speed/wps": 12887.863606213403, "speed/FLOPS": 202421607716028.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07439422607421875, "optim/lr": 0.002990988391456896, "optim/total_tokens": 3628597248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.0266168117523193, "created_at": "2025-01-16T01:06:02.080997+00:00"} {"global_step": 6922, "acc_step": 0, "speed/wps": 12887.726211095578, "speed/FLOPS": 202419449736902.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07020888477563858, "optim/lr": 0.002990979011342885, "optim/total_tokens": 3629121536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419297, "loss/out": 3.126347064971924, "created_at": "2025-01-16T01:06:12.257390+00:00"} {"global_step": 6923, "acc_step": 0, "speed/wps": 12892.896136748326, "speed/FLOPS": 202500650523501.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05583307892084122, "optim/lr": 0.0029909696263642933, "optim/total_tokens": 3629645824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.108487844467163, "created_at": "2025-01-16T01:06:22.427361+00:00"} {"global_step": 6924, "acc_step": 0, "speed/wps": 12888.20169949681, "speed/FLOPS": 202426917935633.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06638344377279282, "optim/lr": 0.0029909602365211507, "optim/total_tokens": 3630170112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.0534520149230957, "created_at": "2025-01-16T01:06:32.601015+00:00"} {"global_step": 6925, "acc_step": 0, "speed/wps": 12892.93798841546, "speed/FLOPS": 202501307861443.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05743712559342384, "optim/lr": 0.002990950841813488, "optim/total_tokens": 3630694400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.967759132385254, "created_at": "2025-01-16T01:06:42.770679+00:00"} {"global_step": 6926, "acc_step": 0, "speed/wps": 12887.786823485443, "speed/FLOPS": 202420401737775.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06865467131137848, "optim/lr": 0.002990941442241337, "optim/total_tokens": 3631218688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 3.102914333343506, "created_at": "2025-01-16T01:06:52.945344+00:00"} {"global_step": 6927, "acc_step": 0, "speed/wps": 12889.836510231997, "speed/FLOPS": 202452594884695.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07055850327014923, "optim/lr": 0.002990932037804726, "optim/total_tokens": 3631742976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.928107738494873, "created_at": "2025-01-16T01:07:03.118978+00:00"} {"global_step": 6928, "acc_step": 0, "speed/wps": 12891.584046319651, "speed/FLOPS": 202480042340317.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.068955197930336, "optim/lr": 0.0029909226285036875, "optim/total_tokens": 3632267264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0258028507232666, "created_at": "2025-01-16T01:07:13.290259+00:00"} {"global_step": 6929, "acc_step": 0, "speed/wps": 12892.577845486332, "speed/FLOPS": 202495651321853.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07000187784433365, "optim/lr": 0.0029909132143382516, "optim/total_tokens": 3632791552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 3.104116678237915, "created_at": "2025-01-16T01:07:23.462096+00:00"} {"global_step": 6930, "acc_step": 0, "speed/wps": 12892.232218139026, "speed/FLOPS": 202490222769422.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06432311236858368, "optim/lr": 0.00299090379530845, "optim/total_tokens": 3633315840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.0770580768585205, "created_at": "2025-01-16T01:07:33.632867+00:00"} {"global_step": 6931, "acc_step": 0, "speed/wps": 12892.611972129927, "speed/FLOPS": 202496187327686.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053505558520555496, "optim/lr": 0.0029908943714143117, "optim/total_tokens": 3633840128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.898332118988037, "created_at": "2025-01-16T01:07:43.805678+00:00"} {"global_step": 6932, "acc_step": 0, "speed/wps": 12886.301761630055, "speed/FLOPS": 202397076800647.94, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050430431962013245, "optim/lr": 0.0029908849426558684, "optim/total_tokens": 3634364416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0622448921203613, "created_at": "2025-01-16T01:07:53.981403+00:00"} {"global_step": 6933, "acc_step": 0, "speed/wps": 12893.018509406154, "speed/FLOPS": 202502572554250.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05123983323574066, "optim/lr": 0.0029908755090331507, "optim/total_tokens": 3634888704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.0289793014526367, "created_at": "2025-01-16T01:08:04.151686+00:00"} {"global_step": 6934, "acc_step": 0, "speed/wps": 12892.662437847937, "speed/FLOPS": 202496979961134.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04984874278306961, "optim/lr": 0.00299086607054619, "optim/total_tokens": 3635412992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.0553431510925293, "created_at": "2025-01-16T01:08:14.329880+00:00"} {"global_step": 6935, "acc_step": 0, "speed/wps": 12892.287737927467, "speed/FLOPS": 202491094783992.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056872930377721786, "optim/lr": 0.002990856627195016, "optim/total_tokens": 3635937280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.0308103561401367, "created_at": "2025-01-16T01:08:24.502697+00:00"} {"global_step": 6936, "acc_step": 0, "speed/wps": 12895.235618154915, "speed/FLOPS": 202537395293776.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05702129378914833, "optim/lr": 0.0029908471789796606, "optim/total_tokens": 3636461568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 3.057856321334839, "created_at": "2025-01-16T01:08:34.674252+00:00"} {"global_step": 6937, "acc_step": 0, "speed/wps": 12888.979116018403, "speed/FLOPS": 202439128330387.44, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05525884032249451, "optim/lr": 0.002990837725900154, "optim/total_tokens": 3636985856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.0537829399108887, "created_at": "2025-01-16T01:08:44.847366+00:00"} {"global_step": 6938, "acc_step": 0, "speed/wps": 12890.40070525031, "speed/FLOPS": 202461456342742.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05414379760622978, "optim/lr": 0.002990828267956527, "optim/total_tokens": 3637510144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.0011773109436035, "created_at": "2025-01-16T01:08:55.021427+00:00"} {"global_step": 6939, "acc_step": 0, "speed/wps": 12892.07875850349, "speed/FLOPS": 202487812474970.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06271622329950333, "optim/lr": 0.00299081880514881, "optim/total_tokens": 3638034432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.981010913848877, "created_at": "2025-01-16T01:09:05.191744+00:00"} {"global_step": 6940, "acc_step": 0, "speed/wps": 12888.666696727267, "speed/FLOPS": 202434221356111.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05336768552660942, "optim/lr": 0.002990809337477035, "optim/total_tokens": 3638558720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.976569652557373, "created_at": "2025-01-16T01:09:15.388795+00:00"} {"global_step": 6941, "acc_step": 0, "speed/wps": 12897.864593411528, "speed/FLOPS": 202578686962771.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05783615633845329, "optim/lr": 0.002990799864941233, "optim/total_tokens": 3639083008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 3.007382392883301, "created_at": "2025-01-16T01:09:25.561971+00:00"} {"global_step": 6942, "acc_step": 0, "speed/wps": 12891.016801695761, "speed/FLOPS": 202471132984023.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0655665248632431, "optim/lr": 0.0029907903875414335, "optim/total_tokens": 3639607296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0228939056396484, "created_at": "2025-01-16T01:09:35.734437+00:00"} {"global_step": 6943, "acc_step": 0, "speed/wps": 12887.91417520876, "speed/FLOPS": 202422401971589.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0676523894071579, "optim/lr": 0.002990780905277668, "optim/total_tokens": 3640131584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346379, "loss/out": 2.9686899185180664, "created_at": "2025-01-16T01:09:45.908410+00:00"} {"global_step": 6944, "acc_step": 0, "speed/wps": 12889.30399833329, "speed/FLOPS": 202444231053577.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06131497398018837, "optim/lr": 0.002990771418149969, "optim/total_tokens": 3640655872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.1126794815063477, "created_at": "2025-01-16T01:09:56.080930+00:00"} {"global_step": 6945, "acc_step": 0, "speed/wps": 12891.054563945863, "speed/FLOPS": 202471726092052.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06698887795209885, "optim/lr": 0.0029907619261583646, "optim/total_tokens": 3641180160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.123772621154785, "created_at": "2025-01-16T01:10:06.253119+00:00"} {"global_step": 6946, "acc_step": 0, "speed/wps": 12889.85590318443, "speed/FLOPS": 202452899477661.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05522909015417099, "optim/lr": 0.002990752429302888, "optim/total_tokens": 3641704448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.9860830307006836, "created_at": "2025-01-16T01:10:16.429638+00:00"} {"global_step": 6947, "acc_step": 0, "speed/wps": 12888.652882072467, "speed/FLOPS": 202434004377974.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06407348066568375, "optim/lr": 0.0029907429275835688, "optim/total_tokens": 3642228736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.0501694679260254, "created_at": "2025-01-16T01:10:26.603098+00:00"} {"global_step": 6948, "acc_step": 0, "speed/wps": 12888.84760250656, "speed/FLOPS": 202437062729965.4, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06936798244714737, "optim/lr": 0.002990733421000439, "optim/total_tokens": 3642753024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.0681300163269043, "created_at": "2025-01-16T01:10:36.778393+00:00"} {"global_step": 6949, "acc_step": 0, "speed/wps": 12889.156982704215, "speed/FLOPS": 202441921971102.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05626598745584488, "optim/lr": 0.00299072390955353, "optim/total_tokens": 3643277312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 2.9756457805633545, "created_at": "2025-01-16T01:10:46.953158+00:00"} {"global_step": 6950, "acc_step": 0, "speed/wps": 12894.744818829304, "speed/FLOPS": 202529686615938.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05234314501285553, "optim/lr": 0.0029907143932428708, "optim/total_tokens": 3643801600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 3.0601940155029297, "created_at": "2025-01-16T01:10:57.131928+00:00"} {"global_step": 6951, "acc_step": 0, "speed/wps": 12890.159859656358, "speed/FLOPS": 202457673531733.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05203849449753761, "optim/lr": 0.0029907048720684945, "optim/total_tokens": 3644325888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.925356864929199, "created_at": "2025-01-16T01:11:07.311546+00:00"} {"global_step": 6952, "acc_step": 0, "speed/wps": 12888.324441002727, "speed/FLOPS": 202428845759648.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07164429128170013, "optim/lr": 0.0029906953460304316, "optim/total_tokens": 3644850176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.027636766433716, "created_at": "2025-01-16T01:11:17.490315+00:00"} {"global_step": 6953, "acc_step": 0, "speed/wps": 12891.386816576965, "speed/FLOPS": 202476944576186.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06630939245223999, "optim/lr": 0.0029906858151287122, "optim/total_tokens": 3645374464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.9555602073669434, "created_at": "2025-01-16T01:11:27.662645+00:00"} {"global_step": 6954, "acc_step": 0, "speed/wps": 12885.380074489809, "speed/FLOPS": 202382600437579.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05865773186087608, "optim/lr": 0.002990676279363369, "optim/total_tokens": 3645898752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 3.0876309871673584, "created_at": "2025-01-16T01:11:37.843361+00:00"} {"global_step": 6955, "acc_step": 0, "speed/wps": 12895.771926522995, "speed/FLOPS": 202545818753661.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05840664729475975, "optim/lr": 0.0029906667387344316, "optim/total_tokens": 3646423040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0583295822143555, "created_at": "2025-01-16T01:11:48.017465+00:00"} {"global_step": 6956, "acc_step": 0, "speed/wps": 12892.120647379123, "speed/FLOPS": 202488470397324.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0774645283818245, "optim/lr": 0.0029906571932419327, "optim/total_tokens": 3646947328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.031402111053467, "created_at": "2025-01-16T01:11:58.188010+00:00"} {"global_step": 6957, "acc_step": 0, "speed/wps": 12892.00830266637, "speed/FLOPS": 202486705869232.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06587646156549454, "optim/lr": 0.002990647642885902, "optim/total_tokens": 3647471616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 3.045866012573242, "created_at": "2025-01-16T01:12:08.358456+00:00"} {"global_step": 6958, "acc_step": 0, "speed/wps": 12891.042329949174, "speed/FLOPS": 202471533940323.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06309635937213898, "optim/lr": 0.002990638087666371, "optim/total_tokens": 3647995904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.00793194770813, "created_at": "2025-01-16T01:12:18.533012+00:00"} {"global_step": 6959, "acc_step": 0, "speed/wps": 12892.898522595502, "speed/FLOPS": 202500687996509.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05488380044698715, "optim/lr": 0.0029906285275833715, "optim/total_tokens": 3648520192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.0095369815826416, "created_at": "2025-01-16T01:12:28.702665+00:00"} {"global_step": 6960, "acc_step": 0, "speed/wps": 12888.4354349981, "speed/FLOPS": 202430589072867.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08112690597772598, "optim/lr": 0.0029906189626369337, "optim/total_tokens": 3649044480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.016202211380005, "created_at": "2025-01-16T01:12:38.880976+00:00"} {"global_step": 6961, "acc_step": 0, "speed/wps": 12892.385370820077, "speed/FLOPS": 202492628242732.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08050361275672913, "optim/lr": 0.0029906093928270903, "optim/total_tokens": 3649568768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.0441720485687256, "created_at": "2025-01-16T01:12:49.054875+00:00"} {"global_step": 6962, "acc_step": 0, "speed/wps": 12890.429507395302, "speed/FLOPS": 202461908720008.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05824096500873566, "optim/lr": 0.002990599818153871, "optim/total_tokens": 3650093056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 2.9683656692504883, "created_at": "2025-01-16T01:12:59.228515+00:00"} {"global_step": 6963, "acc_step": 0, "speed/wps": 12887.066313025765, "speed/FLOPS": 202409085130921.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05833088979125023, "optim/lr": 0.0029905902386173073, "optim/total_tokens": 3650617344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.984372138977051, "created_at": "2025-01-16T01:13:09.408775+00:00"} {"global_step": 6964, "acc_step": 0, "speed/wps": 12891.35763464976, "speed/FLOPS": 202476486233919.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04873683303594589, "optim/lr": 0.002990580654217432, "optim/total_tokens": 3651141632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376202, "loss/out": 3.0205326080322266, "created_at": "2025-01-16T01:13:19.580161+00:00"} {"global_step": 6965, "acc_step": 0, "speed/wps": 12892.367305983895, "speed/FLOPS": 202492344509656.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06770433485507965, "optim/lr": 0.002990571064954274, "optim/total_tokens": 3651665920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.9789819717407227, "created_at": "2025-01-16T01:13:29.750310+00:00"} {"global_step": 6966, "acc_step": 0, "speed/wps": 12891.822349484515, "speed/FLOPS": 202483785218984.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06117914989590645, "optim/lr": 0.0029905614708278664, "optim/total_tokens": 3652190208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.981586456298828, "created_at": "2025-01-16T01:13:39.921260+00:00"} {"global_step": 6967, "acc_step": 0, "speed/wps": 12891.847923441634, "speed/FLOPS": 202484186893123.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05405193939805031, "optim/lr": 0.0029905518718382398, "optim/total_tokens": 3652714496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 2.9827756881713867, "created_at": "2025-01-16T01:13:50.091829+00:00"} {"global_step": 6968, "acc_step": 0, "speed/wps": 12889.061715845934, "speed/FLOPS": 202440425674182.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0536402203142643, "optim/lr": 0.0029905422679854256, "optim/total_tokens": 3653238784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9843592643737793, "created_at": "2025-01-16T01:14:00.265669+00:00"} {"global_step": 6969, "acc_step": 0, "speed/wps": 12887.31139873795, "speed/FLOPS": 202412934538814.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06683142483234406, "optim/lr": 0.0029905326592694547, "optim/total_tokens": 3653763072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.935790777206421, "created_at": "2025-01-16T01:14:10.442098+00:00"} {"global_step": 6970, "acc_step": 0, "speed/wps": 12889.569052343038, "speed/FLOPS": 202448394091023.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07815027236938477, "optim/lr": 0.0029905230456903596, "optim/total_tokens": 3654287360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.958069324493408, "created_at": "2025-01-16T01:14:20.616273+00:00"} {"global_step": 6971, "acc_step": 0, "speed/wps": 12895.961076647847, "speed/FLOPS": 202548789616292.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057144276797771454, "optim/lr": 0.0029905134272481706, "optim/total_tokens": 3654811648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 3.120173931121826, "created_at": "2025-01-16T01:14:30.786390+00:00"} {"global_step": 6972, "acc_step": 0, "speed/wps": 12891.064189629691, "speed/FLOPS": 202471877276644.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08010197430849075, "optim/lr": 0.0029905038039429194, "optim/total_tokens": 3655335936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.058257579803467, "created_at": "2025-01-16T01:14:40.958794+00:00"} {"global_step": 6973, "acc_step": 0, "speed/wps": 12884.49065774437, "speed/FLOPS": 202368630925405.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06435639411211014, "optim/lr": 0.0029904941757746373, "optim/total_tokens": 3655860224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9618382453918457, "created_at": "2025-01-16T01:14:51.137772+00:00"} {"global_step": 6974, "acc_step": 0, "speed/wps": 12887.336077991398, "speed/FLOPS": 202413322160402.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06898099184036255, "optim/lr": 0.002990484542743356, "optim/total_tokens": 3656384512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369341, "loss/out": 2.9061405658721924, "created_at": "2025-01-16T01:15:01.313957+00:00"} {"global_step": 6975, "acc_step": 0, "speed/wps": 12889.110084499827, "speed/FLOPS": 202441185370357.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0625142976641655, "optim/lr": 0.002990474904849107, "optim/total_tokens": 3656908800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.05058217048645, "created_at": "2025-01-16T01:15:11.486677+00:00"} {"global_step": 6976, "acc_step": 0, "speed/wps": 12892.811436202366, "speed/FLOPS": 202499320185036.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06135236844420433, "optim/lr": 0.002990465262091921, "optim/total_tokens": 3657433088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 3.0469579696655273, "created_at": "2025-01-16T01:15:21.659870+00:00"} {"global_step": 6977, "acc_step": 0, "speed/wps": 12888.71626518944, "speed/FLOPS": 202434999896923.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07814193516969681, "optim/lr": 0.00299045561447183, "optim/total_tokens": 3657957376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.940802574157715, "created_at": "2025-01-16T01:15:31.833374+00:00"} {"global_step": 6978, "acc_step": 0, "speed/wps": 12885.785871471851, "speed/FLOPS": 202388974036806.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0635826364159584, "optim/lr": 0.002990445961988866, "optim/total_tokens": 3658481664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 2.9432950019836426, "created_at": "2025-01-16T01:15:42.009858+00:00"} {"global_step": 6979, "acc_step": 0, "speed/wps": 12890.659682422056, "speed/FLOPS": 202465523935097.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07807455956935883, "optim/lr": 0.0029904363046430594, "optim/total_tokens": 3659005952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.0195908546447754, "created_at": "2025-01-16T01:15:52.183031+00:00"} {"global_step": 6980, "acc_step": 0, "speed/wps": 12892.982318360448, "speed/FLOPS": 202502004124145.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07075771689414978, "optim/lr": 0.002990426642434443, "optim/total_tokens": 3659530240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.8933424949645996, "created_at": "2025-01-16T01:16:02.354704+00:00"} {"global_step": 6981, "acc_step": 0, "speed/wps": 12894.621304037972, "speed/FLOPS": 202527746646413.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0769827663898468, "optim/lr": 0.002990416975363047, "optim/total_tokens": 3660054528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 2.9743638038635254, "created_at": "2025-01-16T01:16:12.526817+00:00"} {"global_step": 6982, "acc_step": 0, "speed/wps": 12887.655493624166, "speed/FLOPS": 202418339021838.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06712308526039124, "optim/lr": 0.0029904073034289036, "optim/total_tokens": 3660578816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 2.9358620643615723, "created_at": "2025-01-16T01:16:22.702021+00:00"} {"global_step": 6983, "acc_step": 0, "speed/wps": 12890.201528274272, "speed/FLOPS": 202458327994636.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06981664896011353, "optim/lr": 0.002990397626632044, "optim/total_tokens": 3661103104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.021998167037964, "created_at": "2025-01-16T01:16:32.875166+00:00"} {"global_step": 6984, "acc_step": 0, "speed/wps": 12890.247541867555, "speed/FLOPS": 202459050701348.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0596163235604763, "optim/lr": 0.0029903879449725, "optim/total_tokens": 3661627392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.979820728302002, "created_at": "2025-01-16T01:16:43.053891+00:00"} {"global_step": 6985, "acc_step": 0, "speed/wps": 12889.532730737144, "speed/FLOPS": 202447823610291.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06142732501029968, "optim/lr": 0.002990378258450304, "optim/total_tokens": 3662151680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455532, "loss/out": 2.9598548412323, "created_at": "2025-01-16T01:16:53.227477+00:00"} {"global_step": 6986, "acc_step": 0, "speed/wps": 12891.748921756955, "speed/FLOPS": 202482631935623.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06236429139971733, "optim/lr": 0.0029903685670654863, "optim/total_tokens": 3662675968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.075988292694092, "created_at": "2025-01-16T01:17:03.401326+00:00"} {"global_step": 6987, "acc_step": 0, "speed/wps": 12890.561753337672, "speed/FLOPS": 202463985824258.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05999276414513588, "optim/lr": 0.0029903588708180794, "optim/total_tokens": 3663200256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.99965238571167, "created_at": "2025-01-16T01:17:13.582133+00:00"} {"global_step": 6988, "acc_step": 0, "speed/wps": 12892.010358699921, "speed/FLOPS": 202486738162064.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06336954981088638, "optim/lr": 0.0029903491697081137, "optim/total_tokens": 3663724544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 2.92744779586792, "created_at": "2025-01-16T01:17:23.756065+00:00"} {"global_step": 6989, "acc_step": 0, "speed/wps": 12890.547389145078, "speed/FLOPS": 202463760214874.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0677558183670044, "optim/lr": 0.0029903394637356227, "optim/total_tokens": 3664248832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.959750175476074, "created_at": "2025-01-16T01:17:33.929643+00:00"} {"global_step": 6990, "acc_step": 0, "speed/wps": 12888.508189979088, "speed/FLOPS": 202431731789819.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0757811963558197, "optim/lr": 0.0029903297529006376, "optim/total_tokens": 3664773120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9613866806030273, "created_at": "2025-01-16T01:17:44.102773+00:00"} {"global_step": 6991, "acc_step": 0, "speed/wps": 12888.912321370748, "speed/FLOPS": 202438079228660.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05930570513010025, "optim/lr": 0.0029903200372031883, "optim/total_tokens": 3665297408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.003018379211426, "created_at": "2025-01-16T01:17:54.276740+00:00"} {"global_step": 6992, "acc_step": 0, "speed/wps": 12891.13545107347, "speed/FLOPS": 202472996535542.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10580918937921524, "optim/lr": 0.0029903103166433083, "optim/total_tokens": 3665821696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0043272972106934, "created_at": "2025-01-16T01:18:04.450399+00:00"} {"global_step": 6993, "acc_step": 0, "speed/wps": 12891.304040524175, "speed/FLOPS": 202475644464531.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05804285779595375, "optim/lr": 0.0029903005912210287, "optim/total_tokens": 3666345984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 3.080627202987671, "created_at": "2025-01-16T01:18:14.621723+00:00"} {"global_step": 6994, "acc_step": 0, "speed/wps": 12890.032557995013, "speed/FLOPS": 202455674084211.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07996461540460587, "optim/lr": 0.0029902908609363817, "optim/total_tokens": 3666870272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.9514389038085938, "created_at": "2025-01-16T01:18:24.799749+00:00"} {"global_step": 6995, "acc_step": 0, "speed/wps": 12894.104775901036, "speed/FLOPS": 202519633862238.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06259262561798096, "optim/lr": 0.0029902811257893984, "optim/total_tokens": 3667394560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9975032806396484, "created_at": "2025-01-16T01:18:34.968604+00:00"} {"global_step": 6996, "acc_step": 0, "speed/wps": 12888.685802878019, "speed/FLOPS": 202434521444463.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.060089051723480225, "optim/lr": 0.002990271385780111, "optim/total_tokens": 3667918848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 2.97424578666687, "created_at": "2025-01-16T01:18:45.141604+00:00"} {"global_step": 6997, "acc_step": 0, "speed/wps": 12891.213228977876, "speed/FLOPS": 202474218144408.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057234715670347214, "optim/lr": 0.0029902616409085508, "optim/total_tokens": 3668443136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.0688247680664062, "created_at": "2025-01-16T01:18:55.315272+00:00"} {"global_step": 6998, "acc_step": 0, "speed/wps": 12890.25328207869, "speed/FLOPS": 202459140859251.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05351698398590088, "optim/lr": 0.00299025189117475, "optim/total_tokens": 3668967424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9704108238220215, "created_at": "2025-01-16T01:19:05.488041+00:00"} {"global_step": 6999, "acc_step": 0, "speed/wps": 12888.9810447836, "speed/FLOPS": 202439158624295.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06612744927406311, "optim/lr": 0.0029902421365787407, "optim/total_tokens": 3669491712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.051222801208496, "created_at": "2025-01-16T01:19:15.664047+00:00"} {"global_step": 7000, "acc_step": 0, "speed/wps": 12892.630074264736, "speed/FLOPS": 202496471646588.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06405601650476456, "optim/lr": 0.002990232377120554, "optim/total_tokens": 3670016000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315550, "loss/out": 3.078429937362671, "created_at": "2025-01-16T01:19:25.835702+00:00"} {"global_step": 7001, "acc_step": 0, "speed/wps": 12596.398287794744, "speed/FLOPS": 197843744374945.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.2423, "optim/grad_norm": 0.059044428169727325, "optim/lr": 0.0029902226128002223, "optim/total_tokens": 3670540288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.1401820182800293, "created_at": "2025-01-16T01:19:36.248726+00:00"} {"global_step": 7002, "acc_step": 0, "speed/wps": 12905.773783735394, "speed/FLOPS": 202702911665174.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061401691287755966, "optim/lr": 0.002990212843617777, "optim/total_tokens": 3671064576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.0678744316101074, "created_at": "2025-01-16T01:19:46.408316+00:00"} {"global_step": 7003, "acc_step": 0, "speed/wps": 12902.09235779791, "speed/FLOPS": 202645089812017.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054470304399728775, "optim/lr": 0.0029902030695732507, "optim/total_tokens": 3671588864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309147, "loss/out": 2.9856674671173096, "created_at": "2025-01-16T01:19:56.571106+00:00"} {"global_step": 7004, "acc_step": 0, "speed/wps": 12899.672354548706, "speed/FLOPS": 202607080335555.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05055394023656845, "optim/lr": 0.0029901932906666747, "optim/total_tokens": 3672113152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.052022933959961, "created_at": "2025-01-16T01:20:06.736099+00:00"} {"global_step": 7005, "acc_step": 0, "speed/wps": 12904.758065261683, "speed/FLOPS": 202686958410801.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06230999901890755, "optim/lr": 0.002990183506898081, "optim/total_tokens": 3672637440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9183692932128906, "created_at": "2025-01-16T01:20:16.898874+00:00"} {"global_step": 7006, "acc_step": 0, "speed/wps": 12894.446577878425, "speed/FLOPS": 202525002331978.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052905093878507614, "optim/lr": 0.002990173718267501, "optim/total_tokens": 3673161728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 2.9954605102539062, "created_at": "2025-01-16T01:20:27.070531+00:00"} {"global_step": 7007, "acc_step": 0, "speed/wps": 12901.532621855313, "speed/FLOPS": 202636298389880.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06320612132549286, "optim/lr": 0.0029901639247749676, "optim/total_tokens": 3673686016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.023062229156494, "created_at": "2025-01-16T01:20:37.242120+00:00"} {"global_step": 7008, "acc_step": 0, "speed/wps": 12901.172508332209, "speed/FLOPS": 202630642312152.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06303154677152634, "optim/lr": 0.0029901541264205126, "optim/total_tokens": 3674210304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283142, "loss/out": 2.977168321609497, "created_at": "2025-01-16T01:20:47.405436+00:00"} {"global_step": 7009, "acc_step": 0, "speed/wps": 12899.56074128708, "speed/FLOPS": 202605327295908.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05833360552787781, "optim/lr": 0.0029901443232041676, "optim/total_tokens": 3674734592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.7888190746307373, "created_at": "2025-01-16T01:20:57.573588+00:00"} {"global_step": 7010, "acc_step": 0, "speed/wps": 12903.352156871815, "speed/FLOPS": 202664876687618.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06529662013053894, "optim/lr": 0.0029901345151259646, "optim/total_tokens": 3675258880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.00356388092041, "created_at": "2025-01-16T01:21:07.739878+00:00"} {"global_step": 7011, "acc_step": 0, "speed/wps": 12896.581475819348, "speed/FLOPS": 202558533837798.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06398414075374603, "optim/lr": 0.0029901247021859355, "optim/total_tokens": 3675783168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 2.9941694736480713, "created_at": "2025-01-16T01:21:17.908122+00:00"} {"global_step": 7012, "acc_step": 0, "speed/wps": 12901.693126227314, "speed/FLOPS": 202638819331599.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07083265483379364, "optim/lr": 0.0029901148843841127, "optim/total_tokens": 3676307456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.9057235717773438, "created_at": "2025-01-16T01:21:28.074883+00:00"} {"global_step": 7013, "acc_step": 0, "speed/wps": 12898.985749591913, "speed/FLOPS": 202596296261214.88, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05748040974140167, "optim/lr": 0.0029901050617205283, "optim/total_tokens": 3676831744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9276576042175293, "created_at": "2025-01-16T01:21:38.242753+00:00"} {"global_step": 7014, "acc_step": 0, "speed/wps": 12901.17458266402, "speed/FLOPS": 202630674892384.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0628160908818245, "optim/lr": 0.0029900952341952135, "optim/total_tokens": 3677356032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.007545232772827, "created_at": "2025-01-16T01:21:48.413660+00:00"} {"global_step": 7015, "acc_step": 0, "speed/wps": 12900.238758956617, "speed/FLOPS": 202615976495102.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06304445117712021, "optim/lr": 0.0029900854018082013, "optim/total_tokens": 3677880320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9871137142181396, "created_at": "2025-01-16T01:21:58.580019+00:00"} {"global_step": 7016, "acc_step": 0, "speed/wps": 12906.525808571787, "speed/FLOPS": 202714723248620.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061193741858005524, "optim/lr": 0.0029900755645595233, "optim/total_tokens": 3678404608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.064236640930176, "created_at": "2025-01-16T01:22:08.739542+00:00"} {"global_step": 7017, "acc_step": 0, "speed/wps": 12902.146752531025, "speed/FLOPS": 202645944156046.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08249914646148682, "optim/lr": 0.002990065722449212, "optim/total_tokens": 3678928896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 3.168369770050049, "created_at": "2025-01-16T01:22:18.903346+00:00"} {"global_step": 7018, "acc_step": 0, "speed/wps": 12904.044662177157, "speed/FLOPS": 202675753435040.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.073482446372509, "optim/lr": 0.0029900558754772993, "optim/total_tokens": 3679453184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484762, "loss/out": 2.9365625381469727, "created_at": "2025-01-16T01:22:29.066393+00:00"} {"global_step": 7019, "acc_step": 0, "speed/wps": 12895.051162169442, "speed/FLOPS": 202534498159055.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.07502717524766922, "optim/lr": 0.002990046023643817, "optim/total_tokens": 3679977472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 3.0476999282836914, "created_at": "2025-01-16T01:22:39.235998+00:00"} {"global_step": 7020, "acc_step": 0, "speed/wps": 12894.226193340797, "speed/FLOPS": 202521540889975.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06748965382575989, "optim/lr": 0.002990036166948798, "optim/total_tokens": 3680501760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.1107757091522217, "created_at": "2025-01-16T01:22:49.406851+00:00"} {"global_step": 7021, "acc_step": 0, "speed/wps": 12899.151096971038, "speed/FLOPS": 202598893269015.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06700973212718964, "optim/lr": 0.0029900263053922733, "optim/total_tokens": 3681026048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.939976215362549, "created_at": "2025-01-16T01:22:59.574175+00:00"} {"global_step": 7022, "acc_step": 0, "speed/wps": 12888.18313297477, "speed/FLOPS": 202426626322891.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.059067267924547195, "optim/lr": 0.002990016438974276, "optim/total_tokens": 3681550336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8807830810546875, "created_at": "2025-01-16T01:23:09.747617+00:00"} {"global_step": 7023, "acc_step": 0, "speed/wps": 12894.072662605691, "speed/FLOPS": 202519129478807.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059268392622470856, "optim/lr": 0.002990006567694838, "optim/total_tokens": 3682074624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9202816486358643, "created_at": "2025-01-16T01:23:19.922542+00:00"} {"global_step": 7024, "acc_step": 0, "speed/wps": 12901.718697965945, "speed/FLOPS": 202639220970893.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0667208880186081, "optim/lr": 0.0029899966915539916, "optim/total_tokens": 3682598912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 2.9892215728759766, "created_at": "2025-01-16T01:23:30.087247+00:00"} {"global_step": 7025, "acc_step": 0, "speed/wps": 12893.909896588486, "speed/FLOPS": 202516573014843.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0566435381770134, "optim/lr": 0.0029899868105517694, "optim/total_tokens": 3683123200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9775094985961914, "created_at": "2025-01-16T01:23:40.256530+00:00"} {"global_step": 7026, "acc_step": 0, "speed/wps": 12890.981966774134, "speed/FLOPS": 202470585853711.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06633266061544418, "optim/lr": 0.0029899769246882024, "optim/total_tokens": 3683647488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.950410842895508, "created_at": "2025-01-16T01:23:50.431754+00:00"} {"global_step": 7027, "acc_step": 0, "speed/wps": 12898.292972678493, "speed/FLOPS": 202585415247814.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05108289048075676, "optim/lr": 0.0029899670339633244, "optim/total_tokens": 3684171776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.0429091453552246, "created_at": "2025-01-16T01:24:00.597314+00:00"} {"global_step": 7028, "acc_step": 0, "speed/wps": 12904.062697045467, "speed/FLOPS": 202676036697429.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054959774017333984, "optim/lr": 0.002989957138377166, "optim/total_tokens": 3684696064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9775843620300293, "created_at": "2025-01-16T01:24:10.759562+00:00"} {"global_step": 7029, "acc_step": 0, "speed/wps": 12904.80465378895, "speed/FLOPS": 202687690147641.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05158144235610962, "optim/lr": 0.002989947237929761, "optim/total_tokens": 3685220352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 3.1619343757629395, "created_at": "2025-01-16T01:24:20.920860+00:00"} {"global_step": 7030, "acc_step": 0, "speed/wps": 12895.539342765338, "speed/FLOPS": 202542165706144.25, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05046285688877106, "optim/lr": 0.0029899373326211414, "optim/total_tokens": 3685744640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.083775520324707, "created_at": "2025-01-16T01:24:31.095276+00:00"} {"global_step": 7031, "acc_step": 0, "speed/wps": 12901.023773147728, "speed/FLOPS": 202628306221696.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053048692643642426, "optim/lr": 0.0029899274224513394, "optim/total_tokens": 3686268928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.9231951236724854, "created_at": "2025-01-16T01:24:41.261341+00:00"} {"global_step": 7032, "acc_step": 0, "speed/wps": 12904.55508629184, "speed/FLOPS": 202683770347157.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05511007830500603, "optim/lr": 0.0029899175074203864, "optim/total_tokens": 3686793216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 3.021167755126953, "created_at": "2025-01-16T01:24:51.422100+00:00"} {"global_step": 7033, "acc_step": 0, "speed/wps": 12894.390125101845, "speed/FLOPS": 202524115663548.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055910900235176086, "optim/lr": 0.002989907587528316, "optim/total_tokens": 3687317504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.0397520065307617, "created_at": "2025-01-16T01:25:01.595187+00:00"} {"global_step": 7034, "acc_step": 0, "speed/wps": 12900.167535178522, "speed/FLOPS": 202614857827796.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05318005010485649, "optim/lr": 0.00298989766277516, "optim/total_tokens": 3687841792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 2.980254650115967, "created_at": "2025-01-16T01:25:11.762053+00:00"} {"global_step": 7035, "acc_step": 0, "speed/wps": 12894.198781215631, "speed/FLOPS": 202521110344881.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06038260832428932, "optim/lr": 0.0029898877331609505, "optim/total_tokens": 3688366080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9467806816101074, "created_at": "2025-01-16T01:25:21.932476+00:00"} {"global_step": 7036, "acc_step": 0, "speed/wps": 12907.884698930862, "speed/FLOPS": 202736066489020.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06479060649871826, "optim/lr": 0.0029898777986857206, "optim/total_tokens": 3688890368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.016141891479492, "created_at": "2025-01-16T01:25:32.090848+00:00"} {"global_step": 7037, "acc_step": 0, "speed/wps": 12901.29367573675, "speed/FLOPS": 202632545412742.4, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06594523787498474, "optim/lr": 0.0029898678593495024, "optim/total_tokens": 3689414656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 3.0885043144226074, "created_at": "2025-01-16T01:25:42.256482+00:00"} {"global_step": 7038, "acc_step": 0, "speed/wps": 12895.26622250049, "speed/FLOPS": 202537875977078.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06232338398694992, "optim/lr": 0.002989857915152328, "optim/total_tokens": 3689938944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.956996440887451, "created_at": "2025-01-16T01:25:52.424409+00:00"} {"global_step": 7039, "acc_step": 0, "speed/wps": 12903.679442982391, "speed/FLOPS": 202670017165722.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07639071345329285, "optim/lr": 0.0029898479660942306, "optim/total_tokens": 3690463232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.976505994796753, "created_at": "2025-01-16T01:26:02.589182+00:00"} {"global_step": 7040, "acc_step": 0, "speed/wps": 12901.331003868845, "speed/FLOPS": 202633131702351.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06130388006567955, "optim/lr": 0.002989838012175242, "optim/total_tokens": 3690987520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.006537914276123, "created_at": "2025-01-16T01:26:12.752311+00:00"} {"global_step": 7041, "acc_step": 0, "speed/wps": 12898.70052914881, "speed/FLOPS": 202591816482221.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0634290874004364, "optim/lr": 0.002989828053395395, "optim/total_tokens": 3691511808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.908186435699463, "created_at": "2025-01-16T01:26:22.917464+00:00"} {"global_step": 7042, "acc_step": 0, "speed/wps": 12901.030541181679, "speed/FLOPS": 202628412522970.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05694648250937462, "optim/lr": 0.002989818089754722, "optim/total_tokens": 3692036096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.9968528747558594, "created_at": "2025-01-16T01:26:33.080768+00:00"} {"global_step": 7043, "acc_step": 0, "speed/wps": 12900.689649474118, "speed/FLOPS": 202623058350268.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05843498930335045, "optim/lr": 0.0029898081212532553, "optim/total_tokens": 3692560384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.959829330444336, "created_at": "2025-01-16T01:26:43.249544+00:00"} {"global_step": 7044, "acc_step": 0, "speed/wps": 12899.554751587624, "speed/FLOPS": 202605233219448.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06731065362691879, "optim/lr": 0.002989798147891028, "optim/total_tokens": 3693084672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.1489484310150146, "created_at": "2025-01-16T01:26:53.417522+00:00"} {"global_step": 7045, "acc_step": 0, "speed/wps": 12899.389697711616, "speed/FLOPS": 202602640821517.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05883476510643959, "optim/lr": 0.002989788169668072, "optim/total_tokens": 3693608960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.961855888366699, "created_at": "2025-01-16T01:27:03.582089+00:00"} {"global_step": 7046, "acc_step": 0, "speed/wps": 12902.279748804463, "speed/FLOPS": 202648033045277.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054561421275138855, "optim/lr": 0.0029897781865844196, "optim/total_tokens": 3694133248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.973405122756958, "created_at": "2025-01-16T01:27:13.746736+00:00"} {"global_step": 7047, "acc_step": 0, "speed/wps": 12905.14987517553, "speed/FLOPS": 202693112323901.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06131347268819809, "optim/lr": 0.0029897681986401047, "optim/total_tokens": 3694657536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.963149309158325, "created_at": "2025-01-16T01:27:23.908243+00:00"} {"global_step": 7048, "acc_step": 0, "speed/wps": 12898.066074698307, "speed/FLOPS": 202581851503245.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0553552582859993, "optim/lr": 0.0029897582058351587, "optim/total_tokens": 3695181824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.0351643562316895, "created_at": "2025-01-16T01:27:34.074318+00:00"} {"global_step": 7049, "acc_step": 0, "speed/wps": 12898.46229672085, "speed/FLOPS": 202588074714575.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.05703100934624672, "optim/lr": 0.0029897482081696143, "optim/total_tokens": 3695706112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 2.8376030921936035, "created_at": "2025-01-16T01:27:44.239929+00:00"} {"global_step": 7050, "acc_step": 0, "speed/wps": 12902.384281257766, "speed/FLOPS": 202649674871099.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06637072563171387, "optim/lr": 0.002989738205643505, "optim/total_tokens": 3696230400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.9340853691101074, "created_at": "2025-01-16T01:27:54.413412+00:00"} {"global_step": 7051, "acc_step": 0, "speed/wps": 12898.219825849757, "speed/FLOPS": 202584266376354.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08670160174369812, "optim/lr": 0.0029897281982568624, "optim/total_tokens": 3696754688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.0726630687713623, "created_at": "2025-01-16T01:28:04.582870+00:00"} {"global_step": 7052, "acc_step": 0, "speed/wps": 12896.649419508023, "speed/FLOPS": 202559600986793.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08971482515335083, "optim/lr": 0.0029897181860097195, "optim/total_tokens": 3697278976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.108757495880127, "created_at": "2025-01-16T01:28:14.753867+00:00"} {"global_step": 7053, "acc_step": 0, "speed/wps": 12903.3537847196, "speed/FLOPS": 202664902255204.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061926547437906265, "optim/lr": 0.002989708168902109, "optim/total_tokens": 3697803264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.966235876083374, "created_at": "2025-01-16T01:28:24.915440+00:00"} {"global_step": 7054, "acc_step": 0, "speed/wps": 12900.41186129666, "speed/FLOPS": 202618695305220.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06712689250707626, "optim/lr": 0.002989698146934064, "optim/total_tokens": 3698327552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414437, "loss/out": 2.9694817066192627, "created_at": "2025-01-16T01:28:35.081826+00:00"} {"global_step": 7055, "acc_step": 0, "speed/wps": 12898.696891594409, "speed/FLOPS": 202591759349431.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07535549253225327, "optim/lr": 0.002989688120105617, "optim/total_tokens": 3698851840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.950943946838379, "created_at": "2025-01-16T01:28:45.247537+00:00"} {"global_step": 7056, "acc_step": 0, "speed/wps": 12901.381309613294, "speed/FLOPS": 202633921823195.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051253076642751694, "optim/lr": 0.0029896780884168, "optim/total_tokens": 3699376128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.0629312992095947, "created_at": "2025-01-16T01:28:55.413754+00:00"} {"global_step": 7057, "acc_step": 0, "speed/wps": 12901.549455177255, "speed/FLOPS": 202636562780333.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05970363691449165, "optim/lr": 0.002989668051867647, "optim/total_tokens": 3699900416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434742, "loss/out": 2.9571971893310547, "created_at": "2025-01-16T01:29:05.576815+00:00"} {"global_step": 7058, "acc_step": 0, "speed/wps": 12897.353283648748, "speed/FLOPS": 202570656140334.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055623993277549744, "optim/lr": 0.002989658010458189, "optim/total_tokens": 3700424704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.926149368286133, "created_at": "2025-01-16T01:29:15.743017+00:00"} {"global_step": 7059, "acc_step": 0, "speed/wps": 12894.115262608828, "speed/FLOPS": 202519798570395.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0702212005853653, "optim/lr": 0.0029896479641884606, "optim/total_tokens": 3700948992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 2.9455301761627197, "created_at": "2025-01-16T01:29:25.916218+00:00"} {"global_step": 7060, "acc_step": 0, "speed/wps": 12899.734197876569, "speed/FLOPS": 202608051670001.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05943018198013306, "optim/lr": 0.0029896379130584932, "optim/total_tokens": 3701473280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 3.0761897563934326, "created_at": "2025-01-16T01:29:36.080714+00:00"} {"global_step": 7061, "acc_step": 0, "speed/wps": 12894.401960237534, "speed/FLOPS": 202524301550617.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08408573269844055, "optim/lr": 0.0029896278570683207, "optim/total_tokens": 3701997568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454695, "loss/out": 2.8876194953918457, "created_at": "2025-01-16T01:29:46.252665+00:00"} {"global_step": 7062, "acc_step": 0, "speed/wps": 12898.571172023909, "speed/FLOPS": 202589784750820.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11972630023956299, "optim/lr": 0.0029896177962179748, "optim/total_tokens": 3702521856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.0638203620910645, "created_at": "2025-01-16T01:29:56.424335+00:00"} {"global_step": 7063, "acc_step": 0, "speed/wps": 12898.496546730681, "speed/FLOPS": 202588612658044.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058034345507621765, "optim/lr": 0.002989607730507489, "optim/total_tokens": 3703046144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.028423547744751, "created_at": "2025-01-16T01:30:06.593885+00:00"} {"global_step": 7064, "acc_step": 0, "speed/wps": 12899.871379738515, "speed/FLOPS": 202610206299650.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0929446816444397, "optim/lr": 0.002989597659936896, "optim/total_tokens": 3703570432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.964484453201294, "created_at": "2025-01-16T01:30:16.759072+00:00"} {"global_step": 7065, "acc_step": 0, "speed/wps": 12903.297614744233, "speed/FLOPS": 202664020028555.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07556895166635513, "optim/lr": 0.002989587584506229, "optim/total_tokens": 3704094720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420607, "loss/out": 2.943441152572632, "created_at": "2025-01-16T01:30:26.920758+00:00"} {"global_step": 7066, "acc_step": 0, "speed/wps": 12898.119663595911, "speed/FLOPS": 202582693190520.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0594804510474205, "optim/lr": 0.0029895775042155207, "optim/total_tokens": 3704619008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.931518793106079, "created_at": "2025-01-16T01:30:37.089802+00:00"} {"global_step": 7067, "acc_step": 0, "speed/wps": 12894.722131465996, "speed/FLOPS": 202529330279722.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05985088646411896, "optim/lr": 0.0029895674190648035, "optim/total_tokens": 3705143296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.924522876739502, "created_at": "2025-01-16T01:30:47.259676+00:00"} {"global_step": 7068, "acc_step": 0, "speed/wps": 12898.822633713025, "speed/FLOPS": 202593734302192.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0548432394862175, "optim/lr": 0.0029895573290541107, "optim/total_tokens": 3705667584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.009873151779175, "created_at": "2025-01-16T01:30:57.424680+00:00"} {"global_step": 7069, "acc_step": 0, "speed/wps": 12899.216995217055, "speed/FLOPS": 202599928291522.53, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06198795884847641, "optim/lr": 0.002989547234183475, "optim/total_tokens": 3706191872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.960670232772827, "created_at": "2025-01-16T01:31:07.595735+00:00"} {"global_step": 7070, "acc_step": 0, "speed/wps": 12900.585876458827, "speed/FLOPS": 202621428452464.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07057646661996841, "optim/lr": 0.0029895371344529297, "optim/total_tokens": 3706716160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.057258129119873, "created_at": "2025-01-16T01:31:17.761453+00:00"} {"global_step": 7071, "acc_step": 0, "speed/wps": 12902.26050297163, "speed/FLOPS": 202647730763026.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07257897406816483, "optim/lr": 0.0029895270298625075, "optim/total_tokens": 3707240448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 2.9425601959228516, "created_at": "2025-01-16T01:31:27.925966+00:00"} {"global_step": 7072, "acc_step": 0, "speed/wps": 12894.551837606685, "speed/FLOPS": 202526655580653.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06146387383341789, "optim/lr": 0.0029895169204122414, "optim/total_tokens": 3707764736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 3.1316285133361816, "created_at": "2025-01-16T01:31:38.095525+00:00"} {"global_step": 7073, "acc_step": 0, "speed/wps": 12899.735662827516, "speed/FLOPS": 202608074679068.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057797569781541824, "optim/lr": 0.0029895068061021645, "optim/total_tokens": 3708289024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365889, "loss/out": 2.922438383102417, "created_at": "2025-01-16T01:31:48.261932+00:00"} {"global_step": 7074, "acc_step": 0, "speed/wps": 12895.283920910248, "speed/FLOPS": 202538153954923.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05895369499921799, "optim/lr": 0.0029894966869323096, "optim/total_tokens": 3708813312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.9764890670776367, "created_at": "2025-01-16T01:31:58.429847+00:00"} {"global_step": 7075, "acc_step": 0, "speed/wps": 12893.482673062228, "speed/FLOPS": 202509862882296.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06929966062307358, "optim/lr": 0.0029894865629027097, "optim/total_tokens": 3709337600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.936424732208252, "created_at": "2025-01-16T01:32:08.607883+00:00"} {"global_step": 7076, "acc_step": 0, "speed/wps": 12896.428938393503, "speed/FLOPS": 202556138027917.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05743707716464996, "optim/lr": 0.0029894764340133986, "optim/total_tokens": 3709861888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 3.03350830078125, "created_at": "2025-01-16T01:32:18.779218+00:00"} {"global_step": 7077, "acc_step": 0, "speed/wps": 12898.132737297703, "speed/FLOPS": 202582898530972.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07090575248003006, "optim/lr": 0.002989466300264408, "optim/total_tokens": 3710386176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.9878182411193848, "created_at": "2025-01-16T01:32:28.944805+00:00"} {"global_step": 7078, "acc_step": 0, "speed/wps": 12902.25039807483, "speed/FLOPS": 202647572051736.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06350845098495483, "optim/lr": 0.002989456161655772, "optim/total_tokens": 3710910464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.034271717071533, "created_at": "2025-01-16T01:32:39.107227+00:00"} {"global_step": 7079, "acc_step": 0, "speed/wps": 12900.266200316319, "speed/FLOPS": 202616407499365.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053643085062503815, "optim/lr": 0.0029894460181875233, "optim/total_tokens": 3711434752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333702, "loss/out": 3.15720534324646, "created_at": "2025-01-16T01:32:49.272839+00:00"} {"global_step": 7080, "acc_step": 0, "speed/wps": 12900.843432434063, "speed/FLOPS": 202625473722974.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06106232479214668, "optim/lr": 0.0029894358698596946, "optim/total_tokens": 3711959040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 3.1000890731811523, "created_at": "2025-01-16T01:32:59.436615+00:00"} {"global_step": 7081, "acc_step": 0, "speed/wps": 12899.338407277673, "speed/FLOPS": 202601835234772.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0525670200586319, "optim/lr": 0.00298942571667232, "optim/total_tokens": 3712483328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.9668385982513428, "created_at": "2025-01-16T01:33:09.601232+00:00"} {"global_step": 7082, "acc_step": 0, "speed/wps": 12897.059341055005, "speed/FLOPS": 202566039367980.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.047595471143722534, "optim/lr": 0.002989415558625432, "optim/total_tokens": 3713007616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.0065369606018066, "created_at": "2025-01-16T01:33:19.771103+00:00"} {"global_step": 7083, "acc_step": 0, "speed/wps": 12902.42673511323, "speed/FLOPS": 202650341667236.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050599999725818634, "optim/lr": 0.002989405395719064, "optim/total_tokens": 3713531904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.996936082839966, "created_at": "2025-01-16T01:33:29.935216+00:00"} {"global_step": 7084, "acc_step": 0, "speed/wps": 12899.585688575195, "speed/FLOPS": 202605719127349.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05728877708315849, "optim/lr": 0.0029893952279532485, "optim/total_tokens": 3714056192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 3.025369644165039, "created_at": "2025-01-16T01:33:40.099887+00:00"} {"global_step": 7085, "acc_step": 0, "speed/wps": 12900.278728523619, "speed/FLOPS": 202616604272076.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05554501339793205, "optim/lr": 0.0029893850553280197, "optim/total_tokens": 3714580480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.0351850986480713, "created_at": "2025-01-16T01:33:50.263860+00:00"} {"global_step": 7086, "acc_step": 0, "speed/wps": 12893.375030589777, "speed/FLOPS": 202508172209350.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06034151092171669, "optim/lr": 0.0029893748778434094, "optim/total_tokens": 3715104768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 2.9380106925964355, "created_at": "2025-01-16T01:34:00.443329+00:00"} {"global_step": 7087, "acc_step": 0, "speed/wps": 12897.350108988183, "speed/FLOPS": 202570606277927.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05724630504846573, "optim/lr": 0.0029893646954994527, "optim/total_tokens": 3715629056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.193403720855713, "created_at": "2025-01-16T01:34:10.610390+00:00"} {"global_step": 7088, "acc_step": 0, "speed/wps": 12897.940405249368, "speed/FLOPS": 202579877691861.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05458683893084526, "optim/lr": 0.0029893545082961812, "optim/total_tokens": 3716153344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.9972519874572754, "created_at": "2025-01-16T01:34:20.778905+00:00"} {"global_step": 7089, "acc_step": 0, "speed/wps": 12898.955445606594, "speed/FLOPS": 202595820295483.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07322760671377182, "optim/lr": 0.002989344316233629, "optim/total_tokens": 3716677632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 3.0577845573425293, "created_at": "2025-01-16T01:34:30.944311+00:00"} {"global_step": 7090, "acc_step": 0, "speed/wps": 12895.203856050102, "speed/FLOPS": 202536896426274.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06262025982141495, "optim/lr": 0.0029893341193118284, "optim/total_tokens": 3717201920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.0414044857025146, "created_at": "2025-01-16T01:34:41.112587+00:00"} {"global_step": 7091, "acc_step": 0, "speed/wps": 12898.019231025935, "speed/FLOPS": 202581115759000.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055630456656217575, "optim/lr": 0.0029893239175308134, "optim/total_tokens": 3717726208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.0045204162597656, "created_at": "2025-01-16T01:34:51.280469+00:00"} {"global_step": 7092, "acc_step": 0, "speed/wps": 12898.188929693695, "speed/FLOPS": 202583781109768.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0680350661277771, "optim/lr": 0.002989313710890618, "optim/total_tokens": 3718250496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.9980146884918213, "created_at": "2025-01-16T01:35:01.446334+00:00"} {"global_step": 7093, "acc_step": 0, "speed/wps": 12890.88631280931, "speed/FLOPS": 202469083476751.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05989091470837593, "optim/lr": 0.0029893034993912744, "optim/total_tokens": 3718774784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.1130716800689697, "created_at": "2025-01-16T01:35:11.627828+00:00"} {"global_step": 7094, "acc_step": 0, "speed/wps": 12896.419149886082, "speed/FLOPS": 202555984285957.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0598805770277977, "optim/lr": 0.0029892932830328156, "optim/total_tokens": 3719299072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 2.9695076942443848, "created_at": "2025-01-16T01:35:21.794826+00:00"} {"global_step": 7095, "acc_step": 0, "speed/wps": 12898.708595140426, "speed/FLOPS": 202591943169704.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0624842643737793, "optim/lr": 0.0029892830618152766, "optim/total_tokens": 3719823360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.952549934387207, "created_at": "2025-01-16T01:35:31.961055+00:00"} {"global_step": 7096, "acc_step": 0, "speed/wps": 12898.698981680664, "speed/FLOPS": 202591792177108.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06142483279109001, "optim/lr": 0.002989272835738689, "optim/total_tokens": 3720347648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.122480869293213, "created_at": "2025-01-16T01:35:42.128792+00:00"} {"global_step": 7097, "acc_step": 0, "speed/wps": 12897.51249027109, "speed/FLOPS": 202573156699110.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06459186226129532, "optim/lr": 0.0029892626048030873, "optim/total_tokens": 3720871936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284736, "loss/out": 2.9731926918029785, "created_at": "2025-01-16T01:35:52.297726+00:00"} {"global_step": 7098, "acc_step": 0, "speed/wps": 12898.43630641266, "speed/FLOPS": 202587666501070.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06166050583124161, "optim/lr": 0.002989252369008504, "optim/total_tokens": 3721396224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0653605461120605, "created_at": "2025-01-16T01:36:02.463255+00:00"} {"global_step": 7099, "acc_step": 0, "speed/wps": 12891.628864105049, "speed/FLOPS": 202480746265223.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05944019928574562, "optim/lr": 0.002989242128354973, "optim/total_tokens": 3721920512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.900303602218628, "created_at": "2025-01-16T01:36:12.634224+00:00"} {"global_step": 7100, "acc_step": 0, "speed/wps": 12895.009413750537, "speed/FLOPS": 202533842442768.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06707789748907089, "optim/lr": 0.002989231882842528, "optim/total_tokens": 3722444800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.042726516723633, "created_at": "2025-01-16T01:36:22.802247+00:00"} {"global_step": 7101, "acc_step": 0, "speed/wps": 12896.104295929388, "speed/FLOPS": 202551039071912.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07923818379640579, "optim/lr": 0.0029892216324712015, "optim/total_tokens": 3722969088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.934443235397339, "created_at": "2025-01-16T01:36:32.970873+00:00"} {"global_step": 7102, "acc_step": 0, "speed/wps": 12896.237063667431, "speed/FLOPS": 202553124371680.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05973958969116211, "optim/lr": 0.002989211377241028, "optim/total_tokens": 3723493376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407300, "loss/out": 2.9675774574279785, "created_at": "2025-01-16T01:36:43.138384+00:00"} {"global_step": 7103, "acc_step": 0, "speed/wps": 12898.374656392507, "speed/FLOPS": 202586698202788.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07180207222700119, "optim/lr": 0.0029892011171520406, "optim/total_tokens": 3724017664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.0711710453033447, "created_at": "2025-01-16T01:36:53.304819+00:00"} {"global_step": 7104, "acc_step": 0, "speed/wps": 12900.9994021348, "speed/FLOPS": 202627923441449.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05630454421043396, "optim/lr": 0.0029891908522042725, "optim/total_tokens": 3724541952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.920215606689453, "created_at": "2025-01-16T01:37:03.468251+00:00"} {"global_step": 7105, "acc_step": 0, "speed/wps": 12896.479637261844, "speed/FLOPS": 202556934323311.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056623946875333786, "optim/lr": 0.0029891805823977573, "optim/total_tokens": 3725066240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 3.029059648513794, "created_at": "2025-01-16T01:37:13.635005+00:00"} {"global_step": 7106, "acc_step": 0, "speed/wps": 12901.009427020614, "speed/FLOPS": 202628080896056.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06541799008846283, "optim/lr": 0.0029891703077325293, "optim/total_tokens": 3725590528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.991628646850586, "created_at": "2025-01-16T01:37:23.801557+00:00"} {"global_step": 7107, "acc_step": 0, "speed/wps": 12898.884912222356, "speed/FLOPS": 202594712471760.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08065769076347351, "optim/lr": 0.0029891600282086206, "optim/total_tokens": 3726114816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.850318431854248, "created_at": "2025-01-16T01:37:33.969504+00:00"} {"global_step": 7108, "acc_step": 0, "speed/wps": 12898.145351199122, "speed/FLOPS": 202583096649627.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05963568016886711, "optim/lr": 0.0029891497438260654, "optim/total_tokens": 3726639104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.060145854949951, "created_at": "2025-01-16T01:37:44.143500+00:00"} {"global_step": 7109, "acc_step": 0, "speed/wps": 12898.299170438351, "speed/FLOPS": 202585512592149.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047689907252788544, "optim/lr": 0.0029891394545848975, "optim/total_tokens": 3727163392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9369125366210938, "created_at": "2025-01-16T01:37:54.310111+00:00"} {"global_step": 7110, "acc_step": 0, "speed/wps": 12898.518305209896, "speed/FLOPS": 202588954404859.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07020889967679977, "optim/lr": 0.00298912916048515, "optim/total_tokens": 3727687680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 3.0250654220581055, "created_at": "2025-01-16T01:38:04.476314+00:00"} {"global_step": 7111, "acc_step": 0, "speed/wps": 12896.67738095707, "speed/FLOPS": 202560040159773.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.059804096817970276, "optim/lr": 0.002989118861526857, "optim/total_tokens": 3728211968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9747724533081055, "created_at": "2025-01-16T01:38:14.643064+00:00"} {"global_step": 7112, "acc_step": 0, "speed/wps": 12895.632481606868, "speed/FLOPS": 202543628579636.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05199314281344414, "optim/lr": 0.002989108557710052, "optim/total_tokens": 3728736256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9096667766571045, "created_at": "2025-01-16T01:38:24.812410+00:00"} {"global_step": 7113, "acc_step": 0, "speed/wps": 12899.015410584469, "speed/FLOPS": 202596762127861.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07091884315013885, "optim/lr": 0.002989098249034768, "optim/total_tokens": 3729260544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9796266555786133, "created_at": "2025-01-16T01:38:34.978499+00:00"} {"global_step": 7114, "acc_step": 0, "speed/wps": 12896.563112048514, "speed/FLOPS": 202558245409542.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06085428223013878, "optim/lr": 0.0029890879355010396, "optim/total_tokens": 3729784832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0330605506896973, "created_at": "2025-01-16T01:38:45.147765+00:00"} {"global_step": 7115, "acc_step": 0, "speed/wps": 12894.739328628924, "speed/FLOPS": 202529600384797.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0536823645234108, "optim/lr": 0.0029890776171088993, "optim/total_tokens": 3730309120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 3.1300582885742188, "created_at": "2025-01-16T01:38:55.317119+00:00"} {"global_step": 7116, "acc_step": 0, "speed/wps": 12896.590865128393, "speed/FLOPS": 202558681309799.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06856252253055573, "optim/lr": 0.002989067293858382, "optim/total_tokens": 3730833408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.986243963241577, "created_at": "2025-01-16T01:39:05.483977+00:00"} {"global_step": 7117, "acc_step": 0, "speed/wps": 12893.017901121832, "speed/FLOPS": 202502563000309.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06269340217113495, "optim/lr": 0.0029890569657495206, "optim/total_tokens": 3731357696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.960947275161743, "created_at": "2025-01-16T01:39:15.656643+00:00"} {"global_step": 7118, "acc_step": 0, "speed/wps": 12890.565440616669, "speed/FLOPS": 202464043738041.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.07039470225572586, "optim/lr": 0.002989046632782349, "optim/total_tokens": 3731881984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428597, "loss/out": 3.0471599102020264, "created_at": "2025-01-16T01:39:25.830358+00:00"} {"global_step": 7119, "acc_step": 0, "speed/wps": 12893.788946984223, "speed/FLOPS": 202514673335106.0, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06251628696918488, "optim/lr": 0.0029890362949569005, "optim/total_tokens": 3732406272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9674336910247803, "created_at": "2025-01-16T01:39:36.000978+00:00"} {"global_step": 7120, "acc_step": 0, "speed/wps": 12899.536718676856, "speed/FLOPS": 202604949987804.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05824283882975578, "optim/lr": 0.00298902595227321, "optim/total_tokens": 3732930560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471388, "loss/out": 2.92791485786438, "created_at": "2025-01-16T01:39:46.167545+00:00"} {"global_step": 7121, "acc_step": 0, "speed/wps": 12894.66539627617, "speed/FLOPS": 202528439175603.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07502832263708115, "optim/lr": 0.00298901560473131, "optim/total_tokens": 3733454848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 2.98186993598938, "created_at": "2025-01-16T01:39:56.338782+00:00"} {"global_step": 7122, "acc_step": 0, "speed/wps": 12893.09969048403, "speed/FLOPS": 202503847614632.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060488663613796234, "optim/lr": 0.0029890052523312348, "optim/total_tokens": 3733979136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.0081849098205566, "created_at": "2025-01-16T01:40:06.512236+00:00"} {"global_step": 7123, "acc_step": 0, "speed/wps": 12894.971550348228, "speed/FLOPS": 202533247746005.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06260287016630173, "optim/lr": 0.0029889948950730183, "optim/total_tokens": 3734503424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.0363125801086426, "created_at": "2025-01-16T01:40:16.687691+00:00"} {"global_step": 7124, "acc_step": 0, "speed/wps": 12901.43507981082, "speed/FLOPS": 202634766358035.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07124792784452438, "optim/lr": 0.002988984532956694, "optim/total_tokens": 3735027712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 2.9822990894317627, "created_at": "2025-01-16T01:40:26.856278+00:00"} {"global_step": 7125, "acc_step": 0, "speed/wps": 12900.197648572557, "speed/FLOPS": 202615330800029.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06732781231403351, "optim/lr": 0.0029889741659822958, "optim/total_tokens": 3735552000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.067802667617798, "created_at": "2025-01-16T01:40:37.026203+00:00"} {"global_step": 7126, "acc_step": 0, "speed/wps": 12897.856825162786, "speed/FLOPS": 202578564951750.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06047017499804497, "optim/lr": 0.0029889637941498577, "optim/total_tokens": 3736076288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.947911262512207, "created_at": "2025-01-16T01:40:47.197412+00:00"} {"global_step": 7127, "acc_step": 0, "speed/wps": 12899.525971313025, "speed/FLOPS": 202604781185687.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0635184496641159, "optim/lr": 0.002988953417459413, "optim/total_tokens": 3736600576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.9948391914367676, "created_at": "2025-01-16T01:40:57.364561+00:00"} {"global_step": 7128, "acc_step": 0, "speed/wps": 12894.406419979367, "speed/FLOPS": 202524371596991.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055970657616853714, "optim/lr": 0.002988943035910996, "optim/total_tokens": 3737124864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.838758707046509, "created_at": "2025-01-16T01:41:07.537260+00:00"} {"global_step": 7129, "acc_step": 0, "speed/wps": 12902.171328957103, "speed/FLOPS": 202646330162589.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05811842903494835, "optim/lr": 0.0029889326495046407, "optim/total_tokens": 3737649152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 3.0671000480651855, "created_at": "2025-01-16T01:41:17.699899+00:00"} {"global_step": 7130, "acc_step": 0, "speed/wps": 12897.389143474928, "speed/FLOPS": 202571219368180.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06126246973872185, "optim/lr": 0.0029889222582403807, "optim/total_tokens": 3738173440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.027395486831665, "created_at": "2025-01-16T01:41:27.870103+00:00"} {"global_step": 7131, "acc_step": 0, "speed/wps": 12900.73832775891, "speed/FLOPS": 202623822909622.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05422820895910263, "optim/lr": 0.0029889118621182504, "optim/total_tokens": 3738697728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.039411783218384, "created_at": "2025-01-16T01:41:38.033675+00:00"} {"global_step": 7132, "acc_step": 0, "speed/wps": 12902.60628615694, "speed/FLOPS": 202653161763107.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07062772661447525, "optim/lr": 0.002988901461138283, "optim/total_tokens": 3739222016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.0786919593811035, "created_at": "2025-01-16T01:41:48.196366+00:00"} {"global_step": 7133, "acc_step": 0, "speed/wps": 12892.584364907312, "speed/FLOPS": 202495753718318.88, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07625555247068405, "optim/lr": 0.002988891055300512, "optim/total_tokens": 3739746304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.0098319053649902, "created_at": "2025-01-16T01:41:58.366273+00:00"} {"global_step": 7134, "acc_step": 0, "speed/wps": 12902.798513702224, "speed/FLOPS": 202656180960854.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06226632744073868, "optim/lr": 0.0029888806446049733, "optim/total_tokens": 3740270592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 3.04226016998291, "created_at": "2025-01-16T01:42:08.530611+00:00"} {"global_step": 7135, "acc_step": 0, "speed/wps": 12896.504656539799, "speed/FLOPS": 202557327285452.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06779599189758301, "optim/lr": 0.002988870229051699, "optim/total_tokens": 3740794880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.0113277435302734, "created_at": "2025-01-16T01:42:18.697502+00:00"} {"global_step": 7136, "acc_step": 0, "speed/wps": 12897.597604824738, "speed/FLOPS": 202574493540135.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05855564400553703, "optim/lr": 0.0029888598086407242, "optim/total_tokens": 3741319168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9518027305603027, "created_at": "2025-01-16T01:42:28.863507+00:00"} {"global_step": 7137, "acc_step": 0, "speed/wps": 12898.056364085462, "speed/FLOPS": 202581698984726.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06037390977144241, "optim/lr": 0.0029888493833720824, "optim/total_tokens": 3741843456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 3.003967046737671, "created_at": "2025-01-16T01:42:39.030341+00:00"} {"global_step": 7138, "acc_step": 0, "speed/wps": 12900.566029047299, "speed/FLOPS": 202621116721593.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06628639996051788, "optim/lr": 0.0029888389532458075, "optim/total_tokens": 3742367744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.909231662750244, "created_at": "2025-01-16T01:42:49.226862+00:00"} {"global_step": 7139, "acc_step": 0, "speed/wps": 12899.142237131544, "speed/FLOPS": 202598754113061.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06398393213748932, "optim/lr": 0.002988828518261934, "optim/total_tokens": 3742892032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9282472133636475, "created_at": "2025-01-16T01:42:59.392820+00:00"} {"global_step": 7140, "acc_step": 0, "speed/wps": 12897.148239952534, "speed/FLOPS": 202567435647325.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06637230515480042, "optim/lr": 0.0029888180784204948, "optim/total_tokens": 3743416320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 2.942073345184326, "created_at": "2025-01-16T01:43:09.561723+00:00"} {"global_step": 7141, "acc_step": 0, "speed/wps": 12899.364085547992, "speed/FLOPS": 202602238547293.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061674922704696655, "optim/lr": 0.0029888076337215254, "optim/total_tokens": 3743940608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.969998836517334, "created_at": "2025-01-16T01:43:19.728578+00:00"} {"global_step": 7142, "acc_step": 0, "speed/wps": 12900.042615816474, "speed/FLOPS": 202612895797558.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07303165644407272, "optim/lr": 0.002988797184165059, "optim/total_tokens": 3744464896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.938295841217041, "created_at": "2025-01-16T01:43:29.893470+00:00"} {"global_step": 7143, "acc_step": 0, "speed/wps": 12897.99289852779, "speed/FLOPS": 202580702170932.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05534639582037926, "optim/lr": 0.0029887867297511304, "optim/total_tokens": 3744989184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308327, "loss/out": 3.0197174549102783, "created_at": "2025-01-16T01:43:40.059163+00:00"} {"global_step": 7144, "acc_step": 0, "speed/wps": 12897.548436084056, "speed/FLOPS": 202573721277498.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05707382410764694, "optim/lr": 0.002988776270479773, "optim/total_tokens": 3745513472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.065615177154541, "created_at": "2025-01-16T01:43:50.228473+00:00"} {"global_step": 7145, "acc_step": 0, "speed/wps": 12899.402776352792, "speed/FLOPS": 202602846239549.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0702163577079773, "optim/lr": 0.0029887658063510215, "optim/total_tokens": 3746037760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.911452531814575, "created_at": "2025-01-16T01:44:00.398391+00:00"} {"global_step": 7146, "acc_step": 0, "speed/wps": 12894.42699744535, "speed/FLOPS": 202524694794371.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05192554369568825, "optim/lr": 0.0029887553373649095, "optim/total_tokens": 3746562048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.093108654022217, "created_at": "2025-01-16T01:44:10.570925+00:00"} {"global_step": 7147, "acc_step": 0, "speed/wps": 12887.356267066998, "speed/FLOPS": 202413639257577.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08382928371429443, "optim/lr": 0.0029887448635214716, "optim/total_tokens": 3747086336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 2.8465490341186523, "created_at": "2025-01-16T01:44:20.749170+00:00"} {"global_step": 7148, "acc_step": 0, "speed/wps": 12898.167490304753, "speed/FLOPS": 202583444374703.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07677856832742691, "optim/lr": 0.0029887343848207415, "optim/total_tokens": 3747610624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0599796772003174, "created_at": "2025-01-16T01:44:30.919137+00:00"} {"global_step": 7149, "acc_step": 0, "speed/wps": 12900.737819262637, "speed/FLOPS": 202623814922989.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05007755011320114, "optim/lr": 0.002988723901262754, "optim/total_tokens": 3748134912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 3.0412373542785645, "created_at": "2025-01-16T01:44:41.082869+00:00"} {"global_step": 7150, "acc_step": 0, "speed/wps": 12898.120058407132, "speed/FLOPS": 202582699391573.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08826285600662231, "optim/lr": 0.0029887134128475427, "optim/total_tokens": 3748659200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.982409954071045, "created_at": "2025-01-16T01:44:51.248440+00:00"} {"global_step": 7151, "acc_step": 0, "speed/wps": 12899.232347019462, "speed/FLOPS": 202600169412674.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06403697282075882, "optim/lr": 0.002988702919575142, "optim/total_tokens": 3749183488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.020879030227661, "created_at": "2025-01-16T01:45:01.416405+00:00"} {"global_step": 7152, "acc_step": 0, "speed/wps": 12894.078334163032, "speed/FLOPS": 202519218558408.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05843173339962959, "optim/lr": 0.0029886924214455866, "optim/total_tokens": 3749707776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.9291255474090576, "created_at": "2025-01-16T01:45:11.586689+00:00"} {"global_step": 7153, "acc_step": 0, "speed/wps": 12896.98857430517, "speed/FLOPS": 202564927878930.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06441432982683182, "optim/lr": 0.00298868191845891, "optim/total_tokens": 3750232064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.08602237701416, "created_at": "2025-01-16T01:45:21.756510+00:00"} {"global_step": 7154, "acc_step": 0, "speed/wps": 12890.410259118285, "speed/FLOPS": 202461606399367.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06040121987462044, "optim/lr": 0.002988671410615147, "optim/total_tokens": 3750756352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 3.072265625, "created_at": "2025-01-16T01:45:31.928596+00:00"} {"global_step": 7155, "acc_step": 0, "speed/wps": 12897.10994205605, "speed/FLOPS": 202566834126234.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05248509347438812, "optim/lr": 0.002988660897914332, "optim/total_tokens": 3751280640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9775991439819336, "created_at": "2025-01-16T01:45:42.101951+00:00"} {"global_step": 7156, "acc_step": 0, "speed/wps": 12896.19384224069, "speed/FLOPS": 202552445519781.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0669146403670311, "optim/lr": 0.002988650380356499, "optim/total_tokens": 3751804928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.0810866355895996, "created_at": "2025-01-16T01:45:52.272104+00:00"} {"global_step": 7157, "acc_step": 0, "speed/wps": 12894.117195274195, "speed/FLOPS": 202519828925560.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07571574300527573, "optim/lr": 0.0029886398579416825, "optim/total_tokens": 3752329216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364255, "loss/out": 3.003800868988037, "created_at": "2025-01-16T01:46:02.441134+00:00"} {"global_step": 7158, "acc_step": 0, "speed/wps": 12893.155637055906, "speed/FLOPS": 202504726332422.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058608099818229675, "optim/lr": 0.002988629330669916, "optim/total_tokens": 3752853504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9205527305603027, "created_at": "2025-01-16T01:46:12.610774+00:00"} {"global_step": 7159, "acc_step": 0, "speed/wps": 12891.368029507965, "speed/FLOPS": 202476649499452.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062374260276556015, "optim/lr": 0.0029886187985412355, "optim/total_tokens": 3753377792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9821102619171143, "created_at": "2025-01-16T01:46:22.782403+00:00"} {"global_step": 7160, "acc_step": 0, "speed/wps": 12899.273197060886, "speed/FLOPS": 202600811018709.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060370102524757385, "optim/lr": 0.002988608261555673, "optim/total_tokens": 3753902080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.047682285308838, "created_at": "2025-01-16T01:46:32.947668+00:00"} {"global_step": 7161, "acc_step": 0, "speed/wps": 12899.157120581167, "speed/FLOPS": 202598987878089.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06834375113248825, "optim/lr": 0.002988597719713265, "optim/total_tokens": 3754426368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9943313598632812, "created_at": "2025-01-16T01:46:43.115562+00:00"} {"global_step": 7162, "acc_step": 0, "speed/wps": 12895.958498235985, "speed/FLOPS": 202548749118790.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07157789915800095, "optim/lr": 0.0029885871730140454, "optim/total_tokens": 3754950656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414480, "loss/out": 2.9415669441223145, "created_at": "2025-01-16T01:46:53.283315+00:00"} {"global_step": 7163, "acc_step": 0, "speed/wps": 12893.280961690187, "speed/FLOPS": 202506694728018.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06912799179553986, "optim/lr": 0.0029885766214580484, "optim/total_tokens": 3755474944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.930023193359375, "created_at": "2025-01-16T01:47:03.455188+00:00"} {"global_step": 7164, "acc_step": 0, "speed/wps": 12889.265366905312, "speed/FLOPS": 202443624293916.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05417061969637871, "optim/lr": 0.0029885660650453083, "optim/total_tokens": 3755999232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9185843467712402, "created_at": "2025-01-16T01:47:13.627842+00:00"} {"global_step": 7165, "acc_step": 0, "speed/wps": 12897.86914420067, "speed/FLOPS": 202578758439168.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08317263424396515, "optim/lr": 0.0029885555037758595, "optim/total_tokens": 3756523520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.997124671936035, "created_at": "2025-01-16T01:47:23.794913+00:00"} {"global_step": 7166, "acc_step": 0, "speed/wps": 12903.707063432134, "speed/FLOPS": 202670450982840.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07557246088981628, "optim/lr": 0.0029885449376497366, "optim/total_tokens": 3757047808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.0757555961608887, "created_at": "2025-01-16T01:47:33.956141+00:00"} {"global_step": 7167, "acc_step": 0, "speed/wps": 12893.190984540708, "speed/FLOPS": 202505281513244.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0572507306933403, "optim/lr": 0.002988534366666974, "optim/total_tokens": 3757572096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.066770553588867, "created_at": "2025-01-16T01:47:44.125721+00:00"} {"global_step": 7168, "acc_step": 0, "speed/wps": 12893.471297926439, "speed/FLOPS": 202509684220157.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07080302387475967, "optim/lr": 0.0029885237908276064, "optim/total_tokens": 3758096384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 2.94291090965271, "created_at": "2025-01-16T01:47:54.295447+00:00"} {"global_step": 7169, "acc_step": 0, "speed/wps": 12890.264754251884, "speed/FLOPS": 202459321045496.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05865549296140671, "optim/lr": 0.0029885132101316685, "optim/total_tokens": 3758620672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.9987902641296387, "created_at": "2025-01-16T01:48:04.467311+00:00"} {"global_step": 7170, "acc_step": 0, "speed/wps": 12898.607173388962, "speed/FLOPS": 202590350201729.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06601499766111374, "optim/lr": 0.002988502624579194, "optim/total_tokens": 3759144960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.0220813751220703, "created_at": "2025-01-16T01:48:14.634733+00:00"} {"global_step": 7171, "acc_step": 0, "speed/wps": 12898.205715282558, "speed/FLOPS": 202584044750507.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058311644941568375, "optim/lr": 0.002988492034170218, "optim/total_tokens": 3759669248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.9484829902648926, "created_at": "2025-01-16T01:48:24.800586+00:00"} {"global_step": 7172, "acc_step": 0, "speed/wps": 12891.567404314055, "speed/FLOPS": 202479780954751.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05741344392299652, "optim/lr": 0.0029884814389047752, "optim/total_tokens": 3760193536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.085516929626465, "created_at": "2025-01-16T01:48:34.971580+00:00"} {"global_step": 7173, "acc_step": 0, "speed/wps": 12897.330640292334, "speed/FLOPS": 202570300495306.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05251854658126831, "optim/lr": 0.0029884708387829, "optim/total_tokens": 3760717824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 2.9666736125946045, "created_at": "2025-01-16T01:48:45.139553+00:00"} {"global_step": 7174, "acc_step": 0, "speed/wps": 12896.011279075841, "speed/FLOPS": 202549578114409.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047239068895578384, "optim/lr": 0.0029884602338046268, "optim/total_tokens": 3761242112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9587132930755615, "created_at": "2025-01-16T01:48:55.306796+00:00"} {"global_step": 7175, "acc_step": 0, "speed/wps": 12900.04893704747, "speed/FLOPS": 202612995081177.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057298943400382996, "optim/lr": 0.00298844962396999, "optim/total_tokens": 3761766400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.004551410675049, "created_at": "2025-01-16T01:49:05.470791+00:00"} {"global_step": 7176, "acc_step": 0, "speed/wps": 12900.043149088808, "speed/FLOPS": 202612904173332.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062442924827337265, "optim/lr": 0.002988439009279025, "optim/total_tokens": 3762290688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.028463363647461, "created_at": "2025-01-16T01:49:15.640438+00:00"} {"global_step": 7177, "acc_step": 0, "speed/wps": 12896.993949416586, "speed/FLOPS": 202565012302441.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05060455575585365, "optim/lr": 0.002988428389731766, "optim/total_tokens": 3762814976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.986422538757324, "created_at": "2025-01-16T01:49:25.810117+00:00"} {"global_step": 7178, "acc_step": 0, "speed/wps": 12892.491285444024, "speed/FLOPS": 202494291777443.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051654741168022156, "optim/lr": 0.0029884177653282475, "optim/total_tokens": 3763339264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.018455982208252, "created_at": "2025-01-16T01:49:35.983789+00:00"} {"global_step": 7179, "acc_step": 0, "speed/wps": 12900.68382664685, "speed/FLOPS": 202622966894765.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06862783432006836, "optim/lr": 0.0029884071360685043, "optim/total_tokens": 3763863552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 2.978975534439087, "created_at": "2025-01-16T01:49:46.147754+00:00"} {"global_step": 7180, "acc_step": 0, "speed/wps": 12897.960826146551, "speed/FLOPS": 202580198430113.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05726471170783043, "optim/lr": 0.002988396501952571, "optim/total_tokens": 3764387840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9445652961730957, "created_at": "2025-01-16T01:49:56.315836+00:00"} {"global_step": 7181, "acc_step": 0, "speed/wps": 12897.481083354274, "speed/FLOPS": 202572663410324.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07329101115465164, "optim/lr": 0.0029883858629804823, "optim/total_tokens": 3764912128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303224, "loss/out": 2.943387508392334, "created_at": "2025-01-16T01:50:06.489660+00:00"} {"global_step": 7182, "acc_step": 0, "speed/wps": 12893.215420554756, "speed/FLOPS": 202505665314423.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05180232971906662, "optim/lr": 0.0029883752191522733, "optim/total_tokens": 3765436416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.0695900917053223, "created_at": "2025-01-16T01:50:16.666669+00:00"} {"global_step": 7183, "acc_step": 0, "speed/wps": 12900.162998139114, "speed/FLOPS": 202614786567357.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05865740776062012, "optim/lr": 0.0029883645704679787, "optim/total_tokens": 3765960704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.050173282623291, "created_at": "2025-01-16T01:50:26.833819+00:00"} {"global_step": 7184, "acc_step": 0, "speed/wps": 12898.086360056137, "speed/FLOPS": 202582170112664.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0588872954249382, "optim/lr": 0.002988353916927632, "optim/total_tokens": 3766484992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471752, "loss/out": 2.9661707878112793, "created_at": "2025-01-16T01:50:37.000427+00:00"} {"global_step": 7185, "acc_step": 0, "speed/wps": 12898.573091839446, "speed/FLOPS": 202589814904161.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061054717749357224, "optim/lr": 0.0029883432585312694, "optim/total_tokens": 3767009280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.005890130996704, "created_at": "2025-01-16T01:50:47.165770+00:00"} {"global_step": 7186, "acc_step": 0, "speed/wps": 12900.187637955241, "speed/FLOPS": 202615173569528.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058918897062540054, "optim/lr": 0.0029883325952789253, "optim/total_tokens": 3767533568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486334, "loss/out": 2.8988566398620605, "created_at": "2025-01-16T01:50:57.332145+00:00"} {"global_step": 7187, "acc_step": 0, "speed/wps": 12895.665911906877, "speed/FLOPS": 202544153648436.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06424393504858017, "optim/lr": 0.002988321927170634, "optim/total_tokens": 3768057856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0043463706970215, "created_at": "2025-01-16T01:51:07.499662+00:00"} {"global_step": 7188, "acc_step": 0, "speed/wps": 12891.583804240925, "speed/FLOPS": 202480038538138.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057102885097265244, "optim/lr": 0.002988311254206431, "optim/total_tokens": 3768582144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.98553204536438, "created_at": "2025-01-16T01:51:17.672317+00:00"} {"global_step": 7189, "acc_step": 0, "speed/wps": 12897.533024463595, "speed/FLOPS": 202573479216821.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07884103804826736, "optim/lr": 0.002988300576386351, "optim/total_tokens": 3769106432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.0738000869750977, "created_at": "2025-01-16T01:51:27.839454+00:00"} {"global_step": 7190, "acc_step": 0, "speed/wps": 12890.125808565008, "speed/FLOPS": 202457138712552.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08560260385274887, "optim/lr": 0.0029882898937104285, "optim/total_tokens": 3769630720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.968585968017578, "created_at": "2025-01-16T01:51:38.015735+00:00"} {"global_step": 7191, "acc_step": 0, "speed/wps": 12897.517460407238, "speed/FLOPS": 202573234761928.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09252975136041641, "optim/lr": 0.0029882792061786986, "optim/total_tokens": 3770155008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.0043203830718994, "created_at": "2025-01-16T01:51:48.181739+00:00"} {"global_step": 7192, "acc_step": 0, "speed/wps": 12899.129987779159, "speed/FLOPS": 202598561720149.62, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06282659620046616, "optim/lr": 0.0029882685137911963, "optim/total_tokens": 3770679296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398431, "loss/out": 2.9776599407196045, "created_at": "2025-01-16T01:51:58.352521+00:00"} {"global_step": 7193, "acc_step": 0, "speed/wps": 12899.909805182304, "speed/FLOPS": 202610809824047.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08802428841590881, "optim/lr": 0.0029882578165479554, "optim/total_tokens": 3771203584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.038553237915039, "created_at": "2025-01-16T01:52:08.520359+00:00"} {"global_step": 7194, "acc_step": 0, "speed/wps": 12896.181535856547, "speed/FLOPS": 202552252231107.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09503018856048584, "optim/lr": 0.002988247114449012, "optim/total_tokens": 3771727872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.058701992034912, "created_at": "2025-01-16T01:52:18.690368+00:00"} {"global_step": 7195, "acc_step": 0, "speed/wps": 12897.10780520937, "speed/FLOPS": 202566800564121.06, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056960318237543106, "optim/lr": 0.002988236407494401, "optim/total_tokens": 3772252160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292143, "loss/out": 2.9676742553710938, "created_at": "2025-01-16T01:52:28.860093+00:00"} {"global_step": 7196, "acc_step": 0, "speed/wps": 12898.337801660193, "speed/FLOPS": 202586119348573.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0816257894039154, "optim/lr": 0.0029882256956841567, "optim/total_tokens": 3772776448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.997234582901001, "created_at": "2025-01-16T01:52:39.026458+00:00"} {"global_step": 7197, "acc_step": 0, "speed/wps": 12897.827385807175, "speed/FLOPS": 202578102566216.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054873138666152954, "optim/lr": 0.0029882149790183142, "optim/total_tokens": 3773300736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.207911729812622, "created_at": "2025-01-16T01:52:49.195912+00:00"} {"global_step": 7198, "acc_step": 0, "speed/wps": 12899.937864721547, "speed/FLOPS": 202611250537669.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07077595591545105, "optim/lr": 0.002988204257496909, "optim/total_tokens": 3773825024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326001, "loss/out": 2.7991726398468018, "created_at": "2025-01-16T01:52:59.361303+00:00"} {"global_step": 7199, "acc_step": 0, "speed/wps": 12899.277954561114, "speed/FLOPS": 202600885741787.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06745412945747375, "optim/lr": 0.0029881935311199762, "optim/total_tokens": 3774349312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.1628811359405518, "created_at": "2025-01-16T01:53:09.531872+00:00"} {"global_step": 7200, "acc_step": 0, "speed/wps": 12896.148520026103, "speed/FLOPS": 202551733672121.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05566541105508804, "optim/lr": 0.0029881827998875492, "optim/total_tokens": 3774873600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.993788719177246, "created_at": "2025-01-16T01:53:19.703132+00:00"} {"global_step": 7201, "acc_step": 0, "speed/wps": 12900.957048731128, "speed/FLOPS": 202627258223045.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07160407304763794, "optim/lr": 0.002988172063799665, "optim/total_tokens": 3775397888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.918428659439087, "created_at": "2025-01-16T01:53:29.873922+00:00"} {"global_step": 7202, "acc_step": 0, "speed/wps": 12900.027769343193, "speed/FLOPS": 202612662613293.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060127630829811096, "optim/lr": 0.002988161322856357, "optim/total_tokens": 3775922176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9821977615356445, "created_at": "2025-01-16T01:53:40.041141+00:00"} {"global_step": 7203, "acc_step": 0, "speed/wps": 12896.606857120849, "speed/FLOPS": 202558932486016.0, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08061611652374268, "optim/lr": 0.0029881505770576614, "optim/total_tokens": 3776446464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 3.0017402172088623, "created_at": "2025-01-16T01:53:50.208349+00:00"} {"global_step": 7204, "acc_step": 0, "speed/wps": 12897.264612397496, "speed/FLOPS": 202569263436486.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05503937974572182, "optim/lr": 0.002988139826403613, "optim/total_tokens": 3776970752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0864317417144775, "created_at": "2025-01-16T01:54:00.376052+00:00"} {"global_step": 7205, "acc_step": 0, "speed/wps": 12894.249466094998, "speed/FLOPS": 202521906420560.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06317164748907089, "optim/lr": 0.0029881290708942466, "optim/total_tokens": 3777495040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.9561967849731445, "created_at": "2025-01-16T01:54:10.545166+00:00"} {"global_step": 7206, "acc_step": 0, "speed/wps": 12896.684014432745, "speed/FLOPS": 202560144347624.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05604967102408409, "optim/lr": 0.0029881183105295974, "optim/total_tokens": 3778019328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.0065438747406006, "created_at": "2025-01-16T01:54:20.715886+00:00"} {"global_step": 7207, "acc_step": 0, "speed/wps": 12897.144155337011, "speed/FLOPS": 202567371492825.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056268371641635895, "optim/lr": 0.002988107545309701, "optim/total_tokens": 3778543616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.0026893615722656, "created_at": "2025-01-16T01:54:30.887513+00:00"} {"global_step": 7208, "acc_step": 0, "speed/wps": 12897.561862036066, "speed/FLOPS": 202573932150523.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0561395063996315, "optim/lr": 0.002988096775234591, "optim/total_tokens": 3779067904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372023, "loss/out": 2.9673495292663574, "created_at": "2025-01-16T01:54:41.053531+00:00"} {"global_step": 7209, "acc_step": 0, "speed/wps": 12897.391060574744, "speed/FLOPS": 202571249478867.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06561306864023209, "optim/lr": 0.002988086000304304, "optim/total_tokens": 3779592192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.9374327659606934, "created_at": "2025-01-16T01:54:51.220123+00:00"} {"global_step": 7210, "acc_step": 0, "speed/wps": 12898.450892700717, "speed/FLOPS": 202587895598768.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06194235011935234, "optim/lr": 0.002988075220518875, "optim/total_tokens": 3780116480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.984692096710205, "created_at": "2025-01-16T01:55:01.385658+00:00"} {"global_step": 7211, "acc_step": 0, "speed/wps": 12895.047325816966, "speed/FLOPS": 202534437903867.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05937390774488449, "optim/lr": 0.002988064435878339, "optim/total_tokens": 3780640768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 2.9899940490722656, "created_at": "2025-01-16T01:55:11.555854+00:00"} {"global_step": 7212, "acc_step": 0, "speed/wps": 12897.972804954832, "speed/FLOPS": 202580386573757.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06429397314786911, "optim/lr": 0.0029880536463827304, "optim/total_tokens": 3781165056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9696948528289795, "created_at": "2025-01-16T01:55:21.721667+00:00"} {"global_step": 7213, "acc_step": 0, "speed/wps": 12902.203531100931, "speed/FLOPS": 202646835941509.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06610094010829926, "optim/lr": 0.002988042852032086, "optim/total_tokens": 3781689344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 3.02730393409729, "created_at": "2025-01-16T01:55:31.884250+00:00"} {"global_step": 7214, "acc_step": 0, "speed/wps": 12897.810817511358, "speed/FLOPS": 202577842338363.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05616679787635803, "optim/lr": 0.0029880320528264397, "optim/total_tokens": 3782213632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.0062644481658936, "created_at": "2025-01-16T01:55:42.050935+00:00"} {"global_step": 7215, "acc_step": 0, "speed/wps": 12897.000434303014, "speed/FLOPS": 202565114156493.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07045506685972214, "optim/lr": 0.0029880212487658274, "optim/total_tokens": 3782737920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.980792284011841, "created_at": "2025-01-16T01:55:52.220694+00:00"} {"global_step": 7216, "acc_step": 0, "speed/wps": 12902.969582628075, "speed/FLOPS": 202658867833408.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049504172056913376, "optim/lr": 0.0029880104398502835, "optim/total_tokens": 3783262208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 3.0708842277526855, "created_at": "2025-01-16T01:56:02.384748+00:00"} {"global_step": 7217, "acc_step": 0, "speed/wps": 12895.051292135226, "speed/FLOPS": 202534500200346.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0757206454873085, "optim/lr": 0.0029879996260798445, "optim/total_tokens": 3783786496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.005401134490967, "created_at": "2025-01-16T01:56:12.552840+00:00"} {"global_step": 7218, "acc_step": 0, "speed/wps": 12895.672484079414, "speed/FLOPS": 202544256873437.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0646742582321167, "optim/lr": 0.002987988807454545, "optim/total_tokens": 3784310784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.990445137023926, "created_at": "2025-01-16T01:56:22.720801+00:00"} {"global_step": 7219, "acc_step": 0, "speed/wps": 12897.440395951706, "speed/FLOPS": 202572024358757.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06912640482187271, "optim/lr": 0.00298797798397442, "optim/total_tokens": 3784835072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.912868022918701, "created_at": "2025-01-16T01:56:32.892043+00:00"} {"global_step": 7220, "acc_step": 0, "speed/wps": 12894.169694058743, "speed/FLOPS": 202520653491112.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058202359825372696, "optim/lr": 0.002987967155639505, "optim/total_tokens": 3785359360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.045341968536377, "created_at": "2025-01-16T01:56:43.068880+00:00"} {"global_step": 7221, "acc_step": 0, "speed/wps": 12897.593434822518, "speed/FLOPS": 202574428044520.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0853826180100441, "optim/lr": 0.002987956322449836, "optim/total_tokens": 3785883648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430246, "loss/out": 2.98150897026062, "created_at": "2025-01-16T01:56:53.237134+00:00"} {"global_step": 7222, "acc_step": 0, "speed/wps": 12895.25248585517, "speed/FLOPS": 202537660224187.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06280305981636047, "optim/lr": 0.0029879454844054477, "optim/total_tokens": 3786407936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 3.103947162628174, "created_at": "2025-01-16T01:57:03.406193+00:00"} {"global_step": 7223, "acc_step": 0, "speed/wps": 12898.522628306679, "speed/FLOPS": 202589022305034.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0697614774107933, "optim/lr": 0.0029879346415063756, "optim/total_tokens": 3786932224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.021428108215332, "created_at": "2025-01-16T01:57:13.572357+00:00"} {"global_step": 7224, "acc_step": 0, "speed/wps": 12895.129397319522, "speed/FLOPS": 202535726949593.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05572551488876343, "optim/lr": 0.0029879237937526547, "optim/total_tokens": 3787456512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 3.1326375007629395, "created_at": "2025-01-16T01:57:23.740626+00:00"} {"global_step": 7225, "acc_step": 0, "speed/wps": 12898.538907623302, "speed/FLOPS": 202589277994072.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05748322606086731, "optim/lr": 0.002987912941144321, "optim/total_tokens": 3787980800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.9383344650268555, "created_at": "2025-01-16T01:57:33.909327+00:00"} {"global_step": 7226, "acc_step": 0, "speed/wps": 12881.240258263204, "speed/FLOPS": 202317578935037.06, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06320162862539291, "optim/lr": 0.0029879020836814096, "optim/total_tokens": 3788505088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.084937810897827, "created_at": "2025-01-16T01:57:44.089981+00:00"} {"global_step": 7227, "acc_step": 0, "speed/wps": 12885.150632986031, "speed/FLOPS": 202378996743477.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06425628066062927, "optim/lr": 0.0029878912213639566, "optim/total_tokens": 3789029376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.038693904876709, "created_at": "2025-01-16T01:57:54.266423+00:00"} {"global_step": 7228, "acc_step": 0, "speed/wps": 12884.310905167986, "speed/FLOPS": 202365807664187.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05276679992675781, "optim/lr": 0.0029878803541919965, "optim/total_tokens": 3789553664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.1053245067596436, "created_at": "2025-01-16T01:58:04.444085+00:00"} {"global_step": 7229, "acc_step": 0, "speed/wps": 12891.845259286212, "speed/FLOPS": 202484145048901.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05252131447196007, "optim/lr": 0.002987869482165565, "optim/total_tokens": 3790077952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.9879353046417236, "created_at": "2025-01-16T01:58:14.616300+00:00"} {"global_step": 7230, "acc_step": 0, "speed/wps": 12890.915879888516, "speed/FLOPS": 202469547868360.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06916941702365875, "optim/lr": 0.0029878586052846977, "optim/total_tokens": 3790602240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441962, "loss/out": 2.962913751602173, "created_at": "2025-01-16T01:58:24.790640+00:00"} {"global_step": 7231, "acc_step": 0, "speed/wps": 12900.594804502873, "speed/FLOPS": 202621568679664.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05047902464866638, "optim/lr": 0.0029878477235494296, "optim/total_tokens": 3791126528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8931920528411865, "created_at": "2025-01-16T01:58:34.955413+00:00"} {"global_step": 7232, "acc_step": 0, "speed/wps": 12898.989514989962, "speed/FLOPS": 202596355401965.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05370316281914711, "optim/lr": 0.0029878368369597973, "optim/total_tokens": 3791650816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.129119634628296, "created_at": "2025-01-16T01:58:45.123846+00:00"} {"global_step": 7233, "acc_step": 0, "speed/wps": 12898.764625865671, "speed/FLOPS": 202592823209238.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051774073392152786, "optim/lr": 0.0029878259455158357, "optim/total_tokens": 3792175104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.0084228515625, "created_at": "2025-01-16T01:58:55.289125+00:00"} {"global_step": 7234, "acc_step": 0, "speed/wps": 12900.696980047442, "speed/FLOPS": 202623173486996.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05396099016070366, "optim/lr": 0.00298781504921758, "optim/total_tokens": 3792699392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.985520839691162, "created_at": "2025-01-16T01:59:05.453738+00:00"} {"global_step": 7235, "acc_step": 0, "speed/wps": 12891.305645259517, "speed/FLOPS": 202475669669105.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06589648127555847, "optim/lr": 0.002987804148065066, "optim/total_tokens": 3793223680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463795, "loss/out": 3.1876790523529053, "created_at": "2025-01-16T01:59:15.628120+00:00"} {"global_step": 7236, "acc_step": 0, "speed/wps": 12900.8257353164, "speed/FLOPS": 202625195765423.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.061338234692811966, "optim/lr": 0.00298779324205833, "optim/total_tokens": 3793747968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.04864239692688, "created_at": "2025-01-16T01:59:25.792385+00:00"} {"global_step": 7237, "acc_step": 0, "speed/wps": 12897.27925985907, "speed/FLOPS": 202569493494997.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07162854820489883, "optim/lr": 0.002987782331197406, "optim/total_tokens": 3794272256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.003937244415283, "created_at": "2025-01-16T01:59:35.959231+00:00"} {"global_step": 7238, "acc_step": 0, "speed/wps": 12902.803220571452, "speed/FLOPS": 202656254888703.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06597984582185745, "optim/lr": 0.002987771415482331, "optim/total_tokens": 3794796544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 2.972388982772827, "created_at": "2025-01-16T01:59:46.121845+00:00"} {"global_step": 7239, "acc_step": 0, "speed/wps": 12899.185935860109, "speed/FLOPS": 202599440461641.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050276245921850204, "optim/lr": 0.0029877604949131403, "optim/total_tokens": 3795320832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455791, "loss/out": 2.9422504901885986, "created_at": "2025-01-16T01:59:56.287161+00:00"} {"global_step": 7240, "acc_step": 0, "speed/wps": 12901.911896135292, "speed/FLOPS": 202642255413625.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05587325245141983, "optim/lr": 0.002987749569489869, "optim/total_tokens": 3795845120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 3.012207508087158, "created_at": "2025-01-16T02:00:06.453264+00:00"} {"global_step": 7241, "acc_step": 0, "speed/wps": 12892.834335346362, "speed/FLOPS": 202499679847559.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05833953991532326, "optim/lr": 0.0029877386392125533, "optim/total_tokens": 3796369408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9510269165039062, "created_at": "2025-01-16T02:00:16.623916+00:00"} {"global_step": 7242, "acc_step": 0, "speed/wps": 12900.125525413654, "speed/FLOPS": 202614198006711.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06075167655944824, "optim/lr": 0.002987727704081229, "optim/total_tokens": 3796893696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.928941249847412, "created_at": "2025-01-16T02:00:26.790093+00:00"} {"global_step": 7243, "acc_step": 0, "speed/wps": 12898.129315200506, "speed/FLOPS": 202582844782233.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06906326860189438, "optim/lr": 0.0029877167640959307, "optim/total_tokens": 3797417984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 2.9930198192596436, "created_at": "2025-01-16T02:00:36.962778+00:00"} {"global_step": 7244, "acc_step": 0, "speed/wps": 12895.844129809122, "speed/FLOPS": 202546952805488.2, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07154424488544464, "optim/lr": 0.0029877058192566953, "optim/total_tokens": 3797942272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9542574882507324, "created_at": "2025-01-16T02:00:47.134774+00:00"} {"global_step": 7245, "acc_step": 0, "speed/wps": 12900.352851784572, "speed/FLOPS": 202617768479745.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04883524030447006, "optim/lr": 0.0029876948695635583, "optim/total_tokens": 3798466560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.964942216873169, "created_at": "2025-01-16T02:00:57.299749+00:00"} {"global_step": 7246, "acc_step": 0, "speed/wps": 12900.734229043574, "speed/FLOPS": 202623758533665.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06746044009923935, "optim/lr": 0.0029876839150165544, "optim/total_tokens": 3798990848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.028722047805786, "created_at": "2025-01-16T02:01:07.463636+00:00"} {"global_step": 7247, "acc_step": 0, "speed/wps": 12898.774928639781, "speed/FLOPS": 202592985028463.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0535145066678524, "optim/lr": 0.002987672955615721, "optim/total_tokens": 3799515136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.898350477218628, "created_at": "2025-01-16T02:01:17.628847+00:00"} {"global_step": 7248, "acc_step": 0, "speed/wps": 12896.899765065335, "speed/FLOPS": 202563533007781.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07112958282232285, "optim/lr": 0.002987661991361092, "optim/total_tokens": 3800039424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425394, "loss/out": 3.0003855228424072, "created_at": "2025-01-16T02:01:27.801786+00:00"} {"global_step": 7249, "acc_step": 0, "speed/wps": 12892.726945820665, "speed/FLOPS": 202497993147489.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07135777175426483, "optim/lr": 0.0029876510222527053, "optim/total_tokens": 3800563712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 3.025099992752075, "created_at": "2025-01-16T02:01:37.974846+00:00"} {"global_step": 7250, "acc_step": 0, "speed/wps": 12895.13953950768, "speed/FLOPS": 202535886246595.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07443366199731827, "optim/lr": 0.0029876400482905944, "optim/total_tokens": 3801088000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.951491117477417, "created_at": "2025-01-16T02:01:48.144889+00:00"} {"global_step": 7251, "acc_step": 0, "speed/wps": 12901.108060097607, "speed/FLOPS": 202629630064067.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09258133172988892, "optim/lr": 0.002987629069474797, "optim/total_tokens": 3801612288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.9479618072509766, "created_at": "2025-01-16T02:01:58.310842+00:00"} {"global_step": 7252, "acc_step": 0, "speed/wps": 12899.225858706062, "speed/FLOPS": 202600067504797.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048848580569028854, "optim/lr": 0.0029876180858053477, "optim/total_tokens": 3802136576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.9553656578063965, "created_at": "2025-01-16T02:02:08.475687+00:00"} {"global_step": 7253, "acc_step": 0, "speed/wps": 12898.766472443136, "speed/FLOPS": 202592852212274.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07146744430065155, "optim/lr": 0.002987607097282283, "optim/total_tokens": 3802660864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.0676279067993164, "created_at": "2025-01-16T02:02:18.641129+00:00"} {"global_step": 7254, "acc_step": 0, "speed/wps": 12899.638496745467, "speed/FLOPS": 202606548552230.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06376723200082779, "optim/lr": 0.0029875961039056388, "optim/total_tokens": 3803185152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 3.0145692825317383, "created_at": "2025-01-16T02:02:28.806029+00:00"} {"global_step": 7255, "acc_step": 0, "speed/wps": 12899.880727725278, "speed/FLOPS": 202610353122628.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06526995450258255, "optim/lr": 0.0029875851056754507, "optim/total_tokens": 3803709440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 2.9855949878692627, "created_at": "2025-01-16T02:02:38.970359+00:00"} {"global_step": 7256, "acc_step": 0, "speed/wps": 12900.088113209844, "speed/FLOPS": 202613610396640.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07966076582670212, "optim/lr": 0.0029875741025917548, "optim/total_tokens": 3804233728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9585134983062744, "created_at": "2025-01-16T02:02:49.136732+00:00"} {"global_step": 7257, "acc_step": 0, "speed/wps": 12901.447865747352, "speed/FLOPS": 202634967178738.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07873465120792389, "optim/lr": 0.0029875630946545858, "optim/total_tokens": 3804758016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 2.977663516998291, "created_at": "2025-01-16T02:02:59.305906+00:00"} {"global_step": 7258, "acc_step": 0, "speed/wps": 12899.225566684821, "speed/FLOPS": 202600062918202.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08557363599538803, "optim/lr": 0.0029875520818639814, "optim/total_tokens": 3805282304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.097595691680908, "created_at": "2025-01-16T02:03:09.472908+00:00"} {"global_step": 7259, "acc_step": 0, "speed/wps": 12895.59395166414, "speed/FLOPS": 202543023413940.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07486217468976974, "optim/lr": 0.0029875410642199764, "optim/total_tokens": 3805806592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.917184829711914, "created_at": "2025-01-16T02:03:19.641326+00:00"} {"global_step": 7260, "acc_step": 0, "speed/wps": 12899.892457067575, "speed/FLOPS": 202610537348066.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07826324552297592, "optim/lr": 0.0029875300417226075, "optim/total_tokens": 3806330880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.968146324157715, "created_at": "2025-01-16T02:03:29.807859+00:00"} {"global_step": 7261, "acc_step": 0, "speed/wps": 12895.021727349762, "speed/FLOPS": 202534035844764.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06610667705535889, "optim/lr": 0.0029875190143719096, "optim/total_tokens": 3806855168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.864349365234375, "created_at": "2025-01-16T02:03:39.979553+00:00"} {"global_step": 7262, "acc_step": 0, "speed/wps": 12898.664400770504, "speed/FLOPS": 202591249036395.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07354315370321274, "optim/lr": 0.00298750798216792, "optim/total_tokens": 3807379456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.9399232864379883, "created_at": "2025-01-16T02:03:50.150150+00:00"} {"global_step": 7263, "acc_step": 0, "speed/wps": 12894.06249846917, "speed/FLOPS": 202518969837076.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06534372270107269, "optim/lr": 0.0029874969451106734, "optim/total_tokens": 3807903744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0762381553649902, "created_at": "2025-01-16T02:04:00.322632+00:00"} {"global_step": 7264, "acc_step": 0, "speed/wps": 12893.290484392575, "speed/FLOPS": 202506844295144.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054388344287872314, "optim/lr": 0.002987485903200207, "optim/total_tokens": 3808428032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.117922306060791, "created_at": "2025-01-16T02:04:10.495777+00:00"} {"global_step": 7265, "acc_step": 0, "speed/wps": 12899.84969946113, "speed/FLOPS": 202609865781102.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0756760835647583, "optim/lr": 0.002987474856436556, "optim/total_tokens": 3808952320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 2.9686732292175293, "created_at": "2025-01-16T02:04:20.660405+00:00"} {"global_step": 7266, "acc_step": 0, "speed/wps": 12900.411552666863, "speed/FLOPS": 202618690457765.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06710866838693619, "optim/lr": 0.002987463804819757, "optim/total_tokens": 3809476608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.033568859100342, "created_at": "2025-01-16T02:04:30.825636+00:00"} {"global_step": 7267, "acc_step": 0, "speed/wps": 12899.965842796712, "speed/FLOPS": 202611689971785.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07674027234315872, "optim/lr": 0.0029874527483498454, "optim/total_tokens": 3810000896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 3.053769588470459, "created_at": "2025-01-16T02:04:40.992757+00:00"} {"global_step": 7268, "acc_step": 0, "speed/wps": 12900.180367083116, "speed/FLOPS": 202615059370490.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055547647178173065, "optim/lr": 0.0029874416870268577, "optim/total_tokens": 3810525184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.8787460327148438, "created_at": "2025-01-16T02:04:51.165105+00:00"} {"global_step": 7269, "acc_step": 0, "speed/wps": 12899.445862044073, "speed/FLOPS": 202603522959536.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0698307678103447, "optim/lr": 0.00298743062085083, "optim/total_tokens": 3811049472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.101132869720459, "created_at": "2025-01-16T02:05:01.330022+00:00"} {"global_step": 7270, "acc_step": 0, "speed/wps": 12896.222094781255, "speed/FLOPS": 202552889264753.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06068718433380127, "optim/lr": 0.002987419549821798, "optim/total_tokens": 3811573760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 3.183729648590088, "created_at": "2025-01-16T02:05:11.497991+00:00"} {"global_step": 7271, "acc_step": 0, "speed/wps": 12893.640159922683, "speed/FLOPS": 202512336429849.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052129726856946945, "optim/lr": 0.002987408473939799, "optim/total_tokens": 3812098048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.936685085296631, "created_at": "2025-01-16T02:05:21.668611+00:00"} {"global_step": 7272, "acc_step": 0, "speed/wps": 12903.056062214142, "speed/FLOPS": 202660226114142.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04880897328257561, "optim/lr": 0.0029873973932048677, "optim/total_tokens": 3812622336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.955789566040039, "created_at": "2025-01-16T02:05:31.833907+00:00"} {"global_step": 7273, "acc_step": 0, "speed/wps": 12897.391247320626, "speed/FLOPS": 202571252411968.6, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05800953134894371, "optim/lr": 0.002987386307617041, "optim/total_tokens": 3813146624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 3.090282917022705, "created_at": "2025-01-16T02:05:42.000331+00:00"} {"global_step": 7274, "acc_step": 0, "speed/wps": 12899.141746494939, "speed/FLOPS": 202598746406939.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05810995772480965, "optim/lr": 0.002987375217176355, "optim/total_tokens": 3813670912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.8340811729431152, "created_at": "2025-01-16T02:05:52.165626+00:00"} {"global_step": 7275, "acc_step": 0, "speed/wps": 12904.482527483025, "speed/FLOPS": 202682630711359.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06440059840679169, "optim/lr": 0.0029873641218828456, "optim/total_tokens": 3814195200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.9701008796691895, "created_at": "2025-01-16T02:06:02.330481+00:00"} {"global_step": 7276, "acc_step": 0, "speed/wps": 12899.059129734078, "speed/FLOPS": 202597448797182.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05409307777881622, "optim/lr": 0.002987353021736549, "optim/total_tokens": 3814719488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0519652366638184, "created_at": "2025-01-16T02:06:12.497830+00:00"} {"global_step": 7277, "acc_step": 0, "speed/wps": 12899.062818329214, "speed/FLOPS": 202597506731637.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0581531897187233, "optim/lr": 0.002987341916737502, "optim/total_tokens": 3815243776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.851395606994629, "created_at": "2025-01-16T02:06:22.663474+00:00"} {"global_step": 7278, "acc_step": 0, "speed/wps": 12895.338442573486, "speed/FLOPS": 202539010292565.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06121979281306267, "optim/lr": 0.0029873308068857404, "optim/total_tokens": 3815768064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 2.961077928543091, "created_at": "2025-01-16T02:06:32.831621+00:00"} {"global_step": 7279, "acc_step": 0, "speed/wps": 12904.664297581581, "speed/FLOPS": 202685485660535.28, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05596347525715828, "optim/lr": 0.0029873196921813006, "optim/total_tokens": 3816292352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.050262928009033, "created_at": "2025-01-16T02:06:42.995618+00:00"} {"global_step": 7280, "acc_step": 0, "speed/wps": 12900.309099032951, "speed/FLOPS": 202617081282659.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053169045597314835, "optim/lr": 0.0029873085726242186, "optim/total_tokens": 3816816640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.9302752017974854, "created_at": "2025-01-16T02:06:53.161427+00:00"} {"global_step": 7281, "acc_step": 0, "speed/wps": 12901.706392862468, "speed/FLOPS": 202639027702334.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05812758952379227, "optim/lr": 0.0029872974482145305, "optim/total_tokens": 3817340928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.0613574981689453, "created_at": "2025-01-16T02:07:03.324336+00:00"} {"global_step": 7282, "acc_step": 0, "speed/wps": 12901.095205060312, "speed/FLOPS": 202629428158041.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05906260013580322, "optim/lr": 0.0029872863189522735, "optim/total_tokens": 3817865216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.001920700073242, "created_at": "2025-01-16T02:07:13.490533+00:00"} {"global_step": 7283, "acc_step": 0, "speed/wps": 12896.808842755732, "speed/FLOPS": 202562104947969.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0611877404153347, "optim/lr": 0.002987275184837483, "optim/total_tokens": 3818389504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 2.925662040710449, "created_at": "2025-01-16T02:07:23.658487+00:00"} {"global_step": 7284, "acc_step": 0, "speed/wps": 12889.361034590689, "speed/FLOPS": 202445126886376.28, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.052612774074077606, "optim/lr": 0.002987264045870196, "optim/total_tokens": 3818913792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.039929151535034, "created_at": "2025-01-16T02:07:33.832082+00:00"} {"global_step": 7285, "acc_step": 0, "speed/wps": 12885.926066264283, "speed/FLOPS": 202391175988668.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06948911398649216, "optim/lr": 0.0029872529020504484, "optim/total_tokens": 3819438080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.986943244934082, "created_at": "2025-01-16T02:07:44.008786+00:00"} {"global_step": 7286, "acc_step": 0, "speed/wps": 12884.296688190581, "speed/FLOPS": 202365584367021.38, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055027686059474945, "optim/lr": 0.0029872417533782767, "optim/total_tokens": 3819962368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 2.9160449504852295, "created_at": "2025-01-16T02:07:54.188129+00:00"} {"global_step": 7287, "acc_step": 0, "speed/wps": 12886.493344365832, "speed/FLOPS": 202400085870775.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06008383259177208, "optim/lr": 0.0029872305998537166, "optim/total_tokens": 3820486656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.9280588626861572, "created_at": "2025-01-16T02:08:04.363157+00:00"} {"global_step": 7288, "acc_step": 0, "speed/wps": 12896.91033786742, "speed/FLOPS": 202563699068167.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07145436108112335, "optim/lr": 0.002987219441476806, "optim/total_tokens": 3821010944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.922766923904419, "created_at": "2025-01-16T02:08:14.529780+00:00"} {"global_step": 7289, "acc_step": 0, "speed/wps": 12900.385882050923, "speed/FLOPS": 202618287265467.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055545542389154434, "optim/lr": 0.00298720827824758, "optim/total_tokens": 3821535232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.8878679275512695, "created_at": "2025-01-16T02:08:24.697641+00:00"} {"global_step": 7290, "acc_step": 0, "speed/wps": 12891.57355900522, "speed/FLOPS": 202479877622633.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06193554028868675, "optim/lr": 0.0029871971101660755, "optim/total_tokens": 3822059520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.9653689861297607, "created_at": "2025-01-16T02:08:34.871340+00:00"} {"global_step": 7291, "acc_step": 0, "speed/wps": 12897.848678677046, "speed/FLOPS": 202578436999997.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07144152373075485, "optim/lr": 0.002987185937232329, "optim/total_tokens": 3822583808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0589935779571533, "created_at": "2025-01-16T02:08:45.039332+00:00"} {"global_step": 7292, "acc_step": 0, "speed/wps": 12893.663604205638, "speed/FLOPS": 202512704654529.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05473417043685913, "optim/lr": 0.002987174759446377, "optim/total_tokens": 3823108096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.0121066570281982, "created_at": "2025-01-16T02:08:55.216904+00:00"} {"global_step": 7293, "acc_step": 0, "speed/wps": 12898.508736955131, "speed/FLOPS": 202588804122270.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06560028344392776, "optim/lr": 0.0029871635768082554, "optim/total_tokens": 3823632384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.999645471572876, "created_at": "2025-01-16T02:09:05.383622+00:00"} {"global_step": 7294, "acc_step": 0, "speed/wps": 12895.619548858456, "speed/FLOPS": 202543425453051.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05906397104263306, "optim/lr": 0.0029871523893180017, "optim/total_tokens": 3824156672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 2.979619026184082, "created_at": "2025-01-16T02:09:15.555138+00:00"} {"global_step": 7295, "acc_step": 0, "speed/wps": 12896.775852077004, "speed/FLOPS": 202561586784025.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059941668063402176, "optim/lr": 0.0029871411969756515, "optim/total_tokens": 3824680960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9956743717193604, "created_at": "2025-01-16T02:09:25.721803+00:00"} {"global_step": 7296, "acc_step": 0, "speed/wps": 12893.86317996605, "speed/FLOPS": 202515839266092.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08092816919088364, "optim/lr": 0.002987129999781241, "optim/total_tokens": 3825205248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.0464344024658203, "created_at": "2025-01-16T02:09:35.898329+00:00"} {"global_step": 7297, "acc_step": 0, "speed/wps": 12899.039225543658, "speed/FLOPS": 202597136174520.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06757032126188278, "optim/lr": 0.0029871187977348087, "optim/total_tokens": 3825729536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395039, "loss/out": 2.9821317195892334, "created_at": "2025-01-16T02:09:46.064121+00:00"} {"global_step": 7298, "acc_step": 0, "speed/wps": 12900.433967978926, "speed/FLOPS": 202619042521043.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06287512928247452, "optim/lr": 0.002987107590836389, "optim/total_tokens": 3826253824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.971269130706787, "created_at": "2025-01-16T02:09:56.230217+00:00"} {"global_step": 7299, "acc_step": 0, "speed/wps": 12902.151205803711, "speed/FLOPS": 202646014100813.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07379617542028427, "optim/lr": 0.0029870963790860194, "optim/total_tokens": 3826778112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381526, "loss/out": 2.9092211723327637, "created_at": "2025-01-16T02:10:06.395140+00:00"} {"global_step": 7300, "acc_step": 0, "speed/wps": 12897.526457486267, "speed/FLOPS": 202573376073418.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05039316043257713, "optim/lr": 0.0029870851624837365, "optim/total_tokens": 3827302400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.987517833709717, "created_at": "2025-01-16T02:10:16.561255+00:00"} {"global_step": 7301, "acc_step": 0, "speed/wps": 12902.46616347535, "speed/FLOPS": 202650960943844.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07480643689632416, "optim/lr": 0.002987073941029576, "optim/total_tokens": 3827826688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.0038974285125732, "created_at": "2025-01-16T02:10:26.725397+00:00"} {"global_step": 7302, "acc_step": 0, "speed/wps": 12897.805510275712, "speed/FLOPS": 202577758980935.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.061208296567201614, "optim/lr": 0.0029870627147235763, "optim/total_tokens": 3828350976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.9540491104125977, "created_at": "2025-01-16T02:10:36.891303+00:00"} {"global_step": 7303, "acc_step": 0, "speed/wps": 12901.125101635482, "speed/FLOPS": 202629897724836.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05029290169477463, "optim/lr": 0.002987051483565772, "optim/total_tokens": 3828875264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.0192651748657227, "created_at": "2025-01-16T02:10:47.055871+00:00"} {"global_step": 7304, "acc_step": 0, "speed/wps": 12903.57384134227, "speed/FLOPS": 202668358546852.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0552934892475605, "optim/lr": 0.0029870402475562015, "optim/total_tokens": 3829399552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.94254994392395, "created_at": "2025-01-16T02:10:57.219416+00:00"} {"global_step": 7305, "acc_step": 0, "speed/wps": 12898.453392906576, "speed/FLOPS": 202587934867936.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05200543627142906, "optim/lr": 0.0029870290066949007, "optim/total_tokens": 3829923840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.9793500900268555, "created_at": "2025-01-16T02:11:07.386010+00:00"} {"global_step": 7306, "acc_step": 0, "speed/wps": 12900.936370002191, "speed/FLOPS": 202626933435191.62, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05219820514321327, "optim/lr": 0.002987017760981906, "optim/total_tokens": 3830448128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.027549982070923, "created_at": "2025-01-16T02:11:17.557488+00:00"} {"global_step": 7307, "acc_step": 0, "speed/wps": 12901.279334922601, "speed/FLOPS": 202632320170549.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055615708231925964, "optim/lr": 0.0029870065104172544, "optim/total_tokens": 3830972416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 2.8483006954193115, "created_at": "2025-01-16T02:11:27.721226+00:00"} {"global_step": 7308, "acc_step": 0, "speed/wps": 12902.421091695038, "speed/FLOPS": 202650253029599.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061450228095054626, "optim/lr": 0.0029869952550009827, "optim/total_tokens": 3831496704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.009610176086426, "created_at": "2025-01-16T02:11:37.884396+00:00"} {"global_step": 7309, "acc_step": 0, "speed/wps": 12897.081240056774, "speed/FLOPS": 202566383321895.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04790477827191353, "optim/lr": 0.002986983994733128, "optim/total_tokens": 3832020992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9778380393981934, "created_at": "2025-01-16T02:11:48.054452+00:00"} {"global_step": 7310, "acc_step": 0, "speed/wps": 12892.964399491933, "speed/FLOPS": 202501722683692.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05678139999508858, "optim/lr": 0.002986972729613725, "optim/total_tokens": 3832545280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 3.051827907562256, "created_at": "2025-01-16T02:11:58.223995+00:00"} {"global_step": 7311, "acc_step": 0, "speed/wps": 12899.06462243132, "speed/FLOPS": 202597535067540.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07314110547304153, "optim/lr": 0.002986961459642813, "optim/total_tokens": 3833069568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.0573668479919434, "created_at": "2025-01-16T02:12:08.391804+00:00"} {"global_step": 7312, "acc_step": 0, "speed/wps": 12900.222384299901, "speed/FLOPS": 202615719308618.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06553525477647781, "optim/lr": 0.002986950184820428, "optim/total_tokens": 3833593856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 3.0077171325683594, "created_at": "2025-01-16T02:12:18.558926+00:00"} {"global_step": 7313, "acc_step": 0, "speed/wps": 12899.787054557788, "speed/FLOPS": 202608881856812.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06177305802702904, "optim/lr": 0.0029869389051466063, "optim/total_tokens": 3834118144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.933680772781372, "created_at": "2025-01-16T02:12:28.723672+00:00"} {"global_step": 7314, "acc_step": 0, "speed/wps": 12896.132052943105, "speed/FLOPS": 202551475033955.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07305794954299927, "optim/lr": 0.002986927620621385, "optim/total_tokens": 3834642432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.009901523590088, "created_at": "2025-01-16T02:12:38.895124+00:00"} {"global_step": 7315, "acc_step": 0, "speed/wps": 12902.107191281782, "speed/FLOPS": 202645322792264.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05959306284785271, "optim/lr": 0.0029869163312448003, "optim/total_tokens": 3835166720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 3.0374598503112793, "created_at": "2025-01-16T02:12:49.063421+00:00"} {"global_step": 7316, "acc_step": 0, "speed/wps": 12898.11220592382, "speed/FLOPS": 202582576057532.7, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06008337810635567, "optim/lr": 0.00298690503701689, "optim/total_tokens": 3835691008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9723024368286133, "created_at": "2025-01-16T02:12:59.229057+00:00"} {"global_step": 7317, "acc_step": 0, "speed/wps": 12900.591755958761, "speed/FLOPS": 202621520798090.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06869015097618103, "optim/lr": 0.002986893737937691, "optim/total_tokens": 3836215296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.96462345123291, "created_at": "2025-01-16T02:13:09.394340+00:00"} {"global_step": 7318, "acc_step": 0, "speed/wps": 12896.831752556167, "speed/FLOPS": 202562464777866.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05895759537816048, "optim/lr": 0.002986882434007239, "optim/total_tokens": 3836739584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 3.0027103424072266, "created_at": "2025-01-16T02:13:19.563709+00:00"} {"global_step": 7319, "acc_step": 0, "speed/wps": 12903.139405669855, "speed/FLOPS": 202661535137640.53, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.078143410384655, "optim/lr": 0.002986871125225572, "optim/total_tokens": 3837263872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.1203274726867676, "created_at": "2025-01-16T02:13:29.729771+00:00"} {"global_step": 7320, "acc_step": 0, "speed/wps": 12896.557981925342, "speed/FLOPS": 202558164833909.0, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06623091548681259, "optim/lr": 0.0029868598115927263, "optim/total_tokens": 3837788160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.920056104660034, "created_at": "2025-01-16T02:13:39.902165+00:00"} {"global_step": 7321, "acc_step": 0, "speed/wps": 12898.468995868026, "speed/FLOPS": 202588179933887.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07126083970069885, "optim/lr": 0.0029868484931087395, "optim/total_tokens": 3838312448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.980276584625244, "created_at": "2025-01-16T02:13:50.068713+00:00"} {"global_step": 7322, "acc_step": 0, "speed/wps": 12893.286177795335, "speed/FLOPS": 202506776654117.25, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05876999348402023, "optim/lr": 0.002986837169773647, "optim/total_tokens": 3838836736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.917905807495117, "created_at": "2025-01-16T02:14:00.238615+00:00"} {"global_step": 7323, "acc_step": 0, "speed/wps": 12888.032612870713, "speed/FLOPS": 202424262197821.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07776353508234024, "optim/lr": 0.002986825841587488, "optim/total_tokens": 3839361024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 2.9389820098876953, "created_at": "2025-01-16T02:14:10.413429+00:00"} {"global_step": 7324, "acc_step": 0, "speed/wps": 12887.94446274306, "speed/FLOPS": 202422877678935.56, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0651162788271904, "optim/lr": 0.0029868145085502973, "optim/total_tokens": 3839885312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 2.9672117233276367, "created_at": "2025-01-16T02:14:20.587137+00:00"} {"global_step": 7325, "acc_step": 0, "speed/wps": 12883.259669489682, "speed/FLOPS": 202349296563310.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06051437556743622, "optim/lr": 0.0029868031706621135, "optim/total_tokens": 3840409600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9294593334198, "created_at": "2025-01-16T02:14:30.772177+00:00"} {"global_step": 7326, "acc_step": 0, "speed/wps": 12892.995580079665, "speed/FLOPS": 202502212417669.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06350625306367874, "optim/lr": 0.0029867918279229723, "optim/total_tokens": 3840933888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371206, "loss/out": 2.994568109512329, "created_at": "2025-01-16T02:14:40.942518+00:00"} {"global_step": 7327, "acc_step": 0, "speed/wps": 12897.943654668888, "speed/FLOPS": 202579928728460.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05758069455623627, "optim/lr": 0.0029867804803329115, "optim/total_tokens": 3841458176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.984241485595703, "created_at": "2025-01-16T02:14:51.114082+00:00"} {"global_step": 7328, "acc_step": 0, "speed/wps": 12893.697673990755, "speed/FLOPS": 202513239767321.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09643485397100449, "optim/lr": 0.002986769127891968, "optim/total_tokens": 3841982464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9627954959869385, "created_at": "2025-01-16T02:15:01.285308+00:00"} {"global_step": 7329, "acc_step": 0, "speed/wps": 12897.088068958177, "speed/FLOPS": 202566490579175.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06566494703292847, "optim/lr": 0.0029867577706001783, "optim/total_tokens": 3842506752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452340, "loss/out": 2.937001943588257, "created_at": "2025-01-16T02:15:11.457665+00:00"} {"global_step": 7330, "acc_step": 0, "speed/wps": 12895.582316394568, "speed/FLOPS": 202542840666043.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0810890644788742, "optim/lr": 0.002986746408457581, "optim/total_tokens": 3843031040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 3.000349998474121, "created_at": "2025-01-16T02:15:21.627784+00:00"} {"global_step": 7331, "acc_step": 0, "speed/wps": 12891.110273144393, "speed/FLOPS": 202472601081567.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07622149586677551, "optim/lr": 0.0029867350414642113, "optim/total_tokens": 3843555328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.0391242504119873, "created_at": "2025-01-16T02:15:31.802935+00:00"} {"global_step": 7332, "acc_step": 0, "speed/wps": 12894.751014361234, "speed/FLOPS": 202529783925281.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06860776245594025, "optim/lr": 0.002986723669620107, "optim/total_tokens": 3844079616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 2.9283390045166016, "created_at": "2025-01-16T02:15:41.975640+00:00"} {"global_step": 7333, "acc_step": 0, "speed/wps": 12893.270269632772, "speed/FLOPS": 202506526794564.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0659102350473404, "optim/lr": 0.002986712292925306, "optim/total_tokens": 3844603904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.0619935989379883, "created_at": "2025-01-16T02:15:52.145331+00:00"} {"global_step": 7334, "acc_step": 0, "speed/wps": 12895.432477693032, "speed/FLOPS": 202540487243335.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0733136311173439, "optim/lr": 0.0029867009113798437, "optim/total_tokens": 3845128192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9958410263061523, "created_at": "2025-01-16T02:16:02.313046+00:00"} {"global_step": 7335, "acc_step": 0, "speed/wps": 12902.665258324207, "speed/FLOPS": 202654088002030.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04977669566869736, "optim/lr": 0.002986689524983759, "optim/total_tokens": 3845652480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.948836088180542, "created_at": "2025-01-16T02:16:12.475685+00:00"} {"global_step": 7336, "acc_step": 0, "speed/wps": 12899.69064086319, "speed/FLOPS": 202607367547252.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0759459137916565, "optim/lr": 0.0029866781337370882, "optim/total_tokens": 3846176768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.953284740447998, "created_at": "2025-01-16T02:16:22.643602+00:00"} {"global_step": 7337, "acc_step": 0, "speed/wps": 12896.374672220552, "speed/FLOPS": 202555285703101.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.066799096763134, "optim/lr": 0.0029866667376398687, "optim/total_tokens": 3846701056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.84390926361084, "created_at": "2025-01-16T02:16:32.814431+00:00"} {"global_step": 7338, "acc_step": 0, "speed/wps": 12900.569114237325, "speed/FLOPS": 202621165178742.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05983971804380417, "optim/lr": 0.0029866553366921372, "optim/total_tokens": 3847225344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 2.9746854305267334, "created_at": "2025-01-16T02:16:42.978915+00:00"} {"global_step": 7339, "acc_step": 0, "speed/wps": 12896.482274970573, "speed/FLOPS": 202556975752152.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0747557058930397, "optim/lr": 0.002986643930893931, "optim/total_tokens": 3847749632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.9558935165405273, "created_at": "2025-01-16T02:16:53.146983+00:00"} {"global_step": 7340, "acc_step": 0, "speed/wps": 12894.747838618723, "speed/FLOPS": 202529734045881.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06409516930580139, "optim/lr": 0.002986632520245288, "optim/total_tokens": 3848273920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.996459484100342, "created_at": "2025-01-16T02:17:03.317548+00:00"} {"global_step": 7341, "acc_step": 0, "speed/wps": 12892.326254020878, "speed/FLOPS": 202491699732165.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054780472069978714, "optim/lr": 0.002986621104746245, "optim/total_tokens": 3848798208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.7382493019104004, "created_at": "2025-01-16T02:17:13.488236+00:00"} {"global_step": 7342, "acc_step": 0, "speed/wps": 12896.844645189994, "speed/FLOPS": 202562667274396.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0585358589887619, "optim/lr": 0.002986609684396839, "optim/total_tokens": 3849322496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9601263999938965, "created_at": "2025-01-16T02:17:23.660523+00:00"} {"global_step": 7343, "acc_step": 0, "speed/wps": 12896.867213565878, "speed/FLOPS": 202563021741752.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059276800602674484, "optim/lr": 0.0029865982591971074, "optim/total_tokens": 3849846784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.978330135345459, "created_at": "2025-01-16T02:17:33.831157+00:00"} {"global_step": 7344, "acc_step": 0, "speed/wps": 12900.427835569395, "speed/FLOPS": 202618946203125.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0481778122484684, "optim/lr": 0.002986586829147088, "optim/total_tokens": 3850371072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.956068992614746, "created_at": "2025-01-16T02:17:43.996248+00:00"} {"global_step": 7345, "acc_step": 0, "speed/wps": 12888.886997495496, "speed/FLOPS": 202437681482401.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058018285781145096, "optim/lr": 0.0029865753942468175, "optim/total_tokens": 3850895360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.941096782684326, "created_at": "2025-01-16T02:17:54.170836+00:00"} {"global_step": 7346, "acc_step": 0, "speed/wps": 12896.81669956128, "speed/FLOPS": 202562228349896.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07487493753433228, "optim/lr": 0.0029865639544963333, "optim/total_tokens": 3851419648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.024855375289917, "created_at": "2025-01-16T02:18:04.340835+00:00"} {"global_step": 7347, "acc_step": 0, "speed/wps": 12893.240953973302, "speed/FLOPS": 202506066351848.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05445341765880585, "optim/lr": 0.002986552509895673, "optim/total_tokens": 3851943936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.9871826171875, "created_at": "2025-01-16T02:18:14.511504+00:00"} {"global_step": 7348, "acc_step": 0, "speed/wps": 12901.012296006964, "speed/FLOPS": 202628125957429.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052013467997312546, "optim/lr": 0.0029865410604448736, "optim/total_tokens": 3852468224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446459, "loss/out": 2.9501936435699463, "created_at": "2025-01-16T02:18:24.679800+00:00"} {"global_step": 7349, "acc_step": 0, "speed/wps": 12899.566106719383, "speed/FLOPS": 202605411567396.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059353169053792953, "optim/lr": 0.002986529606143972, "optim/total_tokens": 3852992512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 2.9412126541137695, "created_at": "2025-01-16T02:18:34.850259+00:00"} {"global_step": 7350, "acc_step": 0, "speed/wps": 12898.099807223214, "speed/FLOPS": 202582381318902.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05587630346417427, "optim/lr": 0.002986518146993007, "optim/total_tokens": 3853516800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9983906745910645, "created_at": "2025-01-16T02:18:45.027251+00:00"} {"global_step": 7351, "acc_step": 0, "speed/wps": 12903.800755680046, "speed/FLOPS": 202671922548340.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05383938178420067, "optim/lr": 0.002986506682992015, "optim/total_tokens": 3854041088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.988081932067871, "created_at": "2025-01-16T02:18:55.188804+00:00"} {"global_step": 7352, "acc_step": 0, "speed/wps": 12899.752648715654, "speed/FLOPS": 202608341465783.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05891192704439163, "optim/lr": 0.002986495214141033, "optim/total_tokens": 3854565376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.95816707611084, "created_at": "2025-01-16T02:19:05.353494+00:00"} {"global_step": 7353, "acc_step": 0, "speed/wps": 12900.172481347348, "speed/FLOPS": 202614935514174.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05388568714261055, "optim/lr": 0.0029864837404400993, "optim/total_tokens": 3855089664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.035238742828369, "created_at": "2025-01-16T02:19:15.522185+00:00"} {"global_step": 7354, "acc_step": 0, "speed/wps": 12900.382727181961, "speed/FLOPS": 202618237713914.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07475495338439941, "optim/lr": 0.0029864722618892507, "optim/total_tokens": 3855613952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.8551197052001953, "created_at": "2025-01-16T02:19:25.689703+00:00"} {"global_step": 7355, "acc_step": 0, "speed/wps": 12896.44346341723, "speed/FLOPS": 202556366163374.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05441022664308548, "optim/lr": 0.0029864607784885256, "optim/total_tokens": 3856138240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.930511474609375, "created_at": "2025-01-16T02:19:35.857982+00:00"} {"global_step": 7356, "acc_step": 0, "speed/wps": 12897.912499788941, "speed/FLOPS": 202579439398259.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06198963150382042, "optim/lr": 0.00298644929023796, "optim/total_tokens": 3856662528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.976632833480835, "created_at": "2025-01-16T02:19:46.024912+00:00"} {"global_step": 7357, "acc_step": 0, "speed/wps": 12895.002822168368, "speed/FLOPS": 202533738912912.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05531977489590645, "optim/lr": 0.002986437797137593, "optim/total_tokens": 3857186816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.0447001457214355, "created_at": "2025-01-16T02:19:56.195895+00:00"} {"global_step": 7358, "acc_step": 0, "speed/wps": 12900.493068103457, "speed/FLOPS": 202619970769711.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05069814994931221, "optim/lr": 0.002986426299187461, "optim/total_tokens": 3857711104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.90376877784729, "created_at": "2025-01-16T02:20:06.361512+00:00"} {"global_step": 7359, "acc_step": 0, "speed/wps": 12891.376076884997, "speed/FLOPS": 202476775894566.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060338571667671204, "optim/lr": 0.002986414796387602, "optim/total_tokens": 3858235392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.8666749000549316, "created_at": "2025-01-16T02:20:16.532602+00:00"} {"global_step": 7360, "acc_step": 0, "speed/wps": 12897.061431792477, "speed/FLOPS": 202566072205885.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05283210426568985, "optim/lr": 0.0029864032887380533, "optim/total_tokens": 3858759680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.974707841873169, "created_at": "2025-01-16T02:20:26.701468+00:00"} {"global_step": 7361, "acc_step": 0, "speed/wps": 12896.607133660233, "speed/FLOPS": 202558936829447.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05225396156311035, "optim/lr": 0.002986391776238852, "optim/total_tokens": 3859283968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9396185874938965, "created_at": "2025-01-16T02:20:36.870937+00:00"} {"global_step": 7362, "acc_step": 0, "speed/wps": 12890.626642908142, "speed/FLOPS": 202465005004129.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060704972594976425, "optim/lr": 0.0029863802588900364, "optim/total_tokens": 3859808256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330478, "loss/out": 2.9757308959960938, "created_at": "2025-01-16T02:20:47.043707+00:00"} {"global_step": 7363, "acc_step": 0, "speed/wps": 12899.337764110376, "speed/FLOPS": 202601825132945.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05874726176261902, "optim/lr": 0.002986368736691644, "optim/total_tokens": 3860332544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.1262598037719727, "created_at": "2025-01-16T02:20:57.212630+00:00"} {"global_step": 7364, "acc_step": 0, "speed/wps": 12893.625559706916, "speed/FLOPS": 202512107113398.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04782986268401146, "optim/lr": 0.0029863572096437124, "optim/total_tokens": 3860856832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.887399673461914, "created_at": "2025-01-16T02:21:07.397876+00:00"} {"global_step": 7365, "acc_step": 0, "speed/wps": 12895.956979785851, "speed/FLOPS": 202548725269444.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05482932925224304, "optim/lr": 0.0029863456777462784, "optim/total_tokens": 3861381120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.051382541656494, "created_at": "2025-01-16T02:21:17.565943+00:00"} {"global_step": 7366, "acc_step": 0, "speed/wps": 12899.846336754477, "speed/FLOPS": 202609812965173.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0661936104297638, "optim/lr": 0.002986334140999381, "optim/total_tokens": 3861905408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9837985038757324, "created_at": "2025-01-16T02:21:27.731569+00:00"} {"global_step": 7367, "acc_step": 0, "speed/wps": 12894.742829266155, "speed/FLOPS": 202529655367115.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061621274799108505, "optim/lr": 0.0029863225994030563, "optim/total_tokens": 3862429696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392005, "loss/out": 3.022432804107666, "created_at": "2025-01-16T02:21:37.904814+00:00"} {"global_step": 7368, "acc_step": 0, "speed/wps": 12880.166294473995, "speed/FLOPS": 202300710857946.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04952128976583481, "optim/lr": 0.002986311052957343, "optim/total_tokens": 3862953984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 2.9367194175720215, "created_at": "2025-01-16T02:21:48.085193+00:00"} {"global_step": 7369, "acc_step": 0, "speed/wps": 12894.871782346363, "speed/FLOPS": 202531680752440.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05843217298388481, "optim/lr": 0.002986299501662279, "optim/total_tokens": 3863478272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9694573879241943, "created_at": "2025-01-16T02:21:58.256049+00:00"} {"global_step": 7370, "acc_step": 0, "speed/wps": 12891.576672959995, "speed/FLOPS": 202479926531572.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05053030699491501, "optim/lr": 0.002986287945517901, "optim/total_tokens": 3864002560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.9468648433685303, "created_at": "2025-01-16T02:22:08.446766+00:00"} {"global_step": 7371, "acc_step": 0, "speed/wps": 12896.199239142315, "speed/FLOPS": 202552530285537.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06964598596096039, "optim/lr": 0.0029862763845242475, "optim/total_tokens": 3864526848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 3.095292568206787, "created_at": "2025-01-16T02:22:18.617265+00:00"} {"global_step": 7372, "acc_step": 0, "speed/wps": 12899.45779158908, "speed/FLOPS": 202603710329434.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07474437355995178, "optim/lr": 0.002986264818681356, "optim/total_tokens": 3865051136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.956277370452881, "created_at": "2025-01-16T02:22:28.782467+00:00"} {"global_step": 7373, "acc_step": 0, "speed/wps": 12894.735991033604, "speed/FLOPS": 202529547963276.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05307161808013916, "optim/lr": 0.002986253247989264, "optim/total_tokens": 3865575424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 2.9729056358337402, "created_at": "2025-01-16T02:22:38.954610+00:00"} {"global_step": 7374, "acc_step": 0, "speed/wps": 12898.05825183218, "speed/FLOPS": 202581728634382.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051978155970573425, "optim/lr": 0.0029862416724480098, "optim/total_tokens": 3866099712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.031869888305664, "created_at": "2025-01-16T02:22:49.121112+00:00"} {"global_step": 7375, "acc_step": 0, "speed/wps": 12897.361011105313, "speed/FLOPS": 202570777510658.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04804220795631409, "optim/lr": 0.0029862300920576305, "optim/total_tokens": 3866624000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9250168800354004, "created_at": "2025-01-16T02:22:59.287561+00:00"} {"global_step": 7376, "acc_step": 0, "speed/wps": 12895.67353335893, "speed/FLOPS": 202544273353813.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05466829985380173, "optim/lr": 0.002986218506818164, "optim/total_tokens": 3867148288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 3.0391647815704346, "created_at": "2025-01-16T02:23:09.457904+00:00"} {"global_step": 7377, "acc_step": 0, "speed/wps": 12895.686194442214, "speed/FLOPS": 202544472213524.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060772012919187546, "optim/lr": 0.0029862069167296483, "optim/total_tokens": 3867672576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9428553581237793, "created_at": "2025-01-16T02:23:19.628698+00:00"} {"global_step": 7378, "acc_step": 0, "speed/wps": 12897.730361942558, "speed/FLOPS": 202576578673098.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06578294932842255, "optim/lr": 0.0029861953217921212, "optim/total_tokens": 3868196864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.921541690826416, "created_at": "2025-01-16T02:23:29.795307+00:00"} {"global_step": 7379, "acc_step": 0, "speed/wps": 12896.277102979408, "speed/FLOPS": 202553753244095.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05823862925171852, "optim/lr": 0.002986183722005621, "optim/total_tokens": 3868721152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.918689250946045, "created_at": "2025-01-16T02:23:39.966616+00:00"} {"global_step": 7380, "acc_step": 0, "speed/wps": 12893.957230191496, "speed/FLOPS": 202517316454121.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0674554631114006, "optim/lr": 0.0029861721173701844, "optim/total_tokens": 3869245440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.871593952178955, "created_at": "2025-01-16T02:23:50.136478+00:00"} {"global_step": 7381, "acc_step": 0, "speed/wps": 12884.637335741007, "speed/FLOPS": 202370934704899.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0583500899374485, "optim/lr": 0.0029861605078858504, "optim/total_tokens": 3869769728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.0100491046905518, "created_at": "2025-01-16T02:24:00.315345+00:00"} {"global_step": 7382, "acc_step": 0, "speed/wps": 12882.106221311991, "speed/FLOPS": 202331180074676.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05030512064695358, "optim/lr": 0.002986148893552656, "optim/total_tokens": 3870294016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374860, "loss/out": 3.008434295654297, "created_at": "2025-01-16T02:24:10.496055+00:00"} {"global_step": 7383, "acc_step": 0, "speed/wps": 12891.828191476214, "speed/FLOPS": 202483876975492.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05553160980343819, "optim/lr": 0.00298613727437064, "optim/total_tokens": 3870818304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.0540218353271484, "created_at": "2025-01-16T02:24:20.672228+00:00"} {"global_step": 7384, "acc_step": 0, "speed/wps": 12893.819768906866, "speed/FLOPS": 202515157435754.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0651208832859993, "optim/lr": 0.0029861256503398393, "optim/total_tokens": 3871342592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.014026403427124, "created_at": "2025-01-16T02:24:30.845038+00:00"} {"global_step": 7385, "acc_step": 0, "speed/wps": 12882.971987107869, "speed/FLOPS": 202344778116187.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057822100818157196, "optim/lr": 0.0029861140214602925, "optim/total_tokens": 3871866880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.9274239540100098, "created_at": "2025-01-16T02:24:41.024896+00:00"} {"global_step": 7386, "acc_step": 0, "speed/wps": 12896.812036059118, "speed/FLOPS": 202562155103186.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05690306797623634, "optim/lr": 0.0029861023877320375, "optim/total_tokens": 3872391168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.025254249572754, "created_at": "2025-01-16T02:24:51.192860+00:00"} {"global_step": 7387, "acc_step": 0, "speed/wps": 12896.715411558698, "speed/FLOPS": 202560637482631.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07066985219717026, "optim/lr": 0.0029860907491551123, "optim/total_tokens": 3872915456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.9742705821990967, "created_at": "2025-01-16T02:25:01.364590+00:00"} {"global_step": 7388, "acc_step": 0, "speed/wps": 12895.471285524694, "speed/FLOPS": 202541096773659.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056428100913763046, "optim/lr": 0.002986079105729554, "optim/total_tokens": 3873439744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379737, "loss/out": 2.8422374725341797, "created_at": "2025-01-16T02:25:11.534699+00:00"} {"global_step": 7389, "acc_step": 0, "speed/wps": 12896.924705607587, "speed/FLOPS": 202563924733269.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05543331056833267, "optim/lr": 0.002986067457455402, "optim/total_tokens": 3873964032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9569272994995117, "created_at": "2025-01-16T02:25:21.702454+00:00"} {"global_step": 7390, "acc_step": 0, "speed/wps": 12896.097527154689, "speed/FLOPS": 202550932759004.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06170482560992241, "optim/lr": 0.0029860558043326934, "optim/total_tokens": 3874488320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9321131706237793, "created_at": "2025-01-16T02:25:31.870531+00:00"} {"global_step": 7391, "acc_step": 0, "speed/wps": 12898.246810607598, "speed/FLOPS": 202584690209057.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06034846603870392, "optim/lr": 0.0029860441463614665, "optim/total_tokens": 3875012608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294366, "loss/out": 3.0438904762268066, "created_at": "2025-01-16T02:25:42.039102+00:00"} {"global_step": 7392, "acc_step": 0, "speed/wps": 12893.98621952224, "speed/FLOPS": 202517771771396.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05727549269795418, "optim/lr": 0.002986032483541759, "optim/total_tokens": 3875536896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.9257733821868896, "created_at": "2025-01-16T02:25:52.211456+00:00"} {"global_step": 7393, "acc_step": 0, "speed/wps": 12892.724741923437, "speed/FLOPS": 202497958532255.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05803941935300827, "optim/lr": 0.0029860208158736094, "optim/total_tokens": 3876061184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.979125499725342, "created_at": "2025-01-16T02:26:02.381444+00:00"} {"global_step": 7394, "acc_step": 0, "speed/wps": 12898.231690476397, "speed/FLOPS": 202584452726619.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05841654911637306, "optim/lr": 0.002986009143357056, "optim/total_tokens": 3876585472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.0477614402770996, "created_at": "2025-01-16T02:26:12.550956+00:00"} {"global_step": 7395, "acc_step": 0, "speed/wps": 12902.372788684836, "speed/FLOPS": 202649494364448.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05176450312137604, "optim/lr": 0.0029859974659921357, "optim/total_tokens": 3877109760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.024247646331787, "created_at": "2025-01-16T02:26:22.719543+00:00"} {"global_step": 7396, "acc_step": 0, "speed/wps": 12902.967083037009, "speed/FLOPS": 202658828573895.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04827174171805382, "optim/lr": 0.0029859857837788877, "optim/total_tokens": 3877634048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490054, "loss/out": 2.971992015838623, "created_at": "2025-01-16T02:26:32.883681+00:00"} {"global_step": 7397, "acc_step": 0, "speed/wps": 12899.83460158413, "speed/FLOPS": 202609628648197.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04687007889151573, "optim/lr": 0.0029859740967173496, "optim/total_tokens": 3878158336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.942383050918579, "created_at": "2025-01-16T02:26:43.049231+00:00"} {"global_step": 7398, "acc_step": 0, "speed/wps": 12893.773475833237, "speed/FLOPS": 202514430339420.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058482054620981216, "optim/lr": 0.00298596240480756, "optim/total_tokens": 3878682624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.158703565597534, "created_at": "2025-01-16T02:26:53.221483+00:00"} {"global_step": 7399, "acc_step": 0, "speed/wps": 12895.05220189579, "speed/FLOPS": 202534514489386.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.051681507378816605, "optim/lr": 0.0029859507080495567, "optim/total_tokens": 3879206912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 2.9798166751861572, "created_at": "2025-01-16T02:27:03.391990+00:00"} {"global_step": 7400, "acc_step": 0, "speed/wps": 12897.487016758272, "speed/FLOPS": 202572756602587.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051145825535058975, "optim/lr": 0.0029859390064433775, "optim/total_tokens": 3879731200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9100890159606934, "created_at": "2025-01-16T02:27:13.558635+00:00"} {"global_step": 7401, "acc_step": 0, "speed/wps": 12891.959348949227, "speed/FLOPS": 202485936983835.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047232285141944885, "optim/lr": 0.002985927299989061, "optim/total_tokens": 3880255488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.073857069015503, "created_at": "2025-01-16T02:27:23.731314+00:00"} {"global_step": 7402, "acc_step": 0, "speed/wps": 12897.216675375697, "speed/FLOPS": 202568510519686.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05399112403392792, "optim/lr": 0.0029859155886866463, "optim/total_tokens": 3880779776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 2.908834457397461, "created_at": "2025-01-16T02:27:33.902591+00:00"} {"global_step": 7403, "acc_step": 0, "speed/wps": 12894.44205076927, "speed/FLOPS": 202524931227508.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06074649095535278, "optim/lr": 0.00298590387253617, "optim/total_tokens": 3881304064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.060774326324463, "created_at": "2025-01-16T02:27:44.072168+00:00"} {"global_step": 7404, "acc_step": 0, "speed/wps": 12895.487592587593, "speed/FLOPS": 202541352898490.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0680689737200737, "optim/lr": 0.002985892151537671, "optim/total_tokens": 3881828352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0757203102111816, "created_at": "2025-01-16T02:27:54.241850+00:00"} {"global_step": 7405, "acc_step": 0, "speed/wps": 12900.453824444085, "speed/FLOPS": 202619354394114.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060783516615629196, "optim/lr": 0.0029858804256911874, "optim/total_tokens": 3882352640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.1067161560058594, "created_at": "2025-01-16T02:28:04.408646+00:00"} {"global_step": 7406, "acc_step": 0, "speed/wps": 12895.366954855197, "speed/FLOPS": 202539458117129.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07195856422185898, "optim/lr": 0.002985868694996758, "optim/total_tokens": 3882876928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.0975570678710938, "created_at": "2025-01-16T02:28:14.577038+00:00"} {"global_step": 7407, "acc_step": 0, "speed/wps": 12895.478630255693, "speed/FLOPS": 202541212132752.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06472919136285782, "optim/lr": 0.0029858569594544207, "optim/total_tokens": 3883401216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342804, "loss/out": 3.1270618438720703, "created_at": "2025-01-16T02:28:24.745953+00:00"} {"global_step": 7408, "acc_step": 0, "speed/wps": 12900.488044826052, "speed/FLOPS": 202619891872237.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05695128068327904, "optim/lr": 0.002985845219064214, "optim/total_tokens": 3883925504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.9872710704803467, "created_at": "2025-01-16T02:28:34.909826+00:00"} {"global_step": 7409, "acc_step": 0, "speed/wps": 12895.91570525768, "speed/FLOPS": 202548076996262.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05977531895041466, "optim/lr": 0.0029858334738261756, "optim/total_tokens": 3884449792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9722650051116943, "created_at": "2025-01-16T02:28:45.078879+00:00"} {"global_step": 7410, "acc_step": 0, "speed/wps": 12898.22703540757, "speed/FLOPS": 202584379612367.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06399036943912506, "optim/lr": 0.002985821723740344, "optim/total_tokens": 3884974080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 3.057583808898926, "created_at": "2025-01-16T02:28:55.244410+00:00"} {"global_step": 7411, "acc_step": 0, "speed/wps": 12896.991517142658, "speed/FLOPS": 202564974100236.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06715556234121323, "optim/lr": 0.002985809968806758, "optim/total_tokens": 3885498368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.0104455947875977, "created_at": "2025-01-16T02:29:05.411724+00:00"} {"global_step": 7412, "acc_step": 0, "speed/wps": 12896.423910008323, "speed/FLOPS": 202556059050218.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0643286481499672, "optim/lr": 0.002985798209025456, "optim/total_tokens": 3886022656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343206, "loss/out": 2.932528495788574, "created_at": "2025-01-16T02:29:15.579012+00:00"} {"global_step": 7413, "acc_step": 0, "speed/wps": 12894.832967489716, "speed/FLOPS": 202531071111779.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061540666967630386, "optim/lr": 0.0029857864443964756, "optim/total_tokens": 3886546944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.055708885192871, "created_at": "2025-01-16T02:29:25.755004+00:00"} {"global_step": 7414, "acc_step": 0, "speed/wps": 12899.646671504006, "speed/FLOPS": 202606676948046.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0714290663599968, "optim/lr": 0.002985774674919856, "optim/total_tokens": 3887071232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.026946783065796, "created_at": "2025-01-16T02:29:35.922151+00:00"} {"global_step": 7415, "acc_step": 0, "speed/wps": 12900.267313014067, "speed/FLOPS": 202616424975812.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0598270408809185, "optim/lr": 0.002985762900595635, "optim/total_tokens": 3887595520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 2.9666459560394287, "created_at": "2025-01-16T02:29:46.088052+00:00"} {"global_step": 7416, "acc_step": 0, "speed/wps": 12899.28180407182, "speed/FLOPS": 202600946203643.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06616628915071487, "optim/lr": 0.0029857511214238515, "optim/total_tokens": 3888119808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.018246650695801, "created_at": "2025-01-16T02:29:56.253036+00:00"} {"global_step": 7417, "acc_step": 0, "speed/wps": 12895.724520514077, "speed/FLOPS": 202545074177149.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05890295282006264, "optim/lr": 0.002985739337404544, "optim/total_tokens": 3888644096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.9473533630371094, "created_at": "2025-01-16T02:30:06.423301+00:00"} {"global_step": 7418, "acc_step": 0, "speed/wps": 12891.624310633117, "speed/FLOPS": 202480674746689.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06144692748785019, "optim/lr": 0.00298572754853775, "optim/total_tokens": 3889168384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.8575756549835205, "created_at": "2025-01-16T02:30:16.594557+00:00"} {"global_step": 7419, "acc_step": 0, "speed/wps": 12893.051234006269, "speed/FLOPS": 202503086539063.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05262650549411774, "optim/lr": 0.0029857157548235087, "optim/total_tokens": 3889692672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.0146610736846924, "created_at": "2025-01-16T02:30:26.766466+00:00"} {"global_step": 7420, "acc_step": 0, "speed/wps": 12892.85479913341, "speed/FLOPS": 202500001259455.3, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07354943454265594, "optim/lr": 0.002985703956261859, "optim/total_tokens": 3890216960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9979121685028076, "created_at": "2025-01-16T02:30:36.938115+00:00"} {"global_step": 7421, "acc_step": 0, "speed/wps": 12900.708768818818, "speed/FLOPS": 202623358645849.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06265072524547577, "optim/lr": 0.0029856921528528383, "optim/total_tokens": 3890741248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.0277106761932373, "created_at": "2025-01-16T02:30:47.106182+00:00"} {"global_step": 7422, "acc_step": 0, "speed/wps": 12895.302331845094, "speed/FLOPS": 202538443123954.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05229286476969719, "optim/lr": 0.002985680344596486, "optim/total_tokens": 3891265536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.998831272125244, "created_at": "2025-01-16T02:30:57.277730+00:00"} {"global_step": 7423, "acc_step": 0, "speed/wps": 12890.266603119004, "speed/FLOPS": 202459350084495.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.053701601922512054, "optim/lr": 0.0029856685314928405, "optim/total_tokens": 3891789824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9965312480926514, "created_at": "2025-01-16T02:31:07.453075+00:00"} {"global_step": 7424, "acc_step": 0, "speed/wps": 12896.168193134537, "speed/FLOPS": 202552042665323.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06132727116346359, "optim/lr": 0.00298565671354194, "optim/total_tokens": 3892314112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 2.964289903640747, "created_at": "2025-01-16T02:31:17.624107+00:00"} {"global_step": 7425, "acc_step": 0, "speed/wps": 12900.000455516183, "speed/FLOPS": 202612233612107.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06019197404384613, "optim/lr": 0.0029856448907438233, "optim/total_tokens": 3892838400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.063931703567505, "created_at": "2025-01-16T02:31:27.788906+00:00"} {"global_step": 7426, "acc_step": 0, "speed/wps": 12896.74302702146, "speed/FLOPS": 202561071221422.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054898347705602646, "optim/lr": 0.0029856330630985283, "optim/total_tokens": 3893362688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 3.0338478088378906, "created_at": "2025-01-16T02:31:37.955938+00:00"} {"global_step": 7427, "acc_step": 0, "speed/wps": 12897.784235278494, "speed/FLOPS": 202577424827868.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06639927625656128, "optim/lr": 0.002985621230606095, "optim/total_tokens": 3893886976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.983769416809082, "created_at": "2025-01-16T02:31:48.123145+00:00"} {"global_step": 7428, "acc_step": 0, "speed/wps": 12884.048165678565, "speed/FLOPS": 202361680979464.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055782947689294815, "optim/lr": 0.0029856093932665606, "optim/total_tokens": 3894411264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.074748992919922, "created_at": "2025-01-16T02:31:58.305021+00:00"} {"global_step": 7429, "acc_step": 0, "speed/wps": 12887.136741997685, "speed/FLOPS": 202410191314704.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051459744572639465, "optim/lr": 0.002985597551079964, "optim/total_tokens": 3894935552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8958821296691895, "created_at": "2025-01-16T02:32:08.479902+00:00"} {"global_step": 7430, "acc_step": 0, "speed/wps": 12888.373127520963, "speed/FLOPS": 202429610448319.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07682467252016068, "optim/lr": 0.0029855857040463453, "optim/total_tokens": 3895459840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.102640390396118, "created_at": "2025-01-16T02:32:18.653557+00:00"} {"global_step": 7431, "acc_step": 0, "speed/wps": 12895.175722579195, "speed/FLOPS": 202536454551454.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0713641494512558, "optim/lr": 0.0029855738521657412, "optim/total_tokens": 3895984128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.937124729156494, "created_at": "2025-01-16T02:32:28.825115+00:00"} {"global_step": 7432, "acc_step": 0, "speed/wps": 12889.862742144805, "speed/FLOPS": 202453006892932.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.071648970246315, "optim/lr": 0.002985561995438192, "optim/total_tokens": 3896508416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284736, "loss/out": 3.0579590797424316, "created_at": "2025-01-16T02:32:38.999817+00:00"} {"global_step": 7433, "acc_step": 0, "speed/wps": 12894.913421177756, "speed/FLOPS": 202532334747504.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07522542029619217, "optim/lr": 0.0029855501338637348, "optim/total_tokens": 3897032704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.902435302734375, "created_at": "2025-01-16T02:32:49.169870+00:00"} {"global_step": 7434, "acc_step": 0, "speed/wps": 12894.27674730413, "speed/FLOPS": 202522334909438.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06021778658032417, "optim/lr": 0.002985538267442409, "optim/total_tokens": 3897556992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.9394609928131104, "created_at": "2025-01-16T02:32:59.339077+00:00"} {"global_step": 7435, "acc_step": 0, "speed/wps": 12899.368886860308, "speed/FLOPS": 202602313958500.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06475469470024109, "optim/lr": 0.0029855263961742536, "optim/total_tokens": 3898081280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358144, "loss/out": 3.065659999847412, "created_at": "2025-01-16T02:33:09.504011+00:00"} {"global_step": 7436, "acc_step": 0, "speed/wps": 12893.706633847301, "speed/FLOPS": 202513380494180.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07509565353393555, "optim/lr": 0.002985514520059307, "optim/total_tokens": 3898605568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494928, "loss/out": 2.998786449432373, "created_at": "2025-01-16T02:33:19.679017+00:00"} {"global_step": 7437, "acc_step": 0, "speed/wps": 12898.1736891261, "speed/FLOPS": 202583541735711.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05785132572054863, "optim/lr": 0.0029855026390976083, "optim/total_tokens": 3899129856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.0668039321899414, "created_at": "2025-01-16T02:33:29.852553+00:00"} {"global_step": 7438, "acc_step": 0, "speed/wps": 12894.157235600123, "speed/FLOPS": 202520457813900.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0701623260974884, "optim/lr": 0.002985490753289196, "optim/total_tokens": 3899654144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 2.9367899894714355, "created_at": "2025-01-16T02:33:40.030087+00:00"} {"global_step": 7439, "acc_step": 0, "speed/wps": 12895.679738054687, "speed/FLOPS": 202544370807086.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06422652304172516, "optim/lr": 0.002985478862634109, "optim/total_tokens": 3900178432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.1451210975646973, "created_at": "2025-01-16T02:33:50.200922+00:00"} {"global_step": 7440, "acc_step": 0, "speed/wps": 12898.218783248361, "speed/FLOPS": 202584250000867.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06614286452531815, "optim/lr": 0.002985466967132386, "optim/total_tokens": 3900702720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0272574424743652, "created_at": "2025-01-16T02:34:00.368841+00:00"} {"global_step": 7441, "acc_step": 0, "speed/wps": 12893.982825613823, "speed/FLOPS": 202517718465401.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09134472906589508, "optim/lr": 0.0029854550667840654, "optim/total_tokens": 3901227008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.976555824279785, "created_at": "2025-01-16T02:34:10.542644+00:00"} {"global_step": 7442, "acc_step": 0, "speed/wps": 12891.329875748992, "speed/FLOPS": 202476050242238.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06153931841254234, "optim/lr": 0.0029854431615891864, "optim/total_tokens": 3901751296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9847826957702637, "created_at": "2025-01-16T02:34:20.714912+00:00"} {"global_step": 7443, "acc_step": 0, "speed/wps": 12900.216328951024, "speed/FLOPS": 202615624201042.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06195759028196335, "optim/lr": 0.002985431251547788, "optim/total_tokens": 3902275584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.965083122253418, "created_at": "2025-01-16T02:34:30.881175+00:00"} {"global_step": 7444, "acc_step": 0, "speed/wps": 12892.731984315229, "speed/FLOPS": 202498072283970.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.054884620010852814, "optim/lr": 0.002985419336659909, "optim/total_tokens": 3902799872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330478, "loss/out": 3.028578758239746, "created_at": "2025-01-16T02:34:41.055147+00:00"} {"global_step": 7445, "acc_step": 0, "speed/wps": 12894.53454879368, "speed/FLOPS": 202526384036087.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056934233754873276, "optim/lr": 0.0029854074169255885, "optim/total_tokens": 3903324160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 2.8432817459106445, "created_at": "2025-01-16T02:34:51.224234+00:00"} {"global_step": 7446, "acc_step": 0, "speed/wps": 12900.510915669436, "speed/FLOPS": 202620251090260.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050532348453998566, "optim/lr": 0.0029853954923448646, "optim/total_tokens": 3903848448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.8714795112609863, "created_at": "2025-01-16T02:35:01.388041+00:00"} {"global_step": 7447, "acc_step": 0, "speed/wps": 12896.311249209703, "speed/FLOPS": 202554289557563.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05477260798215866, "optim/lr": 0.002985383562917777, "optim/total_tokens": 3904372736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0430922508239746, "created_at": "2025-01-16T02:35:11.559238+00:00"} {"global_step": 7448, "acc_step": 0, "speed/wps": 12902.271451242428, "speed/FLOPS": 202647902720663.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05583490803837776, "optim/lr": 0.002985371628644364, "optim/total_tokens": 3904897024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.992283344268799, "created_at": "2025-01-16T02:35:21.723645+00:00"} {"global_step": 7449, "acc_step": 0, "speed/wps": 12895.42961945891, "speed/FLOPS": 202540442350841.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05604858696460724, "optim/lr": 0.002985359689524665, "optim/total_tokens": 3905421312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 2.941652774810791, "created_at": "2025-01-16T02:35:31.891916+00:00"} {"global_step": 7450, "acc_step": 0, "speed/wps": 12896.334933647026, "speed/FLOPS": 202554661554197.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051733795553445816, "optim/lr": 0.0029853477455587195, "optim/total_tokens": 3905945600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.979372501373291, "created_at": "2025-01-16T02:35:42.062834+00:00"} {"global_step": 7451, "acc_step": 0, "speed/wps": 12895.045696519825, "speed/FLOPS": 202534412313517.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05402425676584244, "optim/lr": 0.0029853357967465654, "optim/total_tokens": 3906469888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.9936704635620117, "created_at": "2025-01-16T02:35:52.235852+00:00"} {"global_step": 7452, "acc_step": 0, "speed/wps": 12900.631309097622, "speed/FLOPS": 202622142034487.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0547991506755352, "optim/lr": 0.002985323843088242, "optim/total_tokens": 3906994176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.983564853668213, "created_at": "2025-01-16T02:36:02.399854+00:00"} {"global_step": 7453, "acc_step": 0, "speed/wps": 12898.977289572167, "speed/FLOPS": 202596163384979.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04886240139603615, "optim/lr": 0.002985311884583788, "optim/total_tokens": 3907518464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.954800605773926, "created_at": "2025-01-16T02:36:12.569099+00:00"} {"global_step": 7454, "acc_step": 0, "speed/wps": 12889.306087737541, "speed/FLOPS": 202444263870542.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06032206490635872, "optim/lr": 0.002985299921233243, "optim/total_tokens": 3908042752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.951709270477295, "created_at": "2025-01-16T02:36:22.745090+00:00"} {"global_step": 7455, "acc_step": 0, "speed/wps": 12892.28670337087, "speed/FLOPS": 202491078534859.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05217671021819115, "optim/lr": 0.0029852879530366462, "optim/total_tokens": 3908567040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 2.9062795639038086, "created_at": "2025-01-16T02:36:32.917526+00:00"} {"global_step": 7456, "acc_step": 0, "speed/wps": 12897.044103156746, "speed/FLOPS": 202565800035848.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05512024089694023, "optim/lr": 0.002985275979994036, "optim/total_tokens": 3909091328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.957228899002075, "created_at": "2025-01-16T02:36:43.086408+00:00"} {"global_step": 7457, "acc_step": 0, "speed/wps": 12887.145972412101, "speed/FLOPS": 202410336291047.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05850103497505188, "optim/lr": 0.0029852640021054517, "optim/total_tokens": 3909615616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.9692161083221436, "created_at": "2025-01-16T02:36:53.260864+00:00"} {"global_step": 7458, "acc_step": 0, "speed/wps": 12894.679181291162, "speed/FLOPS": 202528655688206.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06160194054245949, "optim/lr": 0.0029852520193709327, "optim/total_tokens": 3910139904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9780468940734863, "created_at": "2025-01-16T02:37:03.429948+00:00"} {"global_step": 7459, "acc_step": 0, "speed/wps": 12898.838612325604, "speed/FLOPS": 202593985268259.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055827923119068146, "optim/lr": 0.0029852400317905174, "optim/total_tokens": 3910664192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.056534767150879, "created_at": "2025-01-16T02:37:13.598367+00:00"} {"global_step": 7460, "acc_step": 0, "speed/wps": 12896.75315874805, "speed/FLOPS": 202561230354110.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05248022824525833, "optim/lr": 0.0029852280393642453, "optim/total_tokens": 3911188480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407300, "loss/out": 2.9937117099761963, "created_at": "2025-01-16T02:37:23.766119+00:00"} {"global_step": 7461, "acc_step": 0, "speed/wps": 12900.830933876634, "speed/FLOPS": 202625277415955.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07222440093755722, "optim/lr": 0.002985216042092156, "optim/total_tokens": 3911712768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9904229640960693, "created_at": "2025-01-16T02:37:33.929691+00:00"} {"global_step": 7462, "acc_step": 0, "speed/wps": 12892.266771654768, "speed/FLOPS": 202490765479868.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07866863161325455, "optim/lr": 0.0029852040399742878, "optim/total_tokens": 3912237056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.078444480895996, "created_at": "2025-01-16T02:37:44.101016+00:00"} {"global_step": 7463, "acc_step": 0, "speed/wps": 12899.710025642804, "speed/FLOPS": 202607672011853.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05343632400035858, "optim/lr": 0.0029851920330106804, "optim/total_tokens": 3912761344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418499, "loss/out": 3.0352327823638916, "created_at": "2025-01-16T02:37:54.268236+00:00"} {"global_step": 7464, "acc_step": 0, "speed/wps": 12892.1700758939, "speed/FLOPS": 202489246740070.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06922603398561478, "optim/lr": 0.0029851800212013727, "optim/total_tokens": 3913285632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.0410895347595215, "created_at": "2025-01-16T02:38:04.440642+00:00"} {"global_step": 7465, "acc_step": 0, "speed/wps": 12892.522242888235, "speed/FLOPS": 202494778006644.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0455787219107151, "optim/lr": 0.0029851680045464043, "optim/total_tokens": 3913809920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 2.986506462097168, "created_at": "2025-01-16T02:38:14.610954+00:00"} {"global_step": 7466, "acc_step": 0, "speed/wps": 12895.652649420163, "speed/FLOPS": 202543945342857.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045162133872509, "optim/lr": 0.0029851559830458138, "optim/total_tokens": 3914334208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 2.975243330001831, "created_at": "2025-01-16T02:38:24.781717+00:00"} {"global_step": 7467, "acc_step": 0, "speed/wps": 12894.361640767218, "speed/FLOPS": 202523668277931.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051898591220378876, "optim/lr": 0.002985143956699641, "optim/total_tokens": 3914858496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.830097198486328, "created_at": "2025-01-16T02:38:34.950423+00:00"} {"global_step": 7468, "acc_step": 0, "speed/wps": 12892.879164187712, "speed/FLOPS": 202500383946114.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05397878587245941, "optim/lr": 0.002985131925507925, "optim/total_tokens": 3915382784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9206483364105225, "created_at": "2025-01-16T02:38:45.122654+00:00"} {"global_step": 7469, "acc_step": 0, "speed/wps": 12892.453732085916, "speed/FLOPS": 202493701950350.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04845202714204788, "optim/lr": 0.0029851198894707046, "optim/total_tokens": 3915907072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 2.9926929473876953, "created_at": "2025-01-16T02:38:55.293821+00:00"} {"global_step": 7470, "acc_step": 0, "speed/wps": 12898.289686437842, "speed/FLOPS": 202585363632889.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06356456130743027, "optim/lr": 0.002985107848588019, "optim/total_tokens": 3916431360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.97756028175354, "created_at": "2025-01-16T02:39:05.461112+00:00"} {"global_step": 7471, "acc_step": 0, "speed/wps": 12897.749698279007, "speed/FLOPS": 202576882376833.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06491220742464066, "optim/lr": 0.0029850958028599093, "optim/total_tokens": 3916955648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306453, "loss/out": 2.8607423305511475, "created_at": "2025-01-16T02:39:15.628638+00:00"} {"global_step": 7472, "acc_step": 0, "speed/wps": 12899.046865170421, "speed/FLOPS": 202597256165356.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09698601812124252, "optim/lr": 0.0029850837522864123, "optim/total_tokens": 3917479936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 2.975856065750122, "created_at": "2025-01-16T02:39:25.797130+00:00"} {"global_step": 7473, "acc_step": 0, "speed/wps": 12889.111019391481, "speed/FLOPS": 202441200054115.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16883394122123718, "optim/lr": 0.0029850716968675686, "optim/total_tokens": 3918004224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9857964515686035, "created_at": "2025-01-16T02:39:35.975827+00:00"} {"global_step": 7474, "acc_step": 0, "speed/wps": 12890.883800186493, "speed/FLOPS": 202469044012557.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08012790977954865, "optim/lr": 0.002985059636603417, "optim/total_tokens": 3918528512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 2.9607770442962646, "created_at": "2025-01-16T02:39:46.147356+00:00"} {"global_step": 7475, "acc_step": 0, "speed/wps": 12899.073332019168, "speed/FLOPS": 202597671863585.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09400314837694168, "optim/lr": 0.002985047571493998, "optim/total_tokens": 3919052800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.967170238494873, "created_at": "2025-01-16T02:39:56.312792+00:00"} {"global_step": 7476, "acc_step": 0, "speed/wps": 12893.109857827289, "speed/FLOPS": 202504007306730.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06830041110515594, "optim/lr": 0.0029850355015393496, "optim/total_tokens": 3919577088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9209635257720947, "created_at": "2025-01-16T02:40:06.483929+00:00"} {"global_step": 7477, "acc_step": 0, "speed/wps": 12897.456978661186, "speed/FLOPS": 202572284812996.4, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.069169782102108, "optim/lr": 0.0029850234267395125, "optim/total_tokens": 3920101376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.952407121658325, "created_at": "2025-01-16T02:40:16.653082+00:00"} {"global_step": 7478, "acc_step": 0, "speed/wps": 12893.96162465379, "speed/FLOPS": 202517385475190.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08117883652448654, "optim/lr": 0.0029850113470945247, "optim/total_tokens": 3920625664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.014822006225586, "created_at": "2025-01-16T02:40:26.829253+00:00"} {"global_step": 7479, "acc_step": 0, "speed/wps": 12894.789126937761, "speed/FLOPS": 202530382535667.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05501627177000046, "optim/lr": 0.002984999262604426, "optim/total_tokens": 3921149952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.008296489715576, "created_at": "2025-01-16T02:40:36.999013+00:00"} {"global_step": 7480, "acc_step": 0, "speed/wps": 12893.795493618198, "speed/FLOPS": 202514776158988.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058858055621385574, "optim/lr": 0.0029849871732692563, "optim/total_tokens": 3921674240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.082897186279297, "created_at": "2025-01-16T02:40:47.168089+00:00"} {"global_step": 7481, "acc_step": 0, "speed/wps": 12896.825610570731, "speed/FLOPS": 202562368309545.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05976200848817825, "optim/lr": 0.0029849750790890547, "optim/total_tokens": 3922198528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.907832622528076, "created_at": "2025-01-16T02:40:57.337042+00:00"} {"global_step": 7482, "acc_step": 0, "speed/wps": 12895.682719288461, "speed/FLOPS": 202544417631459.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05034544691443443, "optim/lr": 0.002984962980063861, "optim/total_tokens": 3922722816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.984259843826294, "created_at": "2025-01-16T02:41:07.504999+00:00"} {"global_step": 7483, "acc_step": 0, "speed/wps": 12897.575064623756, "speed/FLOPS": 202574139515305.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0580255463719368, "optim/lr": 0.0029849508761937145, "optim/total_tokens": 3923247104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 2.8941452503204346, "created_at": "2025-01-16T02:41:17.671412+00:00"} {"global_step": 7484, "acc_step": 0, "speed/wps": 12896.94042896542, "speed/FLOPS": 202564171690210.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05203915387392044, "optim/lr": 0.0029849387674786548, "optim/total_tokens": 3923771392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0314249992370605, "created_at": "2025-01-16T02:41:27.839426+00:00"} {"global_step": 7485, "acc_step": 0, "speed/wps": 12894.267527936603, "speed/FLOPS": 202522190106602.25, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06901159882545471, "optim/lr": 0.0029849266539187214, "optim/total_tokens": 3924295680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 3.019598960876465, "created_at": "2025-01-16T02:41:38.008391+00:00"} {"global_step": 7486, "acc_step": 0, "speed/wps": 12896.714236837566, "speed/FLOPS": 202560619032022.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06133585423231125, "optim/lr": 0.002984914535513953, "optim/total_tokens": 3924819968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9336671829223633, "created_at": "2025-01-16T02:41:48.175139+00:00"} {"global_step": 7487, "acc_step": 0, "speed/wps": 12894.615882465554, "speed/FLOPS": 202527661493168.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05497903376817703, "optim/lr": 0.00298490241226439, "optim/total_tokens": 3925344256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.0092673301696777, "created_at": "2025-01-16T02:41:58.345547+00:00"} {"global_step": 7488, "acc_step": 0, "speed/wps": 12898.836459537451, "speed/FLOPS": 202593951455763.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054246917366981506, "optim/lr": 0.0029848902841700716, "optim/total_tokens": 3925868544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 2.9345405101776123, "created_at": "2025-01-16T02:42:08.510707+00:00"} {"global_step": 7489, "acc_step": 0, "speed/wps": 12899.354309136601, "speed/FLOPS": 202602084995318.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06975433230400085, "optim/lr": 0.002984878151231038, "optim/total_tokens": 3926392832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.0768167972564697, "created_at": "2025-01-16T02:42:18.679204+00:00"} {"global_step": 7490, "acc_step": 0, "speed/wps": 12897.312500134942, "speed/FLOPS": 202570015579207.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06330849230289459, "optim/lr": 0.0029848660134473284, "optim/total_tokens": 3926917120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.9125452041625977, "created_at": "2025-01-16T02:42:28.849645+00:00"} {"global_step": 7491, "acc_step": 0, "speed/wps": 12897.038988015229, "speed/FLOPS": 202565719695522.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05816197395324707, "optim/lr": 0.002984853870818982, "optim/total_tokens": 3927441408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 2.9300522804260254, "created_at": "2025-01-16T02:42:39.021682+00:00"} {"global_step": 7492, "acc_step": 0, "speed/wps": 12891.93806371075, "speed/FLOPS": 202485602669915.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06156459450721741, "optim/lr": 0.0029848417233460387, "optim/total_tokens": 3927965696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8799386024475098, "created_at": "2025-01-16T02:42:49.193714+00:00"} {"global_step": 7493, "acc_step": 0, "speed/wps": 12891.525652470826, "speed/FLOPS": 202479125184680.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054181311279535294, "optim/lr": 0.0029848295710285383, "optim/total_tokens": 3928489984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 2.858224391937256, "created_at": "2025-01-16T02:42:59.367151+00:00"} {"global_step": 7494, "acc_step": 0, "speed/wps": 12895.014550912276, "speed/FLOPS": 202533923128952.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10956387966871262, "optim/lr": 0.0029848174138665203, "optim/total_tokens": 3929014272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.1171960830688477, "created_at": "2025-01-16T02:43:09.536563+00:00"} {"global_step": 7495, "acc_step": 0, "speed/wps": 12895.354569702875, "speed/FLOPS": 202539263591293.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0765194371342659, "optim/lr": 0.0029848052518600247, "optim/total_tokens": 3929538560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0716400146484375, "created_at": "2025-01-16T02:43:19.704356+00:00"} {"global_step": 7496, "acc_step": 0, "speed/wps": 12896.228421722351, "speed/FLOPS": 202552988638057.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061813682317733765, "optim/lr": 0.0029847930850090904, "optim/total_tokens": 3930062848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0051441192626953, "created_at": "2025-01-16T02:43:29.872687+00:00"} {"global_step": 7497, "acc_step": 0, "speed/wps": 12885.295135269127, "speed/FLOPS": 202381266350400.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07343427836894989, "optim/lr": 0.0029847809133137578, "optim/total_tokens": 3930587136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.0652568340301514, "created_at": "2025-01-16T02:43:40.049578+00:00"} {"global_step": 7498, "acc_step": 0, "speed/wps": 12884.892297075683, "speed/FLOPS": 202374939223014.28, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0594555102288723, "optim/lr": 0.0029847687367740663, "optim/total_tokens": 3931111424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.9470601081848145, "created_at": "2025-01-16T02:43:50.227002+00:00"} {"global_step": 7499, "acc_step": 0, "speed/wps": 12895.166768951487, "speed/FLOPS": 202536313922427.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06281022727489471, "optim/lr": 0.0029847565553900554, "optim/total_tokens": 3931635712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 2.9979159832000732, "created_at": "2025-01-16T02:44:00.396597+00:00"} {"global_step": 7500, "acc_step": 0, "speed/wps": 12893.538319381792, "speed/FLOPS": 202510736884211.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07537510246038437, "optim/lr": 0.002984744369161766, "optim/total_tokens": 3932160000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.048192024230957, "created_at": "2025-01-16T02:44:10.565879+00:00"} {"global_step": 7501, "acc_step": 0, "speed/wps": 5553.065461290554, "speed/FLOPS": 87218523781153.39, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 0.061048563569784164, "optim/lr": 0.0029847321780892364, "optim/total_tokens": 3932684288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 3.0723507404327393, "created_at": "2025-01-16T02:44:34.176724+00:00"} {"global_step": 7502, "acc_step": 0, "speed/wps": 12931.89027399388, "speed/FLOPS": 203113107032502.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10916470736265182, "optim/lr": 0.002984719982172507, "optim/total_tokens": 3933208576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 2.872516393661499, "created_at": "2025-01-16T02:44:44.316659+00:00"} {"global_step": 7503, "acc_step": 0, "speed/wps": 12912.590016115555, "speed/FLOPS": 202809969961189.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05750122666358948, "optim/lr": 0.0029847077814116178, "optim/total_tokens": 3933732864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.963243246078491, "created_at": "2025-01-16T02:44:54.471917+00:00"} {"global_step": 7504, "acc_step": 0, "speed/wps": 12895.534811344045, "speed/FLOPS": 202542094533946.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07395968586206436, "optim/lr": 0.002984695575806608, "optim/total_tokens": 3934257152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.8706812858581543, "created_at": "2025-01-16T02:45:04.643033+00:00"} {"global_step": 7505, "acc_step": 0, "speed/wps": 12891.77193521743, "speed/FLOPS": 202482993393643.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07666554301977158, "optim/lr": 0.0029846833653575182, "optim/total_tokens": 3934781440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 2.986614227294922, "created_at": "2025-01-16T02:45:14.813932+00:00"} {"global_step": 7506, "acc_step": 0, "speed/wps": 12894.113932425054, "speed/FLOPS": 202519777678031.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050160497426986694, "optim/lr": 0.0029846711500643875, "optim/total_tokens": 3935305728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.058004856109619, "created_at": "2025-01-16T02:45:24.988429+00:00"} {"global_step": 7507, "acc_step": 0, "speed/wps": 12892.680319263862, "speed/FLOPS": 202497260813343.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06109481304883957, "optim/lr": 0.0029846589299272566, "optim/total_tokens": 3935830016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.001398801803589, "created_at": "2025-01-16T02:45:35.162948+00:00"} {"global_step": 7508, "acc_step": 0, "speed/wps": 12892.77200422767, "speed/FLOPS": 202498700851688.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12740828096866608, "optim/lr": 0.0029846467049461645, "optim/total_tokens": 3936354304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.1613638401031494, "created_at": "2025-01-16T02:45:45.335881+00:00"} {"global_step": 7509, "acc_step": 0, "speed/wps": 12892.0637225066, "speed/FLOPS": 202487576313978.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08151403069496155, "optim/lr": 0.0029846344751211515, "optim/total_tokens": 3936878592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 2.9814887046813965, "created_at": "2025-01-16T02:45:55.510402+00:00"} {"global_step": 7510, "acc_step": 0, "speed/wps": 12895.288881058103, "speed/FLOPS": 202538231860861.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061314038932323456, "optim/lr": 0.0029846222404522576, "optim/total_tokens": 3937402880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.0202040672302246, "created_at": "2025-01-16T02:46:05.681496+00:00"} {"global_step": 7511, "acc_step": 0, "speed/wps": 12890.276302005966, "speed/FLOPS": 202459502418842.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07149873673915863, "optim/lr": 0.002984610000939523, "optim/total_tokens": 3937927168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.1608924865722656, "created_at": "2025-01-16T02:46:15.855773+00:00"} {"global_step": 7512, "acc_step": 0, "speed/wps": 12890.92238818278, "speed/FLOPS": 202469650090064.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06870361417531967, "optim/lr": 0.0029845977565829862, "optim/total_tokens": 3938451456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 3.035750150680542, "created_at": "2025-01-16T02:46:26.027485+00:00"} {"global_step": 7513, "acc_step": 0, "speed/wps": 12890.494926936886, "speed/FLOPS": 202462936223804.7, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07234197854995728, "optim/lr": 0.0029845855073826885, "optim/total_tokens": 3938975744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.92440128326416, "created_at": "2025-01-16T02:46:36.201601+00:00"} {"global_step": 7514, "acc_step": 0, "speed/wps": 12895.899684256028, "speed/FLOPS": 202547825364416.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07769106328487396, "optim/lr": 0.00298457325333867, "optim/total_tokens": 3939500032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 2.963123321533203, "created_at": "2025-01-16T02:46:46.373249+00:00"} {"global_step": 7515, "acc_step": 0, "speed/wps": 12889.365908731012, "speed/FLOPS": 202445203441447.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05652520805597305, "optim/lr": 0.00298456099445097, "optim/total_tokens": 3940024320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9717905521392822, "created_at": "2025-01-16T02:46:56.548261+00:00"} {"global_step": 7516, "acc_step": 0, "speed/wps": 12889.130282675336, "speed/FLOPS": 202441502610459.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05830395966768265, "optim/lr": 0.002984548730719628, "optim/total_tokens": 3940548608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.0985333919525146, "created_at": "2025-01-16T02:47:06.724419+00:00"} {"global_step": 7517, "acc_step": 0, "speed/wps": 12883.946805678628, "speed/FLOPS": 202360088981382.2, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05388333648443222, "optim/lr": 0.002984536462144686, "optim/total_tokens": 3941072896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.0050911903381348, "created_at": "2025-01-16T02:47:16.902641+00:00"} {"global_step": 7518, "acc_step": 0, "speed/wps": 12890.584890074226, "speed/FLOPS": 202464349218498.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052861884236335754, "optim/lr": 0.002984524188726182, "optim/total_tokens": 3941597184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.999657154083252, "created_at": "2025-01-16T02:47:27.075272+00:00"} {"global_step": 7519, "acc_step": 0, "speed/wps": 12889.728947519874, "speed/FLOPS": 202450905464494.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05050051212310791, "optim/lr": 0.002984511910464157, "optim/total_tokens": 3942121472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 3.0240278244018555, "created_at": "2025-01-16T02:47:37.251218+00:00"} {"global_step": 7520, "acc_step": 0, "speed/wps": 12891.160303201683, "speed/FLOPS": 202473386872366.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04525032639503479, "optim/lr": 0.0029844996273586505, "optim/total_tokens": 3942645760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.009310722351074, "created_at": "2025-01-16T02:47:47.422499+00:00"} {"global_step": 7521, "acc_step": 0, "speed/wps": 12887.127431352714, "speed/FLOPS": 202410045078231.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051782023161649704, "optim/lr": 0.002984487339409703, "optim/total_tokens": 3943170048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 3.0469977855682373, "created_at": "2025-01-16T02:47:57.598597+00:00"} {"global_step": 7522, "acc_step": 0, "speed/wps": 12887.813335020532, "speed/FLOPS": 202420818137864.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053218793123960495, "optim/lr": 0.002984475046617355, "optim/total_tokens": 3943694336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.094599962234497, "created_at": "2025-01-16T02:48:07.773453+00:00"} {"global_step": 7523, "acc_step": 0, "speed/wps": 12884.442534896556, "speed/FLOPS": 202367875089953.1, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0495922677218914, "optim/lr": 0.0029844627489816456, "optim/total_tokens": 3944218624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 3.0144882202148438, "created_at": "2025-01-16T02:48:17.953772+00:00"} {"global_step": 7524, "acc_step": 0, "speed/wps": 12891.106381278001, "speed/FLOPS": 202472539954458.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05019216239452362, "optim/lr": 0.0029844504465026157, "optim/total_tokens": 3944742912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.8215646743774414, "created_at": "2025-01-16T02:48:28.128853+00:00"} {"global_step": 7525, "acc_step": 0, "speed/wps": 12887.024519530512, "speed/FLOPS": 202408428706648.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05792294442653656, "optim/lr": 0.002984438139180305, "optim/total_tokens": 3945267200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.915072441101074, "created_at": "2025-01-16T02:48:38.304879+00:00"} {"global_step": 7526, "acc_step": 0, "speed/wps": 12892.872226356387, "speed/FLOPS": 202500274977939.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05868682265281677, "optim/lr": 0.002984425827014754, "optim/total_tokens": 3945791488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.047729015350342, "created_at": "2025-01-16T02:48:48.475865+00:00"} {"global_step": 7527, "acc_step": 0, "speed/wps": 12886.837341288172, "speed/FLOPS": 202405488815142.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05523175746202469, "optim/lr": 0.0029844135100060027, "optim/total_tokens": 3946315776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 2.9150402545928955, "created_at": "2025-01-16T02:48:58.650818+00:00"} {"global_step": 7528, "acc_step": 0, "speed/wps": 12891.290213130245, "speed/FLOPS": 202475427286309.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054399751126766205, "optim/lr": 0.0029844011881540915, "optim/total_tokens": 3946840064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9062540531158447, "created_at": "2025-01-16T02:49:08.821935+00:00"} {"global_step": 7529, "acc_step": 0, "speed/wps": 12890.00386387628, "speed/FLOPS": 202455223403646.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051814258098602295, "optim/lr": 0.00298438886145906, "optim/total_tokens": 3947364352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.971025228500366, "created_at": "2025-01-16T02:49:18.996568+00:00"} {"global_step": 7530, "acc_step": 0, "speed/wps": 12888.764751646877, "speed/FLOPS": 202435761443365.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.06104882061481476, "optim/lr": 0.002984376529920949, "optim/total_tokens": 3947888640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348536, "loss/out": 3.027656078338623, "created_at": "2025-01-16T02:49:29.171566+00:00"} {"global_step": 7531, "acc_step": 0, "speed/wps": 12893.450640825662, "speed/FLOPS": 202509359772004.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0570177361369133, "optim/lr": 0.002984364193539798, "optim/total_tokens": 3948412928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9790971279144287, "created_at": "2025-01-16T02:49:39.343282+00:00"} {"global_step": 7532, "acc_step": 0, "speed/wps": 12895.369276630818, "speed/FLOPS": 202539494583806.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053834397345781326, "optim/lr": 0.002984351852315649, "optim/total_tokens": 3948937216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.968660831451416, "created_at": "2025-01-16T02:49:49.514169+00:00"} {"global_step": 7533, "acc_step": 0, "speed/wps": 12889.55996720037, "speed/FLOPS": 202448251396373.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06157241389155388, "optim/lr": 0.0029843395062485404, "optim/total_tokens": 3949461504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.054410696029663, "created_at": "2025-01-16T02:49:59.686901+00:00"} {"global_step": 7534, "acc_step": 0, "speed/wps": 12888.598703746704, "speed/FLOPS": 202433153432919.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05092332512140274, "optim/lr": 0.0029843271553385135, "optim/total_tokens": 3949985792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340756, "loss/out": 2.9964025020599365, "created_at": "2025-01-16T02:50:09.863290+00:00"} {"global_step": 7535, "acc_step": 0, "speed/wps": 12889.140935996023, "speed/FLOPS": 202441669935499.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07128800451755524, "optim/lr": 0.002984314799585608, "optim/total_tokens": 3950510080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.930635929107666, "created_at": "2025-01-16T02:50:20.036115+00:00"} {"global_step": 7536, "acc_step": 0, "speed/wps": 12887.94506936855, "speed/FLOPS": 202422887206822.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06267642229795456, "optim/lr": 0.002984302438989864, "optim/total_tokens": 3951034368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.910792112350464, "created_at": "2025-01-16T02:50:30.209745+00:00"} {"global_step": 7537, "acc_step": 0, "speed/wps": 12884.039724062113, "speed/FLOPS": 202361548392278.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05972229689359665, "optim/lr": 0.0029842900735513227, "optim/total_tokens": 3951558656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 2.9920852184295654, "created_at": "2025-01-16T02:50:40.388869+00:00"} {"global_step": 7538, "acc_step": 0, "speed/wps": 12888.525591289706, "speed/FLOPS": 202432005101314.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06562726944684982, "optim/lr": 0.002984277703270024, "optim/total_tokens": 3952082944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.931704044342041, "created_at": "2025-01-16T02:50:50.562079+00:00"} {"global_step": 7539, "acc_step": 0, "speed/wps": 12886.32518277374, "speed/FLOPS": 202397444661893.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06922879815101624, "optim/lr": 0.0029842653281460085, "optim/total_tokens": 3952607232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.0095338821411133, "created_at": "2025-01-16T02:51:00.743724+00:00"} {"global_step": 7540, "acc_step": 0, "speed/wps": 12883.93419240537, "speed/FLOPS": 202359890872593.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07165581732988358, "optim/lr": 0.002984252948179316, "optim/total_tokens": 3953131520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.1642279624938965, "created_at": "2025-01-16T02:51:10.922007+00:00"} {"global_step": 7541, "acc_step": 0, "speed/wps": 12887.39590430216, "speed/FLOPS": 202414261814822.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.060567278414964676, "optim/lr": 0.0029842405633699874, "optim/total_tokens": 3953655808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.032838821411133, "created_at": "2025-01-16T02:51:21.099220+00:00"} {"global_step": 7542, "acc_step": 0, "speed/wps": 12893.853467411656, "speed/FLOPS": 202515686717079.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07185295969247818, "optim/lr": 0.002984228173718063, "optim/total_tokens": 3954180096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.957068920135498, "created_at": "2025-01-16T02:51:31.271505+00:00"} {"global_step": 7543, "acc_step": 0, "speed/wps": 12888.304942893, "speed/FLOPS": 202428539515042.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05656881257891655, "optim/lr": 0.0029842157792235834, "optim/total_tokens": 3954704384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 3.0795223712921143, "created_at": "2025-01-16T02:51:41.445073+00:00"} {"global_step": 7544, "acc_step": 0, "speed/wps": 12888.489138648943, "speed/FLOPS": 202431432562500.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06156671792268753, "optim/lr": 0.0029842033798865883, "optim/total_tokens": 3955228672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.028580904006958, "created_at": "2025-01-16T02:51:51.618295+00:00"} {"global_step": 7545, "acc_step": 0, "speed/wps": 12889.413746093618, "speed/FLOPS": 202445954792962.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06802904605865479, "optim/lr": 0.002984190975707119, "optim/total_tokens": 3955752960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.8393523693084717, "created_at": "2025-01-16T02:52:01.790623+00:00"} {"global_step": 7546, "acc_step": 0, "speed/wps": 12892.232438984274, "speed/FLOPS": 202490226238100.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05663691833615303, "optim/lr": 0.0029841785666852158, "optim/total_tokens": 3956277248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.2079296112060547, "created_at": "2025-01-16T02:52:11.962763+00:00"} {"global_step": 7547, "acc_step": 0, "speed/wps": 12884.115577519487, "speed/FLOPS": 202362739775059.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052857644855976105, "optim/lr": 0.002984166152820919, "optim/total_tokens": 3956801536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.011220932006836, "created_at": "2025-01-16T02:52:22.143702+00:00"} {"global_step": 7548, "acc_step": 0, "speed/wps": 12891.790712835515, "speed/FLOPS": 202483288321939.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04842852056026459, "optim/lr": 0.0029841537341142687, "optim/total_tokens": 3957325824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.0236754417419434, "created_at": "2025-01-16T02:52:32.314294+00:00"} {"global_step": 7549, "acc_step": 0, "speed/wps": 12886.937193866535, "speed/FLOPS": 202407057137097.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06369391083717346, "optim/lr": 0.0029841413105653066, "optim/total_tokens": 3957850112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.973702907562256, "created_at": "2025-01-16T02:52:42.489444+00:00"} {"global_step": 7550, "acc_step": 0, "speed/wps": 12889.137445482253, "speed/FLOPS": 202441615112184.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08016965538263321, "optim/lr": 0.002984128882174072, "optim/total_tokens": 3958374400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.9972970485687256, "created_at": "2025-01-16T02:52:52.668052+00:00"} {"global_step": 7551, "acc_step": 0, "speed/wps": 12889.20522930777, "speed/FLOPS": 202442679750309.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05261499062180519, "optim/lr": 0.0029841164489406062, "optim/total_tokens": 3958898688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 3.0352745056152344, "created_at": "2025-01-16T02:53:02.842052+00:00"} {"global_step": 7552, "acc_step": 0, "speed/wps": 12886.438138776579, "speed/FLOPS": 202399218791135.94, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048307936638593674, "optim/lr": 0.0029841040108649496, "optim/total_tokens": 3959422976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399773, "loss/out": 2.888671875, "created_at": "2025-01-16T02:53:13.018313+00:00"} {"global_step": 7553, "acc_step": 0, "speed/wps": 12890.304219297554, "speed/FLOPS": 202459940898269.2, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05044199153780937, "optim/lr": 0.0029840915679471424, "optim/total_tokens": 3959947264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9649124145507812, "created_at": "2025-01-16T02:53:23.190201+00:00"} {"global_step": 7554, "acc_step": 0, "speed/wps": 12891.733430702314, "speed/FLOPS": 202482388627323.25, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062243957072496414, "optim/lr": 0.0029840791201872253, "optim/total_tokens": 3960471552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470861, "loss/out": 3.0526657104492188, "created_at": "2025-01-16T02:53:33.362224+00:00"} {"global_step": 7555, "acc_step": 0, "speed/wps": 12887.982636592607, "speed/FLOPS": 202423477251699.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06700073927640915, "optim/lr": 0.0029840666675852398, "optim/total_tokens": 3960995840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9513204097747803, "created_at": "2025-01-16T02:53:43.542023+00:00"} {"global_step": 7556, "acc_step": 0, "speed/wps": 12888.416243428343, "speed/FLOPS": 202430287642892.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05829960107803345, "optim/lr": 0.002984054210141225, "optim/total_tokens": 3961520128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.9617884159088135, "created_at": "2025-01-16T02:53:53.716005+00:00"} {"global_step": 7557, "acc_step": 0, "speed/wps": 12891.226420932817, "speed/FLOPS": 202474425342188.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07018174231052399, "optim/lr": 0.002984041747855223, "optim/total_tokens": 3962044416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.985520601272583, "created_at": "2025-01-16T02:54:03.890306+00:00"} {"global_step": 7558, "acc_step": 0, "speed/wps": 12889.181560427056, "speed/FLOPS": 202442307998013.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0904899314045906, "optim/lr": 0.002984029280727274, "optim/total_tokens": 3962568704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0473527908325195, "created_at": "2025-01-16T02:54:14.066557+00:00"} {"global_step": 7559, "acc_step": 0, "speed/wps": 12891.544631299317, "speed/FLOPS": 202479423273261.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06142977997660637, "optim/lr": 0.0029840168087574177, "optim/total_tokens": 3963092992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 2.86501407623291, "created_at": "2025-01-16T02:54:24.237412+00:00"} {"global_step": 7560, "acc_step": 0, "speed/wps": 12888.2969925911, "speed/FLOPS": 202428414644625.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07442628592252731, "optim/lr": 0.0029840043319456964, "optim/total_tokens": 3963617280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.0313284397125244, "created_at": "2025-01-16T02:54:34.411181+00:00"} {"global_step": 7561, "acc_step": 0, "speed/wps": 12890.594224606783, "speed/FLOPS": 202464495830159.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06860290467739105, "optim/lr": 0.0029839918502921497, "optim/total_tokens": 3964141568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.935142993927002, "created_at": "2025-01-16T02:54:44.584081+00:00"} {"global_step": 7562, "acc_step": 0, "speed/wps": 12893.105487568579, "speed/FLOPS": 202503938665811.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06676450371742249, "optim/lr": 0.002983979363796818, "optim/total_tokens": 3964665856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 3.063727855682373, "created_at": "2025-01-16T02:54:54.754857+00:00"} {"global_step": 7563, "acc_step": 0, "speed/wps": 12886.630255371527, "speed/FLOPS": 202402236246255.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08138098567724228, "optim/lr": 0.0029839668724597436, "optim/total_tokens": 3965190144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.9786243438720703, "created_at": "2025-01-16T02:55:04.932207+00:00"} {"global_step": 7564, "acc_step": 0, "speed/wps": 12893.977663285266, "speed/FLOPS": 202517637383937.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04674172401428223, "optim/lr": 0.002983954376280966, "optim/total_tokens": 3965714432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.017996311187744, "created_at": "2025-01-16T02:55:15.104497+00:00"} {"global_step": 7565, "acc_step": 0, "speed/wps": 12885.460298187103, "speed/FLOPS": 202383860460984.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.06895849108695984, "optim/lr": 0.0029839418752605268, "optim/total_tokens": 3966238720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.026244640350342, "created_at": "2025-01-16T02:55:25.280032+00:00"} {"global_step": 7566, "acc_step": 0, "speed/wps": 12886.711127105622, "speed/FLOPS": 202403506447968.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04975973814725876, "optim/lr": 0.0029839293693984656, "optim/total_tokens": 3966763008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.0118093490600586, "created_at": "2025-01-16T02:55:35.455749+00:00"} {"global_step": 7567, "acc_step": 0, "speed/wps": 12891.84817143488, "speed/FLOPS": 202484190788198.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07816573232412338, "optim/lr": 0.0029839168586948245, "optim/total_tokens": 3967287296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.954610586166382, "created_at": "2025-01-16T02:55:45.629841+00:00"} {"global_step": 7568, "acc_step": 0, "speed/wps": 12894.016200108617, "speed/FLOPS": 202518242657703.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06789009273052216, "optim/lr": 0.002983904343149643, "optim/total_tokens": 3967811584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.9518699645996094, "created_at": "2025-01-16T02:55:55.802953+00:00"} {"global_step": 7569, "acc_step": 0, "speed/wps": 12893.517350222573, "speed/FLOPS": 202510407534751.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06214222311973572, "optim/lr": 0.002983891822762963, "optim/total_tokens": 3968335872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 2.9349424839019775, "created_at": "2025-01-16T02:56:05.972693+00:00"} {"global_step": 7570, "acc_step": 0, "speed/wps": 12891.049891621446, "speed/FLOPS": 202471652706777.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06376012414693832, "optim/lr": 0.0029838792975348245, "optim/total_tokens": 3968860160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.092167377471924, "created_at": "2025-01-16T02:56:16.152098+00:00"} {"global_step": 7571, "acc_step": 0, "speed/wps": 12891.824132667189, "speed/FLOPS": 202483813226319.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05491937696933746, "optim/lr": 0.0029838667674652697, "optim/total_tokens": 3969384448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.052454948425293, "created_at": "2025-01-16T02:56:26.324558+00:00"} {"global_step": 7572, "acc_step": 0, "speed/wps": 12892.621211570344, "speed/FLOPS": 202496332445794.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07945467531681061, "optim/lr": 0.002983854232554339, "optim/total_tokens": 3969908736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.932378053665161, "created_at": "2025-01-16T02:56:36.500074+00:00"} {"global_step": 7573, "acc_step": 0, "speed/wps": 12888.911335751482, "speed/FLOPS": 202438063748155.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09771043062210083, "optim/lr": 0.002983841692802072, "optim/total_tokens": 3970433024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291617, "loss/out": 2.903228282928467, "created_at": "2025-01-16T02:56:46.672963+00:00"} {"global_step": 7574, "acc_step": 0, "speed/wps": 12893.541583120854, "speed/FLOPS": 202510788145718.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08172152936458588, "optim/lr": 0.002983829148208511, "optim/total_tokens": 3970957312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0433082580566406, "created_at": "2025-01-16T02:56:56.843372+00:00"} {"global_step": 7575, "acc_step": 0, "speed/wps": 12890.829839250247, "speed/FLOPS": 202468196481903.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05295659601688385, "optim/lr": 0.002983816598773696, "optim/total_tokens": 3971481600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.9682483673095703, "created_at": "2025-01-16T02:57:07.016660+00:00"} {"global_step": 7576, "acc_step": 0, "speed/wps": 12896.696564046673, "speed/FLOPS": 202560341456556.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05561422184109688, "optim/lr": 0.002983804044497668, "optim/total_tokens": 3972005888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 2.9069314002990723, "created_at": "2025-01-16T02:57:17.187109+00:00"} {"global_step": 7577, "acc_step": 0, "speed/wps": 12891.318259981048, "speed/FLOPS": 202475867800641.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06496327370405197, "optim/lr": 0.0029837914853804697, "optim/total_tokens": 3972530176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.915229320526123, "created_at": "2025-01-16T02:57:27.358276+00:00"} {"global_step": 7578, "acc_step": 0, "speed/wps": 12895.05591891886, "speed/FLOPS": 202534572870341.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07770702987909317, "optim/lr": 0.0029837789214221403, "optim/total_tokens": 3973054464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.829317569732666, "created_at": "2025-01-16T02:57:37.529040+00:00"} {"global_step": 7579, "acc_step": 0, "speed/wps": 12890.699824071306, "speed/FLOPS": 202466154414858.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05722828209400177, "optim/lr": 0.002983766352622721, "optim/total_tokens": 3973578752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411310, "loss/out": 2.964207649230957, "created_at": "2025-01-16T02:57:47.700504+00:00"} {"global_step": 7580, "acc_step": 0, "speed/wps": 12890.465874198138, "speed/FLOPS": 202462479910620.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06335221230983734, "optim/lr": 0.0029837537789822533, "optim/total_tokens": 3974103040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.08896541595459, "created_at": "2025-01-16T02:57:57.872724+00:00"} {"global_step": 7581, "acc_step": 0, "speed/wps": 12893.692573315951, "speed/FLOPS": 202513159654214.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056708693504333496, "optim/lr": 0.0029837412005007777, "optim/total_tokens": 3974627328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.054372787475586, "created_at": "2025-01-16T02:58:08.044539+00:00"} {"global_step": 7582, "acc_step": 0, "speed/wps": 12892.642999836613, "speed/FLOPS": 202496674660456.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05976339429616928, "optim/lr": 0.002983728617178336, "optim/total_tokens": 3975151616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.987978935241699, "created_at": "2025-01-16T02:58:18.221862+00:00"} {"global_step": 7583, "acc_step": 0, "speed/wps": 12885.981767720743, "speed/FLOPS": 202392050856584.5, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05378013104200363, "optim/lr": 0.0029837160290149684, "optim/total_tokens": 3975675904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.977660655975342, "created_at": "2025-01-16T02:58:28.398073+00:00"} {"global_step": 7584, "acc_step": 0, "speed/wps": 12869.941197205044, "speed/FLOPS": 202140111654581.72, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06831780821084976, "optim/lr": 0.002983703436010717, "optim/total_tokens": 3976200192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.896466016769409, "created_at": "2025-01-16T02:58:38.590938+00:00"} {"global_step": 7585, "acc_step": 0, "speed/wps": 12875.827835764687, "speed/FLOPS": 202232569402249.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053847841918468475, "optim/lr": 0.002983690838165621, "optim/total_tokens": 3976724480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.0433688163757324, "created_at": "2025-01-16T02:58:48.777593+00:00"} {"global_step": 7586, "acc_step": 0, "speed/wps": 12886.42485868585, "speed/FLOPS": 202399010209062.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06356610357761383, "optim/lr": 0.0029836782354797234, "optim/total_tokens": 3977248768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9746899604797363, "created_at": "2025-01-16T02:58:58.955310+00:00"} {"global_step": 7587, "acc_step": 0, "speed/wps": 12894.276574824811, "speed/FLOPS": 202522332200413.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057004597038030624, "optim/lr": 0.0029836656279530653, "optim/total_tokens": 3977773056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.890756130218506, "created_at": "2025-01-16T02:59:09.126346+00:00"} {"global_step": 7588, "acc_step": 0, "speed/wps": 12888.199173757403, "speed/FLOPS": 202426878265425.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06469079107046127, "optim/lr": 0.0029836530155856866, "optim/total_tokens": 3978297344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296188, "loss/out": 2.996967315673828, "created_at": "2025-01-16T02:59:19.300634+00:00"} {"global_step": 7589, "acc_step": 0, "speed/wps": 12892.52404043706, "speed/FLOPS": 202494806239619.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053716521710157394, "optim/lr": 0.0029836403983776287, "optim/total_tokens": 3978821632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8847126960754395, "created_at": "2025-01-16T02:59:29.476416+00:00"} {"global_step": 7590, "acc_step": 0, "speed/wps": 12891.49775468631, "speed/FLOPS": 202478687011639.12, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07180413603782654, "optim/lr": 0.0029836277763289334, "optim/total_tokens": 3979345920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9614181518554688, "created_at": "2025-01-16T02:59:39.648026+00:00"} {"global_step": 7591, "acc_step": 0, "speed/wps": 12881.979077494574, "speed/FLOPS": 202329183106309.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07218718528747559, "optim/lr": 0.0029836151494396414, "optim/total_tokens": 3979870208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374042, "loss/out": 2.974325656890869, "created_at": "2025-01-16T02:59:49.833464+00:00"} {"global_step": 7592, "acc_step": 0, "speed/wps": 12891.404450020962, "speed/FLOPS": 202477221533655.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05792487412691116, "optim/lr": 0.002983602517709795, "optim/total_tokens": 3980394496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 3.0200376510620117, "created_at": "2025-01-16T03:00:00.006891+00:00"} {"global_step": 7593, "acc_step": 0, "speed/wps": 12886.605817173222, "speed/FLOPS": 202401852410769.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08790746331214905, "optim/lr": 0.0029835898811394336, "optim/total_tokens": 3980918784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8652796745300293, "created_at": "2025-01-16T03:00:10.182950+00:00"} {"global_step": 7594, "acc_step": 0, "speed/wps": 12890.789167825884, "speed/FLOPS": 202467557681295.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05866627022624016, "optim/lr": 0.0029835772397286, "optim/total_tokens": 3981443072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9635703563690186, "created_at": "2025-01-16T03:00:20.355185+00:00"} {"global_step": 7595, "acc_step": 0, "speed/wps": 12889.8806193372, "speed/FLOPS": 202453287678804.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09757154434919357, "optim/lr": 0.0029835645934773343, "optim/total_tokens": 3981967360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 2.9780921936035156, "created_at": "2025-01-16T03:00:30.528635+00:00"} {"global_step": 7596, "acc_step": 0, "speed/wps": 12886.182321020957, "speed/FLOPS": 202395200821753.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08659212291240692, "optim/lr": 0.0029835519423856783, "optim/total_tokens": 3982491648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.0694241523742676, "created_at": "2025-01-16T03:00:40.709033+00:00"} {"global_step": 7597, "acc_step": 0, "speed/wps": 12888.714554867536, "speed/FLOPS": 202434973033967.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08614503592252731, "optim/lr": 0.002983539286453673, "optim/total_tokens": 3983015936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0962777137756348, "created_at": "2025-01-16T03:00:50.888038+00:00"} {"global_step": 7598, "acc_step": 0, "speed/wps": 12883.287449075704, "speed/FLOPS": 202349732879882.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09683609753847122, "optim/lr": 0.0029835266256813603, "optim/total_tokens": 3983540224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 3.045555591583252, "created_at": "2025-01-16T03:01:01.065641+00:00"} {"global_step": 7599, "acc_step": 0, "speed/wps": 12885.397449658569, "speed/FLOPS": 202382873338480.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06306450068950653, "optim/lr": 0.0029835139600687807, "optim/total_tokens": 3984064512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 3.037667989730835, "created_at": "2025-01-16T03:01:11.244983+00:00"} {"global_step": 7600, "acc_step": 0, "speed/wps": 12886.375939103085, "speed/FLOPS": 202398241859792.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08600127696990967, "optim/lr": 0.002983501289615976, "optim/total_tokens": 3984588800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9306280612945557, "created_at": "2025-01-16T03:01:21.424177+00:00"} {"global_step": 7601, "acc_step": 0, "speed/wps": 12891.614269624815, "speed/FLOPS": 202480517038856.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.062373969703912735, "optim/lr": 0.002983488614322988, "optim/total_tokens": 3985113088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.003448724746704, "created_at": "2025-01-16T03:01:31.595974+00:00"} {"global_step": 7602, "acc_step": 0, "speed/wps": 12887.489148590434, "speed/FLOPS": 202415726344501.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0740169882774353, "optim/lr": 0.0029834759341898576, "optim/total_tokens": 3985637376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.05106782913208, "created_at": "2025-01-16T03:01:41.770867+00:00"} {"global_step": 7603, "acc_step": 0, "speed/wps": 12888.74472576798, "speed/FLOPS": 202435446909418.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06642910093069077, "optim/lr": 0.002983463249216625, "optim/total_tokens": 3986161664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 2.9760990142822266, "created_at": "2025-01-16T03:01:51.946348+00:00"} {"global_step": 7604, "acc_step": 0, "speed/wps": 12891.188170005713, "speed/FLOPS": 202473824558816.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05248897895216942, "optim/lr": 0.002983450559403334, "optim/total_tokens": 3986685952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.0187528133392334, "created_at": "2025-01-16T03:02:02.117570+00:00"} {"global_step": 7605, "acc_step": 0, "speed/wps": 12888.165374974431, "speed/FLOPS": 202426347409094.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060443006455898285, "optim/lr": 0.002983437864750024, "optim/total_tokens": 3987210240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9249253273010254, "created_at": "2025-01-16T03:02:12.295181+00:00"} {"global_step": 7606, "acc_step": 0, "speed/wps": 12886.836184885906, "speed/FLOPS": 202405470652255.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05713162198662758, "optim/lr": 0.002983425165256737, "optim/total_tokens": 3987734528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 2.9887077808380127, "created_at": "2025-01-16T03:02:22.472659+00:00"} {"global_step": 7607, "acc_step": 0, "speed/wps": 12882.90592218101, "speed/FLOPS": 202343740475728.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07544084638357162, "optim/lr": 0.0029834124609235155, "optim/total_tokens": 3988258816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 3.015270709991455, "created_at": "2025-01-16T03:02:32.651698+00:00"} {"global_step": 7608, "acc_step": 0, "speed/wps": 12883.977421335301, "speed/FLOPS": 202360569842340.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06419263035058975, "optim/lr": 0.002983399751750399, "optim/total_tokens": 3988783104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9937288761138916, "created_at": "2025-01-16T03:02:42.829544+00:00"} {"global_step": 7609, "acc_step": 0, "speed/wps": 12885.234890789208, "speed/FLOPS": 202380320128058.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05230562388896942, "optim/lr": 0.0029833870377374305, "optim/total_tokens": 3989307392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.04537034034729, "created_at": "2025-01-16T03:02:53.005442+00:00"} {"global_step": 7610, "acc_step": 0, "speed/wps": 12889.46433669838, "speed/FLOPS": 202446749387929.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06561796367168427, "optim/lr": 0.0029833743188846505, "optim/total_tokens": 3989831680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 3.0275166034698486, "created_at": "2025-01-16T03:03:03.181038+00:00"} {"global_step": 7611, "acc_step": 0, "speed/wps": 12890.96394837794, "speed/FLOPS": 202470302850039.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.061556149274110794, "optim/lr": 0.0029833615951921012, "optim/total_tokens": 3990355968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.082782745361328, "created_at": "2025-01-16T03:03:13.353470+00:00"} {"global_step": 7612, "acc_step": 0, "speed/wps": 12888.18043492637, "speed/FLOPS": 202426583946334.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05878248065710068, "optim/lr": 0.002983348866659824, "optim/total_tokens": 3990880256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9794278144836426, "created_at": "2025-01-16T03:03:23.528649+00:00"} {"global_step": 7613, "acc_step": 0, "speed/wps": 12887.413053606659, "speed/FLOPS": 202414531168215.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05020196735858917, "optim/lr": 0.00298333613328786, "optim/total_tokens": 3991404544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0440425872802734, "created_at": "2025-01-16T03:03:33.707479+00:00"} {"global_step": 7614, "acc_step": 0, "speed/wps": 12888.288953778887, "speed/FLOPS": 202428288384033.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05176783353090286, "optim/lr": 0.0029833233950762516, "optim/total_tokens": 3991928832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 2.9730629920959473, "created_at": "2025-01-16T03:03:43.883947+00:00"} {"global_step": 7615, "acc_step": 0, "speed/wps": 12890.2803267827, "speed/FLOPS": 202459565633491.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06618398427963257, "optim/lr": 0.0029833106520250393, "optim/total_tokens": 3992453120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.828331470489502, "created_at": "2025-01-16T03:03:54.056875+00:00"} {"global_step": 7616, "acc_step": 0, "speed/wps": 12890.880560226937, "speed/FLOPS": 202468993124540.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05396705120801926, "optim/lr": 0.002983297904134265, "optim/total_tokens": 3992977408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9822726249694824, "created_at": "2025-01-16T03:04:04.231646+00:00"} {"global_step": 7617, "acc_step": 0, "speed/wps": 12890.552284254296, "speed/FLOPS": 202463837099291.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06083934009075165, "optim/lr": 0.002983285151403971, "optim/total_tokens": 3993501696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 3.0549659729003906, "created_at": "2025-01-16T03:04:14.406034+00:00"} {"global_step": 7618, "acc_step": 0, "speed/wps": 12891.992880664224, "speed/FLOPS": 202486463645496.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.061281967908144, "optim/lr": 0.002983272393834198, "optim/total_tokens": 3994025984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.0338845252990723, "created_at": "2025-01-16T03:04:24.582154+00:00"} {"global_step": 7619, "acc_step": 0, "speed/wps": 12893.70389568547, "speed/FLOPS": 202513337487586.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061427172273397446, "optim/lr": 0.002983259631424988, "optim/total_tokens": 3994550272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9580469131469727, "created_at": "2025-01-16T03:04:34.753678+00:00"} {"global_step": 7620, "acc_step": 0, "speed/wps": 12888.656400651847, "speed/FLOPS": 202434059642098.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0630226656794548, "optim/lr": 0.0029832468641763825, "optim/total_tokens": 3995074560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.954930067062378, "created_at": "2025-01-16T03:04:44.928515+00:00"} {"global_step": 7621, "acc_step": 0, "speed/wps": 12894.151599413637, "speed/FLOPS": 202520369289846.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06503176689147949, "optim/lr": 0.0029832340920884237, "optim/total_tokens": 3995598848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.9502410888671875, "created_at": "2025-01-16T03:04:55.101285+00:00"} {"global_step": 7622, "acc_step": 0, "speed/wps": 12891.041914317228, "speed/FLOPS": 202471527412252.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04806823655962944, "optim/lr": 0.0029832213151611527, "optim/total_tokens": 3996123136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.907320261001587, "created_at": "2025-01-16T03:05:05.282191+00:00"} {"global_step": 7623, "acc_step": 0, "speed/wps": 12891.305379574831, "speed/FLOPS": 202475665496162.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05944618582725525, "optim/lr": 0.0029832085333946117, "optim/total_tokens": 3996647424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.9037575721740723, "created_at": "2025-01-16T03:05:15.453771+00:00"} {"global_step": 7624, "acc_step": 0, "speed/wps": 12889.50050445509, "speed/FLOPS": 202447317452248.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06104596331715584, "optim/lr": 0.0029831957467888414, "optim/total_tokens": 3997171712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.9344935417175293, "created_at": "2025-01-16T03:05:25.631216+00:00"} {"global_step": 7625, "acc_step": 0, "speed/wps": 12889.562705963443, "speed/FLOPS": 202448294412410.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05323035642504692, "optim/lr": 0.0029831829553438843, "optim/total_tokens": 3997696000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.0901594161987305, "created_at": "2025-01-16T03:05:35.804719+00:00"} {"global_step": 7626, "acc_step": 0, "speed/wps": 12885.837671232326, "speed/FLOPS": 202389787623224.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059724606573581696, "optim/lr": 0.002983170159059782, "optim/total_tokens": 3998220288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482329, "loss/out": 2.956235647201538, "created_at": "2025-01-16T03:05:45.983253+00:00"} {"global_step": 7627, "acc_step": 0, "speed/wps": 12889.622863423998, "speed/FLOPS": 202449239267993.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0710475817322731, "optim/lr": 0.002983157357936577, "optim/total_tokens": 3998744576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.972113847732544, "created_at": "2025-01-16T03:05:56.156531+00:00"} {"global_step": 7628, "acc_step": 0, "speed/wps": 12886.601260169238, "speed/FLOPS": 202401780836759.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048256222158670425, "optim/lr": 0.002983144551974309, "optim/total_tokens": 3999268864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 3.1284127235412598, "created_at": "2025-01-16T03:06:06.332792+00:00"} {"global_step": 7629, "acc_step": 0, "speed/wps": 12890.619170284537, "speed/FLOPS": 202464887636307.78, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05487207695841789, "optim/lr": 0.0029831317411730222, "optim/total_tokens": 3999793152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.0424866676330566, "created_at": "2025-01-16T03:06:16.505914+00:00"} {"global_step": 7630, "acc_step": 0, "speed/wps": 12891.294098022725, "speed/FLOPS": 202475488303883.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059768542647361755, "optim/lr": 0.0029831189255327567, "optim/total_tokens": 4000317440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.216125249862671, "created_at": "2025-01-16T03:06:26.682292+00:00"} {"global_step": 7631, "acc_step": 0, "speed/wps": 12889.529116059146, "speed/FLOPS": 202447766836806.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04372581094503403, "optim/lr": 0.0029831061050535554, "optim/total_tokens": 4000841728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 2.9100723266601562, "created_at": "2025-01-16T03:06:36.854996+00:00"} {"global_step": 7632, "acc_step": 0, "speed/wps": 12887.84315472695, "speed/FLOPS": 202421286497330.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060110222548246384, "optim/lr": 0.002983093279735459, "optim/total_tokens": 4001366016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.99031400680542, "created_at": "2025-01-16T03:06:47.029149+00:00"} {"global_step": 7633, "acc_step": 0, "speed/wps": 12891.258060716373, "speed/FLOPS": 202474922288467.22, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.058464594185352325, "optim/lr": 0.0029830804495785106, "optim/total_tokens": 4001890304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9910216331481934, "created_at": "2025-01-16T03:06:57.208637+00:00"} {"global_step": 7634, "acc_step": 0, "speed/wps": 12891.651682290985, "speed/FLOPS": 202481104656189.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05263066291809082, "optim/lr": 0.0029830676145827513, "optim/total_tokens": 4002414592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9962656497955322, "created_at": "2025-01-16T03:07:07.379403+00:00"} {"global_step": 7635, "acc_step": 0, "speed/wps": 12890.556805090497, "speed/FLOPS": 202463908105236.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0664905458688736, "optim/lr": 0.0029830547747482228, "optim/total_tokens": 4002938880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.908846139907837, "created_at": "2025-01-16T03:07:17.556556+00:00"} {"global_step": 7636, "acc_step": 0, "speed/wps": 12892.190421836016, "speed/FLOPS": 202489566301049.66, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04999923333525658, "optim/lr": 0.0029830419300749676, "optim/total_tokens": 4003463168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 2.870450735092163, "created_at": "2025-01-16T03:07:27.729321+00:00"} {"global_step": 7637, "acc_step": 0, "speed/wps": 12888.626535870406, "speed/FLOPS": 202433590574667.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08164579421281815, "optim/lr": 0.002983029080563027, "optim/total_tokens": 4003987456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.899160146713257, "created_at": "2025-01-16T03:07:37.906127+00:00"} {"global_step": 7638, "acc_step": 0, "speed/wps": 12891.539775585763, "speed/FLOPS": 202479347007607.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058444246649742126, "optim/lr": 0.002983016226212444, "optim/total_tokens": 4004511744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354275, "loss/out": 3.015397787094116, "created_at": "2025-01-16T03:07:48.077509+00:00"} {"global_step": 7639, "acc_step": 0, "speed/wps": 12894.470869795572, "speed/FLOPS": 202525383869918.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06046455726027489, "optim/lr": 0.0029830033670232596, "optim/total_tokens": 4005036032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.9349632263183594, "created_at": "2025-01-16T03:07:58.248616+00:00"} {"global_step": 7640, "acc_step": 0, "speed/wps": 12891.028526732069, "speed/FLOPS": 202471317141830.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06204897165298462, "optim/lr": 0.0029829905029955155, "optim/total_tokens": 4005560320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478894, "loss/out": 2.990673065185547, "created_at": "2025-01-16T03:08:08.422058+00:00"} {"global_step": 7641, "acc_step": 0, "speed/wps": 12891.302596380792, "speed/FLOPS": 202475621782275.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06411148607730865, "optim/lr": 0.002982977634129254, "optim/total_tokens": 4006084608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.053271770477295, "created_at": "2025-01-16T03:08:18.596580+00:00"} {"global_step": 7642, "acc_step": 0, "speed/wps": 12887.784424187457, "speed/FLOPS": 202420364053503.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06412612646818161, "optim/lr": 0.0029829647604245175, "optim/total_tokens": 4006608896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9712936878204346, "created_at": "2025-01-16T03:08:28.771570+00:00"} {"global_step": 7643, "acc_step": 0, "speed/wps": 12889.441300783414, "speed/FLOPS": 202446387577230.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06807971000671387, "optim/lr": 0.0029829518818813476, "optim/total_tokens": 4007133184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9208903312683105, "created_at": "2025-01-16T03:08:38.944330+00:00"} {"global_step": 7644, "acc_step": 0, "speed/wps": 12890.1271945951, "speed/FLOPS": 202457160482059.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0579475499689579, "optim/lr": 0.002982938998499787, "optim/total_tokens": 4007657472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.948298931121826, "created_at": "2025-01-16T03:08:49.117430+00:00"} {"global_step": 7645, "acc_step": 0, "speed/wps": 12886.58136962796, "speed/FLOPS": 202401468428476.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056785698980093, "optim/lr": 0.0029829261102798763, "optim/total_tokens": 4008181760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399303, "loss/out": 2.911773204803467, "created_at": "2025-01-16T03:08:59.293563+00:00"} {"global_step": 7646, "acc_step": 0, "speed/wps": 12890.799140247418, "speed/FLOPS": 202467714311879.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05363914370536804, "optim/lr": 0.002982913217221659, "optim/total_tokens": 4008706048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9908199310302734, "created_at": "2025-01-16T03:09:09.467770+00:00"} {"global_step": 7647, "acc_step": 0, "speed/wps": 12887.491459269311, "speed/FLOPS": 202415762636888.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07826483994722366, "optim/lr": 0.002982900319325176, "optim/total_tokens": 4009230336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.9533615112304688, "created_at": "2025-01-16T03:09:19.643503+00:00"} {"global_step": 7648, "acc_step": 0, "speed/wps": 12892.341811455846, "speed/FLOPS": 202491944083060.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08765961229801178, "optim/lr": 0.00298288741659047, "optim/total_tokens": 4009754624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.0101869106292725, "created_at": "2025-01-16T03:09:29.813740+00:00"} {"global_step": 7649, "acc_step": 0, "speed/wps": 12887.104973754784, "speed/FLOPS": 202409692350795.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04998955503106117, "optim/lr": 0.0029828745090175832, "optim/total_tokens": 4010278912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.979027271270752, "created_at": "2025-01-16T03:09:39.988139+00:00"} {"global_step": 7650, "acc_step": 0, "speed/wps": 12891.47881260227, "speed/FLOPS": 202478389500180.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06720254570245743, "optim/lr": 0.0029828615966065583, "optim/total_tokens": 4010803200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.0410265922546387, "created_at": "2025-01-16T03:09:50.159784+00:00"} {"global_step": 7651, "acc_step": 0, "speed/wps": 12889.444019427901, "speed/FLOPS": 202446430277277.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054435499012470245, "optim/lr": 0.0029828486793574355, "optim/total_tokens": 4011327488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.957426071166992, "created_at": "2025-01-16T03:10:00.332212+00:00"} {"global_step": 7652, "acc_step": 0, "speed/wps": 12885.771318533569, "speed/FLOPS": 202388745462912.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05331781506538391, "optim/lr": 0.002982835757270259, "optim/total_tokens": 4011851776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411794, "loss/out": 3.047779083251953, "created_at": "2025-01-16T03:10:10.508610+00:00"} {"global_step": 7653, "acc_step": 0, "speed/wps": 12889.07965219796, "speed/FLOPS": 202440707389238.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08043640851974487, "optim/lr": 0.00298282283034507, "optim/total_tokens": 4012376064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 2.807018280029297, "created_at": "2025-01-16T03:10:20.682363+00:00"} {"global_step": 7654, "acc_step": 0, "speed/wps": 12892.374858581514, "speed/FLOPS": 202492463133580.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05185965821146965, "optim/lr": 0.0029828098985819106, "optim/total_tokens": 4012900352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.0839595794677734, "created_at": "2025-01-16T03:10:30.856080+00:00"} {"global_step": 7655, "acc_step": 0, "speed/wps": 12892.044828359483, "speed/FLOPS": 202487279555434.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09965039789676666, "optim/lr": 0.0029827969619808234, "optim/total_tokens": 4013424640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.0516443252563477, "created_at": "2025-01-16T03:10:41.028263+00:00"} {"global_step": 7656, "acc_step": 0, "speed/wps": 12889.966510357875, "speed/FLOPS": 202454636715310.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10019272565841675, "optim/lr": 0.00298278402054185, "optim/total_tokens": 4013948928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.9715373516082764, "created_at": "2025-01-16T03:10:51.200635+00:00"} {"global_step": 7657, "acc_step": 0, "speed/wps": 12890.863778348788, "speed/FLOPS": 202468729542083.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050913747400045395, "optim/lr": 0.0029827710742650333, "optim/total_tokens": 4014473216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.016408920288086, "created_at": "2025-01-16T03:11:01.378181+00:00"} {"global_step": 7658, "acc_step": 0, "speed/wps": 12890.113173745653, "speed/FLOPS": 202456940265352.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.11497876048088074, "optim/lr": 0.0029827581231504155, "optim/total_tokens": 4014997504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.01991605758667, "created_at": "2025-01-16T03:11:11.550487+00:00"} {"global_step": 7659, "acc_step": 0, "speed/wps": 12889.772594297756, "speed/FLOPS": 202451590997118.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08626881241798401, "optim/lr": 0.002982745167198038, "optim/total_tokens": 4015521792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 2.918527603149414, "created_at": "2025-01-16T03:11:21.724875+00:00"} {"global_step": 7660, "acc_step": 0, "speed/wps": 12893.250937275705, "speed/FLOPS": 202506223153330.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07685112953186035, "optim/lr": 0.0029827322064079433, "optim/total_tokens": 4016046080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 2.9625003337860107, "created_at": "2025-01-16T03:11:31.894328+00:00"} {"global_step": 7661, "acc_step": 0, "speed/wps": 12892.04402767118, "speed/FLOPS": 202487266979524.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06344462931156158, "optim/lr": 0.0029827192407801753, "optim/total_tokens": 4016570368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.0057737827301025, "created_at": "2025-01-16T03:11:42.064711+00:00"} {"global_step": 7662, "acc_step": 0, "speed/wps": 12889.96275022587, "speed/FLOPS": 202454577657270.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08120189607143402, "optim/lr": 0.002982706270314774, "optim/total_tokens": 4017094656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332595, "loss/out": 2.981113910675049, "created_at": "2025-01-16T03:11:52.238896+00:00"} {"global_step": 7663, "acc_step": 0, "speed/wps": 12890.23376632619, "speed/FLOPS": 202458834337541.4, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07180768996477127, "optim/lr": 0.0029826932950117835, "optim/total_tokens": 4017618944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 2.9407033920288086, "created_at": "2025-01-16T03:12:02.411894+00:00"} {"global_step": 7664, "acc_step": 0, "speed/wps": 12889.413528886353, "speed/FLOPS": 202445951381424.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06016680598258972, "optim/lr": 0.002982680314871245, "optim/total_tokens": 4018143232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.0212466716766357, "created_at": "2025-01-16T03:12:12.586914+00:00"} {"global_step": 7665, "acc_step": 0, "speed/wps": 12892.37085611703, "speed/FLOPS": 202492400269375.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06721975654363632, "optim/lr": 0.002982667329893201, "optim/total_tokens": 4018667520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.9127068519592285, "created_at": "2025-01-16T03:12:22.765688+00:00"} {"global_step": 7666, "acc_step": 0, "speed/wps": 12895.057114605057, "speed/FLOPS": 202534591650236.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07296784967184067, "optim/lr": 0.0029826543400776947, "optim/total_tokens": 4019191808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344789, "loss/out": 3.0069143772125244, "created_at": "2025-01-16T03:12:32.935515+00:00"} {"global_step": 7667, "acc_step": 0, "speed/wps": 12886.813487583431, "speed/FLOPS": 202405114159930.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05840218439698219, "optim/lr": 0.0029826413454247676, "optim/total_tokens": 4019716096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.971041202545166, "created_at": "2025-01-16T03:12:43.111784+00:00"} {"global_step": 7668, "acc_step": 0, "speed/wps": 12889.728411558357, "speed/FLOPS": 202450897046482.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05963897705078125, "optim/lr": 0.0029826283459344623, "optim/total_tokens": 4020240384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 3.037957191467285, "created_at": "2025-01-16T03:12:53.287355+00:00"} {"global_step": 7669, "acc_step": 0, "speed/wps": 12888.60061705382, "speed/FLOPS": 202433183484036.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061631981283426285, "optim/lr": 0.0029826153416068215, "optim/total_tokens": 4020764672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405621, "loss/out": 2.999807596206665, "created_at": "2025-01-16T03:13:03.463252+00:00"} {"global_step": 7670, "acc_step": 0, "speed/wps": 12890.557447382445, "speed/FLOPS": 202463918193314.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053577955812215805, "optim/lr": 0.002982602332441888, "optim/total_tokens": 4021288960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 3.122603416442871, "created_at": "2025-01-16T03:13:13.634882+00:00"} {"global_step": 7671, "acc_step": 0, "speed/wps": 12871.326695585103, "speed/FLOPS": 202161872810514.8, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06510830670595169, "optim/lr": 0.002982589318439703, "optim/total_tokens": 4021813248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.016406297683716, "created_at": "2025-01-16T03:13:23.847941+00:00"} {"global_step": 7672, "acc_step": 0, "speed/wps": 12887.553978907046, "speed/FLOPS": 202416744593709.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058868326246738434, "optim/lr": 0.0029825762996003096, "optim/total_tokens": 4022337536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 3.0156702995300293, "created_at": "2025-01-16T03:13:34.022080+00:00"} {"global_step": 7673, "acc_step": 0, "speed/wps": 12885.765930356327, "speed/FLOPS": 202388660834184.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06244992837309837, "optim/lr": 0.00298256327592375, "optim/total_tokens": 4022861824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.8887252807617188, "created_at": "2025-01-16T03:13:44.197634+00:00"} {"global_step": 7674, "acc_step": 0, "speed/wps": 12884.870463040354, "speed/FLOPS": 202374596289486.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06778406351804733, "optim/lr": 0.0029825502474100677, "optim/total_tokens": 4023386112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0331778526306152, "created_at": "2025-01-16T03:13:54.374331+00:00"} {"global_step": 7675, "acc_step": 0, "speed/wps": 12888.498602340875, "speed/FLOPS": 202431581202786.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0520746149122715, "optim/lr": 0.002982537214059305, "optim/total_tokens": 4023910400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 3.031759738922119, "created_at": "2025-01-16T03:14:04.547594+00:00"} {"global_step": 7676, "acc_step": 0, "speed/wps": 12882.057322721312, "speed/FLOPS": 202330412055115.88, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06797781586647034, "optim/lr": 0.0029825241758715025, "optim/total_tokens": 4024434688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 3.075018882751465, "created_at": "2025-01-16T03:14:14.726344+00:00"} {"global_step": 7677, "acc_step": 0, "speed/wps": 12887.257838250403, "speed/FLOPS": 202412093297757.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06421345472335815, "optim/lr": 0.0029825111328467056, "optim/total_tokens": 4024958976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.062953233718872, "created_at": "2025-01-16T03:14:24.901780+00:00"} {"global_step": 7678, "acc_step": 0, "speed/wps": 12885.279378947982, "speed/FLOPS": 202381018875724.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06424146145582199, "optim/lr": 0.002982498084984955, "optim/total_tokens": 4025483264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 2.9184155464172363, "created_at": "2025-01-16T03:14:35.094252+00:00"} {"global_step": 7679, "acc_step": 0, "speed/wps": 12889.599298074663, "speed/FLOPS": 202448869141800.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052573010325431824, "optim/lr": 0.002982485032286293, "optim/total_tokens": 4026007552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 3.053387403488159, "created_at": "2025-01-16T03:14:45.267110+00:00"} {"global_step": 7680, "acc_step": 0, "speed/wps": 12888.021227285673, "speed/FLOPS": 202424083371562.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0752195492386818, "optim/lr": 0.0029824719747507636, "optim/total_tokens": 4026531840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9456241130828857, "created_at": "2025-01-16T03:14:55.443847+00:00"} {"global_step": 7681, "acc_step": 0, "speed/wps": 12893.423846226538, "speed/FLOPS": 202508938926005.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06942300498485565, "optim/lr": 0.002982458912378409, "optim/total_tokens": 4027056128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9806418418884277, "created_at": "2025-01-16T03:15:05.613945+00:00"} {"global_step": 7682, "acc_step": 0, "speed/wps": 12886.695936069958, "speed/FLOPS": 202403267851879.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06324737519025803, "optim/lr": 0.0029824458451692707, "optim/total_tokens": 4027580416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367469, "loss/out": 3.0369949340820312, "created_at": "2025-01-16T03:15:15.788762+00:00"} {"global_step": 7683, "acc_step": 0, "speed/wps": 12891.344663241103, "speed/FLOPS": 202476282500122.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05948261544108391, "optim/lr": 0.0029824327731233923, "optim/total_tokens": 4028104704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9874777793884277, "created_at": "2025-01-16T03:15:25.963128+00:00"} {"global_step": 7684, "acc_step": 0, "speed/wps": 12884.299875300849, "speed/FLOPS": 202365634424967.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05726577341556549, "optim/lr": 0.0029824196962408168, "optim/total_tokens": 4028628992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.9000678062438965, "created_at": "2025-01-16T03:15:36.171209+00:00"} {"global_step": 7685, "acc_step": 0, "speed/wps": 12888.800972135006, "speed/FLOPS": 202436330335903.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05692075937986374, "optim/lr": 0.0029824066145215863, "optim/total_tokens": 4029153280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.9792232513427734, "created_at": "2025-01-16T03:15:46.346426+00:00"} {"global_step": 7686, "acc_step": 0, "speed/wps": 12890.336331534298, "speed/FLOPS": 202460445265074.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055061254650354385, "optim/lr": 0.0029823935279657435, "optim/total_tokens": 4029677568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0994675159454346, "created_at": "2025-01-16T03:15:56.520441+00:00"} {"global_step": 7687, "acc_step": 0, "speed/wps": 12893.193013784674, "speed/FLOPS": 202505313385309.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061437588185071945, "optim/lr": 0.002982380436573331, "optim/total_tokens": 4030201856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418214, "loss/out": 3.046416759490967, "created_at": "2025-01-16T03:16:06.692578+00:00"} {"global_step": 7688, "acc_step": 0, "speed/wps": 12888.452688574976, "speed/FLOPS": 202430860064001.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05280206725001335, "optim/lr": 0.0029823673403443916, "optim/total_tokens": 4030726144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.994645357131958, "created_at": "2025-01-16T03:16:16.866998+00:00"} {"global_step": 7689, "acc_step": 0, "speed/wps": 12891.750179417773, "speed/FLOPS": 202482651688914.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05378930643200874, "optim/lr": 0.0029823542392789686, "optim/total_tokens": 4031250432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9402337074279785, "created_at": "2025-01-16T03:16:27.037773+00:00"} {"global_step": 7690, "acc_step": 0, "speed/wps": 12895.769252474034, "speed/FLOPS": 202545776754048.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06875678151845932, "optim/lr": 0.002982341133377104, "optim/total_tokens": 4031774720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 3.0760412216186523, "created_at": "2025-01-16T03:16:37.209373+00:00"} {"global_step": 7691, "acc_step": 0, "speed/wps": 12884.673693461225, "speed/FLOPS": 202371505752856.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06980421394109726, "optim/lr": 0.0029823280226388403, "optim/total_tokens": 4032299008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.981337308883667, "created_at": "2025-01-16T03:16:47.385597+00:00"} {"global_step": 7692, "acc_step": 0, "speed/wps": 12888.353196101138, "speed/FLOPS": 202429297397982.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054611627012491226, "optim/lr": 0.002982314907064221, "optim/total_tokens": 4032823296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9596798419952393, "created_at": "2025-01-16T03:16:57.561441+00:00"} {"global_step": 7693, "acc_step": 0, "speed/wps": 12887.424347248709, "speed/FLOPS": 202414708550383.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06739195436239243, "optim/lr": 0.0029823017866532887, "optim/total_tokens": 4033347584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9156508445739746, "created_at": "2025-01-16T03:17:07.739670+00:00"} {"global_step": 7694, "acc_step": 0, "speed/wps": 12893.073102862712, "speed/FLOPS": 202503430019504.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05114321410655975, "optim/lr": 0.002982288661406086, "optim/total_tokens": 4033871872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306979, "loss/out": 2.9817214012145996, "created_at": "2025-01-16T03:17:17.910862+00:00"} {"global_step": 7695, "acc_step": 0, "speed/wps": 12885.7578451494, "speed/FLOPS": 202388533844899.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07545565813779831, "optim/lr": 0.002982275531322657, "optim/total_tokens": 4034396160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.045480251312256, "created_at": "2025-01-16T03:17:28.087477+00:00"} {"global_step": 7696, "acc_step": 0, "speed/wps": 12892.782655301593, "speed/FLOPS": 202498868141440.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07303910702466965, "optim/lr": 0.0029822623964030426, "optim/total_tokens": 4034920448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.85394287109375, "created_at": "2025-01-16T03:17:38.259006+00:00"} {"global_step": 7697, "acc_step": 0, "speed/wps": 12885.934671990459, "speed/FLOPS": 202391311153424.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0521969236433506, "optim/lr": 0.002982249256647286, "optim/total_tokens": 4035444736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 2.8849010467529297, "created_at": "2025-01-16T03:17:48.438619+00:00"} {"global_step": 7698, "acc_step": 0, "speed/wps": 12891.738364501161, "speed/FLOPS": 202482466119413.88, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05604808032512665, "optim/lr": 0.0029822361120554314, "optim/total_tokens": 4035969024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.965141534805298, "created_at": "2025-01-16T03:17:58.609328+00:00"} {"global_step": 7699, "acc_step": 0, "speed/wps": 12889.670203333426, "speed/FLOPS": 202449982806323.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0816529169678688, "optim/lr": 0.0029822229626275203, "optim/total_tokens": 4036493312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.014756679534912, "created_at": "2025-01-16T03:18:08.783412+00:00"} {"global_step": 7700, "acc_step": 0, "speed/wps": 12890.234941047673, "speed/FLOPS": 202458852788156.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06344754993915558, "optim/lr": 0.002982209808363596, "optim/total_tokens": 4037017600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9276986122131348, "created_at": "2025-01-16T03:18:18.955797+00:00"} {"global_step": 7701, "acc_step": 0, "speed/wps": 12891.188661218253, "speed/FLOPS": 202473832273984.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07744111120700836, "optim/lr": 0.0029821966492637017, "optim/total_tokens": 4037541888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 3.0572874546051025, "created_at": "2025-01-16T03:18:29.129539+00:00"} {"global_step": 7702, "acc_step": 0, "speed/wps": 12886.836243886015, "speed/FLOPS": 202405471578933.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08919364213943481, "optim/lr": 0.0029821834853278814, "optim/total_tokens": 4038066176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.1634795665740967, "created_at": "2025-01-16T03:18:39.305578+00:00"} {"global_step": 7703, "acc_step": 0, "speed/wps": 12886.595822919608, "speed/FLOPS": 202401695437282.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049938932061195374, "optim/lr": 0.0029821703165561755, "optim/total_tokens": 4038590464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499005, "loss/out": 2.7377724647521973, "created_at": "2025-01-16T03:18:49.482645+00:00"} {"global_step": 7704, "acc_step": 0, "speed/wps": 12887.254053736793, "speed/FLOPS": 202412033856770.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07261299341917038, "optim/lr": 0.0029821571429486287, "optim/total_tokens": 4039114752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 2.845738410949707, "created_at": "2025-01-16T03:18:59.656809+00:00"} {"global_step": 7705, "acc_step": 0, "speed/wps": 12884.65144612639, "speed/FLOPS": 202371156327891.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06350391358137131, "optim/lr": 0.002982143964505284, "optim/total_tokens": 4039639040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.912050247192383, "created_at": "2025-01-16T03:19:09.833838+00:00"} {"global_step": 7706, "acc_step": 0, "speed/wps": 12888.367479865989, "speed/FLOPS": 202429521744137.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055707428604364395, "optim/lr": 0.0029821307812261835, "optim/total_tokens": 4040163328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0526485443115234, "created_at": "2025-01-16T03:19:20.008849+00:00"} {"global_step": 7707, "acc_step": 0, "speed/wps": 12884.352623266368, "speed/FLOPS": 202366462904249.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06112530827522278, "optim/lr": 0.0029821175931113706, "optim/total_tokens": 4040687616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.057640314102173, "created_at": "2025-01-16T03:19:30.186022+00:00"} {"global_step": 7708, "acc_step": 0, "speed/wps": 12886.415042881606, "speed/FLOPS": 202398856038368.3, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060973841696977615, "optim/lr": 0.0029821044001608885, "optim/total_tokens": 4041211904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 3.0282766819000244, "created_at": "2025-01-16T03:19:40.360800+00:00"} {"global_step": 7709, "acc_step": 0, "speed/wps": 12891.778584904114, "speed/FLOPS": 202483097836110.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05846940353512764, "optim/lr": 0.00298209120237478, "optim/total_tokens": 4041736192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507696, "loss/out": 2.895745277404785, "created_at": "2025-01-16T03:19:50.534651+00:00"} {"global_step": 7710, "acc_step": 0, "speed/wps": 12889.20954026184, "speed/FLOPS": 202442747459766.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062387581914663315, "optim/lr": 0.0029820779997530894, "optim/total_tokens": 4042260480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.0215392112731934, "created_at": "2025-01-16T03:20:00.712709+00:00"} {"global_step": 7711, "acc_step": 0, "speed/wps": 12888.476368942487, "speed/FLOPS": 202431231996713.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05084792524576187, "optim/lr": 0.0029820647922958574, "optim/total_tokens": 4042784768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.945560932159424, "created_at": "2025-01-16T03:20:10.887338+00:00"} {"global_step": 7712, "acc_step": 0, "speed/wps": 12887.156964670503, "speed/FLOPS": 202410508939570.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052753057330846786, "optim/lr": 0.002982051580003129, "optim/total_tokens": 4043309056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 2.979431390762329, "created_at": "2025-01-16T03:20:21.064577+00:00"} {"global_step": 7713, "acc_step": 0, "speed/wps": 12888.021760744792, "speed/FLOPS": 202424091750271.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061816468834877014, "optim/lr": 0.0029820383628749473, "optim/total_tokens": 4043833344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.93009877204895, "created_at": "2025-01-16T03:20:31.240191+00:00"} {"global_step": 7714, "acc_step": 0, "speed/wps": 12889.28560570283, "speed/FLOPS": 202443942172042.3, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06377674639225006, "optim/lr": 0.0029820251409113543, "optim/total_tokens": 4044357632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.046192169189453, "created_at": "2025-01-16T03:20:41.417013+00:00"} {"global_step": 7715, "acc_step": 0, "speed/wps": 12883.091343445993, "speed/FLOPS": 202346652771490.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09221921116113663, "optim/lr": 0.0029820119141123935, "optim/total_tokens": 4044881920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 3.025175094604492, "created_at": "2025-01-16T03:20:51.597643+00:00"} {"global_step": 7716, "acc_step": 0, "speed/wps": 12888.581668138433, "speed/FLOPS": 202432885865281.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07449129223823547, "optim/lr": 0.0029819986824781083, "optim/total_tokens": 4045406208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457361, "loss/out": 2.9595680236816406, "created_at": "2025-01-16T03:21:01.774070+00:00"} {"global_step": 7717, "acc_step": 0, "speed/wps": 12888.780857616675, "speed/FLOPS": 202436014409752.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05525406077504158, "optim/lr": 0.0029819854460085425, "optim/total_tokens": 4045930496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.963418960571289, "created_at": "2025-01-16T03:21:11.957461+00:00"} {"global_step": 7718, "acc_step": 0, "speed/wps": 12885.352301215114, "speed/FLOPS": 202382164220135.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07947022467851639, "optim/lr": 0.0029819722047037373, "optim/total_tokens": 4046454784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.016799211502075, "created_at": "2025-01-16T03:21:22.133689+00:00"} {"global_step": 7719, "acc_step": 0, "speed/wps": 12886.261976558499, "speed/FLOPS": 202396451921428.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05677342042326927, "optim/lr": 0.0029819589585637378, "optim/total_tokens": 4046979072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9547929763793945, "created_at": "2025-01-16T03:21:32.308725+00:00"} {"global_step": 7720, "acc_step": 0, "speed/wps": 12884.86221380741, "speed/FLOPS": 202374466723947.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06296167522668839, "optim/lr": 0.0029819457075885873, "optim/total_tokens": 4047503360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 2.945687770843506, "created_at": "2025-01-16T03:21:42.485950+00:00"} {"global_step": 7721, "acc_step": 0, "speed/wps": 12887.62620349194, "speed/FLOPS": 202417878980062.7, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07070380449295044, "optim/lr": 0.0029819324517783275, "optim/total_tokens": 4048027648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9318790435791016, "created_at": "2025-01-16T03:21:52.659786+00:00"} {"global_step": 7722, "acc_step": 0, "speed/wps": 12890.415980583815, "speed/FLOPS": 202461696262846.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055076513439416885, "optim/lr": 0.0029819191911330026, "optim/total_tokens": 4048551936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.0410518646240234, "created_at": "2025-01-16T03:22:02.834574+00:00"} {"global_step": 7723, "acc_step": 0, "speed/wps": 12882.036406217836, "speed/FLOPS": 202330083532686.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050926730036735535, "optim/lr": 0.0029819059256526565, "optim/total_tokens": 4049076224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 2.905470371246338, "created_at": "2025-01-16T03:22:13.015575+00:00"} {"global_step": 7724, "acc_step": 0, "speed/wps": 12883.560552784144, "speed/FLOPS": 202354022348907.22, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051796652376651764, "optim/lr": 0.0029818926553373307, "optim/total_tokens": 4049600512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 2.9616587162017822, "created_at": "2025-01-16T03:22:23.195383+00:00"} {"global_step": 7725, "acc_step": 0, "speed/wps": 12885.291689307907, "speed/FLOPS": 202381212226843.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04861147329211235, "optim/lr": 0.0029818793801870704, "optim/total_tokens": 4050124800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.9132776260375977, "created_at": "2025-01-16T03:22:33.375769+00:00"} {"global_step": 7726, "acc_step": 0, "speed/wps": 12888.744021098022, "speed/FLOPS": 202435435841608.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056837376207113266, "optim/lr": 0.0029818661002019176, "optim/total_tokens": 4050649088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9641916751861572, "created_at": "2025-01-16T03:22:43.548901+00:00"} {"global_step": 7727, "acc_step": 0, "speed/wps": 12884.982377851924, "speed/FLOPS": 202376354065389.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07228463143110275, "optim/lr": 0.002981852815381916, "optim/total_tokens": 4051173376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9142160415649414, "created_at": "2025-01-16T03:22:53.724898+00:00"} {"global_step": 7728, "acc_step": 0, "speed/wps": 12885.433817567407, "speed/FLOPS": 202383444546463.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05328430235385895, "optim/lr": 0.0029818395257271094, "optim/total_tokens": 4051697664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.9589319229125977, "created_at": "2025-01-16T03:23:03.901872+00:00"} {"global_step": 7729, "acc_step": 0, "speed/wps": 12880.606693879063, "speed/FLOPS": 202307627935775.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06045663729310036, "optim/lr": 0.0029818262312375403, "optim/total_tokens": 4052221952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9498753547668457, "created_at": "2025-01-16T03:23:14.084381+00:00"} {"global_step": 7730, "acc_step": 0, "speed/wps": 12880.422949691405, "speed/FLOPS": 202304741980817.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.05541235953569412, "optim/lr": 0.0029818129319132526, "optim/total_tokens": 4052746240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9163198471069336, "created_at": "2025-01-16T03:23:24.266563+00:00"} {"global_step": 7731, "acc_step": 0, "speed/wps": 12889.081003771096, "speed/FLOPS": 202440728617551.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06288089603185654, "optim/lr": 0.00298179962775429, "optim/total_tokens": 4053270528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.0345938205718994, "created_at": "2025-01-16T03:23:34.439580+00:00"} {"global_step": 7732, "acc_step": 0, "speed/wps": 12887.153563735734, "speed/FLOPS": 202410455523216.38, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053735293447971344, "optim/lr": 0.0029817863187606955, "optim/total_tokens": 4053794816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 3.0202770233154297, "created_at": "2025-01-16T03:23:44.615928+00:00"} {"global_step": 7733, "acc_step": 0, "speed/wps": 12887.910158990668, "speed/FLOPS": 202422338891365.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052989594638347626, "optim/lr": 0.002981773004932513, "optim/total_tokens": 4054319104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.07724666595459, "created_at": "2025-01-16T03:23:54.810878+00:00"} {"global_step": 7734, "acc_step": 0, "speed/wps": 12886.541836518749, "speed/FLOPS": 202400847506672.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06047045812010765, "optim/lr": 0.0029817596862697845, "optim/total_tokens": 4054843392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.862367868423462, "created_at": "2025-01-16T03:24:04.985736+00:00"} {"global_step": 7735, "acc_step": 0, "speed/wps": 12883.499054318594, "speed/FLOPS": 202353056430998.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.063571996986866, "optim/lr": 0.002981746362772555, "optim/total_tokens": 4055367680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0223684310913086, "created_at": "2025-01-16T03:24:15.169782+00:00"} {"global_step": 7736, "acc_step": 0, "speed/wps": 12891.548065254594, "speed/FLOPS": 202479477208247.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053718727082014084, "optim/lr": 0.0029817330344408676, "optim/total_tokens": 4055891968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323796, "loss/out": 2.955742359161377, "created_at": "2025-01-16T03:24:25.343288+00:00"} {"global_step": 7737, "acc_step": 0, "speed/wps": 12888.163454715064, "speed/FLOPS": 202426317248782.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07909655570983887, "optim/lr": 0.002981719701274765, "optim/total_tokens": 4056416256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 3.0686917304992676, "created_at": "2025-01-16T03:24:35.518093+00:00"} {"global_step": 7738, "acc_step": 0, "speed/wps": 12881.33171338632, "speed/FLOPS": 202319015363418.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07827359437942505, "optim/lr": 0.002981706363274292, "optim/total_tokens": 4056940544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.0546116828918457, "created_at": "2025-01-16T03:24:45.697011+00:00"} {"global_step": 7739, "acc_step": 0, "speed/wps": 12886.141204607975, "speed/FLOPS": 202394555031987.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07178007811307907, "optim/lr": 0.0029816930204394913, "optim/total_tokens": 4057464832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 2.997805595397949, "created_at": "2025-01-16T03:24:55.874313+00:00"} {"global_step": 7740, "acc_step": 0, "speed/wps": 12884.684922142587, "speed/FLOPS": 202371682114727.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05400928482413292, "optim/lr": 0.0029816796727704068, "optim/total_tokens": 4057989120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487913, "loss/out": 3.0043959617614746, "created_at": "2025-01-16T03:25:06.050651+00:00"} {"global_step": 7741, "acc_step": 0, "speed/wps": 12889.13386879949, "speed/FLOPS": 202441558935467.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052915554493665695, "optim/lr": 0.002981666320267082, "optim/total_tokens": 4058513408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.0615234375, "created_at": "2025-01-16T03:25:16.225388+00:00"} {"global_step": 7742, "acc_step": 0, "speed/wps": 12890.099692460857, "speed/FLOPS": 202456728523249.56, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05546167865395546, "optim/lr": 0.0029816529629295594, "optim/total_tokens": 4059037696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.023503065109253, "created_at": "2025-01-16T03:25:26.399546+00:00"} {"global_step": 7743, "acc_step": 0, "speed/wps": 12886.182822471024, "speed/FLOPS": 202395208697715.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06385020911693573, "optim/lr": 0.002981639600757884, "optim/total_tokens": 4059561984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.068087100982666, "created_at": "2025-01-16T03:25:36.575445+00:00"} {"global_step": 7744, "acc_step": 0, "speed/wps": 12890.899924440922, "speed/FLOPS": 202469297266130.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06390444189310074, "optim/lr": 0.0029816262337520986, "optim/total_tokens": 4060086272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 2.9367523193359375, "created_at": "2025-01-16T03:25:46.746646+00:00"} {"global_step": 7745, "acc_step": 0, "speed/wps": 12887.993548843566, "speed/FLOPS": 202423648643595.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04818979650735855, "optim/lr": 0.0029816128619122475, "optim/total_tokens": 4060610560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 3.014237880706787, "created_at": "2025-01-16T03:25:56.926270+00:00"} {"global_step": 7746, "acc_step": 0, "speed/wps": 12883.10027441606, "speed/FLOPS": 202346793044647.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05213760584592819, "optim/lr": 0.0029815994852383735, "optim/total_tokens": 4061134848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 3.066295623779297, "created_at": "2025-01-16T03:26:07.105672+00:00"} {"global_step": 7747, "acc_step": 0, "speed/wps": 12887.435138182866, "speed/FLOPS": 202414878036832.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04917052388191223, "optim/lr": 0.002981586103730521, "optim/total_tokens": 4061659136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.912135362625122, "created_at": "2025-01-16T03:26:17.279974+00:00"} {"global_step": 7748, "acc_step": 0, "speed/wps": 12886.385734789123, "speed/FLOPS": 202398395714502.3, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04752485081553459, "optim/lr": 0.002981572717388733, "optim/total_tokens": 4062183424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 3.0629639625549316, "created_at": "2025-01-16T03:26:27.454849+00:00"} {"global_step": 7749, "acc_step": 0, "speed/wps": 12886.220814801396, "speed/FLOPS": 202395805419471.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055606480687856674, "optim/lr": 0.0029815593262130543, "optim/total_tokens": 4062707712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.9234108924865723, "created_at": "2025-01-16T03:26:37.631163+00:00"} {"global_step": 7750, "acc_step": 0, "speed/wps": 12885.048800284467, "speed/FLOPS": 202377397320966.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05467645451426506, "optim/lr": 0.002981545930203527, "optim/total_tokens": 4063232000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.895118236541748, "created_at": "2025-01-16T03:26:47.808889+00:00"} {"global_step": 7751, "acc_step": 0, "speed/wps": 12887.947155971839, "speed/FLOPS": 202422919979794.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056350793689489365, "optim/lr": 0.0029815325293601957, "optim/total_tokens": 4063756288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471773, "loss/out": 3.0349278450012207, "created_at": "2025-01-16T03:26:57.982624+00:00"} {"global_step": 7752, "acc_step": 0, "speed/wps": 12886.600074310367, "speed/FLOPS": 202401762211216.8, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06159306317567825, "optim/lr": 0.0029815191236831043, "optim/total_tokens": 4064280576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.857616901397705, "created_at": "2025-01-16T03:27:08.158378+00:00"} {"global_step": 7753, "acc_step": 0, "speed/wps": 12882.013508861757, "speed/FLOPS": 202329723898245.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0570996031165123, "optim/lr": 0.002981505713172296, "optim/total_tokens": 4064804864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.996333360671997, "created_at": "2025-01-16T03:27:18.337964+00:00"} {"global_step": 7754, "acc_step": 0, "speed/wps": 12889.46541211899, "speed/FLOPS": 202446766278888.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06628066301345825, "optim/lr": 0.0029814922978278146, "optim/total_tokens": 4065329152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.044166326522827, "created_at": "2025-01-16T03:27:28.517221+00:00"} {"global_step": 7755, "acc_step": 0, "speed/wps": 12884.098668171953, "speed/FLOPS": 202362474190520.8, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055002789944410324, "optim/lr": 0.0029814788776497047, "optim/total_tokens": 4065853440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.972830057144165, "created_at": "2025-01-16T03:27:38.696737+00:00"} {"global_step": 7756, "acc_step": 0, "speed/wps": 12891.989294129922, "speed/FLOPS": 202486407314047.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052926406264305115, "optim/lr": 0.002981465452638009, "optim/total_tokens": 4066377728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.996548891067505, "created_at": "2025-01-16T03:27:48.872020+00:00"} {"global_step": 7757, "acc_step": 0, "speed/wps": 12887.620388910836, "speed/FLOPS": 202417787654076.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055310819298028946, "optim/lr": 0.002981452022792772, "optim/total_tokens": 4066902016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.9185471534729004, "created_at": "2025-01-16T03:27:59.045958+00:00"} {"global_step": 7758, "acc_step": 0, "speed/wps": 12885.725709708153, "speed/FLOPS": 202388029113636.2, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06460893899202347, "optim/lr": 0.002981438588114037, "optim/total_tokens": 4067426304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.9798367023468018, "created_at": "2025-01-16T03:28:09.227190+00:00"} {"global_step": 7759, "acc_step": 0, "speed/wps": 12884.067468088891, "speed/FLOPS": 202361984150342.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053249701857566833, "optim/lr": 0.002981425148601848, "optim/total_tokens": 4067950592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458205, "loss/out": 2.898146152496338, "created_at": "2025-01-16T03:28:19.403976+00:00"} {"global_step": 7760, "acc_step": 0, "speed/wps": 12886.204847376373, "speed/FLOPS": 202395554629119.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05851154401898384, "optim/lr": 0.0029814117042562495, "optim/total_tokens": 4068474880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 3.0067710876464844, "created_at": "2025-01-16T03:28:29.581838+00:00"} {"global_step": 7761, "acc_step": 0, "speed/wps": 12883.960630297968, "speed/FLOPS": 202360306116025.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06401588767766953, "optim/lr": 0.0029813982550772845, "optim/total_tokens": 4068999168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9505128860473633, "created_at": "2025-01-16T03:28:39.764570+00:00"} {"global_step": 7762, "acc_step": 0, "speed/wps": 12883.458151984065, "speed/FLOPS": 202352414003628.53, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05669508874416351, "optim/lr": 0.0029813848010649974, "optim/total_tokens": 4069523456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 2.9502768516540527, "created_at": "2025-01-16T03:28:49.948697+00:00"} {"global_step": 7763, "acc_step": 0, "speed/wps": 12886.788475213307, "speed/FLOPS": 202404721306287.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06439491361379623, "optim/lr": 0.0029813713422194316, "optim/total_tokens": 4070047744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 2.8240606784820557, "created_at": "2025-01-16T03:29:00.129310+00:00"} {"global_step": 7764, "acc_step": 0, "speed/wps": 12887.712178946562, "speed/FLOPS": 202419229342720.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05642290413379669, "optim/lr": 0.0029813578785406318, "optim/total_tokens": 4070572032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 3.0060625076293945, "created_at": "2025-01-16T03:29:10.303635+00:00"} {"global_step": 7765, "acc_step": 0, "speed/wps": 12884.219820569759, "speed/FLOPS": 202364377055409.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06264971941709518, "optim/lr": 0.002981344410028641, "optim/total_tokens": 4071096320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.9600260257720947, "created_at": "2025-01-16T03:29:20.480421+00:00"} {"global_step": 7766, "acc_step": 0, "speed/wps": 12883.978355482524, "speed/FLOPS": 202360584514406.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06393381953239441, "optim/lr": 0.0029813309366835032, "optim/total_tokens": 4071620608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.058502674102783, "created_at": "2025-01-16T03:29:30.658386+00:00"} {"global_step": 7767, "acc_step": 0, "speed/wps": 12890.501827970398, "speed/FLOPS": 202463044614019.0, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057489827275276184, "optim/lr": 0.0029813174585052635, "optim/total_tokens": 4072144896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435843, "loss/out": 2.9461073875427246, "created_at": "2025-01-16T03:29:40.831085+00:00"} {"global_step": 7768, "acc_step": 0, "speed/wps": 12884.096586353147, "speed/FLOPS": 202362441492695.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06452162563800812, "optim/lr": 0.002981303975493965, "optim/total_tokens": 4072669184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9993185997009277, "created_at": "2025-01-16T03:29:51.012560+00:00"} {"global_step": 7769, "acc_step": 0, "speed/wps": 12885.694590229012, "speed/FLOPS": 202387540339453.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06616901606321335, "optim/lr": 0.0029812904876496517, "optim/total_tokens": 4073193472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.003556251525879, "created_at": "2025-01-16T03:30:01.188186+00:00"} {"global_step": 7770, "acc_step": 0, "speed/wps": 12884.452702745324, "speed/FLOPS": 202368034789990.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05504949018359184, "optim/lr": 0.002981276994972368, "optim/total_tokens": 4073717760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.068969249725342, "created_at": "2025-01-16T03:30:11.364661+00:00"} {"global_step": 7771, "acc_step": 0, "speed/wps": 12889.837545575947, "speed/FLOPS": 202452611146194.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05863514170050621, "optim/lr": 0.0029812634974621575, "optim/total_tokens": 4074242048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.9236927032470703, "created_at": "2025-01-16T03:30:21.541876+00:00"} {"global_step": 7772, "acc_step": 0, "speed/wps": 12890.262247774512, "speed/FLOPS": 202459281677824.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058337584137916565, "optim/lr": 0.002981249995119064, "optim/total_tokens": 4074766336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.9401021003723145, "created_at": "2025-01-16T03:30:31.717022+00:00"} {"global_step": 7773, "acc_step": 0, "speed/wps": 12886.807704415582, "speed/FLOPS": 202405023327332.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06869412958621979, "optim/lr": 0.002981236487943132, "optim/total_tokens": 4075290624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0671026706695557, "created_at": "2025-01-16T03:30:41.894651+00:00"} {"global_step": 7774, "acc_step": 0, "speed/wps": 12887.13548405678, "speed/FLOPS": 202410171557014.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056279923766851425, "optim/lr": 0.002981222975934406, "optim/total_tokens": 4075814912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9350132942199707, "created_at": "2025-01-16T03:30:52.069338+00:00"} {"global_step": 7775, "acc_step": 0, "speed/wps": 12881.270060528706, "speed/FLOPS": 202318047020569.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05349593982100487, "optim/lr": 0.002981209459092929, "optim/total_tokens": 4076339200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.919783592224121, "created_at": "2025-01-16T03:31:02.251666+00:00"} {"global_step": 7776, "acc_step": 0, "speed/wps": 12877.126434783111, "speed/FLOPS": 202252965684296.34, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05619386211037636, "optim/lr": 0.002981195937418746, "optim/total_tokens": 4076863488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.019649028778076, "created_at": "2025-01-16T03:31:12.439189+00:00"} {"global_step": 7777, "acc_step": 0, "speed/wps": 12882.514873342172, "speed/FLOPS": 202337598516356.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055462777614593506, "optim/lr": 0.0029811824109119006, "optim/total_tokens": 4077387776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8560636043548584, "created_at": "2025-01-16T03:31:22.620173+00:00"} {"global_step": 7778, "acc_step": 0, "speed/wps": 12881.447916119889, "speed/FLOPS": 202320840487028.1, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05496375635266304, "optim/lr": 0.002981168879572438, "optim/total_tokens": 4077912064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.860703229904175, "created_at": "2025-01-16T03:31:32.803901+00:00"} {"global_step": 7779, "acc_step": 0, "speed/wps": 12884.096624097165, "speed/FLOPS": 202362442085517.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05635368078947067, "optim/lr": 0.0029811553434004002, "optim/total_tokens": 4078436352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9120867252349854, "created_at": "2025-01-16T03:31:42.981167+00:00"} {"global_step": 7780, "acc_step": 0, "speed/wps": 12882.785269733376, "speed/FLOPS": 202341845463244.38, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052753038704395294, "optim/lr": 0.0029811418023958335, "optim/total_tokens": 4078960640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 2.851740837097168, "created_at": "2025-01-16T03:31:53.159674+00:00"} {"global_step": 7781, "acc_step": 0, "speed/wps": 12885.913082187248, "speed/FLOPS": 202390972055897.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0571255125105381, "optim/lr": 0.002981128256558781, "optim/total_tokens": 4079484928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 2.9407553672790527, "created_at": "2025-01-16T03:32:03.340688+00:00"} {"global_step": 7782, "acc_step": 0, "speed/wps": 12886.339364111142, "speed/FLOPS": 202397667399284.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0657324567437172, "optim/lr": 0.002981114705889287, "optim/total_tokens": 4080009216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.0555498600006104, "created_at": "2025-01-16T03:32:13.521182+00:00"} {"global_step": 7783, "acc_step": 0, "speed/wps": 12885.624148704332, "speed/FLOPS": 202386433958511.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06299963593482971, "optim/lr": 0.002981101150387396, "optim/total_tokens": 4080533504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390663, "loss/out": 2.9403464794158936, "created_at": "2025-01-16T03:32:23.696715+00:00"} {"global_step": 7784, "acc_step": 0, "speed/wps": 12883.568744918472, "speed/FLOPS": 202354151017634.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09059769660234451, "optim/lr": 0.002981087590053152, "optim/total_tokens": 4081057792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328606, "loss/out": 2.8539280891418457, "created_at": "2025-01-16T03:32:33.874697+00:00"} {"global_step": 7785, "acc_step": 0, "speed/wps": 12889.290341673257, "speed/FLOPS": 202444016556965.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05689194053411484, "optim/lr": 0.002981074024886599, "optim/total_tokens": 4081582080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 3.0059595108032227, "created_at": "2025-01-16T03:32:44.047571+00:00"} {"global_step": 7786, "acc_step": 0, "speed/wps": 12888.556260579644, "speed/FLOPS": 202432486804657.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07705637812614441, "optim/lr": 0.002981060454887782, "optim/total_tokens": 4082106368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.0063529014587402, "created_at": "2025-01-16T03:32:54.221811+00:00"} {"global_step": 7787, "acc_step": 0, "speed/wps": 12887.094634152438, "speed/FLOPS": 202409529953133.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058524008840322495, "optim/lr": 0.0029810468800567456, "optim/total_tokens": 4082630656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 2.9144883155822754, "created_at": "2025-01-16T03:33:04.397988+00:00"} {"global_step": 7788, "acc_step": 0, "speed/wps": 12885.142468332166, "speed/FLOPS": 202378868506368.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050831880420446396, "optim/lr": 0.002981033300393532, "optim/total_tokens": 4083154944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.981829881668091, "created_at": "2025-01-16T03:33:14.574214+00:00"} {"global_step": 7789, "acc_step": 0, "speed/wps": 12888.697737316488, "speed/FLOPS": 202434708891219.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059771690517663956, "optim/lr": 0.002981019715898187, "optim/total_tokens": 4083679232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9523887634277344, "created_at": "2025-01-16T03:33:24.750525+00:00"} {"global_step": 7790, "acc_step": 0, "speed/wps": 12890.789096982311, "speed/FLOPS": 202467556568600.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056940894573926926, "optim/lr": 0.0029810061265707556, "optim/total_tokens": 4084203520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462458, "loss/out": 2.9328196048736572, "created_at": "2025-01-16T03:33:34.923061+00:00"} {"global_step": 7791, "acc_step": 0, "speed/wps": 12887.5942994799, "speed/FLOPS": 202417377883712.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047211285680532455, "optim/lr": 0.0029809925324112804, "optim/total_tokens": 4084727808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 2.8721094131469727, "created_at": "2025-01-16T03:33:45.102061+00:00"} {"global_step": 7792, "acc_step": 0, "speed/wps": 12887.849865271586, "speed/FLOPS": 202421391895655.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.063905730843544, "optim/lr": 0.0029809789334198076, "optim/total_tokens": 4085252096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 2.912245988845825, "created_at": "2025-01-16T03:33:55.275790+00:00"} {"global_step": 7793, "acc_step": 0, "speed/wps": 12889.653362039047, "speed/FLOPS": 202449718290652.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0528724379837513, "optim/lr": 0.0029809653295963796, "optim/total_tokens": 4085776384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 2.885829448699951, "created_at": "2025-01-16T03:34:05.450104+00:00"} {"global_step": 7794, "acc_step": 0, "speed/wps": 12889.00351718344, "speed/FLOPS": 202439511584215.06, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056384045630693436, "optim/lr": 0.0029809517209410425, "optim/total_tokens": 4086300672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 2.947052240371704, "created_at": "2025-01-16T03:34:15.629193+00:00"} {"global_step": 7795, "acc_step": 0, "speed/wps": 12887.90031967999, "speed/FLOPS": 202422184351471.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05179668590426445, "optim/lr": 0.0029809381074538395, "optim/total_tokens": 4086824960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.956674814224243, "created_at": "2025-01-16T03:34:25.804179+00:00"} {"global_step": 7796, "acc_step": 0, "speed/wps": 12887.713523151944, "speed/FLOPS": 202419250455312.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049740277230739594, "optim/lr": 0.0029809244891348154, "optim/total_tokens": 4087349248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.0095772743225098, "created_at": "2025-01-16T03:34:35.979401+00:00"} {"global_step": 7797, "acc_step": 0, "speed/wps": 12886.427861600463, "speed/FLOPS": 202399057373962.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048279643058776855, "optim/lr": 0.0029809108659840156, "optim/total_tokens": 4087873536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8949503898620605, "created_at": "2025-01-16T03:34:46.154593+00:00"} {"global_step": 7798, "acc_step": 0, "speed/wps": 12888.857374802104, "speed/FLOPS": 202437216217295.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061108145862817764, "optim/lr": 0.0029808972380014833, "optim/total_tokens": 4088397824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342101, "loss/out": 3.107398748397827, "created_at": "2025-01-16T03:34:56.328132+00:00"} {"global_step": 7799, "acc_step": 0, "speed/wps": 12887.177084741164, "speed/FLOPS": 202410824952927.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06147027388215065, "optim/lr": 0.0029808836051872633, "optim/total_tokens": 4088922112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.9776997566223145, "created_at": "2025-01-16T03:35:06.505688+00:00"} {"global_step": 7800, "acc_step": 0, "speed/wps": 12883.37599782165, "speed/FLOPS": 202351123659616.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06938069313764572, "optim/lr": 0.0029808699675414, "optim/total_tokens": 4089446400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.849276065826416, "created_at": "2025-01-16T03:35:16.685259+00:00"} {"global_step": 7801, "acc_step": 0, "speed/wps": 12881.614745457016, "speed/FLOPS": 202323460771015.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05996369943022728, "optim/lr": 0.002980856325063938, "optim/total_tokens": 4089970688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.866738796234131, "created_at": "2025-01-16T03:35:26.865189+00:00"} {"global_step": 7802, "acc_step": 0, "speed/wps": 12886.945589719322, "speed/FLOPS": 202407189005502.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0788484662771225, "optim/lr": 0.0029808426777549225, "optim/total_tokens": 4090494976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.900611639022827, "created_at": "2025-01-16T03:35:37.042292+00:00"} {"global_step": 7803, "acc_step": 0, "speed/wps": 12885.609540681793, "speed/FLOPS": 202386204519444.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08365853875875473, "optim/lr": 0.002980829025614397, "optim/total_tokens": 4091019264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9322547912597656, "created_at": "2025-01-16T03:35:47.221598+00:00"} {"global_step": 7804, "acc_step": 0, "speed/wps": 12891.70184054457, "speed/FLOPS": 202481892460486.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05964035168290138, "optim/lr": 0.002980815368642406, "optim/total_tokens": 4091543552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.0000171661376953, "created_at": "2025-01-16T03:35:57.392323+00:00"} {"global_step": 7805, "acc_step": 0, "speed/wps": 12886.728508224225, "speed/FLOPS": 202403779442320.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09323649108409882, "optim/lr": 0.002980801706838995, "optim/total_tokens": 4092067840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.917867422103882, "created_at": "2025-01-16T03:36:07.573319+00:00"} {"global_step": 7806, "acc_step": 0, "speed/wps": 12885.569259661846, "speed/FLOPS": 202385571850674.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05853283032774925, "optim/lr": 0.002980788040204208, "optim/total_tokens": 4092592128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 2.9592764377593994, "created_at": "2025-01-16T03:36:17.752054+00:00"} {"global_step": 7807, "acc_step": 0, "speed/wps": 12885.434019303873, "speed/FLOPS": 202383447715012.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08588425070047379, "optim/lr": 0.0029807743687380894, "optim/total_tokens": 4093116416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.0314254760742188, "created_at": "2025-01-16T03:36:27.929926+00:00"} {"global_step": 7808, "acc_step": 0, "speed/wps": 12885.963838809821, "speed/FLOPS": 202391769258401.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05696149170398712, "optim/lr": 0.002980760692440684, "optim/total_tokens": 4093640704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.0396924018859863, "created_at": "2025-01-16T03:36:38.105858+00:00"} {"global_step": 7809, "acc_step": 0, "speed/wps": 12882.787236740136, "speed/FLOPS": 202341876357788.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05402229353785515, "optim/lr": 0.0029807470113120365, "optim/total_tokens": 4094164992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.0620391368865967, "created_at": "2025-01-16T03:36:48.285500+00:00"} {"global_step": 7810, "acc_step": 0, "speed/wps": 12887.0776237608, "speed/FLOPS": 202409262781557.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05413083732128143, "optim/lr": 0.002980733325352192, "optim/total_tokens": 4094689280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302403, "loss/out": 2.952479839324951, "created_at": "2025-01-16T03:36:58.460474+00:00"} {"global_step": 7811, "acc_step": 0, "speed/wps": 12888.04387691286, "speed/FLOPS": 202424439115081.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053532056510448456, "optim/lr": 0.0029807196345611946, "optim/total_tokens": 4095213568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.1375644207000732, "created_at": "2025-01-16T03:37:08.636654+00:00"} {"global_step": 7812, "acc_step": 0, "speed/wps": 12890.20703704298, "speed/FLOPS": 202458414517418.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056196872144937515, "optim/lr": 0.0029807059389390885, "optim/total_tokens": 4095737856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9724645614624023, "created_at": "2025-01-16T03:37:18.816493+00:00"} {"global_step": 7813, "acc_step": 0, "speed/wps": 12886.992607877126, "speed/FLOPS": 202407927490280.4, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05183256044983864, "optim/lr": 0.002980692238485919, "optim/total_tokens": 4096262144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.993914842605591, "created_at": "2025-01-16T03:37:28.991048+00:00"} {"global_step": 7814, "acc_step": 0, "speed/wps": 12887.55226299293, "speed/FLOPS": 202416717642920.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0549299381673336, "optim/lr": 0.002980678533201731, "optim/total_tokens": 4096786432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.982006072998047, "created_at": "2025-01-16T03:37:39.165669+00:00"} {"global_step": 7815, "acc_step": 0, "speed/wps": 12884.635508541549, "speed/FLOPS": 202370906006221.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05446034297347069, "optim/lr": 0.0029806648230865686, "optim/total_tokens": 4097310720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.91610050201416, "created_at": "2025-01-16T03:37:49.342170+00:00"} {"global_step": 7816, "acc_step": 0, "speed/wps": 12889.015747310643, "speed/FLOPS": 202439703675168.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053935397416353226, "optim/lr": 0.0029806511081404774, "optim/total_tokens": 4097835008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.9895129203796387, "created_at": "2025-01-16T03:37:59.520585+00:00"} {"global_step": 7817, "acc_step": 0, "speed/wps": 12887.399030400302, "speed/FLOPS": 202414310914489.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06003361940383911, "optim/lr": 0.0029806373883635015, "optim/total_tokens": 4098359296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.96699595451355, "created_at": "2025-01-16T03:38:09.694833+00:00"} {"global_step": 7818, "acc_step": 0, "speed/wps": 12888.498486670822, "speed/FLOPS": 202431579386029.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07980351150035858, "optim/lr": 0.0029806236637556854, "optim/total_tokens": 4098883584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 3.080148696899414, "created_at": "2025-01-16T03:38:19.869853+00:00"} {"global_step": 7819, "acc_step": 0, "speed/wps": 12891.13560103376, "speed/FLOPS": 202472998890874.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0518144927918911, "optim/lr": 0.0029806099343170746, "optim/total_tokens": 4099407872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.0691943168640137, "created_at": "2025-01-16T03:38:30.042212+00:00"} {"global_step": 7820, "acc_step": 0, "speed/wps": 12887.617517616685, "speed/FLOPS": 202417742556456.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05617259070277214, "optim/lr": 0.0029805962000477126, "optim/total_tokens": 4099932160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.9337844848632812, "created_at": "2025-01-16T03:38:40.216498+00:00"} {"global_step": 7821, "acc_step": 0, "speed/wps": 12883.644693080761, "speed/FLOPS": 202355343887886.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07389932870864868, "optim/lr": 0.0029805824609476462, "optim/total_tokens": 4100456448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 3.0589489936828613, "created_at": "2025-01-16T03:38:50.396919+00:00"} {"global_step": 7822, "acc_step": 0, "speed/wps": 12886.92263714756, "speed/FLOPS": 202406828503822.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06844553351402283, "optim/lr": 0.002980568717016919, "optim/total_tokens": 4100980736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 2.9400594234466553, "created_at": "2025-01-16T03:39:00.573520+00:00"} {"global_step": 7823, "acc_step": 0, "speed/wps": 12887.342943780526, "speed/FLOPS": 202413429997055.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056935638189315796, "optim/lr": 0.0029805549682555763, "optim/total_tokens": 4101505024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.896355628967285, "created_at": "2025-01-16T03:39:10.752932+00:00"} {"global_step": 7824, "acc_step": 0, "speed/wps": 12885.792645943673, "speed/FLOPS": 202389080439195.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05961687117815018, "optim/lr": 0.002980541214663662, "optim/total_tokens": 4102029312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.0771703720092773, "created_at": "2025-01-16T03:39:20.928396+00:00"} {"global_step": 7825, "acc_step": 0, "speed/wps": 12888.600080005806, "speed/FLOPS": 202433175048959.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06543534249067307, "optim/lr": 0.0029805274562412216, "optim/total_tokens": 4102553600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340756, "loss/out": 2.9240143299102783, "created_at": "2025-01-16T03:39:31.104679+00:00"} {"global_step": 7826, "acc_step": 0, "speed/wps": 12888.975747172864, "speed/FLOPS": 202439075418038.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06587545573711395, "optim/lr": 0.0029805136929883006, "optim/total_tokens": 4103077888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.0223817825317383, "created_at": "2025-01-16T03:39:41.277469+00:00"} {"global_step": 7827, "acc_step": 0, "speed/wps": 12890.633489770591, "speed/FLOPS": 202465112543513.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07163209468126297, "optim/lr": 0.002980499924904943, "optim/total_tokens": 4103602176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.943531036376953, "created_at": "2025-01-16T03:39:51.449035+00:00"} {"global_step": 7828, "acc_step": 0, "speed/wps": 12886.151902542906, "speed/FLOPS": 202394723057756.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06154017895460129, "optim/lr": 0.0029804861519911944, "optim/total_tokens": 4104126464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.8726296424865723, "created_at": "2025-01-16T03:40:01.624366+00:00"} {"global_step": 7829, "acc_step": 0, "speed/wps": 12888.70384323446, "speed/FLOPS": 202434804793051.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06819197535514832, "optim/lr": 0.002980472374247099, "optim/total_tokens": 4104650752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9757349491119385, "created_at": "2025-01-16T03:40:11.797437+00:00"} {"global_step": 7830, "acc_step": 0, "speed/wps": 12886.738155768267, "speed/FLOPS": 202403930970256.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05375601723790169, "optim/lr": 0.0029804585916727027, "optim/total_tokens": 4105175040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9089136123657227, "created_at": "2025-01-16T03:40:21.972154+00:00"} {"global_step": 7831, "acc_step": 0, "speed/wps": 12888.965816525257, "speed/FLOPS": 202438919443572.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056454576551914215, "optim/lr": 0.00298044480426805, "optim/total_tokens": 4105699328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.023135185241699, "created_at": "2025-01-16T03:40:32.147329+00:00"} {"global_step": 7832, "acc_step": 0, "speed/wps": 12887.093160270733, "speed/FLOPS": 202409506803795.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053529467433691025, "optim/lr": 0.002980431012033185, "optim/total_tokens": 4106223616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408061, "loss/out": 3.023014545440674, "created_at": "2025-01-16T03:40:42.323374+00:00"} {"global_step": 7833, "acc_step": 0, "speed/wps": 12887.026426470098, "speed/FLOPS": 202408458657754.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0581972636282444, "optim/lr": 0.0029804172149681544, "optim/total_tokens": 4106747904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8690342903137207, "created_at": "2025-01-16T03:40:52.500889+00:00"} {"global_step": 7834, "acc_step": 0, "speed/wps": 12887.509644978265, "speed/FLOPS": 202416048268437.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05386112257838249, "optim/lr": 0.002980403413073002, "optim/total_tokens": 4107272192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.9288065433502197, "created_at": "2025-01-16T03:41:02.676769+00:00"} {"global_step": 7835, "acc_step": 0, "speed/wps": 12890.013773692435, "speed/FLOPS": 202455379050926.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.058611366897821426, "optim/lr": 0.0029803896063477727, "optim/total_tokens": 4107796480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.938154697418213, "created_at": "2025-01-16T03:41:12.853993+00:00"} {"global_step": 7836, "acc_step": 0, "speed/wps": 12887.970618503294, "speed/FLOPS": 202423288491091.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059379495680332184, "optim/lr": 0.0029803757947925126, "optim/total_tokens": 4108320768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 3.0057055950164795, "created_at": "2025-01-16T03:41:23.033822+00:00"} {"global_step": 7837, "acc_step": 0, "speed/wps": 12887.615651807237, "speed/FLOPS": 202417713251354.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05288832262158394, "optim/lr": 0.002980361978407266, "optim/total_tokens": 4108845056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.97556209564209, "created_at": "2025-01-16T03:41:33.208604+00:00"} {"global_step": 7838, "acc_step": 0, "speed/wps": 12886.93941938597, "speed/FLOPS": 202407092091938.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07720425724983215, "optim/lr": 0.0029803481571920778, "optim/total_tokens": 4109369344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.9793105125427246, "created_at": "2025-01-16T03:41:43.383658+00:00"} {"global_step": 7839, "acc_step": 0, "speed/wps": 12890.823569557404, "speed/FLOPS": 202468098007762.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0654464066028595, "optim/lr": 0.0029803343311469934, "optim/total_tokens": 4109893632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.9647605419158936, "created_at": "2025-01-16T03:41:53.556648+00:00"} {"global_step": 7840, "acc_step": 0, "speed/wps": 12887.882734767296, "speed/FLOPS": 202421908156253.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057885970920324326, "optim/lr": 0.002980320500272059, "optim/total_tokens": 4110417920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.987156629562378, "created_at": "2025-01-16T03:42:03.731403+00:00"} {"global_step": 7841, "acc_step": 0, "speed/wps": 12888.821158718429, "speed/FLOPS": 202436647393935.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06261094659566879, "optim/lr": 0.002980306664567318, "optim/total_tokens": 4110942208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 2.930319309234619, "created_at": "2025-01-16T03:42:13.909141+00:00"} {"global_step": 7842, "acc_step": 0, "speed/wps": 12884.314807086794, "speed/FLOPS": 202365868949184.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05173603072762489, "optim/lr": 0.0029802928240328153, "optim/total_tokens": 4111466496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.03666353225708, "created_at": "2025-01-16T03:42:24.088776+00:00"} {"global_step": 7843, "acc_step": 0, "speed/wps": 12888.982857870627, "speed/FLOPS": 202439187101318.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055595219135284424, "optim/lr": 0.0029802789786685983, "optim/total_tokens": 4111990784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 3.0698471069335938, "created_at": "2025-01-16T03:42:34.261549+00:00"} {"global_step": 7844, "acc_step": 0, "speed/wps": 12888.108361100612, "speed/FLOPS": 202425451927860.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06837823241949081, "optim/lr": 0.0029802651284747106, "optim/total_tokens": 4112515072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.825530529022217, "created_at": "2025-01-16T03:42:44.441579+00:00"} {"global_step": 7845, "acc_step": 0, "speed/wps": 12889.907676644261, "speed/FLOPS": 202453712650992.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05936630442738533, "optim/lr": 0.002980251273451197, "optim/total_tokens": 4113039360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.813897132873535, "created_at": "2025-01-16T03:42:54.616028+00:00"} {"global_step": 7846, "acc_step": 0, "speed/wps": 12890.435972694646, "speed/FLOPS": 202462010266418.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05509551987051964, "optim/lr": 0.002980237413598104, "optim/total_tokens": 4113563648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.965160608291626, "created_at": "2025-01-16T03:43:04.790587+00:00"} {"global_step": 7847, "acc_step": 0, "speed/wps": 12894.522219585118, "speed/FLOPS": 202526190388925.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051682665944099426, "optim/lr": 0.0029802235489154757, "optim/total_tokens": 4114087936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288730, "loss/out": 2.9069106578826904, "created_at": "2025-01-16T03:43:14.960441+00:00"} {"global_step": 7848, "acc_step": 0, "speed/wps": 12888.731791486487, "speed/FLOPS": 202435243758754.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.048918016254901886, "optim/lr": 0.002980209679403358, "optim/total_tokens": 4114612224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9850642681121826, "created_at": "2025-01-16T03:43:25.134764+00:00"} {"global_step": 7849, "acc_step": 0, "speed/wps": 12889.440875810866, "speed/FLOPS": 202446380902452.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046407390385866165, "optim/lr": 0.002980195805061796, "optim/total_tokens": 4115136512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.889268159866333, "created_at": "2025-01-16T03:43:35.314253+00:00"} {"global_step": 7850, "acc_step": 0, "speed/wps": 12890.028301977974, "speed/FLOPS": 202455607237615.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04484722018241882, "optim/lr": 0.002980181925890835, "optim/total_tokens": 4115660800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8122711181640625, "created_at": "2025-01-16T03:43:45.489225+00:00"} {"global_step": 7851, "acc_step": 0, "speed/wps": 12890.191199050481, "speed/FLOPS": 202458165759982.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045913729816675186, "optim/lr": 0.00298016804189052, "optim/total_tokens": 4116185088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 3.1232540607452393, "created_at": "2025-01-16T03:43:55.661009+00:00"} {"global_step": 7852, "acc_step": 0, "speed/wps": 12890.161390908956, "speed/FLOPS": 202457697582159.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056547172367572784, "optim/lr": 0.002980154153060897, "optim/total_tokens": 4116709376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8866126537323, "created_at": "2025-01-16T03:44:05.835953+00:00"} {"global_step": 7853, "acc_step": 0, "speed/wps": 12890.118512431953, "speed/FLOPS": 202457024116756.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051050882786512375, "optim/lr": 0.0029801402594020103, "optim/total_tokens": 4117233664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.0118961334228516, "created_at": "2025-01-16T03:44:16.009196+00:00"} {"global_step": 7854, "acc_step": 0, "speed/wps": 12887.888295843224, "speed/FLOPS": 202421995500592.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047570355236530304, "optim/lr": 0.0029801263609139056, "optim/total_tokens": 4117757952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.854764461517334, "created_at": "2025-01-16T03:44:26.185641+00:00"} {"global_step": 7855, "acc_step": 0, "speed/wps": 12887.34519186192, "speed/FLOPS": 202413465306262.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05410139635205269, "optim/lr": 0.002980112457596629, "optim/total_tokens": 4118282240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.953507423400879, "created_at": "2025-01-16T03:44:36.366569+00:00"} {"global_step": 7856, "acc_step": 0, "speed/wps": 12891.492170420477, "speed/FLOPS": 202478599303070.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051712024956941605, "optim/lr": 0.0029800985494502245, "optim/total_tokens": 4118806528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.033452033996582, "created_at": "2025-01-16T03:44:46.540763+00:00"} {"global_step": 7857, "acc_step": 0, "speed/wps": 12886.672756652835, "speed/FLOPS": 202402903787281.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05558522790670395, "optim/lr": 0.002980084636474739, "optim/total_tokens": 4119330816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.89963960647583, "created_at": "2025-01-16T03:44:56.716296+00:00"} {"global_step": 7858, "acc_step": 0, "speed/wps": 12887.955548419832, "speed/FLOPS": 202423051794722.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06256618350744247, "optim/lr": 0.0029800707186702162, "optim/total_tokens": 4119855104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344099, "loss/out": 2.9474034309387207, "created_at": "2025-01-16T03:45:06.895589+00:00"} {"global_step": 7859, "acc_step": 0, "speed/wps": 12885.394816482236, "speed/FLOPS": 202382831980828.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05625868961215019, "optim/lr": 0.0029800567960367036, "optim/total_tokens": 4120379392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.037015676498413, "created_at": "2025-01-16T03:45:17.072966+00:00"} {"global_step": 7860, "acc_step": 0, "speed/wps": 12891.237146205733, "speed/FLOPS": 202474593797338.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050340212881565094, "optim/lr": 0.0029800428685742447, "optim/total_tokens": 4120903680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9304747581481934, "created_at": "2025-01-16T03:45:27.246865+00:00"} {"global_step": 7861, "acc_step": 0, "speed/wps": 12891.507453064332, "speed/FLOPS": 202478839337992.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054109297692775726, "optim/lr": 0.002980028936282886, "optim/total_tokens": 4121427968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 2.8798282146453857, "created_at": "2025-01-16T03:45:37.423049+00:00"} {"global_step": 7862, "acc_step": 0, "speed/wps": 12878.251533833696, "speed/FLOPS": 202270636910933.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.050404615700244904, "optim/lr": 0.0029800149991626723, "optim/total_tokens": 4121952256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 2.956839084625244, "created_at": "2025-01-16T03:45:47.606756+00:00"} {"global_step": 7863, "acc_step": 0, "speed/wps": 12888.769594646974, "speed/FLOPS": 202435837509336.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05004207789897919, "optim/lr": 0.00298000105721365, "optim/total_tokens": 4122476544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9358630180358887, "created_at": "2025-01-16T03:45:57.781859+00:00"} {"global_step": 7864, "acc_step": 0, "speed/wps": 12895.625928380905, "speed/FLOPS": 202543525652217.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05250801518559456, "optim/lr": 0.002979987110435863, "optim/total_tokens": 4123000832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.92956280708313, "created_at": "2025-01-16T03:46:07.949670+00:00"} {"global_step": 7865, "acc_step": 0, "speed/wps": 12887.89165939954, "speed/FLOPS": 202422048329866.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05920222029089928, "optim/lr": 0.002979973158829359, "optim/total_tokens": 4123525120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405960, "loss/out": 2.978367328643799, "created_at": "2025-01-16T03:46:18.123414+00:00"} {"global_step": 7866, "acc_step": 0, "speed/wps": 12892.127171061613, "speed/FLOPS": 202488572860721.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07401689887046814, "optim/lr": 0.002979959202394182, "optim/total_tokens": 4124049408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 2.9308359622955322, "created_at": "2025-01-16T03:46:28.293700+00:00"} {"global_step": 7867, "acc_step": 0, "speed/wps": 12888.792171361558, "speed/FLOPS": 202436192107662.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05412333458662033, "optim/lr": 0.0029799452411303773, "optim/total_tokens": 4124573696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.007835865020752, "created_at": "2025-01-16T03:46:38.467481+00:00"} {"global_step": 7868, "acc_step": 0, "speed/wps": 12884.188325020586, "speed/FLOPS": 202363882374530.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05076369270682335, "optim/lr": 0.0029799312750379914, "optim/total_tokens": 4125097984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 2.9014644622802734, "created_at": "2025-01-16T03:46:48.646551+00:00"} {"global_step": 7869, "acc_step": 0, "speed/wps": 12894.258357018096, "speed/FLOPS": 202522046064724.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06470825523138046, "optim/lr": 0.0029799173041170695, "optim/total_tokens": 4125622272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.912170886993408, "created_at": "2025-01-16T03:46:58.821496+00:00"} {"global_step": 7870, "acc_step": 0, "speed/wps": 12888.978286200996, "speed/FLOPS": 202439115296964.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07508347928524017, "optim/lr": 0.002979903328367657, "optim/total_tokens": 4126146560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8695485591888428, "created_at": "2025-01-16T03:47:08.997092+00:00"} {"global_step": 7871, "acc_step": 0, "speed/wps": 12891.48639013772, "speed/FLOPS": 202478508515786.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06813392788171768, "optim/lr": 0.0029798893477898003, "optim/total_tokens": 4126670848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.023421287536621, "created_at": "2025-01-16T03:47:19.169976+00:00"} {"global_step": 7872, "acc_step": 0, "speed/wps": 12888.691175792295, "speed/FLOPS": 202434605833465.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07738219201564789, "optim/lr": 0.0029798753623835436, "optim/total_tokens": 4127195136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 3.0002212524414062, "created_at": "2025-01-16T03:47:29.344596+00:00"} {"global_step": 7873, "acc_step": 0, "speed/wps": 12892.70417579554, "speed/FLOPS": 202497635512955.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062245067209005356, "optim/lr": 0.002979861372148934, "optim/total_tokens": 4127719424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.79297137260437, "created_at": "2025-01-16T03:47:39.517418+00:00"} {"global_step": 7874, "acc_step": 0, "speed/wps": 12892.68542149996, "speed/FLOPS": 202497340950972.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05365241691470146, "optim/lr": 0.0029798473770860164, "optim/total_tokens": 4128243712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9218742847442627, "created_at": "2025-01-16T03:47:49.687655+00:00"} {"global_step": 7875, "acc_step": 0, "speed/wps": 12889.178329586135, "speed/FLOPS": 202442257253216.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07730606198310852, "optim/lr": 0.0029798333771948364, "optim/total_tokens": 4128768000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 2.9515931606292725, "created_at": "2025-01-16T03:47:59.860444+00:00"} {"global_step": 7876, "acc_step": 0, "speed/wps": 12892.444020473764, "speed/FLOPS": 202493549416136.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052061256021261215, "optim/lr": 0.0029798193724754393, "optim/total_tokens": 4129292288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.936277389526367, "created_at": "2025-01-16T03:48:10.030509+00:00"} {"global_step": 7877, "acc_step": 0, "speed/wps": 12890.060985427508, "speed/FLOPS": 202456120576101.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06607227772474289, "optim/lr": 0.002979805362927871, "optim/total_tokens": 4129816576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 3.1039376258850098, "created_at": "2025-01-16T03:48:20.203804+00:00"} {"global_step": 7878, "acc_step": 0, "speed/wps": 12888.495000045925, "speed/FLOPS": 202431524623794.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07519324123859406, "optim/lr": 0.0029797913485521787, "optim/total_tokens": 4130340864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.9556469917297363, "created_at": "2025-01-16T03:48:30.377862+00:00"} {"global_step": 7879, "acc_step": 0, "speed/wps": 12888.21719976879, "speed/FLOPS": 202427161388704.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059159863740205765, "optim/lr": 0.0029797773293484065, "optim/total_tokens": 4130865152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 2.9218263626098633, "created_at": "2025-01-16T03:48:40.553004+00:00"} {"global_step": 7880, "acc_step": 0, "speed/wps": 12890.84280489994, "speed/FLOPS": 202468400125248.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0784374549984932, "optim/lr": 0.0029797633053166004, "optim/total_tokens": 4131389440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8831522464752197, "created_at": "2025-01-16T03:48:50.724289+00:00"} {"global_step": 7881, "acc_step": 0, "speed/wps": 12892.693275644437, "speed/FLOPS": 202497464311104.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06745587289333344, "optim/lr": 0.0029797492764568065, "optim/total_tokens": 4131913728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 3.0966525077819824, "created_at": "2025-01-16T03:49:00.895064+00:00"} {"global_step": 7882, "acc_step": 0, "speed/wps": 12894.698070164846, "speed/FLOPS": 202528952363923.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057248152792453766, "optim/lr": 0.0029797352427690703, "optim/total_tokens": 4132438016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457889, "loss/out": 2.889028787612915, "created_at": "2025-01-16T03:49:11.064614+00:00"} {"global_step": 7883, "acc_step": 0, "speed/wps": 12881.633571314933, "speed/FLOPS": 202323756456983.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06311521679162979, "optim/lr": 0.0029797212042534374, "optim/total_tokens": 4132962304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9832096099853516, "created_at": "2025-01-16T03:49:21.244134+00:00"} {"global_step": 7884, "acc_step": 0, "speed/wps": 12879.036580266637, "speed/FLOPS": 202282967143930.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04906866326928139, "optim/lr": 0.0029797071609099545, "optim/total_tokens": 4133486592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.1411972045898438, "created_at": "2025-01-16T03:49:31.425271+00:00"} {"global_step": 7885, "acc_step": 0, "speed/wps": 12887.545653069437, "speed/FLOPS": 202416613824989.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07054860144853592, "optim/lr": 0.0029796931127386663, "optim/total_tokens": 4134010880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.927095890045166, "created_at": "2025-01-16T03:49:41.603213+00:00"} {"global_step": 7886, "acc_step": 0, "speed/wps": 12885.495842980441, "speed/FLOPS": 202384418740807.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0613120011985302, "optim/lr": 0.0029796790597396198, "optim/total_tokens": 4134535168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378926, "loss/out": 3.006535768508911, "created_at": "2025-01-16T03:49:51.779128+00:00"} {"global_step": 7887, "acc_step": 0, "speed/wps": 12888.604349243671, "speed/FLOPS": 202433242103206.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06354981660842896, "optim/lr": 0.0029796650019128595, "optim/total_tokens": 4135059456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9941585063934326, "created_at": "2025-01-16T03:50:01.956532+00:00"} {"global_step": 7888, "acc_step": 0, "speed/wps": 12889.059598191083, "speed/FLOPS": 202440392413503.44, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05150919407606125, "optim/lr": 0.002979650939258432, "optim/total_tokens": 4135583744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 2.964456558227539, "created_at": "2025-01-16T03:50:12.132367+00:00"} {"global_step": 7889, "acc_step": 0, "speed/wps": 12888.331486066689, "speed/FLOPS": 202428956412058.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06545894593000412, "optim/lr": 0.002979636871776383, "optim/total_tokens": 4136108032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.039754629135132, "created_at": "2025-01-16T03:50:22.305670+00:00"} {"global_step": 7890, "acc_step": 0, "speed/wps": 12891.974102457196, "speed/FLOPS": 202486168707951.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05230025202035904, "optim/lr": 0.002979622799466759, "optim/total_tokens": 4136632320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 2.965336561203003, "created_at": "2025-01-16T03:50:32.476136+00:00"} {"global_step": 7891, "acc_step": 0, "speed/wps": 12885.563437498527, "speed/FLOPS": 202385480405598.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0782070979475975, "optim/lr": 0.0029796087223296058, "optim/total_tokens": 4137156608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.973928213119507, "created_at": "2025-01-16T03:50:42.652917+00:00"} {"global_step": 7892, "acc_step": 0, "speed/wps": 12888.756341620016, "speed/FLOPS": 202435629352336.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05939773842692375, "optim/lr": 0.002979594640364968, "optim/total_tokens": 4137680896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.888871669769287, "created_at": "2025-01-16T03:50:52.825895+00:00"} {"global_step": 7893, "acc_step": 0, "speed/wps": 12889.954455512287, "speed/FLOPS": 202454447377395.78, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05699431151151657, "optim/lr": 0.002979580553572893, "optim/total_tokens": 4138205184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332063, "loss/out": 2.897536516189575, "created_at": "2025-01-16T03:51:02.997886+00:00"} {"global_step": 7894, "acc_step": 0, "speed/wps": 12888.498034613393, "speed/FLOPS": 202431572285846.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058949198573827744, "optim/lr": 0.002979566461953427, "optim/total_tokens": 4138729472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8485453128814697, "created_at": "2025-01-16T03:51:13.175607+00:00"} {"global_step": 7895, "acc_step": 0, "speed/wps": 12887.050931166004, "speed/FLOPS": 202408843537676.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06281395256519318, "optim/lr": 0.002979552365506614, "optim/total_tokens": 4139253760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.8692235946655273, "created_at": "2025-01-16T03:51:23.350033+00:00"} {"global_step": 7896, "acc_step": 0, "speed/wps": 12891.61626058441, "speed/FLOPS": 202480548309612.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05701884999871254, "optim/lr": 0.002979538264232502, "optim/total_tokens": 4139778048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 3.0402026176452637, "created_at": "2025-01-16T03:51:33.520815+00:00"} {"global_step": 7897, "acc_step": 0, "speed/wps": 12886.523458850144, "speed/FLOPS": 202400558860133.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04785279557108879, "optim/lr": 0.0029795241581311357, "optim/total_tokens": 4140302336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9381766319274902, "created_at": "2025-01-16T03:51:43.696975+00:00"} {"global_step": 7898, "acc_step": 0, "speed/wps": 12890.14483285758, "speed/FLOPS": 202457437515209.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0630858913064003, "optim/lr": 0.0029795100472025625, "optim/total_tokens": 4140826624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.968433380126953, "created_at": "2025-01-16T03:51:53.869090+00:00"} {"global_step": 7899, "acc_step": 0, "speed/wps": 12889.381333892528, "speed/FLOPS": 202445445714805.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04506387189030647, "optim/lr": 0.0029794959314468268, "optim/total_tokens": 4141350912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9012880325317383, "created_at": "2025-01-16T03:52:04.041638+00:00"} {"global_step": 7900, "acc_step": 0, "speed/wps": 12886.22864927254, "speed/FLOPS": 202395928470606.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053147073835134506, "optim/lr": 0.002979481810863976, "optim/total_tokens": 4141875200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 3.0370750427246094, "created_at": "2025-01-16T03:52:14.216665+00:00"} {"global_step": 7901, "acc_step": 0, "speed/wps": 12889.780858083204, "speed/FLOPS": 202451720791224.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050253745168447495, "optim/lr": 0.002979467685454056, "optim/total_tokens": 4142399488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.9651057720184326, "created_at": "2025-01-16T03:52:24.395621+00:00"} {"global_step": 7902, "acc_step": 0, "speed/wps": 12889.894917112908, "speed/FLOPS": 202453512245019.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050550978630781174, "optim/lr": 0.0029794535552171114, "optim/total_tokens": 4142923776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.013035297393799, "created_at": "2025-01-16T03:52:34.573467+00:00"} {"global_step": 7903, "acc_step": 0, "speed/wps": 12894.042342717246, "speed/FLOPS": 202518653263295.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054878417402505875, "optim/lr": 0.002979439420153191, "optim/total_tokens": 4143448064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474398, "loss/out": 2.8463096618652344, "created_at": "2025-01-16T03:52:44.742317+00:00"} {"global_step": 7904, "acc_step": 0, "speed/wps": 12887.402208424617, "speed/FLOPS": 202414360829728.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050865691155195236, "optim/lr": 0.0029794252802623376, "optim/total_tokens": 4143972352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.932528495788574, "created_at": "2025-01-16T03:52:54.916401+00:00"} {"global_step": 7905, "acc_step": 0, "speed/wps": 12886.21167774347, "speed/FLOPS": 202395661909420.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06537164002656937, "optim/lr": 0.0029794111355446002, "optim/total_tokens": 4144496640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8145668506622314, "created_at": "2025-01-16T03:53:05.095128+00:00"} {"global_step": 7906, "acc_step": 0, "speed/wps": 12882.83913975605, "speed/FLOPS": 202342691565975.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07297056168317795, "optim/lr": 0.0029793969860000234, "optim/total_tokens": 4145020928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.0337533950805664, "created_at": "2025-01-16T03:53:15.277249+00:00"} {"global_step": 7907, "acc_step": 0, "speed/wps": 12892.39851202815, "speed/FLOPS": 202492834643463.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05305140092968941, "optim/lr": 0.002979382831628654, "optim/total_tokens": 4145545216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 2.915510654449463, "created_at": "2025-01-16T03:53:25.449390+00:00"} {"global_step": 7908, "acc_step": 0, "speed/wps": 12891.066949105245, "speed/FLOPS": 202471920617999.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07895652949810028, "optim/lr": 0.002979368672430538, "optim/total_tokens": 4146069504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9649085998535156, "created_at": "2025-01-16T03:53:35.622768+00:00"} {"global_step": 7909, "acc_step": 0, "speed/wps": 12887.150925121376, "speed/FLOPS": 202410414080152.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05763380602002144, "optim/lr": 0.0029793545084057216, "optim/total_tokens": 4146593792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.040961980819702, "created_at": "2025-01-16T03:53:45.800425+00:00"} {"global_step": 7910, "acc_step": 0, "speed/wps": 12891.493660659691, "speed/FLOPS": 202478622709325.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.053080011159181595, "optim/lr": 0.002979340339554251, "optim/total_tokens": 4147118080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8031864166259766, "created_at": "2025-01-16T03:53:55.973026+00:00"} {"global_step": 7911, "acc_step": 0, "speed/wps": 12889.339780359976, "speed/FLOPS": 202444793059476.56, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05508286505937576, "optim/lr": 0.0029793261658761726, "optim/total_tokens": 4147642368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 2.8936238288879395, "created_at": "2025-01-16T03:54:06.145633+00:00"} {"global_step": 7912, "acc_step": 0, "speed/wps": 12889.288504890694, "speed/FLOPS": 202443987707771.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05086860433220863, "optim/lr": 0.002979311987371532, "optim/total_tokens": 4148166656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9538612365722656, "created_at": "2025-01-16T03:54:16.320370+00:00"} {"global_step": 7913, "acc_step": 0, "speed/wps": 12890.80847272881, "speed/FLOPS": 202467860891323.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06773405522108078, "optim/lr": 0.002979297804040376, "optim/total_tokens": 4148690944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.050389289855957, "created_at": "2025-01-16T03:54:26.496208+00:00"} {"global_step": 7914, "acc_step": 0, "speed/wps": 12876.223031930509, "speed/FLOPS": 202238776501090.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05594319850206375, "optim/lr": 0.0029792836158827516, "optim/total_tokens": 4149215232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 3.08231258392334, "created_at": "2025-01-16T03:54:36.680683+00:00"} {"global_step": 7915, "acc_step": 0, "speed/wps": 12888.545675503414, "speed/FLOPS": 202432320551489.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05698322132229805, "optim/lr": 0.0029792694228987036, "optim/total_tokens": 4149739520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.151346445083618, "created_at": "2025-01-16T03:54:46.854004+00:00"} {"global_step": 7916, "acc_step": 0, "speed/wps": 12895.083893544255, "speed/FLOPS": 202535012250274.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07158711552619934, "optim/lr": 0.0029792552250882793, "optim/total_tokens": 4150263808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.905475616455078, "created_at": "2025-01-16T03:54:57.023541+00:00"} {"global_step": 7917, "acc_step": 0, "speed/wps": 12891.031660496643, "speed/FLOPS": 202471366361909.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04640975594520569, "optim/lr": 0.0029792410224515242, "optim/total_tokens": 4150788096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.916090250015259, "created_at": "2025-01-16T03:55:07.197924+00:00"} {"global_step": 7918, "acc_step": 0, "speed/wps": 12891.06403967106, "speed/FLOPS": 202471874921337.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06561116874217987, "optim/lr": 0.0029792268149884853, "optim/total_tokens": 4151312384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 2.8797552585601807, "created_at": "2025-01-16T03:55:17.373049+00:00"} {"global_step": 7919, "acc_step": 0, "speed/wps": 12892.108868393463, "speed/FLOPS": 202488285392168.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05042334273457527, "optim/lr": 0.0029792126026992093, "optim/total_tokens": 4151836672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8601267337799072, "created_at": "2025-01-16T03:55:27.546258+00:00"} {"global_step": 7920, "acc_step": 0, "speed/wps": 12892.596712850449, "speed/FLOPS": 202495947659733.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0540911927819252, "optim/lr": 0.0029791983855837418, "optim/total_tokens": 4152360960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.955281972885132, "created_at": "2025-01-16T03:55:37.716217+00:00"} {"global_step": 7921, "acc_step": 0, "speed/wps": 12888.440921012061, "speed/FLOPS": 202430675238255.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04272070154547691, "optim/lr": 0.002979184163642129, "optim/total_tokens": 4152885248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 2.8091626167297363, "created_at": "2025-01-16T03:55:47.892150+00:00"} {"global_step": 7922, "acc_step": 0, "speed/wps": 12885.743685148524, "speed/FLOPS": 202388311442628.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05454982444643974, "optim/lr": 0.0029791699368744186, "optim/total_tokens": 4153409536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311884, "loss/out": 2.846052408218384, "created_at": "2025-01-16T03:55:58.077157+00:00"} {"global_step": 7923, "acc_step": 0, "speed/wps": 12888.291745113955, "speed/FLOPS": 202428332225786.47, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06101028621196747, "optim/lr": 0.0029791557052806557, "optim/total_tokens": 4153933824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.0503346920013428, "created_at": "2025-01-16T03:56:08.252062+00:00"} {"global_step": 7924, "acc_step": 0, "speed/wps": 12890.372113389696, "speed/FLOPS": 202461007268282.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047160036861896515, "optim/lr": 0.002979141468860888, "optim/total_tokens": 4154458112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.0653586387634277, "created_at": "2025-01-16T03:56:18.429930+00:00"} {"global_step": 7925, "acc_step": 0, "speed/wps": 12886.789874685714, "speed/FLOPS": 202404743286924.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05077437311410904, "optim/lr": 0.00297912722761516, "optim/total_tokens": 4154982400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.97367525100708, "created_at": "2025-01-16T03:56:28.606593+00:00"} {"global_step": 7926, "acc_step": 0, "speed/wps": 12891.073984179204, "speed/FLOPS": 202472031113503.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05443602427840233, "optim/lr": 0.0029791129815435204, "optim/total_tokens": 4155506688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.94397234916687, "created_at": "2025-01-16T03:56:38.781002+00:00"} {"global_step": 7927, "acc_step": 0, "speed/wps": 12883.045621507452, "speed/FLOPS": 202345934645617.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06203082576394081, "optim/lr": 0.002979098730646014, "optim/total_tokens": 4156030976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.890002727508545, "created_at": "2025-01-16T03:56:48.958506+00:00"} {"global_step": 7928, "acc_step": 0, "speed/wps": 12887.6969973523, "speed/FLOPS": 202418990894920.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04565584287047386, "optim/lr": 0.002979084474922688, "optim/total_tokens": 4156555264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474336, "loss/out": 3.06512451171875, "created_at": "2025-01-16T03:56:59.132732+00:00"} {"global_step": 7929, "acc_step": 0, "speed/wps": 12890.050229055047, "speed/FLOPS": 202455951632491.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05836043134331703, "optim/lr": 0.002979070214373589, "optim/total_tokens": 4157079552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 2.986490249633789, "created_at": "2025-01-16T03:57:09.307431+00:00"} {"global_step": 7930, "acc_step": 0, "speed/wps": 12888.566485675923, "speed/FLOPS": 202432647403845.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05678798258304596, "optim/lr": 0.002979055948998763, "optim/total_tokens": 4157603840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.947939872741699, "created_at": "2025-01-16T03:57:19.480677+00:00"} {"global_step": 7931, "acc_step": 0, "speed/wps": 12890.911179326664, "speed/FLOPS": 202469474039576.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04863862693309784, "optim/lr": 0.002979041678798257, "optim/total_tokens": 4158128128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.01894474029541, "created_at": "2025-01-16T03:57:29.651987+00:00"} {"global_step": 7932, "acc_step": 0, "speed/wps": 12885.814025336238, "speed/FLOPS": 202389416231933.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05079551786184311, "optim/lr": 0.002979027403772118, "optim/total_tokens": 4158652416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 3.0173516273498535, "created_at": "2025-01-16T03:57:39.827295+00:00"} {"global_step": 7933, "acc_step": 0, "speed/wps": 12893.764011480946, "speed/FLOPS": 202514281688762.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053828977048397064, "optim/lr": 0.0029790131239203917, "optim/total_tokens": 4159176704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9845242500305176, "created_at": "2025-01-16T03:57:49.996433+00:00"} {"global_step": 7934, "acc_step": 0, "speed/wps": 12881.971020621317, "speed/FLOPS": 202329056562043.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05714024230837822, "optim/lr": 0.002978998839243125, "optim/total_tokens": 4159700992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.910511016845703, "created_at": "2025-01-16T03:58:00.176589+00:00"} {"global_step": 7935, "acc_step": 0, "speed/wps": 12885.303810935899, "speed/FLOPS": 202381402613668.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06056142598390579, "optim/lr": 0.002978984549740365, "optim/total_tokens": 4160225280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.8979132175445557, "created_at": "2025-01-16T03:58:10.353404+00:00"} {"global_step": 7936, "acc_step": 0, "speed/wps": 12886.142756143703, "speed/FLOPS": 202394579400988.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053157076239585876, "optim/lr": 0.0029789702554121568, "optim/total_tokens": 4160749568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 2.960111618041992, "created_at": "2025-01-16T03:58:20.530586+00:00"} {"global_step": 7937, "acc_step": 0, "speed/wps": 12887.639663081274, "speed/FLOPS": 202418090381408.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05569097027182579, "optim/lr": 0.0029789559562585486, "optim/total_tokens": 4161273856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.0404601097106934, "created_at": "2025-01-16T03:58:30.705555+00:00"} {"global_step": 7938, "acc_step": 0, "speed/wps": 12891.447942828092, "speed/FLOPS": 202477904647957.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05605531856417656, "optim/lr": 0.002978941652279587, "optim/total_tokens": 4161798144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.878373146057129, "created_at": "2025-01-16T03:58:40.876625+00:00"} {"global_step": 7939, "acc_step": 0, "speed/wps": 12880.068593126867, "speed/FLOPS": 202299176324034.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06340118497610092, "optim/lr": 0.0029789273434753178, "optim/total_tokens": 4162322432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.12469220161438, "created_at": "2025-01-16T03:58:51.056939+00:00"} {"global_step": 7940, "acc_step": 0, "speed/wps": 12884.33012815663, "speed/FLOPS": 202366109587639.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06607358157634735, "optim/lr": 0.002978913029845788, "optim/total_tokens": 4162846720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.893768310546875, "created_at": "2025-01-16T03:59:01.233418+00:00"} {"global_step": 7941, "acc_step": 0, "speed/wps": 12889.62350916442, "speed/FLOPS": 202449249410234.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04474809020757675, "optim/lr": 0.0029788987113910444, "optim/total_tokens": 4163371008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.9281487464904785, "created_at": "2025-01-16T03:59:11.405778+00:00"} {"global_step": 7942, "acc_step": 0, "speed/wps": 12884.324329524326, "speed/FLOPS": 202366018512151.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06336632370948792, "optim/lr": 0.0029788843881111337, "optim/total_tokens": 4163895296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 2.8754401206970215, "created_at": "2025-01-16T03:59:21.583962+00:00"} {"global_step": 7943, "acc_step": 0, "speed/wps": 12877.319303629636, "speed/FLOPS": 202255994954560.6, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04835157096385956, "optim/lr": 0.0029788700600061027, "optim/total_tokens": 4164419584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.9389309883117676, "created_at": "2025-01-16T03:59:31.769264+00:00"} {"global_step": 7944, "acc_step": 0, "speed/wps": 12881.41532703093, "speed/FLOPS": 202320328630603.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06359101831912994, "optim/lr": 0.0029788557270759987, "optim/total_tokens": 4164943872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 3.038158893585205, "created_at": "2025-01-16T03:59:41.950005+00:00"} {"global_step": 7945, "acc_step": 0, "speed/wps": 12880.351177147071, "speed/FLOPS": 202303614694382.72, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.07128620892763138, "optim/lr": 0.0029788413893208675, "optim/total_tokens": 4165468160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.9777870178222656, "created_at": "2025-01-16T03:59:52.135414+00:00"} {"global_step": 7946, "acc_step": 0, "speed/wps": 12889.068106587814, "speed/FLOPS": 202440526049565.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05321868881583214, "optim/lr": 0.0029788270467407557, "optim/total_tokens": 4165992448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9376559257507324, "created_at": "2025-01-16T04:00:02.316234+00:00"} {"global_step": 7947, "acc_step": 0, "speed/wps": 12889.844091739662, "speed/FLOPS": 202452713962691.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04572049155831337, "optim/lr": 0.0029788126993357113, "optim/total_tokens": 4166516736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350126, "loss/out": 3.0809130668640137, "created_at": "2025-01-16T04:00:12.489220+00:00"} {"global_step": 7948, "acc_step": 0, "speed/wps": 12887.436753756256, "speed/FLOPS": 202414903411632.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05672001093626022, "optim/lr": 0.00297879834710578, "optim/total_tokens": 4167041024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9222073554992676, "created_at": "2025-01-16T04:00:22.663855+00:00"} {"global_step": 7949, "acc_step": 0, "speed/wps": 12889.864437424212, "speed/FLOPS": 202453033519624.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06673599779605865, "optim/lr": 0.0029787839900510095, "optim/total_tokens": 4167565312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.0510454177856445, "created_at": "2025-01-16T04:00:32.839720+00:00"} {"global_step": 7950, "acc_step": 0, "speed/wps": 12882.887245910244, "speed/FLOPS": 202343447139232.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05114490166306496, "optim/lr": 0.002978769628171446, "optim/total_tokens": 4168089600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.972991704940796, "created_at": "2025-01-16T04:00:43.017610+00:00"} {"global_step": 7951, "acc_step": 0, "speed/wps": 12887.810106045681, "speed/FLOPS": 202420767422377.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04981442540884018, "optim/lr": 0.002978755261467136, "optim/total_tokens": 4168613888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 2.922177791595459, "created_at": "2025-01-16T04:00:53.191573+00:00"} {"global_step": 7952, "acc_step": 0, "speed/wps": 12888.321225935051, "speed/FLOPS": 202428795262592.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05390501767396927, "optim/lr": 0.002978740889938128, "optim/total_tokens": 4169138176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.8303768634796143, "created_at": "2025-01-16T04:01:03.367911+00:00"} {"global_step": 7953, "acc_step": 0, "speed/wps": 12885.342101273505, "speed/FLOPS": 202382004016036.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06258628517389297, "optim/lr": 0.002978726513584467, "optim/total_tokens": 4169662464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.0286753177642822, "created_at": "2025-01-16T04:01:13.552388+00:00"} {"global_step": 7954, "acc_step": 0, "speed/wps": 12884.907433134418, "speed/FLOPS": 202375176955615.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060085248202085495, "optim/lr": 0.0029787121324062016, "optim/total_tokens": 4170186752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.1517975330352783, "created_at": "2025-01-16T04:01:23.730473+00:00"} {"global_step": 7955, "acc_step": 0, "speed/wps": 12887.226929684188, "speed/FLOPS": 202411607836252.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07510314136743546, "optim/lr": 0.0029786977464033772, "optim/total_tokens": 4170711040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296240, "loss/out": 3.0631284713745117, "created_at": "2025-01-16T04:01:33.905182+00:00"} {"global_step": 7956, "acc_step": 0, "speed/wps": 12888.40968691192, "speed/FLOPS": 202430184663792.53, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06682787835597992, "optim/lr": 0.0029786833555760415, "optim/total_tokens": 4171235328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.0440502166748047, "created_at": "2025-01-16T04:01:44.079588+00:00"} {"global_step": 7957, "acc_step": 0, "speed/wps": 12886.46969016365, "speed/FLOPS": 202399714349026.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052170466631650925, "optim/lr": 0.002978668959924241, "optim/total_tokens": 4171759616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.969021797180176, "created_at": "2025-01-16T04:01:54.260297+00:00"} {"global_step": 7958, "acc_step": 0, "speed/wps": 12893.506571543734, "speed/FLOPS": 202510238240788.6, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07387188822031021, "optim/lr": 0.0029786545594480238, "optim/total_tokens": 4172283904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.9663050174713135, "created_at": "2025-01-16T04:02:04.432679+00:00"} {"global_step": 7959, "acc_step": 0, "speed/wps": 12891.764600612705, "speed/FLOPS": 202482878193597.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06026211008429527, "optim/lr": 0.0029786401541474355, "optim/total_tokens": 4172808192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 2.901960849761963, "created_at": "2025-01-16T04:02:14.606228+00:00"} {"global_step": 7960, "acc_step": 0, "speed/wps": 12888.072156307451, "speed/FLOPS": 202424883281833.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06533486396074295, "optim/lr": 0.002978625744022524, "optim/total_tokens": 4173332480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.0267348289489746, "created_at": "2025-01-16T04:02:24.779898+00:00"} {"global_step": 7961, "acc_step": 0, "speed/wps": 12887.103536452301, "speed/FLOPS": 202409669775985.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10990568995475769, "optim/lr": 0.002978611329073336, "optim/total_tokens": 4173856768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.030649423599243, "created_at": "2025-01-16T04:02:34.956926+00:00"} {"global_step": 7962, "acc_step": 0, "speed/wps": 12885.709671542532, "speed/FLOPS": 202387777212206.38, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10449954122304916, "optim/lr": 0.0029785969092999184, "optim/total_tokens": 4174381056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312376, "loss/out": 2.9472436904907227, "created_at": "2025-01-16T04:02:45.135508+00:00"} {"global_step": 7963, "acc_step": 0, "speed/wps": 12885.08380487289, "speed/FLOPS": 202377947116128.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06868947297334671, "optim/lr": 0.0029785824847023187, "optim/total_tokens": 4174905344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.992393970489502, "created_at": "2025-01-16T04:02:55.313585+00:00"} {"global_step": 7964, "acc_step": 0, "speed/wps": 12885.955555141256, "speed/FLOPS": 202391639152003.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06757737696170807, "optim/lr": 0.0029785680552805835, "optim/total_tokens": 4175429632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.0219292640686035, "created_at": "2025-01-16T04:03:05.491638+00:00"} {"global_step": 7965, "acc_step": 0, "speed/wps": 12887.013558178953, "speed/FLOPS": 202408256543559.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09295567125082016, "optim/lr": 0.00297855362103476, "optim/total_tokens": 4175953920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 3.0090436935424805, "created_at": "2025-01-16T04:03:15.670219+00:00"} {"global_step": 7966, "acc_step": 0, "speed/wps": 12890.08269066357, "speed/FLOPS": 202456461486660.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05870741233229637, "optim/lr": 0.0029785391819648955, "optim/total_tokens": 4176478208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 3.0208864212036133, "created_at": "2025-01-16T04:03:25.843964+00:00"} {"global_step": 7967, "acc_step": 0, "speed/wps": 12884.533004543577, "speed/FLOPS": 202369296040077.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06294694542884827, "optim/lr": 0.002978524738071037, "optim/total_tokens": 4177002496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.86808443069458, "created_at": "2025-01-16T04:03:36.022063+00:00"} {"global_step": 7968, "acc_step": 0, "speed/wps": 12891.239499557116, "speed/FLOPS": 202474630759955.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05621781945228577, "optim/lr": 0.0029785102893532312, "optim/total_tokens": 4177526784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.986814498901367, "created_at": "2025-01-16T04:03:46.196187+00:00"} {"global_step": 7969, "acc_step": 0, "speed/wps": 12889.05371975484, "speed/FLOPS": 202440300084584.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05617699399590492, "optim/lr": 0.0029784958358115257, "optim/total_tokens": 4178051072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.055325508117676, "created_at": "2025-01-16T04:03:56.370003+00:00"} {"global_step": 7970, "acc_step": 0, "speed/wps": 12889.863775131857, "speed/FLOPS": 202453023117413.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06413998454809189, "optim/lr": 0.002978481377445968, "optim/total_tokens": 4178575360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 2.9537148475646973, "created_at": "2025-01-16T04:04:06.550073+00:00"} {"global_step": 7971, "acc_step": 0, "speed/wps": 12884.20906332795, "speed/FLOPS": 202364208098144.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05255479738116264, "optim/lr": 0.0029784669142566044, "optim/total_tokens": 4179099648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0036823749542236, "created_at": "2025-01-16T04:04:16.729154+00:00"} {"global_step": 7972, "acc_step": 0, "speed/wps": 12893.55170745547, "speed/FLOPS": 202510947162305.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052945058792829514, "optim/lr": 0.002978452446243483, "optim/total_tokens": 4179623936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.926426887512207, "created_at": "2025-01-16T04:04:26.904665+00:00"} {"global_step": 7973, "acc_step": 0, "speed/wps": 12889.826514508255, "speed/FLOPS": 202452437888118.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05514877662062645, "optim/lr": 0.00297843797340665, "optim/total_tokens": 4180148224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 2.8903539180755615, "created_at": "2025-01-16T04:04:37.078493+00:00"} {"global_step": 7974, "acc_step": 0, "speed/wps": 12884.963649553514, "speed/FLOPS": 202376059911727.2, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07021307200193405, "optim/lr": 0.0029784234957461533, "optim/total_tokens": 4180672512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.0177648067474365, "created_at": "2025-01-16T04:04:47.254663+00:00"} {"global_step": 7975, "acc_step": 0, "speed/wps": 12879.95972937506, "speed/FLOPS": 202297466469217.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04939057677984238, "optim/lr": 0.00297840901326204, "optim/total_tokens": 4181196800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9749274253845215, "created_at": "2025-01-16T04:04:57.438524+00:00"} {"global_step": 7976, "acc_step": 0, "speed/wps": 12884.36503329682, "speed/FLOPS": 202366657820830.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06366238743066788, "optim/lr": 0.0029783945259543574, "optim/total_tokens": 4181721088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9458413124084473, "created_at": "2025-01-16T04:05:07.618492+00:00"} {"global_step": 7977, "acc_step": 0, "speed/wps": 12889.342652421601, "speed/FLOPS": 202444838169151.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06024489179253578, "optim/lr": 0.0029783800338231527, "optim/total_tokens": 4182245376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.8432869911193848, "created_at": "2025-01-16T04:05:17.794271+00:00"} {"global_step": 7978, "acc_step": 0, "speed/wps": 12886.63151793366, "speed/FLOPS": 202402256076528.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04755272716283798, "optim/lr": 0.002978365536868473, "optim/total_tokens": 4182769664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.827831268310547, "created_at": "2025-01-16T04:05:27.969034+00:00"} {"global_step": 7979, "acc_step": 0, "speed/wps": 12888.154631207346, "speed/FLOPS": 202426178663468.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07219745218753815, "optim/lr": 0.002978351035090366, "optim/total_tokens": 4183293952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9484329223632812, "created_at": "2025-01-16T04:05:38.142590+00:00"} {"global_step": 7980, "acc_step": 0, "speed/wps": 12887.054166840704, "speed/FLOPS": 202408894358394.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051488421857357025, "optim/lr": 0.0029783365284888792, "optim/total_tokens": 4183818240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9794881343841553, "created_at": "2025-01-16T04:05:48.321497+00:00"} {"global_step": 7981, "acc_step": 0, "speed/wps": 12882.958554955798, "speed/FLOPS": 202344567145781.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051047518849372864, "optim/lr": 0.002978322017064059, "optim/total_tokens": 4184342528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.9256880283355713, "created_at": "2025-01-16T04:05:58.500056+00:00"} {"global_step": 7982, "acc_step": 0, "speed/wps": 12892.55144189878, "speed/FLOPS": 202495236617228.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05578399449586868, "optim/lr": 0.0029783075008159533, "optim/total_tokens": 4184866816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9194159507751465, "created_at": "2025-01-16T04:06:08.671520+00:00"} {"global_step": 7983, "acc_step": 0, "speed/wps": 12881.962391889847, "speed/FLOPS": 202328921035959.1, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05529899150133133, "optim/lr": 0.0029782929797446097, "optim/total_tokens": 4185391104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.0695295333862305, "created_at": "2025-01-16T04:06:18.852889+00:00"} {"global_step": 7984, "acc_step": 0, "speed/wps": 12890.856560489776, "speed/FLOPS": 202468616175689.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.060246389359235764, "optim/lr": 0.002978278453850075, "optim/total_tokens": 4185915392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.94357967376709, "created_at": "2025-01-16T04:06:29.027302+00:00"} {"global_step": 7985, "acc_step": 0, "speed/wps": 12889.398501466661, "speed/FLOPS": 202445715355148.3, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05435483157634735, "optim/lr": 0.0029782639231323974, "optim/total_tokens": 4186439680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 3.0722973346710205, "created_at": "2025-01-16T04:06:39.200672+00:00"} {"global_step": 7986, "acc_step": 0, "speed/wps": 12892.11191647261, "speed/FLOPS": 202488333266439.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043074384331703186, "optim/lr": 0.002978249387591624, "optim/total_tokens": 4186963968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.0081143379211426, "created_at": "2025-01-16T04:06:49.371085+00:00"} {"global_step": 7987, "acc_step": 0, "speed/wps": 12881.747231523006, "speed/FLOPS": 202325541646736.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053837914019823074, "optim/lr": 0.0029782348472278015, "optim/total_tokens": 4187488256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8216757774353027, "created_at": "2025-01-16T04:06:59.552645+00:00"} {"global_step": 7988, "acc_step": 0, "speed/wps": 12886.32577390799, "speed/FLOPS": 202397453946469.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04754573106765747, "optim/lr": 0.002978220302040978, "optim/total_tokens": 4188012544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475206, "loss/out": 2.9969592094421387, "created_at": "2025-01-16T04:07:09.738372+00:00"} {"global_step": 7989, "acc_step": 0, "speed/wps": 12884.5995226085, "speed/FLOPS": 202370340797692.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06146464869379997, "optim/lr": 0.0029782057520312006, "optim/total_tokens": 4188536832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.9947152137756348, "created_at": "2025-01-16T04:07:19.919811+00:00"} {"global_step": 7990, "acc_step": 0, "speed/wps": 12886.34054992203, "speed/FLOPS": 202397686024074.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057932425290346146, "optim/lr": 0.0029781911971985173, "optim/total_tokens": 4189061120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.948577880859375, "created_at": "2025-01-16T04:07:30.096178+00:00"} {"global_step": 7991, "acc_step": 0, "speed/wps": 12878.847187344458, "speed/FLOPS": 202279992467834.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0550105944275856, "optim/lr": 0.0029781766375429754, "optim/total_tokens": 4189585408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.868736743927002, "created_at": "2025-01-16T04:07:40.278138+00:00"} {"global_step": 7992, "acc_step": 0, "speed/wps": 12886.962563181616, "speed/FLOPS": 202407455597051.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05114319920539856, "optim/lr": 0.002978162073064622, "optim/total_tokens": 4190109696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 2.9619393348693848, "created_at": "2025-01-16T04:07:50.456555+00:00"} {"global_step": 7993, "acc_step": 0, "speed/wps": 12882.690928407592, "speed/FLOPS": 202340363703084.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06623471528291702, "optim/lr": 0.0029781475037635053, "optim/total_tokens": 4190633984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 2.9708962440490723, "created_at": "2025-01-16T04:08:00.634451+00:00"} {"global_step": 7994, "acc_step": 0, "speed/wps": 12887.425744499007, "speed/FLOPS": 202414730496119.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047253869473934174, "optim/lr": 0.002978132929639672, "optim/total_tokens": 4191158272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.9981908798217773, "created_at": "2025-01-16T04:08:10.809503+00:00"} {"global_step": 7995, "acc_step": 0, "speed/wps": 12887.025896633486, "speed/FLOPS": 202408450335942.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06624572724103928, "optim/lr": 0.0029781183506931703, "optim/total_tokens": 4191682560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9614410400390625, "created_at": "2025-01-16T04:08:20.988438+00:00"} {"global_step": 7996, "acc_step": 0, "speed/wps": 12889.054720741195, "speed/FLOPS": 202440315806450.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05811941251158714, "optim/lr": 0.002978103766924048, "optim/total_tokens": 4192206848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.942498207092285, "created_at": "2025-01-16T04:08:31.163987+00:00"} {"global_step": 7997, "acc_step": 0, "speed/wps": 12890.628971241997, "speed/FLOPS": 202465041573812.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060167912393808365, "optim/lr": 0.002978089178332352, "optim/total_tokens": 4192731136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354159, "loss/out": 2.9777581691741943, "created_at": "2025-01-16T04:08:41.338562+00:00"} {"global_step": 7998, "acc_step": 0, "speed/wps": 12886.268224116422, "speed/FLOPS": 202396550047910.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.070921890437603, "optim/lr": 0.0029780745849181302, "optim/total_tokens": 4193255424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9167957305908203, "created_at": "2025-01-16T04:08:51.515020+00:00"} {"global_step": 7999, "acc_step": 0, "speed/wps": 12889.743469260693, "speed/FLOPS": 202451133548389.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05809992551803589, "optim/lr": 0.00297805998668143, "optim/total_tokens": 4193779712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.820807933807373, "created_at": "2025-01-16T04:09:01.687218+00:00"} {"global_step": 8000, "acc_step": 0, "speed/wps": 12886.719599362601, "speed/FLOPS": 202403639516406.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0591542087495327, "optim/lr": 0.0029780453836222996, "optim/total_tokens": 4194304000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389608, "loss/out": 2.90338134765625, "created_at": "2025-01-16T04:09:11.865186+00:00"} {"global_step": 8001, "acc_step": 0, "speed/wps": 12581.220618509245, "speed/FLOPS": 197605358222513.3, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.2524, "optim/grad_norm": 0.05918502062559128, "optim/lr": 0.0029780307757407854, "optim/total_tokens": 4194828288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8902926445007324, "created_at": "2025-01-16T04:09:22.289278+00:00"} {"global_step": 8002, "acc_step": 0, "speed/wps": 12893.325992077553, "speed/FLOPS": 202507401992130.03, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04553603380918503, "optim/lr": 0.002978016163036937, "optim/total_tokens": 4195352576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 2.9463982582092285, "created_at": "2025-01-16T04:09:32.458640+00:00"} {"global_step": 8003, "acc_step": 0, "speed/wps": 12891.08199930807, "speed/FLOPS": 202472157002116.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057914186269044876, "optim/lr": 0.0029780015455108007, "optim/total_tokens": 4195876864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.927187919616699, "created_at": "2025-01-16T04:09:42.630901+00:00"} {"global_step": 8004, "acc_step": 0, "speed/wps": 12885.36218972354, "speed/FLOPS": 202382319532749.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06623198091983795, "optim/lr": 0.002977986923162424, "optim/total_tokens": 4196401152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.0165345668792725, "created_at": "2025-01-16T04:09:52.806591+00:00"} {"global_step": 8005, "acc_step": 0, "speed/wps": 12889.280794187087, "speed/FLOPS": 202443866600575.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0530836246907711, "optim/lr": 0.0029779722959918553, "optim/total_tokens": 4196925440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.004256248474121, "created_at": "2025-01-16T04:10:02.979205+00:00"} {"global_step": 8006, "acc_step": 0, "speed/wps": 12884.77737106546, "speed/FLOPS": 202373134152098.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05835231393575668, "optim/lr": 0.002977957663999142, "optim/total_tokens": 4197449728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293649, "loss/out": 2.895421266555786, "created_at": "2025-01-16T04:10:13.160436+00:00"} {"global_step": 8007, "acc_step": 0, "speed/wps": 12886.918312382897, "speed/FLOPS": 202406760577450.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05982515588402748, "optim/lr": 0.0029779430271843323, "optim/total_tokens": 4197974016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9251708984375, "created_at": "2025-01-16T04:10:23.335653+00:00"} {"global_step": 8008, "acc_step": 0, "speed/wps": 12889.139484074272, "speed/FLOPS": 202441647131073.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05846722796559334, "optim/lr": 0.0029779283855474736, "optim/total_tokens": 4198498304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.012866258621216, "created_at": "2025-01-16T04:10:33.511487+00:00"} {"global_step": 8009, "acc_step": 0, "speed/wps": 12888.947560604654, "speed/FLOPS": 202438632709252.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06287413090467453, "optim/lr": 0.0029779137390886137, "optim/total_tokens": 4199022592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399248, "loss/out": 2.9548234939575195, "created_at": "2025-01-16T04:10:43.689263+00:00"} {"global_step": 8010, "acc_step": 0, "speed/wps": 12890.82345974928, "speed/FLOPS": 202468096283074.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04739363491535187, "optim/lr": 0.0029778990878078, "optim/total_tokens": 4199546880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 2.994093418121338, "created_at": "2025-01-16T04:10:53.860641+00:00"} {"global_step": 8011, "acc_step": 0, "speed/wps": 12884.831099678506, "speed/FLOPS": 202373978033797.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059897277504205704, "optim/lr": 0.002977884431705081, "optim/total_tokens": 4200071168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.9661378860473633, "created_at": "2025-01-16T04:11:04.039503+00:00"} {"global_step": 8012, "acc_step": 0, "speed/wps": 12888.235395787815, "speed/FLOPS": 202427447182187.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05367181450128555, "optim/lr": 0.002977869770780504, "optim/total_tokens": 4200595456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.866286039352417, "created_at": "2025-01-16T04:11:14.214159+00:00"} {"global_step": 8013, "acc_step": 0, "speed/wps": 12888.475088314295, "speed/FLOPS": 202431211882687.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05125727877020836, "optim/lr": 0.002977855105034117, "optim/total_tokens": 4201119744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 2.985900402069092, "created_at": "2025-01-16T04:11:24.387415+00:00"} {"global_step": 8014, "acc_step": 0, "speed/wps": 12890.754477001603, "speed/FLOPS": 202467012814230.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0759555920958519, "optim/lr": 0.002977840434465968, "optim/total_tokens": 4201644032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.025178909301758, "created_at": "2025-01-16T04:11:34.562795+00:00"} {"global_step": 8015, "acc_step": 0, "speed/wps": 12884.890350659674, "speed/FLOPS": 202374908651876.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050040990114212036, "optim/lr": 0.002977825759076105, "optim/total_tokens": 4202168320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.915393352508545, "created_at": "2025-01-16T04:11:44.740706+00:00"} {"global_step": 8016, "acc_step": 0, "speed/wps": 12890.508739637373, "speed/FLOPS": 202463153171246.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07288876920938492, "optim/lr": 0.002977811078864575, "optim/total_tokens": 4202692608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.012113094329834, "created_at": "2025-01-16T04:11:54.912617+00:00"} {"global_step": 8017, "acc_step": 0, "speed/wps": 12890.140781011487, "speed/FLOPS": 202457373875398.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07350371032953262, "optim/lr": 0.0029777963938314262, "optim/total_tokens": 4203216896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 3.0690691471099854, "created_at": "2025-01-16T04:12:05.084557+00:00"} {"global_step": 8018, "acc_step": 0, "speed/wps": 12888.772758002648, "speed/FLOPS": 202435887194184.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05728453770279884, "optim/lr": 0.0029777817039767077, "optim/total_tokens": 4203741184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.998680591583252, "created_at": "2025-01-16T04:12:15.259936+00:00"} {"global_step": 8019, "acc_step": 0, "speed/wps": 12893.462770722024, "speed/FLOPS": 202509550288695.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04921279847621918, "optim/lr": 0.0029777670093004667, "optim/total_tokens": 4204265472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9807534217834473, "created_at": "2025-01-16T04:12:25.432071+00:00"} {"global_step": 8020, "acc_step": 0, "speed/wps": 12885.12351306816, "speed/FLOPS": 202378570787900.2, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048127658665180206, "optim/lr": 0.00297775230980275, "optim/total_tokens": 4204789760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 2.8856232166290283, "created_at": "2025-01-16T04:12:35.608172+00:00"} {"global_step": 8021, "acc_step": 0, "speed/wps": 12887.669777022727, "speed/FLOPS": 202418563362240.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04994292929768562, "optim/lr": 0.0029777376054836075, "optim/total_tokens": 4205314048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.969339370727539, "created_at": "2025-01-16T04:12:45.782114+00:00"} {"global_step": 8022, "acc_step": 0, "speed/wps": 12890.001228816936, "speed/FLOPS": 202455182016418.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05417352542281151, "optim/lr": 0.0029777228963430856, "optim/total_tokens": 4205838336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.019207715988159, "created_at": "2025-01-16T04:12:55.954773+00:00"} {"global_step": 8023, "acc_step": 0, "speed/wps": 12887.29678924977, "speed/FLOPS": 202412705076726.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04753135144710541, "optim/lr": 0.0029777081823812333, "optim/total_tokens": 4206362624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 2.980832815170288, "created_at": "2025-01-16T04:13:06.129730+00:00"} {"global_step": 8024, "acc_step": 0, "speed/wps": 12889.686904189417, "speed/FLOPS": 202450245116216.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045373231172561646, "optim/lr": 0.0029776934635980983, "optim/total_tokens": 4206886912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.903555393218994, "created_at": "2025-01-16T04:13:16.303942+00:00"} {"global_step": 8025, "acc_step": 0, "speed/wps": 12894.58327869725, "speed/FLOPS": 202527149406185.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04907368868589401, "optim/lr": 0.0029776787399937283, "optim/total_tokens": 4207411200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.0380842685699463, "created_at": "2025-01-16T04:13:26.472530+00:00"} {"global_step": 8026, "acc_step": 0, "speed/wps": 12889.446707381048, "speed/FLOPS": 202446472495275.62, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0475785993039608, "optim/lr": 0.0029776640115681714, "optim/total_tokens": 4207935488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333415, "loss/out": 2.9502131938934326, "created_at": "2025-01-16T04:13:36.644966+00:00"} {"global_step": 8027, "acc_step": 0, "speed/wps": 12891.36397335035, "speed/FLOPS": 202476585791923.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05772055685520172, "optim/lr": 0.0029776492783214763, "optim/total_tokens": 4208459776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.016134738922119, "created_at": "2025-01-16T04:13:46.819497+00:00"} {"global_step": 8028, "acc_step": 0, "speed/wps": 12893.079318048955, "speed/FLOPS": 202503527637545.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.057608433067798615, "optim/lr": 0.0029776345402536906, "optim/total_tokens": 4208984064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9521055221557617, "created_at": "2025-01-16T04:13:56.996106+00:00"} {"global_step": 8029, "acc_step": 0, "speed/wps": 12892.340778071579, "speed/FLOPS": 202491927852340.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06006091088056564, "optim/lr": 0.002977619797364862, "optim/total_tokens": 4209508352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.9733052253723145, "created_at": "2025-01-16T04:14:07.170829+00:00"} {"global_step": 8030, "acc_step": 0, "speed/wps": 12888.842864500743, "speed/FLOPS": 202436988313073.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07010426372289658, "optim/lr": 0.0029776050496550396, "optim/total_tokens": 4210032640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 2.8832805156707764, "created_at": "2025-01-16T04:14:17.346439+00:00"} {"global_step": 8031, "acc_step": 0, "speed/wps": 12891.925050982232, "speed/FLOPS": 202485398287133.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08492174744606018, "optim/lr": 0.0029775902971242705, "optim/total_tokens": 4210556928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9001097679138184, "created_at": "2025-01-16T04:14:27.519045+00:00"} {"global_step": 8032, "acc_step": 0, "speed/wps": 12891.438223284398, "speed/FLOPS": 202477751989167.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08230366557836533, "optim/lr": 0.0029775755397726035, "optim/total_tokens": 4211081216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.1344799995422363, "created_at": "2025-01-16T04:14:37.694795+00:00"} {"global_step": 8033, "acc_step": 0, "speed/wps": 12887.28316634492, "speed/FLOPS": 202412491110286.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05131553113460541, "optim/lr": 0.002977560777600086, "optim/total_tokens": 4211605504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385871, "loss/out": 2.910956621170044, "created_at": "2025-01-16T04:14:47.872308+00:00"} {"global_step": 8034, "acc_step": 0, "speed/wps": 12894.39109029427, "speed/FLOPS": 202524130823221.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07787119597196579, "optim/lr": 0.002977546010606767, "optim/total_tokens": 4212129792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 2.8295555114746094, "created_at": "2025-01-16T04:14:58.041233+00:00"} {"global_step": 8035, "acc_step": 0, "speed/wps": 12892.507306245345, "speed/FLOPS": 202494543406142.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07222040742635727, "optim/lr": 0.0029775312387926947, "optim/total_tokens": 4212654080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 2.9946775436401367, "created_at": "2025-01-16T04:15:08.216751+00:00"} {"global_step": 8036, "acc_step": 0, "speed/wps": 12889.501734525978, "speed/FLOPS": 202447336772201.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07825587689876556, "optim/lr": 0.0029775164621579165, "optim/total_tokens": 4213178368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.0261874198913574, "created_at": "2025-01-16T04:15:18.394904+00:00"} {"global_step": 8037, "acc_step": 0, "speed/wps": 12893.966421945212, "speed/FLOPS": 202517460823244.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0686793103814125, "optim/lr": 0.002977501680702481, "optim/total_tokens": 4213702656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370586, "loss/out": 2.9669113159179688, "created_at": "2025-01-16T04:15:28.566503+00:00"} {"global_step": 8038, "acc_step": 0, "speed/wps": 12886.043994265417, "speed/FLOPS": 202393028209976.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1087268814444542, "optim/lr": 0.0029774868944264367, "optim/total_tokens": 4214226944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 2.954085350036621, "created_at": "2025-01-16T04:15:38.742964+00:00"} {"global_step": 8039, "acc_step": 0, "speed/wps": 12886.71796272936, "speed/FLOPS": 202403613810832.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06704792380332947, "optim/lr": 0.0029774721033298318, "optim/total_tokens": 4214751232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.9476704597473145, "created_at": "2025-01-16T04:15:48.917929+00:00"} {"global_step": 8040, "acc_step": 0, "speed/wps": 12889.959406838752, "speed/FLOPS": 202454525144781.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08402551710605621, "optim/lr": 0.002977457307412714, "optim/total_tokens": 4215275520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333415, "loss/out": 2.9341254234313965, "created_at": "2025-01-16T04:15:59.092216+00:00"} {"global_step": 8041, "acc_step": 0, "speed/wps": 12888.725457734825, "speed/FLOPS": 202435144278480.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06801827996969223, "optim/lr": 0.002977442506675132, "optim/total_tokens": 4215799808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9349069595336914, "created_at": "2025-01-16T04:16:09.269407+00:00"} {"global_step": 8042, "acc_step": 0, "speed/wps": 12887.591267698264, "speed/FLOPS": 202417330265416.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07260281592607498, "optim/lr": 0.002977427701117134, "optim/total_tokens": 4216324096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.0275583267211914, "created_at": "2025-01-16T04:16:19.447565+00:00"} {"global_step": 8043, "acc_step": 0, "speed/wps": 12894.537919364288, "speed/FLOPS": 202526436975530.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0761716365814209, "optim/lr": 0.002977412890738769, "optim/total_tokens": 4216848384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9069693088531494, "created_at": "2025-01-16T04:16:29.616182+00:00"} {"global_step": 8044, "acc_step": 0, "speed/wps": 12886.410562707624, "speed/FLOPS": 202398785671079.53, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06568768620491028, "optim/lr": 0.0029773980755400838, "optim/total_tokens": 4217372672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402512, "loss/out": 2.9045257568359375, "created_at": "2025-01-16T04:16:39.792005+00:00"} {"global_step": 8045, "acc_step": 0, "speed/wps": 12888.732798324587, "speed/FLOPS": 202435259572530.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0775344967842102, "optim/lr": 0.002977383255521128, "optim/total_tokens": 4217896960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.0388872623443604, "created_at": "2025-01-16T04:16:49.967018+00:00"} {"global_step": 8046, "acc_step": 0, "speed/wps": 12890.021841809907, "speed/FLOPS": 202455505771798.38, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057602833956480026, "optim/lr": 0.0029773684306819495, "optim/total_tokens": 4218421248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.936615467071533, "created_at": "2025-01-16T04:17:00.141251+00:00"} {"global_step": 8047, "acc_step": 0, "speed/wps": 12880.843645894542, "speed/FLOPS": 202311349592784.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07074062526226044, "optim/lr": 0.0029773536010225966, "optim/total_tokens": 4218945536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.0878801345825195, "created_at": "2025-01-16T04:17:10.320512+00:00"} {"global_step": 8048, "acc_step": 0, "speed/wps": 12883.984434519654, "speed/FLOPS": 202360679994037.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05382803454995155, "optim/lr": 0.002977338766543118, "optim/total_tokens": 4219469824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 3.1239914894104004, "created_at": "2025-01-16T04:17:20.499495+00:00"} {"global_step": 8049, "acc_step": 0, "speed/wps": 12890.38965432825, "speed/FLOPS": 202461282772826.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06166291981935501, "optim/lr": 0.002977323927243562, "optim/total_tokens": 4219994112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.912886142730713, "created_at": "2025-01-16T04:17:30.671527+00:00"} {"global_step": 8050, "acc_step": 0, "speed/wps": 12895.895432635338, "speed/FLOPS": 202547758586871.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06138252094388008, "optim/lr": 0.0029773090831239765, "optim/total_tokens": 4220518400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.8613617420196533, "created_at": "2025-01-16T04:17:40.842654+00:00"} {"global_step": 8051, "acc_step": 0, "speed/wps": 12892.477088201025, "speed/FLOPS": 202494068790233.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05113859102129936, "optim/lr": 0.002977294234184411, "optim/total_tokens": 4221042688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9469141960144043, "created_at": "2025-01-16T04:17:51.012838+00:00"} {"global_step": 8052, "acc_step": 0, "speed/wps": 12890.69658184281, "speed/FLOPS": 202466103491205.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06293980032205582, "optim/lr": 0.002977279380424913, "optim/total_tokens": 4221566976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8839330673217773, "created_at": "2025-01-16T04:18:01.186040+00:00"} {"global_step": 8053, "acc_step": 0, "speed/wps": 12894.13729809985, "speed/FLOPS": 202520144668061.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06704330444335938, "optim/lr": 0.0029772645218455315, "optim/total_tokens": 4222091264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9966399669647217, "created_at": "2025-01-16T04:18:11.360274+00:00"} {"global_step": 8054, "acc_step": 0, "speed/wps": 12878.019991703823, "speed/FLOPS": 202267000223612.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08322004228830338, "optim/lr": 0.0029772496584463144, "optim/total_tokens": 4222615552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.0213489532470703, "created_at": "2025-01-16T04:18:21.544863+00:00"} {"global_step": 8055, "acc_step": 0, "speed/wps": 12889.968130107723, "speed/FLOPS": 202454662155707.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05046848952770233, "optim/lr": 0.0029772347902273107, "optim/total_tokens": 4223139840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 2.9655635356903076, "created_at": "2025-01-16T04:18:31.717062+00:00"} {"global_step": 8056, "acc_step": 0, "speed/wps": 12887.098813889255, "speed/FLOPS": 202409595601643.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05610596388578415, "optim/lr": 0.0029772199171885687, "optim/total_tokens": 4223664128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.915585517883301, "created_at": "2025-01-16T04:18:41.891807+00:00"} {"global_step": 8057, "acc_step": 0, "speed/wps": 12891.14016478758, "speed/FLOPS": 202473070570899.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04791396111249924, "optim/lr": 0.0029772050393301374, "optim/total_tokens": 4224188416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.948134422302246, "created_at": "2025-01-16T04:18:52.064898+00:00"} {"global_step": 8058, "acc_step": 0, "speed/wps": 12889.041807084826, "speed/FLOPS": 202440112979732.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051773011684417725, "optim/lr": 0.0029771901566520644, "optim/total_tokens": 4224712704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9676527976989746, "created_at": "2025-01-16T04:19:02.238004+00:00"} {"global_step": 8059, "acc_step": 0, "speed/wps": 12885.303228153456, "speed/FLOPS": 202381393460269.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07058105617761612, "optim/lr": 0.0029771752691543987, "optim/total_tokens": 4225236992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.903592109680176, "created_at": "2025-01-16T04:19:12.414055+00:00"} {"global_step": 8060, "acc_step": 0, "speed/wps": 12888.896296518578, "speed/FLOPS": 202437827536336.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06863314658403397, "optim/lr": 0.0029771603768371895, "optim/total_tokens": 4225761280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.9411063194274902, "created_at": "2025-01-16T04:19:22.587602+00:00"} {"global_step": 8061, "acc_step": 0, "speed/wps": 12886.788641592388, "speed/FLOPS": 202404723919499.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0659898892045021, "optim/lr": 0.0029771454797004847, "optim/total_tokens": 4226285568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.928253173828125, "created_at": "2025-01-16T04:19:32.764879+00:00"} {"global_step": 8062, "acc_step": 0, "speed/wps": 12886.542601125295, "speed/FLOPS": 202400859515868.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05496116355061531, "optim/lr": 0.0029771305777443326, "optim/total_tokens": 4226809856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 2.7900924682617188, "created_at": "2025-01-16T04:19:42.943365+00:00"} {"global_step": 8063, "acc_step": 0, "speed/wps": 12887.101682592182, "speed/FLOPS": 202409640658564.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06865816563367844, "optim/lr": 0.002977115670968783, "optim/total_tokens": 4227334144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.013247013092041, "created_at": "2025-01-16T04:19:53.117873+00:00"} {"global_step": 8064, "acc_step": 0, "speed/wps": 12893.755078713855, "speed/FLOPS": 202514141387379.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07089436799287796, "optim/lr": 0.002977100759373883, "optim/total_tokens": 4227858432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.0089194774627686, "created_at": "2025-01-16T04:20:03.289424+00:00"} {"global_step": 8065, "acc_step": 0, "speed/wps": 12887.804277136478, "speed/FLOPS": 202420675871349.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.056051623076200485, "optim/lr": 0.0029770858429596825, "optim/total_tokens": 4228382720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0093650817871094, "created_at": "2025-01-16T04:20:13.463651+00:00"} {"global_step": 8066, "acc_step": 0, "speed/wps": 12885.896572802956, "speed/FLOPS": 202390712753330.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06041325256228447, "optim/lr": 0.0029770709217262297, "optim/total_tokens": 4228907008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 2.9549012184143066, "created_at": "2025-01-16T04:20:23.642693+00:00"} {"global_step": 8067, "acc_step": 0, "speed/wps": 12888.400961031015, "speed/FLOPS": 202430047611842.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05187074840068817, "optim/lr": 0.0029770559956735733, "optim/total_tokens": 4229431296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.9573354721069336, "created_at": "2025-01-16T04:20:33.818472+00:00"} {"global_step": 8068, "acc_step": 0, "speed/wps": 12882.15187644653, "speed/FLOPS": 202331897151300.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059501245617866516, "optim/lr": 0.0029770410648017618, "optim/total_tokens": 4229955584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.958446979522705, "created_at": "2025-01-16T04:20:43.996984+00:00"} {"global_step": 8069, "acc_step": 0, "speed/wps": 12893.566880403, "speed/FLOPS": 202511185474296.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06161866337060928, "optim/lr": 0.002977026129110844, "optim/total_tokens": 4230479872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410502, "loss/out": 2.9662182331085205, "created_at": "2025-01-16T04:20:54.167363+00:00"} {"global_step": 8070, "acc_step": 0, "speed/wps": 12895.377124593078, "speed/FLOPS": 202539617846837.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05374644324183464, "optim/lr": 0.002977011188600869, "optim/total_tokens": 4231004160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9942731857299805, "created_at": "2025-01-16T04:21:04.339470+00:00"} {"global_step": 8071, "acc_step": 0, "speed/wps": 12890.183684429563, "speed/FLOPS": 202458047732535.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06262059509754181, "optim/lr": 0.002976996243271886, "optim/total_tokens": 4231528448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.967189311981201, "created_at": "2025-01-16T04:21:14.515948+00:00"} {"global_step": 8072, "acc_step": 0, "speed/wps": 12888.460296782056, "speed/FLOPS": 202430979561348.25, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05573922395706177, "optim/lr": 0.002976981293123942, "optim/total_tokens": 4232052736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.865082263946533, "created_at": "2025-01-16T04:21:24.692394+00:00"} {"global_step": 8073, "acc_step": 0, "speed/wps": 12887.474026527572, "speed/FLOPS": 202415488831724.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054287731647491455, "optim/lr": 0.0029769663381570876, "optim/total_tokens": 4232577024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9305543899536133, "created_at": "2025-01-16T04:21:34.866787+00:00"} {"global_step": 8074, "acc_step": 0, "speed/wps": 12890.983168789513, "speed/FLOPS": 202470604733014.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05274740606546402, "optim/lr": 0.0029769513783713705, "optim/total_tokens": 4233101312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9386355876922607, "created_at": "2025-01-16T04:21:45.038377+00:00"} {"global_step": 8075, "acc_step": 0, "speed/wps": 12889.255623502215, "speed/FLOPS": 202443471260381.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050346724689006805, "optim/lr": 0.0029769364137668397, "optim/total_tokens": 4233625600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.966724395751953, "created_at": "2025-01-16T04:21:55.213101+00:00"} {"global_step": 8076, "acc_step": 0, "speed/wps": 12889.026056944633, "speed/FLOPS": 202439865602137.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05019843205809593, "optim/lr": 0.0029769214443435446, "optim/total_tokens": 4234149888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.9637579917907715, "created_at": "2025-01-16T04:22:05.387177+00:00"} {"global_step": 8077, "acc_step": 0, "speed/wps": 12890.458794922597, "speed/FLOPS": 202462368720869.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05338171124458313, "optim/lr": 0.0029769064701015333, "optim/total_tokens": 4234674176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.0104241371154785, "created_at": "2025-01-16T04:22:15.559615+00:00"} {"global_step": 8078, "acc_step": 0, "speed/wps": 12888.842227101122, "speed/FLOPS": 202436978301836.44, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05890018865466118, "optim/lr": 0.002976891491040855, "optim/total_tokens": 4235198464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.920743942260742, "created_at": "2025-01-16T04:22:25.735236+00:00"} {"global_step": 8079, "acc_step": 0, "speed/wps": 12893.174639582625, "speed/FLOPS": 202505024793217.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05788249522447586, "optim/lr": 0.0029768765071615582, "optim/total_tokens": 4235722752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377784, "loss/out": 3.0380311012268066, "created_at": "2025-01-16T04:22:35.906928+00:00"} {"global_step": 8080, "acc_step": 0, "speed/wps": 12890.819010750674, "speed/FLOPS": 202468026405438.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04788888618350029, "optim/lr": 0.002976861518463693, "optim/total_tokens": 4236247040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.895231246948242, "created_at": "2025-01-16T04:22:46.083690+00:00"} {"global_step": 8081, "acc_step": 0, "speed/wps": 12892.45996202845, "speed/FLOPS": 202493799800158.88, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05890854448080063, "optim/lr": 0.0029768465249473065, "optim/total_tokens": 4236771328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.984987258911133, "created_at": "2025-01-16T04:22:56.254077+00:00"} {"global_step": 8082, "acc_step": 0, "speed/wps": 12885.285001482238, "speed/FLOPS": 202381107185351.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07157136499881744, "optim/lr": 0.0029768315266124486, "optim/total_tokens": 4237295616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.0126209259033203, "created_at": "2025-01-16T04:23:06.432234+00:00"} {"global_step": 8083, "acc_step": 0, "speed/wps": 12890.996408684934, "speed/FLOPS": 202470812683765.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05261020362377167, "optim/lr": 0.002976816523459169, "optim/total_tokens": 4237819904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 2.910553216934204, "created_at": "2025-01-16T04:23:16.604357+00:00"} {"global_step": 8084, "acc_step": 0, "speed/wps": 12889.69110567686, "speed/FLOPS": 202450311106349.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05597425252199173, "optim/lr": 0.002976801515487515, "optim/total_tokens": 4238344192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.836249351501465, "created_at": "2025-01-16T04:23:26.777994+00:00"} {"global_step": 8085, "acc_step": 0, "speed/wps": 12894.151463559278, "speed/FLOPS": 202520367156067.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05178887024521828, "optim/lr": 0.0029767865026975367, "optim/total_tokens": 4238868480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.9193758964538574, "created_at": "2025-01-16T04:23:36.948139+00:00"} {"global_step": 8086, "acc_step": 0, "speed/wps": 12888.713114845303, "speed/FLOPS": 202434950416439.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058188389986753464, "optim/lr": 0.0029767714850892826, "optim/total_tokens": 4239392768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 2.97097110748291, "created_at": "2025-01-16T04:23:47.123630+00:00"} {"global_step": 8087, "acc_step": 0, "speed/wps": 12890.468639320883, "speed/FLOPS": 202462523340672.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06585729867219925, "optim/lr": 0.0029767564626628023, "optim/total_tokens": 4239917056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8504133224487305, "created_at": "2025-01-16T04:23:57.297957+00:00"} {"global_step": 8088, "acc_step": 0, "speed/wps": 12889.965057069547, "speed/FLOPS": 202454613889420.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06156794726848602, "optim/lr": 0.002976741435418144, "optim/total_tokens": 4240441344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.110135078430176, "created_at": "2025-01-16T04:24:07.470361+00:00"} {"global_step": 8089, "acc_step": 0, "speed/wps": 12900.86510540868, "speed/FLOPS": 202625814126822.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06462455540895462, "optim/lr": 0.002976726403355357, "optim/total_tokens": 4240965632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 3.086643695831299, "created_at": "2025-01-16T04:24:17.633961+00:00"} {"global_step": 8090, "acc_step": 0, "speed/wps": 12890.200497594322, "speed/FLOPS": 202458311806391.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06103220209479332, "optim/lr": 0.0029767113664744905, "optim/total_tokens": 4241489920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 3.048572063446045, "created_at": "2025-01-16T04:24:27.808154+00:00"} {"global_step": 8091, "acc_step": 0, "speed/wps": 12889.228129808738, "speed/FLOPS": 202443039434145.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0715639740228653, "optim/lr": 0.002976696324775594, "optim/total_tokens": 4242014208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9331748485565186, "created_at": "2025-01-16T04:24:37.981148+00:00"} {"global_step": 8092, "acc_step": 0, "speed/wps": 12889.78501949346, "speed/FLOPS": 202451786151890.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05899611487984657, "optim/lr": 0.002976681278258716, "optim/total_tokens": 4242538496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 2.992807388305664, "created_at": "2025-01-16T04:24:48.153912+00:00"} {"global_step": 8093, "acc_step": 0, "speed/wps": 12886.288795674325, "speed/FLOPS": 202396873152496.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05762438103556633, "optim/lr": 0.002976666226923906, "optim/total_tokens": 4243062784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.0092382431030273, "created_at": "2025-01-16T04:24:58.330212+00:00"} {"global_step": 8094, "acc_step": 0, "speed/wps": 12890.735421391557, "speed/FLOPS": 202466713519689.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05667854845523834, "optim/lr": 0.002976651170771212, "optim/total_tokens": 4243587072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 2.9252209663391113, "created_at": "2025-01-16T04:25:08.502513+00:00"} {"global_step": 8095, "acc_step": 0, "speed/wps": 12883.378625455467, "speed/FLOPS": 202351164930216.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0514330193400383, "optim/lr": 0.002976636109800684, "optim/total_tokens": 4244111360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.957015037536621, "created_at": "2025-01-16T04:25:18.679759+00:00"} {"global_step": 8096, "acc_step": 0, "speed/wps": 12888.231557588437, "speed/FLOPS": 202427386897991.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048939939588308334, "optim/lr": 0.0029766210440123715, "optim/total_tokens": 4244635648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415616, "loss/out": 2.9790706634521484, "created_at": "2025-01-16T04:25:28.856465+00:00"} {"global_step": 8097, "acc_step": 0, "speed/wps": 12890.15410181928, "speed/FLOPS": 202457583096989.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057558897882699966, "optim/lr": 0.002976605973406323, "optim/total_tokens": 4245159936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9672281742095947, "created_at": "2025-01-16T04:25:39.028719+00:00"} {"global_step": 8098, "acc_step": 0, "speed/wps": 12895.545748203582, "speed/FLOPS": 202542266312353.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05062579736113548, "optim/lr": 0.002976590897982588, "optim/total_tokens": 4245684224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472193, "loss/out": 3.1283063888549805, "created_at": "2025-01-16T04:25:49.201747+00:00"} {"global_step": 8099, "acc_step": 0, "speed/wps": 12889.607516778537, "speed/FLOPS": 202448998227838.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05586814880371094, "optim/lr": 0.0029765758177412154, "optim/total_tokens": 4246208512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.9537932872772217, "created_at": "2025-01-16T04:25:59.376590+00:00"} {"global_step": 8100, "acc_step": 0, "speed/wps": 12892.39976509163, "speed/FLOPS": 202492854324546.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05600481480360031, "optim/lr": 0.0029765607326822548, "optim/total_tokens": 4246732800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.912283420562744, "created_at": "2025-01-16T04:26:09.549720+00:00"} {"global_step": 8101, "acc_step": 0, "speed/wps": 12890.509790438093, "speed/FLOPS": 202463169675515.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06384790688753128, "optim/lr": 0.0029765456428057554, "optim/total_tokens": 4247257088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9886727333068848, "created_at": "2025-01-16T04:26:19.724023+00:00"} {"global_step": 8102, "acc_step": 0, "speed/wps": 12889.869754654435, "speed/FLOPS": 202453117034032.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04944846034049988, "optim/lr": 0.0029765305481117654, "optim/total_tokens": 4247781376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.0375149250030518, "created_at": "2025-01-16T04:26:29.897785+00:00"} {"global_step": 8103, "acc_step": 0, "speed/wps": 12886.750621133991, "speed/FLOPS": 202404126755954.88, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06188088282942772, "optim/lr": 0.0029765154486003356, "optim/total_tokens": 4248305664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 3.0309271812438965, "created_at": "2025-01-16T04:26:40.078024+00:00"} {"global_step": 8104, "acc_step": 0, "speed/wps": 12891.752197580108, "speed/FLOPS": 202482683386927.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051497336477041245, "optim/lr": 0.0029765003442715145, "optim/total_tokens": 4248829952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8764867782592773, "created_at": "2025-01-16T04:26:50.251142+00:00"} {"global_step": 8105, "acc_step": 0, "speed/wps": 12887.086789187884, "speed/FLOPS": 202409406737184.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047500621527433395, "optim/lr": 0.0029764852351253507, "optim/total_tokens": 4249354240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8903627395629883, "created_at": "2025-01-16T04:27:00.434260+00:00"} {"global_step": 8106, "acc_step": 0, "speed/wps": 12891.273323912128, "speed/FLOPS": 202475162017930.38, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04552780091762543, "optim/lr": 0.002976470121161895, "optim/total_tokens": 4249878528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 2.9411487579345703, "created_at": "2025-01-16T04:27:10.607599+00:00"} {"global_step": 8107, "acc_step": 0, "speed/wps": 12888.293506075219, "speed/FLOPS": 202428359884102.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04965629801154137, "optim/lr": 0.0029764550023811957, "optim/total_tokens": 4250402816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 3.0275466442108154, "created_at": "2025-01-16T04:27:20.792170+00:00"} {"global_step": 8108, "acc_step": 0, "speed/wps": 12893.946964644749, "speed/FLOPS": 202517155219603.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043743908405303955, "optim/lr": 0.002976439878783302, "optim/total_tokens": 4250927104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8630144596099854, "created_at": "2025-01-16T04:27:30.961356+00:00"} {"global_step": 8109, "acc_step": 0, "speed/wps": 12893.444711134898, "speed/FLOPS": 202509266638062.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055674802511930466, "optim/lr": 0.002976424750368264, "optim/total_tokens": 4251451392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.98457407951355, "created_at": "2025-01-16T04:27:41.133155+00:00"} {"global_step": 8110, "acc_step": 0, "speed/wps": 12890.637820571286, "speed/FLOPS": 202465180564689.38, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05842311307787895, "optim/lr": 0.002976409617136131, "optim/total_tokens": 4251975680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.110095262527466, "created_at": "2025-01-16T04:27:51.307168+00:00"} {"global_step": 8111, "acc_step": 0, "speed/wps": 12892.604395633318, "speed/FLOPS": 202496068328395.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048669494688510895, "optim/lr": 0.0029763944790869514, "optim/total_tokens": 4252499968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318692, "loss/out": 3.0030555725097656, "created_at": "2025-01-16T04:28:01.479778+00:00"} {"global_step": 8112, "acc_step": 0, "speed/wps": 12887.801137859764, "speed/FLOPS": 202420626564694.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0716114342212677, "optim/lr": 0.0029763793362207755, "optim/total_tokens": 4253024256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.0230350494384766, "created_at": "2025-01-16T04:28:11.661751+00:00"} {"global_step": 8113, "acc_step": 0, "speed/wps": 12893.798713780187, "speed/FLOPS": 202514826736057.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05476026609539986, "optim/lr": 0.002976364188537652, "optim/total_tokens": 4253548544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451288, "loss/out": 2.999361276626587, "created_at": "2025-01-16T04:28:21.831767+00:00"} {"global_step": 8114, "acc_step": 0, "speed/wps": 12889.072408006865, "speed/FLOPS": 202440593609262.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06042880564928055, "optim/lr": 0.0029763490360376313, "optim/total_tokens": 4254072832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 2.9305102825164795, "created_at": "2025-01-16T04:28:32.004885+00:00"} {"global_step": 8115, "acc_step": 0, "speed/wps": 12891.419495107648, "speed/FLOPS": 202477457837416.28, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058702580630779266, "optim/lr": 0.0029763338787207622, "optim/total_tokens": 4254597120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.9649839401245117, "created_at": "2025-01-16T04:28:42.176942+00:00"} {"global_step": 8116, "acc_step": 0, "speed/wps": 12887.108967050066, "speed/FLOPS": 202409755070985.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057543426752090454, "optim/lr": 0.0029763187165870947, "optim/total_tokens": 4255121408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9739861488342285, "created_at": "2025-01-16T04:28:52.351366+00:00"} {"global_step": 8117, "acc_step": 0, "speed/wps": 12893.966102991966, "speed/FLOPS": 202517455813645.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07172541320323944, "optim/lr": 0.0029763035496366765, "optim/total_tokens": 4255645696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 3.0326220989227295, "created_at": "2025-01-16T04:29:02.520656+00:00"} {"global_step": 8118, "acc_step": 0, "speed/wps": 12885.615748674401, "speed/FLOPS": 202386302024498.8, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047860514372587204, "optim/lr": 0.002976288377869559, "optim/total_tokens": 4256169984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.746669292449951, "created_at": "2025-01-16T04:29:12.698847+00:00"} {"global_step": 8119, "acc_step": 0, "speed/wps": 12891.083827155588, "speed/FLOPS": 202472185710973.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07643537223339081, "optim/lr": 0.0029762732012857916, "optim/total_tokens": 4256694272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9517107009887695, "created_at": "2025-01-16T04:29:22.872377+00:00"} {"global_step": 8120, "acc_step": 0, "speed/wps": 12893.014362443677, "speed/FLOPS": 202502507420506.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05056888982653618, "optim/lr": 0.0029762580198854226, "optim/total_tokens": 4257218560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 2.9042606353759766, "created_at": "2025-01-16T04:29:33.044129+00:00"} {"global_step": 8121, "acc_step": 0, "speed/wps": 12886.829565076925, "speed/FLOPS": 202405366679058.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07401765882968903, "optim/lr": 0.0029762428336685036, "optim/total_tokens": 4257742848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9237918853759766, "created_at": "2025-01-16T04:29:43.218964+00:00"} {"global_step": 8122, "acc_step": 0, "speed/wps": 12890.174905379992, "speed/FLOPS": 202457909845497.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056039128452539444, "optim/lr": 0.002976227642635081, "optim/total_tokens": 4258267136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.941859483718872, "created_at": "2025-01-16T04:29:53.395125+00:00"} {"global_step": 8123, "acc_step": 0, "speed/wps": 12890.96773979382, "speed/FLOPS": 202470362399435.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0719514861702919, "optim/lr": 0.002976212446785207, "optim/total_tokens": 4258791424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.988186836242676, "created_at": "2025-01-16T04:30:03.566631+00:00"} {"global_step": 8124, "acc_step": 0, "speed/wps": 12894.897152192814, "speed/FLOPS": 202532079220740.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06036215275526047, "optim/lr": 0.0029761972461189307, "optim/total_tokens": 4259315712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.9381322860717773, "created_at": "2025-01-16T04:30:13.734773+00:00"} {"global_step": 8125, "acc_step": 0, "speed/wps": 12891.356081857999, "speed/FLOPS": 202476461845191.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053911373019218445, "optim/lr": 0.0029761820406363007, "optim/total_tokens": 4259840000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.948234796524048, "created_at": "2025-01-16T04:30:23.905772+00:00"} {"global_step": 8126, "acc_step": 0, "speed/wps": 12889.60031567288, "speed/FLOPS": 202448885124578.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05699888989329338, "optim/lr": 0.0029761668303373676, "optim/total_tokens": 4260364288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 3.0527846813201904, "created_at": "2025-01-16T04:30:34.080872+00:00"} {"global_step": 8127, "acc_step": 0, "speed/wps": 12888.498957613192, "speed/FLOPS": 202431586782826.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06630722433328629, "optim/lr": 0.0029761516152221807, "optim/total_tokens": 4260888576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.8721184730529785, "created_at": "2025-01-16T04:30:44.255255+00:00"} {"global_step": 8128, "acc_step": 0, "speed/wps": 12893.77901718724, "speed/FLOPS": 202514517373999.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07600437104701996, "optim/lr": 0.0029761363952907896, "optim/total_tokens": 4261412864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.769169330596924, "created_at": "2025-01-16T04:30:54.424387+00:00"} {"global_step": 8129, "acc_step": 0, "speed/wps": 12891.615254476637, "speed/FLOPS": 202480532507307.84, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06450837105512619, "optim/lr": 0.0029761211705432436, "optim/total_tokens": 4261937152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.992795467376709, "created_at": "2025-01-16T04:31:04.598867+00:00"} {"global_step": 8130, "acc_step": 0, "speed/wps": 12887.07673400586, "speed/FLOPS": 202409248806733.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057971540838479996, "optim/lr": 0.0029761059409795934, "optim/total_tokens": 4262461440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.0607011318206787, "created_at": "2025-01-16T04:31:14.774455+00:00"} {"global_step": 8131, "acc_step": 0, "speed/wps": 12892.771384155712, "speed/FLOPS": 202498691112606.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.08360572159290314, "optim/lr": 0.0029760907065998874, "optim/total_tokens": 4262985728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9453017711639404, "created_at": "2025-01-16T04:31:24.947282+00:00"} {"global_step": 8132, "acc_step": 0, "speed/wps": 12892.060600049817, "speed/FLOPS": 202487527271503.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050238195806741714, "optim/lr": 0.002976075467404176, "optim/total_tokens": 4263510016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.040748119354248, "created_at": "2025-01-16T04:31:35.119007+00:00"} {"global_step": 8133, "acc_step": 0, "speed/wps": 12892.30804881083, "speed/FLOPS": 202491413794325.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06897741556167603, "optim/lr": 0.0029760602233925093, "optim/total_tokens": 4264034304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 3.0422189235687256, "created_at": "2025-01-16T04:31:45.295190+00:00"} {"global_step": 8134, "acc_step": 0, "speed/wps": 12890.154901092235, "speed/FLOPS": 202457595650669.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05892790108919144, "optim/lr": 0.0029760449745649363, "optim/total_tokens": 4264558592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.03629469871521, "created_at": "2025-01-16T04:31:55.468613+00:00"} {"global_step": 8135, "acc_step": 0, "speed/wps": 12891.660821160658, "speed/FLOPS": 202481248194695.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06091422215104103, "optim/lr": 0.0029760297209215068, "optim/total_tokens": 4265082880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 2.974120616912842, "created_at": "2025-01-16T04:32:05.643001+00:00"} {"global_step": 8136, "acc_step": 0, "speed/wps": 12889.641868524282, "speed/FLOPS": 202449537769209.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0619812086224556, "optim/lr": 0.002976014462462271, "optim/total_tokens": 4265607168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.9263575077056885, "created_at": "2025-01-16T04:32:15.815730+00:00"} {"global_step": 8137, "acc_step": 0, "speed/wps": 12890.264457913718, "speed/FLOPS": 202459316391098.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05145072191953659, "optim/lr": 0.0029759991991872787, "optim/total_tokens": 4266131456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420893, "loss/out": 2.906189441680908, "created_at": "2025-01-16T04:32:25.987931+00:00"} {"global_step": 8138, "acc_step": 0, "speed/wps": 12897.273800593903, "speed/FLOPS": 202569407749736.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08852489292621613, "optim/lr": 0.0029759839310965794, "optim/total_tokens": 4266655744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9148335456848145, "created_at": "2025-01-16T04:32:36.157620+00:00"} {"global_step": 8139, "acc_step": 0, "speed/wps": 12889.201768270064, "speed/FLOPS": 202442625389956.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07937557250261307, "optim/lr": 0.002975968658190223, "optim/total_tokens": 4267180032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9805026054382324, "created_at": "2025-01-16T04:32:46.332983+00:00"} {"global_step": 8140, "acc_step": 0, "speed/wps": 12895.794447423896, "speed/FLOPS": 202546172475357.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06871653348207474, "optim/lr": 0.0029759533804682593, "optim/total_tokens": 4267704320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361365, "loss/out": 3.079146385192871, "created_at": "2025-01-16T04:32:56.503640+00:00"} {"global_step": 8141, "acc_step": 0, "speed/wps": 12892.65104998658, "speed/FLOPS": 202496801099123.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07546042650938034, "optim/lr": 0.002975938097930738, "optim/total_tokens": 4268228608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 3.069101333618164, "created_at": "2025-01-16T04:33:06.676240+00:00"} {"global_step": 8142, "acc_step": 0, "speed/wps": 12890.248834654272, "speed/FLOPS": 202459071006340.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05888739228248596, "optim/lr": 0.0029759228105777098, "optim/total_tokens": 4268752896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.9236462116241455, "created_at": "2025-01-16T04:33:16.850611+00:00"} {"global_step": 8143, "acc_step": 0, "speed/wps": 12893.96497248003, "speed/FLOPS": 202517438057402.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051641374826431274, "optim/lr": 0.0029759075184092234, "optim/total_tokens": 4269277184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 3.051518440246582, "created_at": "2025-01-16T04:33:27.024475+00:00"} {"global_step": 8144, "acc_step": 0, "speed/wps": 12891.496807639829, "speed/FLOPS": 202478672136972.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06257261335849762, "optim/lr": 0.0029758922214253294, "optim/total_tokens": 4269801472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 2.826113224029541, "created_at": "2025-01-16T04:33:37.197155+00:00"} {"global_step": 8145, "acc_step": 0, "speed/wps": 12886.55619178579, "speed/FLOPS": 202401072975867.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.049508195370435715, "optim/lr": 0.0029758769196260772, "optim/total_tokens": 4270325760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.103963613510132, "created_at": "2025-01-16T04:33:47.371992+00:00"} {"global_step": 8146, "acc_step": 0, "speed/wps": 12888.673713837274, "speed/FLOPS": 202434331569466.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05283026769757271, "optim/lr": 0.0029758616130115176, "optim/total_tokens": 4270850048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.982219934463501, "created_at": "2025-01-16T04:33:57.545009+00:00"} {"global_step": 8147, "acc_step": 0, "speed/wps": 12896.080491806233, "speed/FLOPS": 202550665195448.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053375422954559326, "optim/lr": 0.0029758463015816996, "optim/total_tokens": 4271374336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.9475154876708984, "created_at": "2025-01-16T04:34:07.714503+00:00"} {"global_step": 8148, "acc_step": 0, "speed/wps": 12887.623549342132, "speed/FLOPS": 202417837292992.78, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04955163970589638, "optim/lr": 0.002975830985336673, "optim/total_tokens": 4271898624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.975085973739624, "created_at": "2025-01-16T04:34:17.893001+00:00"} {"global_step": 8149, "acc_step": 0, "speed/wps": 12892.818929084351, "speed/FLOPS": 202499437871044.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07351518422365189, "optim/lr": 0.0029758156642764894, "optim/total_tokens": 4272422912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.8821823596954346, "created_at": "2025-01-16T04:34:28.065319+00:00"} {"global_step": 8150, "acc_step": 0, "speed/wps": 12888.174206758593, "speed/FLOPS": 202426486124400.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05503705143928528, "optim/lr": 0.0029758003384011975, "optim/total_tokens": 4272947200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8481743335723877, "created_at": "2025-01-16T04:34:38.241415+00:00"} {"global_step": 8151, "acc_step": 0, "speed/wps": 12893.288645288576, "speed/FLOPS": 202506815409489.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07490953058004379, "optim/lr": 0.0029757850077108475, "optim/total_tokens": 4273471488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310734, "loss/out": 2.93446946144104, "created_at": "2025-01-16T04:34:48.412727+00:00"} {"global_step": 8152, "acc_step": 0, "speed/wps": 12891.710338259345, "speed/FLOPS": 202482025928774.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06001479551196098, "optim/lr": 0.002975769672205489, "optim/total_tokens": 4273995776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.8930883407592773, "created_at": "2025-01-16T04:34:58.583960+00:00"} {"global_step": 8153, "acc_step": 0, "speed/wps": 12887.927883242051, "speed/FLOPS": 202422617275089.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05586351081728935, "optim/lr": 0.0029757543318851727, "optim/total_tokens": 4274520064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396559, "loss/out": 2.932328701019287, "created_at": "2025-01-16T04:35:08.762046+00:00"} {"global_step": 8154, "acc_step": 0, "speed/wps": 12890.640666055335, "speed/FLOPS": 202465225256926.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051961902529001236, "optim/lr": 0.002975738986749948, "optim/total_tokens": 4275044352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9699676036834717, "created_at": "2025-01-16T04:35:18.934189+00:00"} {"global_step": 8155, "acc_step": 0, "speed/wps": 12891.277833445936, "speed/FLOPS": 202475232846355.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05187845975160599, "optim/lr": 0.0029757236367998663, "optim/total_tokens": 4275568640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9635722637176514, "created_at": "2025-01-16T04:35:29.106525+00:00"} {"global_step": 8156, "acc_step": 0, "speed/wps": 12890.570007508855, "speed/FLOPS": 202464115467358.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047037556767463684, "optim/lr": 0.002975708282034976, "optim/total_tokens": 4276092928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.9396772384643555, "created_at": "2025-01-16T04:35:39.278049+00:00"} {"global_step": 8157, "acc_step": 0, "speed/wps": 12889.429434616035, "speed/FLOPS": 202446201202765.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08489440381526947, "optim/lr": 0.0029756929224553285, "optim/total_tokens": 4276617216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384291, "loss/out": 2.9827044010162354, "created_at": "2025-01-16T04:35:49.452575+00:00"} {"global_step": 8158, "acc_step": 0, "speed/wps": 12891.50566170326, "speed/FLOPS": 202478811202205.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061039917171001434, "optim/lr": 0.002975677558060973, "optim/total_tokens": 4277141504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.88796067237854, "created_at": "2025-01-16T04:35:59.625713+00:00"} {"global_step": 8159, "acc_step": 0, "speed/wps": 12894.14588762624, "speed/FLOPS": 202520279578376.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10253545641899109, "optim/lr": 0.0029756621888519597, "optim/total_tokens": 4277665792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497150, "loss/out": 3.0087296962738037, "created_at": "2025-01-16T04:36:09.800282+00:00"} {"global_step": 8160, "acc_step": 0, "speed/wps": 12886.734930872108, "speed/FLOPS": 202403880318831.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09501062333583832, "optim/lr": 0.0029756468148283395, "optim/total_tokens": 4278190080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 3.0378363132476807, "created_at": "2025-01-16T04:36:19.978239+00:00"} {"global_step": 8161, "acc_step": 0, "speed/wps": 12891.247550316111, "speed/FLOPS": 202474757208188.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0637548565864563, "optim/lr": 0.0029756314359901622, "optim/total_tokens": 4278714368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.9641871452331543, "created_at": "2025-01-16T04:36:30.149656+00:00"} {"global_step": 8162, "acc_step": 0, "speed/wps": 12887.50172398049, "speed/FLOPS": 202415923858282.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.09469194710254669, "optim/lr": 0.002975616052337478, "optim/total_tokens": 4279238656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.0080981254577637, "created_at": "2025-01-16T04:36:40.328841+00:00"} {"global_step": 8163, "acc_step": 0, "speed/wps": 12889.193640977026, "speed/FLOPS": 202442497739650.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0702800452709198, "optim/lr": 0.002975600663870337, "optim/total_tokens": 4279762944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.8354952335357666, "created_at": "2025-01-16T04:36:50.503734+00:00"} {"global_step": 8164, "acc_step": 0, "speed/wps": 12893.38014636896, "speed/FLOPS": 202508252559692.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04853192716836929, "optim/lr": 0.002975585270588789, "optim/total_tokens": 4280287232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460071, "loss/out": 2.8814518451690674, "created_at": "2025-01-16T04:37:00.673502+00:00"} {"global_step": 8165, "acc_step": 0, "speed/wps": 12886.209187000717, "speed/FLOPS": 202395622788883.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04691268131136894, "optim/lr": 0.0029755698724928843, "optim/total_tokens": 4280811520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.0070667266845703, "created_at": "2025-01-16T04:37:10.851943+00:00"} {"global_step": 8166, "acc_step": 0, "speed/wps": 12890.589797018822, "speed/FLOPS": 202464426288806.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054235465824604034, "optim/lr": 0.0029755544695826744, "optim/total_tokens": 4281335808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.1459407806396484, "created_at": "2025-01-16T04:37:21.023792+00:00"} {"global_step": 8167, "acc_step": 0, "speed/wps": 12889.152421534896, "speed/FLOPS": 202441850331670.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051524173468351364, "optim/lr": 0.002975539061858208, "optim/total_tokens": 4281860096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9028573036193848, "created_at": "2025-01-16T04:37:31.196521+00:00"} {"global_step": 8168, "acc_step": 0, "speed/wps": 12889.925748755026, "speed/FLOPS": 202453996498325.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04458935558795929, "optim/lr": 0.002975523649319536, "optim/total_tokens": 4282384384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 3.0904083251953125, "created_at": "2025-01-16T04:37:41.373001+00:00"} {"global_step": 8169, "acc_step": 0, "speed/wps": 12889.902355841059, "speed/FLOPS": 202453629080466.7, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06813929975032806, "optim/lr": 0.0029755082319667093, "optim/total_tokens": 4282908672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8726978302001953, "created_at": "2025-01-16T04:37:51.545326+00:00"} {"global_step": 8170, "acc_step": 0, "speed/wps": 12890.690126908416, "speed/FLOPS": 202466002107590.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05139375850558281, "optim/lr": 0.0029754928097997764, "optim/total_tokens": 4283432960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 2.845193862915039, "created_at": "2025-01-16T04:38:01.718420+00:00"} {"global_step": 8171, "acc_step": 0, "speed/wps": 12891.47107211827, "speed/FLOPS": 202478267925242.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05715804919600487, "optim/lr": 0.002975477382818789, "optim/total_tokens": 4283957248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 3.0605785846710205, "created_at": "2025-01-16T04:38:11.889591+00:00"} {"global_step": 8172, "acc_step": 0, "speed/wps": 12884.928443897292, "speed/FLOPS": 202375506958518.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05361544340848923, "optim/lr": 0.0029754619510237983, "optim/total_tokens": 4284481536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9274847507476807, "created_at": "2025-01-16T04:38:22.066633+00:00"} {"global_step": 8173, "acc_step": 0, "speed/wps": 12881.834366596684, "speed/FLOPS": 202326910222804.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06799638271331787, "optim/lr": 0.002975446514414853, "optim/total_tokens": 4285005824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.8838284015655518, "created_at": "2025-01-16T04:38:32.250170+00:00"} {"global_step": 8174, "acc_step": 0, "speed/wps": 12886.315972413578, "speed/FLOPS": 202397300000530.84, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05195556581020355, "optim/lr": 0.0029754310729920037, "optim/total_tokens": 4285530112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9807119369506836, "created_at": "2025-01-16T04:38:42.426426+00:00"} {"global_step": 8175, "acc_step": 0, "speed/wps": 12871.81436232972, "speed/FLOPS": 202169532286863.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07821320742368698, "optim/lr": 0.0029754156267553012, "optim/total_tokens": 4286054400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.807784080505371, "created_at": "2025-01-16T04:38:52.614871+00:00"} {"global_step": 8176, "acc_step": 0, "speed/wps": 12885.779332968883, "speed/FLOPS": 202388871340632.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061905983835458755, "optim/lr": 0.002975400175704796, "optim/total_tokens": 4286578688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 2.9944090843200684, "created_at": "2025-01-16T04:39:02.794228+00:00"} {"global_step": 8177, "acc_step": 0, "speed/wps": 12894.924503487304, "speed/FLOPS": 202532508810405.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05196192488074303, "optim/lr": 0.002975384719840538, "optim/total_tokens": 4287102976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8395497798919678, "created_at": "2025-01-16T04:39:12.965656+00:00"} {"global_step": 8178, "acc_step": 0, "speed/wps": 12886.054272009118, "speed/FLOPS": 202393189636064.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05700138583779335, "optim/lr": 0.0029753692591625784, "optim/total_tokens": 4287627264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9362120628356934, "created_at": "2025-01-16T04:39:23.145559+00:00"} {"global_step": 8179, "acc_step": 0, "speed/wps": 12891.005706028982, "speed/FLOPS": 202470958711330.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07441272586584091, "optim/lr": 0.002975353793670967, "optim/total_tokens": 4288151552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390359, "loss/out": 3.0045320987701416, "created_at": "2025-01-16T04:39:33.320653+00:00"} {"global_step": 8180, "acc_step": 0, "speed/wps": 12889.27516580424, "speed/FLOPS": 202443778199089.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05327964574098587, "optim/lr": 0.0029753383233657544, "optim/total_tokens": 4288675840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.864687442779541, "created_at": "2025-01-16T04:39:43.497952+00:00"} {"global_step": 8181, "acc_step": 0, "speed/wps": 12889.60517818908, "speed/FLOPS": 202448961497077.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07796970009803772, "optim/lr": 0.002975322848246991, "optim/total_tokens": 4289200128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.897336006164551, "created_at": "2025-01-16T04:39:53.670563+00:00"} {"global_step": 8182, "acc_step": 0, "speed/wps": 12893.98949766381, "speed/FLOPS": 202517823259114.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061605509370565414, "optim/lr": 0.0029753073683147276, "optim/total_tokens": 4289724416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.9649291038513184, "created_at": "2025-01-16T04:40:03.839527+00:00"} {"global_step": 8183, "acc_step": 0, "speed/wps": 12891.638526073004, "speed/FLOPS": 202480898019707.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10643840581178665, "optim/lr": 0.0029752918835690145, "optim/total_tokens": 4290248704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 2.8792762756347656, "created_at": "2025-01-16T04:40:14.010932+00:00"} {"global_step": 8184, "acc_step": 0, "speed/wps": 12893.485334350788, "speed/FLOPS": 202509904681490.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06684082001447678, "optim/lr": 0.0029752763940099026, "optim/total_tokens": 4290772992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9703946113586426, "created_at": "2025-01-16T04:40:24.182220+00:00"} {"global_step": 8185, "acc_step": 0, "speed/wps": 12888.91936589704, "speed/FLOPS": 202438189872626.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051565688103437424, "optim/lr": 0.0029752608996374413, "optim/total_tokens": 4291297280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9340014457702637, "created_at": "2025-01-16T04:40:34.357340+00:00"} {"global_step": 8186, "acc_step": 0, "speed/wps": 12886.381656987873, "speed/FLOPS": 202398331667030.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052707403898239136, "optim/lr": 0.0029752454004516827, "optim/total_tokens": 4291821568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369482, "loss/out": 2.902111530303955, "created_at": "2025-01-16T04:40:44.532276+00:00"} {"global_step": 8187, "acc_step": 0, "speed/wps": 12891.41589944564, "speed/FLOPS": 202477401362603.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047776997089385986, "optim/lr": 0.002975229896452676, "optim/total_tokens": 4292345856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9954264163970947, "created_at": "2025-01-16T04:40:54.705487+00:00"} {"global_step": 8188, "acc_step": 0, "speed/wps": 12891.897491576688, "speed/FLOPS": 202484965428798.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050243645906448364, "optim/lr": 0.002975214387640473, "optim/total_tokens": 4292870144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.915367603302002, "created_at": "2025-01-16T04:41:04.878125+00:00"} {"global_step": 8189, "acc_step": 0, "speed/wps": 12888.466273822274, "speed/FLOPS": 202431073438978.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04828180372714996, "optim/lr": 0.002975198874015123, "optim/total_tokens": 4293394432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 3.0127131938934326, "created_at": "2025-01-16T04:41:15.051542+00:00"} {"global_step": 8190, "acc_step": 0, "speed/wps": 12891.247244486272, "speed/FLOPS": 202474752404710.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04577703773975372, "optim/lr": 0.0029751833555766776, "optim/total_tokens": 4293918720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382419, "loss/out": 2.996734857559204, "created_at": "2025-01-16T04:41:25.222631+00:00"} {"global_step": 8191, "acc_step": 0, "speed/wps": 12889.84542104261, "speed/FLOPS": 202452734841220.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05088542401790619, "optim/lr": 0.0029751678323251874, "optim/total_tokens": 4294443008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.883326530456543, "created_at": "2025-01-16T04:41:35.397331+00:00"} {"global_step": 8192, "acc_step": 0, "speed/wps": 12893.392788774869, "speed/FLOPS": 202508451126049.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05371018871665001, "optim/lr": 0.002975152304260702, "optim/total_tokens": 4294967296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.8563499450683594, "created_at": "2025-01-16T04:41:45.568787+00:00"} {"global_step": 8193, "acc_step": 0, "speed/wps": 12889.490504530517, "speed/FLOPS": 202447160389691.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05074351280927658, "optim/lr": 0.002975136771383274, "optim/total_tokens": 4295491584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397090, "loss/out": 2.9113800525665283, "created_at": "2025-01-16T04:41:55.741541+00:00"} {"global_step": 8194, "acc_step": 0, "speed/wps": 12893.737169509954, "speed/FLOPS": 202513860098722.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04707241803407669, "optim/lr": 0.0029751212336929516, "optim/total_tokens": 4296015872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 2.952491283416748, "created_at": "2025-01-16T04:42:05.914297+00:00"} {"global_step": 8195, "acc_step": 0, "speed/wps": 12887.89432664194, "speed/FLOPS": 202422090222573.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.057822104543447495, "optim/lr": 0.0029751056911897877, "optim/total_tokens": 4296540160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.974759817123413, "created_at": "2025-01-16T04:42:16.088101+00:00"} {"global_step": 8196, "acc_step": 0, "speed/wps": 12893.1680061731, "speed/FLOPS": 202504920606405.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04833396524190903, "optim/lr": 0.002975090143873832, "optim/total_tokens": 4297064448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.944735050201416, "created_at": "2025-01-16T04:42:26.259721+00:00"} {"global_step": 8197, "acc_step": 0, "speed/wps": 12891.446958001834, "speed/FLOPS": 202477889179907.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05555069074034691, "optim/lr": 0.002975074591745135, "optim/total_tokens": 4297588736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 2.9636495113372803, "created_at": "2025-01-16T04:42:36.433847+00:00"} {"global_step": 8198, "acc_step": 0, "speed/wps": 12888.456852668585, "speed/FLOPS": 202430925466813.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051477596163749695, "optim/lr": 0.002975059034803749, "optim/total_tokens": 4298113024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.8980813026428223, "created_at": "2025-01-16T04:42:46.607478+00:00"} {"global_step": 8199, "acc_step": 0, "speed/wps": 12892.087329934942, "speed/FLOPS": 202487947101080.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05866819992661476, "optim/lr": 0.002975043473049722, "optim/total_tokens": 4298637312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9873764514923096, "created_at": "2025-01-16T04:42:56.783099+00:00"} {"global_step": 8200, "acc_step": 0, "speed/wps": 12894.797419604532, "speed/FLOPS": 202530512783394.56, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0542195588350296, "optim/lr": 0.0029750279064831063, "optim/total_tokens": 4299161600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 2.868248462677002, "created_at": "2025-01-16T04:43:06.951364+00:00"} {"global_step": 8201, "acc_step": 0, "speed/wps": 12890.930992318466, "speed/FLOPS": 202469785229839.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053766414523124695, "optim/lr": 0.0029750123351039537, "optim/total_tokens": 4299685888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8975563049316406, "created_at": "2025-01-16T04:43:17.128354+00:00"} {"global_step": 8202, "acc_step": 0, "speed/wps": 12894.441007597126, "speed/FLOPS": 202524914843056.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05386281758546829, "optim/lr": 0.002974996758912314, "optim/total_tokens": 4300210176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.953341484069824, "created_at": "2025-01-16T04:43:27.299310+00:00"} {"global_step": 8203, "acc_step": 0, "speed/wps": 12889.896055175188, "speed/FLOPS": 202453530119851.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05872110649943352, "optim/lr": 0.0029749811779082374, "optim/total_tokens": 4300734464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.8549869060516357, "created_at": "2025-01-16T04:43:37.471422+00:00"} {"global_step": 8204, "acc_step": 0, "speed/wps": 12885.683406979424, "speed/FLOPS": 202387364691151.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0591309480369091, "optim/lr": 0.0029749655920917758, "optim/total_tokens": 4301258752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288730, "loss/out": 2.847733974456787, "created_at": "2025-01-16T04:43:47.647415+00:00"} {"global_step": 8205, "acc_step": 0, "speed/wps": 12889.577201358592, "speed/FLOPS": 202448522082510.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0515160858631134, "optim/lr": 0.00297495000146298, "optim/total_tokens": 4301783040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.9718236923217773, "created_at": "2025-01-16T04:43:57.826069+00:00"} {"global_step": 8206, "acc_step": 0, "speed/wps": 12888.034949709356, "speed/FLOPS": 202424298901083.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05799949914216995, "optim/lr": 0.0029749344060218995, "optim/total_tokens": 4302307328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9808356761932373, "created_at": "2025-01-16T04:44:08.000772+00:00"} {"global_step": 8207, "acc_step": 0, "speed/wps": 12891.033447002903, "speed/FLOPS": 202471394421445.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07234694808721542, "optim/lr": 0.002974918805768587, "optim/total_tokens": 4302831616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 3.0011563301086426, "created_at": "2025-01-16T04:44:18.173850+00:00"} {"global_step": 8208, "acc_step": 0, "speed/wps": 12890.959063582879, "speed/FLOPS": 202470226127620.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0474536307156086, "optim/lr": 0.002974903200703092, "optim/total_tokens": 4303355904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.871840238571167, "created_at": "2025-01-16T04:44:28.347403+00:00"} {"global_step": 8209, "acc_step": 0, "speed/wps": 12884.560599707875, "speed/FLOPS": 202369729460052.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06783980131149292, "optim/lr": 0.002974887590825466, "optim/total_tokens": 4303880192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9555087089538574, "created_at": "2025-01-16T04:44:38.527221+00:00"} {"global_step": 8210, "acc_step": 0, "speed/wps": 12885.425249082256, "speed/FLOPS": 202383309966630.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054368503391742706, "optim/lr": 0.0029748719761357604, "optim/total_tokens": 4304404480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9738717079162598, "created_at": "2025-01-16T04:44:48.704220+00:00"} {"global_step": 8211, "acc_step": 0, "speed/wps": 12889.893060088476, "speed/FLOPS": 202453483077898.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05884656682610512, "optim/lr": 0.0029748563566340254, "optim/total_tokens": 4304928768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.954197406768799, "created_at": "2025-01-16T04:44:58.878966+00:00"} {"global_step": 8212, "acc_step": 0, "speed/wps": 12889.110946205514, "speed/FLOPS": 202441198904629.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0591278001666069, "optim/lr": 0.0029748407323203123, "optim/total_tokens": 4305453056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8766415119171143, "created_at": "2025-01-16T04:45:09.051822+00:00"} {"global_step": 8213, "acc_step": 0, "speed/wps": 12888.092058532475, "speed/FLOPS": 202425195873626.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044184498488903046, "optim/lr": 0.002974825103194672, "optim/total_tokens": 4305977344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.0158615112304688, "created_at": "2025-01-16T04:45:19.229486+00:00"} {"global_step": 8214, "acc_step": 0, "speed/wps": 12891.494392790017, "speed/FLOPS": 202478634208438.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05396585538983345, "optim/lr": 0.0029748094692571554, "optim/total_tokens": 4306501632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.002326726913452, "created_at": "2025-01-16T04:45:29.403977+00:00"} {"global_step": 8215, "acc_step": 0, "speed/wps": 12890.097709053265, "speed/FLOPS": 202456697371107.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05571868270635605, "optim/lr": 0.0029747938305078134, "optim/total_tokens": 4307025920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 2.8785624504089355, "created_at": "2025-01-16T04:45:39.576986+00:00"} {"global_step": 8216, "acc_step": 0, "speed/wps": 12895.913783869779, "speed/FLOPS": 202548046818225.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04760637879371643, "optim/lr": 0.0029747781869466977, "optim/total_tokens": 4307550208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.0048556327819824, "created_at": "2025-01-16T04:45:49.744321+00:00"} {"global_step": 8217, "acc_step": 0, "speed/wps": 12891.199748934376, "speed/FLOPS": 202474006421802.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04722798988223076, "optim/lr": 0.0029747625385738584, "optim/total_tokens": 4308074496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.912696361541748, "created_at": "2025-01-16T04:45:59.916542+00:00"} {"global_step": 8218, "acc_step": 0, "speed/wps": 12888.395347579992, "speed/FLOPS": 202429959444880.28, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05415888875722885, "optim/lr": 0.0029747468853893472, "optim/total_tokens": 4308598784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.9187779426574707, "created_at": "2025-01-16T04:46:10.093932+00:00"} {"global_step": 8219, "acc_step": 0, "speed/wps": 12885.348720734119, "speed/FLOPS": 202382107983761.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.050703007727861404, "optim/lr": 0.002974731227393215, "optim/total_tokens": 4309123072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9058403968811035, "created_at": "2025-01-16T04:46:20.272431+00:00"} {"global_step": 8220, "acc_step": 0, "speed/wps": 12888.378764558325, "speed/FLOPS": 202429698985737.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06293091922998428, "optim/lr": 0.0029747155645855135, "optim/total_tokens": 4309647360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.9995040893554688, "created_at": "2025-01-16T04:46:30.449857+00:00"} {"global_step": 8221, "acc_step": 0, "speed/wps": 12881.451303432723, "speed/FLOPS": 202320893689430.8, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.0721120834350586, "optim/lr": 0.0029746998969662924, "optim/total_tokens": 4310171648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.962265729904175, "created_at": "2025-01-16T04:46:40.629112+00:00"} {"global_step": 8222, "acc_step": 0, "speed/wps": 12884.577745006063, "speed/FLOPS": 202369998750521.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05219326168298721, "optim/lr": 0.0029746842245356032, "optim/total_tokens": 4310695936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 2.971662998199463, "created_at": "2025-01-16T04:46:50.810120+00:00"} {"global_step": 8223, "acc_step": 0, "speed/wps": 12894.03687437882, "speed/FLOPS": 202518567375526.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05562755465507507, "optim/lr": 0.0029746685472934985, "optim/total_tokens": 4311220224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9257965087890625, "created_at": "2025-01-16T04:47:00.981273+00:00"} {"global_step": 8224, "acc_step": 0, "speed/wps": 12890.520332694334, "speed/FLOPS": 202463335256136.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06593666970729828, "optim/lr": 0.0029746528652400276, "optim/total_tokens": 4311744512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.0107336044311523, "created_at": "2025-01-16T04:47:11.156718+00:00"} {"global_step": 8225, "acc_step": 0, "speed/wps": 12887.806697685093, "speed/FLOPS": 202420713889391.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05449659749865532, "optim/lr": 0.002974637178375243, "optim/total_tokens": 4312268800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374658, "loss/out": 2.855064630508423, "created_at": "2025-01-16T04:47:21.330587+00:00"} {"global_step": 8226, "acc_step": 0, "speed/wps": 12891.351637215057, "speed/FLOPS": 202476392035966.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08299466967582703, "optim/lr": 0.002974621486699195, "optim/total_tokens": 4312793088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.923478364944458, "created_at": "2025-01-16T04:47:31.502224+00:00"} {"global_step": 8227, "acc_step": 0, "speed/wps": 12890.514160117975, "speed/FLOPS": 202463238307342.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0888209193944931, "optim/lr": 0.002974605790211935, "optim/total_tokens": 4313317376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.968569755554199, "created_at": "2025-01-16T04:47:41.680364+00:00"} {"global_step": 8228, "acc_step": 0, "speed/wps": 12889.666594480925, "speed/FLOPS": 202449926124335.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06234106048941612, "optim/lr": 0.002974590088913515, "optim/total_tokens": 4313841664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0838100910186768, "created_at": "2025-01-16T04:47:51.854406+00:00"} {"global_step": 8229, "acc_step": 0, "speed/wps": 12886.264995915764, "speed/FLOPS": 202396499344583.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07498382031917572, "optim/lr": 0.0029745743828039854, "optim/total_tokens": 4314365952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429193, "loss/out": 2.945915937423706, "created_at": "2025-01-16T04:48:02.029435+00:00"} {"global_step": 8230, "acc_step": 0, "speed/wps": 12891.336709188587, "speed/FLOPS": 202476157570797.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0677134096622467, "optim/lr": 0.0029745586718833974, "optim/total_tokens": 4314890240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.9622673988342285, "created_at": "2025-01-16T04:48:12.201835+00:00"} {"global_step": 8231, "acc_step": 0, "speed/wps": 12889.652002082246, "speed/FLOPS": 202449696930662.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05556989833712578, "optim/lr": 0.0029745429561518024, "optim/total_tokens": 4315414528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.906787872314453, "created_at": "2025-01-16T04:48:22.384324+00:00"} {"global_step": 8232, "acc_step": 0, "speed/wps": 12894.416429895911, "speed/FLOPS": 202524528816485.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057087019085884094, "optim/lr": 0.002974527235609252, "optim/total_tokens": 4315938816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426755, "loss/out": 2.9580342769622803, "created_at": "2025-01-16T04:48:32.553193+00:00"} {"global_step": 8233, "acc_step": 0, "speed/wps": 12885.222799963265, "speed/FLOPS": 202380130225022.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05478415638208389, "optim/lr": 0.002974511510255797, "optim/total_tokens": 4316463104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 2.8999524116516113, "created_at": "2025-01-16T04:48:42.730923+00:00"} {"global_step": 8234, "acc_step": 0, "speed/wps": 12887.129937791176, "speed/FLOPS": 202410084445291.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05646177753806114, "optim/lr": 0.002974495780091489, "optim/total_tokens": 4316987392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.897151470184326, "created_at": "2025-01-16T04:48:52.907442+00:00"} {"global_step": 8235, "acc_step": 0, "speed/wps": 12886.55504841344, "speed/FLOPS": 202401055017633.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051393453031778336, "optim/lr": 0.002974480045116379, "optim/total_tokens": 4317511680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.878293037414551, "created_at": "2025-01-16T04:49:03.082394+00:00"} {"global_step": 8236, "acc_step": 0, "speed/wps": 12889.63921472469, "speed/FLOPS": 202449496087640.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051547080278396606, "optim/lr": 0.0029744643053305187, "optim/total_tokens": 4318035968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9379000663757324, "created_at": "2025-01-16T04:49:13.260286+00:00"} {"global_step": 8237, "acc_step": 0, "speed/wps": 12889.076089712222, "speed/FLOPS": 202440651435504.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04842820763587952, "optim/lr": 0.00297444856073396, "optim/total_tokens": 4318560256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337405, "loss/out": 2.868683099746704, "created_at": "2025-01-16T04:49:23.435462+00:00"} {"global_step": 8238, "acc_step": 0, "speed/wps": 12888.463900237906, "speed/FLOPS": 202431036158574.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05267173796892166, "optim/lr": 0.002974432811326753, "optim/total_tokens": 4319084544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.053173065185547, "created_at": "2025-01-16T04:49:33.611959+00:00"} {"global_step": 8239, "acc_step": 0, "speed/wps": 12884.399419621473, "speed/FLOPS": 202367197905311.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05015572905540466, "optim/lr": 0.0029744170571089495, "optim/total_tokens": 4319608832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8410842418670654, "created_at": "2025-01-16T04:49:43.790810+00:00"} {"global_step": 8240, "acc_step": 0, "speed/wps": 12895.471794787465, "speed/FLOPS": 202541104772331.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06237272918224335, "optim/lr": 0.0029744012980806016, "optim/total_tokens": 4320133120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345322, "loss/out": 2.837920665740967, "created_at": "2025-01-16T04:49:53.958519+00:00"} {"global_step": 8241, "acc_step": 0, "speed/wps": 12892.134641890765, "speed/FLOPS": 202488690200359.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05240991339087486, "optim/lr": 0.0029743855342417606, "optim/total_tokens": 4320657408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.0254766941070557, "created_at": "2025-01-16T04:50:04.130030+00:00"} {"global_step": 8242, "acc_step": 0, "speed/wps": 12889.852493741679, "speed/FLOPS": 202452845927678.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0540846548974514, "optim/lr": 0.0029743697655924767, "optim/total_tokens": 4321181696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.8870718479156494, "created_at": "2025-01-16T04:50:14.302331+00:00"} {"global_step": 8243, "acc_step": 0, "speed/wps": 12892.303638941543, "speed/FLOPS": 202491344531268.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05964936316013336, "optim/lr": 0.002974353992132803, "optim/total_tokens": 4321705984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9814844131469727, "created_at": "2025-01-16T04:50:24.475902+00:00"} {"global_step": 8244, "acc_step": 0, "speed/wps": 12891.39972902855, "speed/FLOPS": 202477147383982.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05945728346705437, "optim/lr": 0.0029743382138627894, "optim/total_tokens": 4322230272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 2.9776294231414795, "created_at": "2025-01-16T04:50:34.647348+00:00"} {"global_step": 8245, "acc_step": 0, "speed/wps": 12890.080285788288, "speed/FLOPS": 202456423714789.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051790010184049606, "optim/lr": 0.002974322430782489, "optim/total_tokens": 4322754560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8750295639038086, "created_at": "2025-01-16T04:50:44.820764+00:00"} {"global_step": 8246, "acc_step": 0, "speed/wps": 12887.988938928675, "speed/FLOPS": 202423576238547.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0475306436419487, "optim/lr": 0.0029743066428919514, "optim/total_tokens": 4323278848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475206, "loss/out": 3.004903793334961, "created_at": "2025-01-16T04:50:54.998167+00:00"} {"global_step": 8247, "acc_step": 0, "speed/wps": 12893.701179968735, "speed/FLOPS": 202513294833523.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06711495667695999, "optim/lr": 0.00297429085019123, "optim/total_tokens": 4323803136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.885484218597412, "created_at": "2025-01-16T04:51:05.167708+00:00"} {"global_step": 8248, "acc_step": 0, "speed/wps": 12885.566758549503, "speed/FLOPS": 202385532567267.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07035117596387863, "optim/lr": 0.0029742750526803754, "optim/total_tokens": 4324327424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.090304374694824, "created_at": "2025-01-16T04:51:15.346527+00:00"} {"global_step": 8249, "acc_step": 0, "speed/wps": 12890.834200881975, "speed/FLOPS": 202468264987323.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08328165858983994, "optim/lr": 0.002974259250359439, "optim/total_tokens": 4324851712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8842358589172363, "created_at": "2025-01-16T04:51:25.520223+00:00"} {"global_step": 8250, "acc_step": 0, "speed/wps": 12889.034043558166, "speed/FLOPS": 202439991042877.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07185442000627518, "optim/lr": 0.0029742434432284727, "optim/total_tokens": 4325376000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9642951488494873, "created_at": "2025-01-16T04:51:35.697772+00:00"} {"global_step": 8251, "acc_step": 0, "speed/wps": 12888.584347469669, "speed/FLOPS": 202432927947860.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05563955381512642, "optim/lr": 0.0029742276312875278, "optim/total_tokens": 4325900288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.95880126953125, "created_at": "2025-01-16T04:51:45.871236+00:00"} {"global_step": 8252, "acc_step": 0, "speed/wps": 12889.754818931164, "speed/FLOPS": 202451311810559.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06354770809412003, "optim/lr": 0.002974211814536656, "optim/total_tokens": 4326424576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 3.0409412384033203, "created_at": "2025-01-16T04:51:56.045927+00:00"} {"global_step": 8253, "acc_step": 0, "speed/wps": 12894.740070579552, "speed/FLOPS": 202529612038151.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0538497157394886, "optim/lr": 0.0029741959929759096, "optim/total_tokens": 4326948864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422232, "loss/out": 2.980100631713867, "created_at": "2025-01-16T04:52:06.215577+00:00"} {"global_step": 8254, "acc_step": 0, "speed/wps": 12889.399437580627, "speed/FLOPS": 202445730058104.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05268057808279991, "optim/lr": 0.002974180166605339, "optim/total_tokens": 4327473152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9607527256011963, "created_at": "2025-01-16T04:52:16.388159+00:00"} {"global_step": 8255, "acc_step": 0, "speed/wps": 12892.23622760338, "speed/FLOPS": 202490285743569.8, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04905442148447037, "optim/lr": 0.0029741643354249964, "optim/total_tokens": 4327997440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.0393660068511963, "created_at": "2025-01-16T04:52:26.561252+00:00"} {"global_step": 8256, "acc_step": 0, "speed/wps": 12890.992826247335, "speed/FLOPS": 202470756416660.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0531054362654686, "optim/lr": 0.0029741484994349338, "optim/total_tokens": 4328521728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.982401132583618, "created_at": "2025-01-16T04:52:36.735928+00:00"} {"global_step": 8257, "acc_step": 0, "speed/wps": 12892.097825171093, "speed/FLOPS": 202488111943185.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060512952506542206, "optim/lr": 0.0029741326586352024, "optim/total_tokens": 4329046016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 2.93735408782959, "created_at": "2025-01-16T04:52:46.908665+00:00"} {"global_step": 8258, "acc_step": 0, "speed/wps": 12893.44594195871, "speed/FLOPS": 202509285969842.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051454633474349976, "optim/lr": 0.0029741168130258543, "optim/total_tokens": 4329570304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.047597885131836, "created_at": "2025-01-16T04:52:57.080586+00:00"} {"global_step": 8259, "acc_step": 0, "speed/wps": 12893.067382581621, "speed/FLOPS": 202503340174629.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05184972286224365, "optim/lr": 0.0029741009626069403, "optim/total_tokens": 4330094592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 2.863933801651001, "created_at": "2025-01-16T04:53:07.251472+00:00"} {"global_step": 8260, "acc_step": 0, "speed/wps": 12885.847007149632, "speed/FLOPS": 202389934256634.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05354462191462517, "optim/lr": 0.002974085107378513, "optim/total_tokens": 4330618880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 2.9148244857788086, "created_at": "2025-01-16T04:53:17.427715+00:00"} {"global_step": 8261, "acc_step": 0, "speed/wps": 12891.580255722105, "speed/FLOPS": 202479982803774.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06849492341279984, "optim/lr": 0.002974069247340624, "optim/total_tokens": 4331143168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.883720874786377, "created_at": "2025-01-16T04:53:27.604694+00:00"} {"global_step": 8262, "acc_step": 0, "speed/wps": 12890.378598702535, "speed/FLOPS": 202461109129032.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056922148913145065, "optim/lr": 0.0029740533824933247, "optim/total_tokens": 4331667456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0089354515075684, "created_at": "2025-01-16T04:53:37.777668+00:00"} {"global_step": 8263, "acc_step": 0, "speed/wps": 12893.541144884352, "speed/FLOPS": 202510781262611.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06288745254278183, "optim/lr": 0.0029740375128366674, "optim/total_tokens": 4332191744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8689610958099365, "created_at": "2025-01-16T04:53:47.946856+00:00"} {"global_step": 8264, "acc_step": 0, "speed/wps": 12892.366440299698, "speed/FLOPS": 202492330912896.5, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05710403248667717, "optim/lr": 0.002974021638370703, "optim/total_tokens": 4332716032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 2.8397104740142822, "created_at": "2025-01-16T04:53:58.117805+00:00"} {"global_step": 8265, "acc_step": 0, "speed/wps": 12890.185674947306, "speed/FLOPS": 202458078996351.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057717036455869675, "optim/lr": 0.002974005759095484, "optim/total_tokens": 4333240320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.940458297729492, "created_at": "2025-01-16T04:54:08.289714+00:00"} {"global_step": 8266, "acc_step": 0, "speed/wps": 12889.661260891038, "speed/FLOPS": 202449842352977.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09164588898420334, "optim/lr": 0.0029739898750110625, "optim/total_tokens": 4333764608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9220173358917236, "created_at": "2025-01-16T04:54:18.464228+00:00"} {"global_step": 8267, "acc_step": 0, "speed/wps": 12888.131938702252, "speed/FLOPS": 202425822246493.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05087314173579216, "optim/lr": 0.0029739739861174897, "optim/total_tokens": 4334288896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 2.826457977294922, "created_at": "2025-01-16T04:54:28.640448+00:00"} {"global_step": 8268, "acc_step": 0, "speed/wps": 12897.07174248284, "speed/FLOPS": 202566234149445.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059407852590084076, "optim/lr": 0.0029739580924148176, "optim/total_tokens": 4334813184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.889251708984375, "created_at": "2025-01-16T04:54:38.808101+00:00"} {"global_step": 8269, "acc_step": 0, "speed/wps": 12891.453068886623, "speed/FLOPS": 202477985159750.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0607408732175827, "optim/lr": 0.0029739421939030983, "optim/total_tokens": 4335337472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9894094467163086, "created_at": "2025-01-16T04:54:48.981868+00:00"} {"global_step": 8270, "acc_step": 0, "speed/wps": 12888.45714420245, "speed/FLOPS": 202430930045753.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04990777373313904, "optim/lr": 0.002973926290582383, "optim/total_tokens": 4335861760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8238019943237305, "created_at": "2025-01-16T04:54:59.157093+00:00"} {"global_step": 8271, "acc_step": 0, "speed/wps": 12889.777069726531, "speed/FLOPS": 202451661289876.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05513271689414978, "optim/lr": 0.002973910382452724, "optim/total_tokens": 4336386048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9397714138031006, "created_at": "2025-01-16T04:55:09.329410+00:00"} {"global_step": 8272, "acc_step": 0, "speed/wps": 12892.115508979985, "speed/FLOPS": 202488389691704.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06258419901132584, "optim/lr": 0.002973894469514173, "optim/total_tokens": 4336910336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9364733695983887, "created_at": "2025-01-16T04:55:19.499934+00:00"} {"global_step": 8273, "acc_step": 0, "speed/wps": 12883.099043208042, "speed/FLOPS": 202346773706833.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05007506161928177, "optim/lr": 0.002973878551766783, "optim/total_tokens": 4337434624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409693, "loss/out": 2.897706985473633, "created_at": "2025-01-16T04:55:29.677676+00:00"} {"global_step": 8274, "acc_step": 0, "speed/wps": 12892.831976688254, "speed/FLOPS": 202499642801592.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06430507451295853, "optim/lr": 0.0029738626292106045, "optim/total_tokens": 4337958912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 3.033006429672241, "created_at": "2025-01-16T04:55:39.847651+00:00"} {"global_step": 8275, "acc_step": 0, "speed/wps": 12882.614847623629, "speed/FLOPS": 202339168749826.25, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0627821758389473, "optim/lr": 0.00297384670184569, "optim/total_tokens": 4338483200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8985495567321777, "created_at": "2025-01-16T04:55:50.029652+00:00"} {"global_step": 8276, "acc_step": 0, "speed/wps": 12890.443566714137, "speed/FLOPS": 202462129540930.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049161288887262344, "optim/lr": 0.002973830769672092, "optim/total_tokens": 4339007488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.0550546646118164, "created_at": "2025-01-16T04:56:00.203437+00:00"} {"global_step": 8277, "acc_step": 0, "speed/wps": 12889.920567243824, "speed/FLOPS": 202453915115572.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06021985039114952, "optim/lr": 0.0029738148326898615, "optim/total_tokens": 4339531776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415534, "loss/out": 2.8454244136810303, "created_at": "2025-01-16T04:56:10.378890+00:00"} {"global_step": 8278, "acc_step": 0, "speed/wps": 12890.861349567256, "speed/FLOPS": 202468691394732.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06980575621128082, "optim/lr": 0.0029737988908990513, "optim/total_tokens": 4340056064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.9217662811279297, "created_at": "2025-01-16T04:56:20.551564+00:00"} {"global_step": 8279, "acc_step": 0, "speed/wps": 12890.238667100344, "speed/FLOPS": 202458911310933.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05780736729502678, "optim/lr": 0.002973782944299713, "optim/total_tokens": 4340580352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9622817039489746, "created_at": "2025-01-16T04:56:30.728481+00:00"} {"global_step": 8280, "acc_step": 0, "speed/wps": 12891.861620949585, "speed/FLOPS": 202484402031308.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061540599912405014, "optim/lr": 0.0029737669928918986, "optim/total_tokens": 4341104640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 3.0481374263763428, "created_at": "2025-01-16T04:56:40.899669+00:00"} {"global_step": 8281, "acc_step": 0, "speed/wps": 12886.574868076203, "speed/FLOPS": 202401366312672.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050176143646240234, "optim/lr": 0.0029737510366756605, "optim/total_tokens": 4341628928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 2.8839595317840576, "created_at": "2025-01-16T04:56:51.075121+00:00"} {"global_step": 8282, "acc_step": 0, "speed/wps": 12891.333486713751, "speed/FLOPS": 202476106957402.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0605105422437191, "optim/lr": 0.0029737350756510504, "optim/total_tokens": 4342153216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.941823720932007, "created_at": "2025-01-16T04:57:01.247125+00:00"} {"global_step": 8283, "acc_step": 0, "speed/wps": 12890.315581761944, "speed/FLOPS": 202460119361386.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055366937071084976, "optim/lr": 0.0029737191098181204, "optim/total_tokens": 4342677504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.978515148162842, "created_at": "2025-01-16T04:57:11.418867+00:00"} {"global_step": 8284, "acc_step": 0, "speed/wps": 12884.01954537643, "speed/FLOPS": 202361231458291.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.061961572617292404, "optim/lr": 0.0029737031391769224, "optim/total_tokens": 4343201792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 2.8867712020874023, "created_at": "2025-01-16T04:57:21.595471+00:00"} {"global_step": 8285, "acc_step": 0, "speed/wps": 12881.275200897015, "speed/FLOPS": 202318127757117.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06726772338151932, "optim/lr": 0.0029736871637275094, "optim/total_tokens": 4343726080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.80487060546875, "created_at": "2025-01-16T04:57:31.775273+00:00"} {"global_step": 8286, "acc_step": 0, "speed/wps": 12887.698991822763, "speed/FLOPS": 202419022220819.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04353916272521019, "optim/lr": 0.0029736711834699326, "optim/total_tokens": 4344250368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.942718982696533, "created_at": "2025-01-16T04:57:41.956474+00:00"} {"global_step": 8287, "acc_step": 0, "speed/wps": 12886.601145713197, "speed/FLOPS": 202401779039070.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061280507594347, "optim/lr": 0.002973655198404245, "optim/total_tokens": 4344774656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486862, "loss/out": 2.9436228275299072, "created_at": "2025-01-16T04:57:52.131186+00:00"} {"global_step": 8288, "acc_step": 0, "speed/wps": 12891.590702897747, "speed/FLOPS": 202480146891023.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07046764343976974, "optim/lr": 0.0029736392085304977, "optim/total_tokens": 4345298944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 2.923882007598877, "created_at": "2025-01-16T04:58:02.302045+00:00"} {"global_step": 8289, "acc_step": 0, "speed/wps": 12885.319638100264, "speed/FLOPS": 202381651201033.4, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0653696209192276, "optim/lr": 0.0029736232138487433, "optim/total_tokens": 4345823232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.97066330909729, "created_at": "2025-01-16T04:58:12.477891+00:00"} {"global_step": 8290, "acc_step": 0, "speed/wps": 12886.455018867438, "speed/FLOPS": 202399483916158.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050801318138837814, "optim/lr": 0.002973607214359034, "optim/total_tokens": 4346347520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.913766860961914, "created_at": "2025-01-16T04:58:22.653225+00:00"} {"global_step": 8291, "acc_step": 0, "speed/wps": 12892.134507259874, "speed/FLOPS": 202488688085796.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07400251924991608, "optim/lr": 0.0029735912100614225, "optim/total_tokens": 4346871808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 2.8613147735595703, "created_at": "2025-01-16T04:58:32.825256+00:00"} {"global_step": 8292, "acc_step": 0, "speed/wps": 12880.737357872937, "speed/FLOPS": 202309680193350.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05099867656826973, "optim/lr": 0.00297357520095596, "optim/total_tokens": 4347396096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 2.949946880340576, "created_at": "2025-01-16T04:58:43.005973+00:00"} {"global_step": 8293, "acc_step": 0, "speed/wps": 12890.693280580512, "speed/FLOPS": 202466051640344.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05357290431857109, "optim/lr": 0.0029735591870427005, "optim/total_tokens": 4347920384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9628725051879883, "created_at": "2025-01-16T04:58:53.177950+00:00"} {"global_step": 8294, "acc_step": 0, "speed/wps": 12894.523735334573, "speed/FLOPS": 202526214195853.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06119147315621376, "optim/lr": 0.0029735431683216938, "optim/total_tokens": 4348444672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 2.7562570571899414, "created_at": "2025-01-16T04:59:03.346615+00:00"} {"global_step": 8295, "acc_step": 0, "speed/wps": 12882.215873653682, "speed/FLOPS": 202332902315380.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09147834032773972, "optim/lr": 0.002973527144792994, "optim/total_tokens": 4348968960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.9399542808532715, "created_at": "2025-01-16T04:59:13.524938+00:00"} {"global_step": 8296, "acc_step": 0, "speed/wps": 12893.918691358036, "speed/FLOPS": 202516711148784.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08754897862672806, "optim/lr": 0.002973511116456653, "optim/total_tokens": 4349493248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.045659303665161, "created_at": "2025-01-16T04:59:23.697009+00:00"} {"global_step": 8297, "acc_step": 0, "speed/wps": 12891.290311138142, "speed/FLOPS": 202475428825657.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049134768545627594, "optim/lr": 0.0029734950833127227, "optim/total_tokens": 4350017536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8861496448516846, "created_at": "2025-01-16T04:59:33.870344+00:00"} {"global_step": 8298, "acc_step": 0, "speed/wps": 12887.247977514258, "speed/FLOPS": 202411938421346.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.08361568301916122, "optim/lr": 0.0029734790453612553, "optim/total_tokens": 4350541824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 2.9435670375823975, "created_at": "2025-01-16T04:59:44.044552+00:00"} {"global_step": 8299, "acc_step": 0, "speed/wps": 12888.19776217637, "speed/FLOPS": 202426856094605.62, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05188840627670288, "optim/lr": 0.002973463002602304, "optim/total_tokens": 4351066112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.893827199935913, "created_at": "2025-01-16T04:59:54.217925+00:00"} {"global_step": 8300, "acc_step": 0, "speed/wps": 12890.550477809795, "speed/FLOPS": 202463808726598.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0967276468873024, "optim/lr": 0.00297344695503592, "optim/total_tokens": 4351590400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8220365047454834, "created_at": "2025-01-16T05:00:04.395152+00:00"} {"global_step": 8301, "acc_step": 0, "speed/wps": 12894.315029557498, "speed/FLOPS": 202522936184832.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05229506641626358, "optim/lr": 0.0029734309026621573, "optim/total_tokens": 4352114688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375182, "loss/out": 3.022976875305176, "created_at": "2025-01-16T05:00:14.567101+00:00"} {"global_step": 8302, "acc_step": 0, "speed/wps": 12890.507368873996, "speed/FLOPS": 202463131641524.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08317472040653229, "optim/lr": 0.002973414845481066, "optim/total_tokens": 4352638976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 3.0155506134033203, "created_at": "2025-01-16T05:00:24.740953+00:00"} {"global_step": 8303, "acc_step": 0, "speed/wps": 12894.286829082735, "speed/FLOPS": 202522493257624.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06137965992093086, "optim/lr": 0.0029733987834927008, "optim/total_tokens": 4353163264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.8292124271392822, "created_at": "2025-01-16T05:00:34.909524+00:00"} {"global_step": 8304, "acc_step": 0, "speed/wps": 12886.299843105886, "speed/FLOPS": 202397046667589.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055798809975385666, "optim/lr": 0.002973382716697112, "optim/total_tokens": 4353687552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414801, "loss/out": 2.9107131958007812, "created_at": "2025-01-16T05:00:45.088166+00:00"} {"global_step": 8305, "acc_step": 0, "speed/wps": 12887.268453073591, "speed/FLOPS": 202412260018141.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06323885917663574, "optim/lr": 0.0029733666450943537, "optim/total_tokens": 4354211840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8886260986328125, "created_at": "2025-01-16T05:00:55.267775+00:00"} {"global_step": 8306, "acc_step": 0, "speed/wps": 12891.839414917207, "speed/FLOPS": 202484053255055.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044283412396907806, "optim/lr": 0.0029733505686844776, "optim/total_tokens": 4354736128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.0131163597106934, "created_at": "2025-01-16T05:01:05.440731+00:00"} {"global_step": 8307, "acc_step": 0, "speed/wps": 12887.175770149292, "speed/FLOPS": 202410804305456.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05453053116798401, "optim/lr": 0.0029733344874675357, "optim/total_tokens": 4355260416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8879284858703613, "created_at": "2025-01-16T05:01:15.615300+00:00"} {"global_step": 8308, "acc_step": 0, "speed/wps": 12890.959246600272, "speed/FLOPS": 202470229002160.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04348297044634819, "optim/lr": 0.0029733184014435815, "optim/total_tokens": 4355784704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369627, "loss/out": 2.8878798484802246, "created_at": "2025-01-16T05:01:25.787069+00:00"} {"global_step": 8309, "acc_step": 0, "speed/wps": 12896.197234949725, "speed/FLOPS": 202552498806938.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05015144497156143, "optim/lr": 0.002973302310612667, "optim/total_tokens": 4356308992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.944366216659546, "created_at": "2025-01-16T05:01:35.957663+00:00"} {"global_step": 8310, "acc_step": 0, "speed/wps": 12892.173357835989, "speed/FLOPS": 202489298287480.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05727149546146393, "optim/lr": 0.0029732862149748442, "optim/total_tokens": 4356833280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.957195281982422, "created_at": "2025-01-16T05:01:46.132019+00:00"} {"global_step": 8311, "acc_step": 0, "speed/wps": 12890.913874976259, "speed/FLOPS": 202469516378457.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045599956065416336, "optim/lr": 0.0029732701145301667, "optim/total_tokens": 4357357568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369048, "loss/out": 2.9854369163513184, "created_at": "2025-01-16T05:01:56.306273+00:00"} {"global_step": 8312, "acc_step": 0, "speed/wps": 12886.893051803201, "speed/FLOPS": 202406363825335.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05871575325727463, "optim/lr": 0.002973254009278686, "optim/total_tokens": 4357881856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 2.839839458465576, "created_at": "2025-01-16T05:02:06.491858+00:00"} {"global_step": 8313, "acc_step": 0, "speed/wps": 12887.606097360745, "speed/FLOPS": 202417563185643.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060899995267391205, "optim/lr": 0.002973237899220455, "optim/total_tokens": 4358406144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.995945453643799, "created_at": "2025-01-16T05:02:16.675279+00:00"} {"global_step": 8314, "acc_step": 0, "speed/wps": 12892.21708730515, "speed/FLOPS": 202489985118884.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05805781111121178, "optim/lr": 0.002973221784355527, "optim/total_tokens": 4358930432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.898831844329834, "created_at": "2025-01-16T05:02:26.849636+00:00"} {"global_step": 8315, "acc_step": 0, "speed/wps": 12890.149583626757, "speed/FLOPS": 202457512132567.66, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056192368268966675, "optim/lr": 0.0029732056646839537, "optim/total_tokens": 4359454720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.035403251647949, "created_at": "2025-01-16T05:02:37.022893+00:00"} {"global_step": 8316, "acc_step": 0, "speed/wps": 12892.729091845646, "speed/FLOPS": 202498026853761.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061858754605054855, "optim/lr": 0.0029731895402057874, "optim/total_tokens": 4359979008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.930781841278076, "created_at": "2025-01-16T05:02:47.193151+00:00"} {"global_step": 8317, "acc_step": 0, "speed/wps": 12891.747806985504, "speed/FLOPS": 202482614426605.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045330654829740524, "optim/lr": 0.0029731734109210814, "optim/total_tokens": 4360503296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.973027229309082, "created_at": "2025-01-16T05:02:57.364110+00:00"} {"global_step": 8318, "acc_step": 0, "speed/wps": 12889.511982371352, "speed/FLOPS": 202447497728695.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06158747896552086, "optim/lr": 0.002973157276829888, "optim/total_tokens": 4361027584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466705, "loss/out": 2.913670539855957, "created_at": "2025-01-16T05:03:07.551314+00:00"} {"global_step": 8319, "acc_step": 0, "speed/wps": 12892.39026732532, "speed/FLOPS": 202492705149075.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055423006415367126, "optim/lr": 0.0029731411379322606, "optim/total_tokens": 4361551872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.083836555480957, "created_at": "2025-01-16T05:03:17.723655+00:00"} {"global_step": 8320, "acc_step": 0, "speed/wps": 12888.535416211063, "speed/FLOPS": 202432159415205.72, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04496816173195839, "optim/lr": 0.0029731249942282502, "optim/total_tokens": 4362076160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8992414474487305, "created_at": "2025-01-16T05:03:27.897594+00:00"} {"global_step": 8321, "acc_step": 0, "speed/wps": 12891.85145675584, "speed/FLOPS": 202484242388678.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06041213870048523, "optim/lr": 0.002973108845717912, "optim/total_tokens": 4362600448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.8855361938476562, "created_at": "2025-01-16T05:03:38.069817+00:00"} {"global_step": 8322, "acc_step": 0, "speed/wps": 12889.647664860822, "speed/FLOPS": 202449628808640.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06651131808757782, "optim/lr": 0.0029730926924012957, "optim/total_tokens": 4363124736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9885189533233643, "created_at": "2025-01-16T05:03:48.247237+00:00"} {"global_step": 8323, "acc_step": 0, "speed/wps": 12888.34542868289, "speed/FLOPS": 202429175400005.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07920010387897491, "optim/lr": 0.0029730765342784564, "optim/total_tokens": 4363649024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 3.011383533477783, "created_at": "2025-01-16T05:03:58.422188+00:00"} {"global_step": 8324, "acc_step": 0, "speed/wps": 12886.642434976553, "speed/FLOPS": 202402427543688.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04935160279273987, "optim/lr": 0.0029730603713494456, "optim/total_tokens": 4364173312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9135162830352783, "created_at": "2025-01-16T05:04:08.599825+00:00"} {"global_step": 8325, "acc_step": 0, "speed/wps": 12887.973720095837, "speed/FLOPS": 202423337205864.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06030901148915291, "optim/lr": 0.002973044203614316, "optim/total_tokens": 4364697600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 2.9649767875671387, "created_at": "2025-01-16T05:04:18.775270+00:00"} {"global_step": 8326, "acc_step": 0, "speed/wps": 12895.188335423698, "speed/FLOPS": 202536652653509.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05063270777463913, "optim/lr": 0.002973028031073122, "optim/total_tokens": 4365221888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 3.0490589141845703, "created_at": "2025-01-16T05:04:28.946144+00:00"} {"global_step": 8327, "acc_step": 0, "speed/wps": 12888.94471586803, "speed/FLOPS": 202438588028755.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06169424578547478, "optim/lr": 0.002973011853725914, "optim/total_tokens": 4365746176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.961911201477051, "created_at": "2025-01-16T05:04:39.122727+00:00"} {"global_step": 8328, "acc_step": 0, "speed/wps": 12894.574638968475, "speed/FLOPS": 202527013707372.6, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0525791272521019, "optim/lr": 0.0029729956715727457, "optim/total_tokens": 4366270464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.0049710273742676, "created_at": "2025-01-16T05:04:49.292471+00:00"} {"global_step": 8329, "acc_step": 0, "speed/wps": 12889.71462886283, "speed/FLOPS": 202450680570309.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04688243195414543, "optim/lr": 0.0029729794846136705, "optim/total_tokens": 4366794752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 3.0106749534606934, "created_at": "2025-01-16T05:04:59.468240+00:00"} {"global_step": 8330, "acc_step": 0, "speed/wps": 12893.564250974112, "speed/FLOPS": 202511144175502.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06047474220395088, "optim/lr": 0.0029729632928487404, "optim/total_tokens": 4367319040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.871199607849121, "created_at": "2025-01-16T05:05:09.638726+00:00"} {"global_step": 8331, "acc_step": 0, "speed/wps": 12894.688118883052, "speed/FLOPS": 202528796065368.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06149827316403389, "optim/lr": 0.0029729470962780087, "optim/total_tokens": 4367843328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.8581066131591797, "created_at": "2025-01-16T05:05:19.810103+00:00"} {"global_step": 8332, "acc_step": 0, "speed/wps": 12896.082846926296, "speed/FLOPS": 202550702185845.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05853939801454544, "optim/lr": 0.0029729308949015287, "optim/total_tokens": 4368367616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439571, "loss/out": 3.037795066833496, "created_at": "2025-01-16T05:05:29.977323+00:00"} {"global_step": 8333, "acc_step": 0, "speed/wps": 12889.666396153467, "speed/FLOPS": 202449923009330.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0435480922460556, "optim/lr": 0.002972914688719352, "optim/total_tokens": 4368891904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 2.997483730316162, "created_at": "2025-01-16T05:05:40.153999+00:00"} {"global_step": 8334, "acc_step": 0, "speed/wps": 12890.63785008875, "speed/FLOPS": 202465181028301.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059102028608322144, "optim/lr": 0.002972898477731532, "optim/total_tokens": 4369416192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.957174777984619, "created_at": "2025-01-16T05:05:50.328659+00:00"} {"global_step": 8335, "acc_step": 0, "speed/wps": 12883.301846625965, "speed/FLOPS": 202349959013193.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05588135868310928, "optim/lr": 0.002972882261938122, "optim/total_tokens": 4369940480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.0359139442443848, "created_at": "2025-01-16T05:06:00.506575+00:00"} {"global_step": 8336, "acc_step": 0, "speed/wps": 12891.104573497414, "speed/FLOPS": 202472511560779.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04573448374867439, "optim/lr": 0.002972866041339175, "optim/total_tokens": 4370464768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.0258102416992188, "created_at": "2025-01-16T05:06:10.677972+00:00"} {"global_step": 8337, "acc_step": 0, "speed/wps": 12889.266650051452, "speed/FLOPS": 202443644447489.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04695573449134827, "optim/lr": 0.0029728498159347434, "optim/total_tokens": 4370989056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.0794968605041504, "created_at": "2025-01-16T05:06:20.852275+00:00"} {"global_step": 8338, "acc_step": 0, "speed/wps": 12891.942812259027, "speed/FLOPS": 202485677252391.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05163365602493286, "optim/lr": 0.0029728335857248807, "optim/total_tokens": 4371513344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.987173080444336, "created_at": "2025-01-16T05:06:31.023694+00:00"} {"global_step": 8339, "acc_step": 0, "speed/wps": 12887.575654458478, "speed/FLOPS": 202417085038030.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05255370959639549, "optim/lr": 0.002972817350709639, "optim/total_tokens": 4372037632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374944, "loss/out": 2.9391651153564453, "created_at": "2025-01-16T05:06:41.197869+00:00"} {"global_step": 8340, "acc_step": 0, "speed/wps": 12889.546904969224, "speed/FLOPS": 202448046236083.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060760319232940674, "optim/lr": 0.002972801110889072, "optim/total_tokens": 4372561920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.050947666168213, "created_at": "2025-01-16T05:06:51.371233+00:00"} {"global_step": 8341, "acc_step": 0, "speed/wps": 12887.530572178499, "speed/FLOPS": 202416376958873.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058396678417921066, "optim/lr": 0.002972784866263232, "optim/total_tokens": 4373086208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.919981002807617, "created_at": "2025-01-16T05:07:01.547399+00:00"} {"global_step": 8342, "acc_step": 0, "speed/wps": 12892.920416918467, "speed/FLOPS": 202501031876937.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06568552553653717, "optim/lr": 0.0029727686168321728, "optim/total_tokens": 4373610496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.974053382873535, "created_at": "2025-01-16T05:07:11.717499+00:00"} {"global_step": 8343, "acc_step": 0, "speed/wps": 12893.305637826308, "speed/FLOPS": 202507082300644.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05764790251851082, "optim/lr": 0.002972752362595947, "optim/total_tokens": 4374134784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 3.0281105041503906, "created_at": "2025-01-16T05:07:21.890415+00:00"} {"global_step": 8344, "acc_step": 0, "speed/wps": 12897.07324701297, "speed/FLOPS": 202566257780158.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05373821780085564, "optim/lr": 0.0029727361035546077, "optim/total_tokens": 4374659072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.00085711479187, "created_at": "2025-01-16T05:07:32.067970+00:00"} {"global_step": 8345, "acc_step": 0, "speed/wps": 12895.148144547722, "speed/FLOPS": 202536021400573.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07419519871473312, "optim/lr": 0.0029727198397082075, "optim/total_tokens": 4375183360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.945680618286133, "created_at": "2025-01-16T05:07:42.237095+00:00"} {"global_step": 8346, "acc_step": 0, "speed/wps": 12892.432942427418, "speed/FLOPS": 202493375420195.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07520800828933716, "optim/lr": 0.0029727035710568005, "optim/total_tokens": 4375707648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.8337535858154297, "created_at": "2025-01-16T05:07:52.409967+00:00"} {"global_step": 8347, "acc_step": 0, "speed/wps": 12889.720115965834, "speed/FLOPS": 202450766752802.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06560519337654114, "optim/lr": 0.0029726872976004385, "optim/total_tokens": 4376231936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.930391550064087, "created_at": "2025-01-16T05:08:02.586095+00:00"} {"global_step": 8348, "acc_step": 0, "speed/wps": 12886.148024288717, "speed/FLOPS": 202394662144445.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05414848402142525, "optim/lr": 0.0029726710193391757, "optim/total_tokens": 4376756224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0253376960754395, "created_at": "2025-01-16T05:08:12.761979+00:00"} {"global_step": 8349, "acc_step": 0, "speed/wps": 12888.340172904358, "speed/FLOPS": 202429092850781.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06424266844987869, "optim/lr": 0.002972654736273065, "optim/total_tokens": 4377280512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 3.072497844696045, "created_at": "2025-01-16T05:08:22.937248+00:00"} {"global_step": 8350, "acc_step": 0, "speed/wps": 12875.897291435364, "speed/FLOPS": 202233660299000.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05359131842851639, "optim/lr": 0.0029726384484021593, "optim/total_tokens": 4377804800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.970522880554199, "created_at": "2025-01-16T05:08:33.123292+00:00"} {"global_step": 8351, "acc_step": 0, "speed/wps": 12885.043259353843, "speed/FLOPS": 202377310293036.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04947032779455185, "optim/lr": 0.002972622155726511, "optim/total_tokens": 4378329088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.0052437782287598, "created_at": "2025-01-16T05:08:43.300208+00:00"} {"global_step": 8352, "acc_step": 0, "speed/wps": 12884.327299612392, "speed/FLOPS": 202366065161465.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058005768805742264, "optim/lr": 0.0029726058582461748, "optim/total_tokens": 4378853376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.932379722595215, "created_at": "2025-01-16T05:08:53.481376+00:00"} {"global_step": 8353, "acc_step": 0, "speed/wps": 12891.76092918031, "speed/FLOPS": 202482820528707.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07312379032373428, "optim/lr": 0.002972589555961203, "optim/total_tokens": 4379377664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411271, "loss/out": 2.964643955230713, "created_at": "2025-01-16T05:09:03.654793+00:00"} {"global_step": 8354, "acc_step": 0, "speed/wps": 12889.269468960332, "speed/FLOPS": 202443688722327.2, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05263080075383186, "optim/lr": 0.002972573248871648, "optim/total_tokens": 4379901952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 3.010462760925293, "created_at": "2025-01-16T05:09:13.832558+00:00"} {"global_step": 8355, "acc_step": 0, "speed/wps": 12888.17771563464, "speed/FLOPS": 202426541236120.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056158553808927536, "optim/lr": 0.0029725569369775653, "optim/total_tokens": 4380426240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 3.03096604347229, "created_at": "2025-01-16T05:09:24.006396+00:00"} {"global_step": 8356, "acc_step": 0, "speed/wps": 12889.244756331693, "speed/FLOPS": 202443300576535.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050531573593616486, "optim/lr": 0.0029725406202790053, "optim/total_tokens": 4380950528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.9158384799957275, "created_at": "2025-01-16T05:09:34.180625+00:00"} {"global_step": 8357, "acc_step": 0, "speed/wps": 12886.811174787668, "speed/FLOPS": 202405077834295.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04702964052557945, "optim/lr": 0.0029725242987760233, "optim/total_tokens": 4381474816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306740, "loss/out": 3.0150256156921387, "created_at": "2025-01-16T05:09:44.355103+00:00"} {"global_step": 8358, "acc_step": 0, "speed/wps": 12890.550033873185, "speed/FLOPS": 202463801753964.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.069234199821949, "optim/lr": 0.002972507972468672, "optim/total_tokens": 4381999104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8908004760742188, "created_at": "2025-01-16T05:09:54.529849+00:00"} {"global_step": 8359, "acc_step": 0, "speed/wps": 12895.383951689968, "speed/FLOPS": 202539725075775.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06028049811720848, "optim/lr": 0.0029724916413570047, "optim/total_tokens": 4382523392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.886821746826172, "created_at": "2025-01-16T05:10:04.697749+00:00"} {"global_step": 8360, "acc_step": 0, "speed/wps": 12893.016526281357, "speed/FLOPS": 202502541406550.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05843457207083702, "optim/lr": 0.002972475305441074, "optim/total_tokens": 4383047680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 2.9281985759735107, "created_at": "2025-01-16T05:10:14.871192+00:00"} {"global_step": 8361, "acc_step": 0, "speed/wps": 12891.108331934041, "speed/FLOPS": 202472570592191.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059997543692588806, "optim/lr": 0.0029724589647209337, "optim/total_tokens": 4383571968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.83994722366333, "created_at": "2025-01-16T05:10:25.042500+00:00"} {"global_step": 8362, "acc_step": 0, "speed/wps": 12896.3881418173, "speed/FLOPS": 202555497261627.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05483630672097206, "optim/lr": 0.0029724426191966376, "optim/total_tokens": 4384096256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 3.015887498855591, "created_at": "2025-01-16T05:10:35.215482+00:00"} {"global_step": 8363, "acc_step": 0, "speed/wps": 12878.917451080906, "speed/FLOPS": 202281096056367.53, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0635855421423912, "optim/lr": 0.0029724262688682382, "optim/total_tokens": 4384620544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.935004234313965, "created_at": "2025-01-16T05:10:45.398117+00:00"} {"global_step": 8364, "acc_step": 0, "speed/wps": 12880.617044292734, "speed/FLOPS": 202307790503245.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06599649041891098, "optim/lr": 0.00297240991373579, "optim/total_tokens": 4385144832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.0777344703674316, "created_at": "2025-01-16T05:10:55.577950+00:00"} {"global_step": 8365, "acc_step": 0, "speed/wps": 12885.703540146083, "speed/FLOPS": 202387680910199.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058774612843990326, "optim/lr": 0.0029723935537993443, "optim/total_tokens": 4385669120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.970289707183838, "created_at": "2025-01-16T05:11:05.757132+00:00"} {"global_step": 8366, "acc_step": 0, "speed/wps": 12879.614132309995, "speed/FLOPS": 202292038392410.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.048580944538116455, "optim/lr": 0.002972377189058957, "optim/total_tokens": 4386193408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9518203735351562, "created_at": "2025-01-16T05:11:15.939359+00:00"} {"global_step": 8367, "acc_step": 0, "speed/wps": 12889.563315102285, "speed/FLOPS": 202448303979773.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0539759136736393, "optim/lr": 0.0029723608195146795, "optim/total_tokens": 4386717696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8992161750793457, "created_at": "2025-01-16T05:11:26.130751+00:00"} {"global_step": 8368, "acc_step": 0, "speed/wps": 12886.014891975046, "speed/FLOPS": 202392571118515.8, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06134547293186188, "optim/lr": 0.0029723444451665664, "optim/total_tokens": 4387241984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9345788955688477, "created_at": "2025-01-16T05:11:36.308404+00:00"} {"global_step": 8369, "acc_step": 0, "speed/wps": 12894.212115137323, "speed/FLOPS": 202521319772444.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0653478130698204, "optim/lr": 0.00297232806601467, "optim/total_tokens": 4387766272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.0641071796417236, "created_at": "2025-01-16T05:11:46.477630+00:00"} {"global_step": 8370, "acc_step": 0, "speed/wps": 12889.789709789713, "speed/FLOPS": 202451859819438.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061819255352020264, "optim/lr": 0.0029723116820590447, "optim/total_tokens": 4388290560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9820480346679688, "created_at": "2025-01-16T05:11:56.649794+00:00"} {"global_step": 8371, "acc_step": 0, "speed/wps": 12893.706167249324, "speed/FLOPS": 202513373165618.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061362605541944504, "optim/lr": 0.002972295293299744, "optim/total_tokens": 4388814848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421736, "loss/out": 2.829768419265747, "created_at": "2025-01-16T05:12:06.821631+00:00"} {"global_step": 8372, "acc_step": 0, "speed/wps": 12890.152276596626, "speed/FLOPS": 202457554429359.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06459882110357285, "optim/lr": 0.002972278899736821, "optim/total_tokens": 4389339136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455791, "loss/out": 2.9924569129943848, "created_at": "2025-01-16T05:12:16.993576+00:00"} {"global_step": 8373, "acc_step": 0, "speed/wps": 12890.339918329599, "speed/FLOPS": 202460501600623.1, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0665016919374466, "optim/lr": 0.002972262501370329, "optim/total_tokens": 4389863424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.884678840637207, "created_at": "2025-01-16T05:12:27.165528+00:00"} {"global_step": 8374, "acc_step": 0, "speed/wps": 12895.611988913928, "speed/FLOPS": 202543306713734.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06074081361293793, "optim/lr": 0.0029722460982003216, "optim/total_tokens": 4390387712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.977475166320801, "created_at": "2025-01-16T05:12:37.337379+00:00"} {"global_step": 8375, "acc_step": 0, "speed/wps": 12876.944137725139, "speed/FLOPS": 202250102458498.5, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05791379511356354, "optim/lr": 0.0029722296902268534, "optim/total_tokens": 4390912000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488197, "loss/out": 2.903644561767578, "created_at": "2025-01-16T05:12:47.520706+00:00"} {"global_step": 8376, "acc_step": 0, "speed/wps": 12885.980381402147, "speed/FLOPS": 202392029082545.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06046156585216522, "optim/lr": 0.0029722132774499758, "optim/total_tokens": 4391436288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 3.0278353691101074, "created_at": "2025-01-16T05:12:57.698133+00:00"} {"global_step": 8377, "acc_step": 0, "speed/wps": 12898.145933961727, "speed/FLOPS": 202583105802715.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05325498431921005, "optim/lr": 0.0029721968598697443, "optim/total_tokens": 4391960576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.9387545585632324, "created_at": "2025-01-16T05:13:07.872965+00:00"} {"global_step": 8378, "acc_step": 0, "speed/wps": 12886.748682416403, "speed/FLOPS": 202404096305731.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051288627088069916, "optim/lr": 0.002972180437486212, "optim/total_tokens": 4392484864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.93642520904541, "created_at": "2025-01-16T05:13:18.048505+00:00"} {"global_step": 8379, "acc_step": 0, "speed/wps": 12889.38770960526, "speed/FLOPS": 202445545854135.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05255371332168579, "optim/lr": 0.0029721640102994312, "optim/total_tokens": 4393009152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 2.9720826148986816, "created_at": "2025-01-16T05:13:28.221707+00:00"} {"global_step": 8380, "acc_step": 0, "speed/wps": 12894.139789545676, "speed/FLOPS": 202520183799641.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04667624086141586, "optim/lr": 0.0029721475783094572, "optim/total_tokens": 4393533440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.8250679969787598, "created_at": "2025-01-16T05:13:38.392269+00:00"} {"global_step": 8381, "acc_step": 0, "speed/wps": 12894.282895131626, "speed/FLOPS": 202522431469516.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04792284220457077, "optim/lr": 0.0029721311415163423, "optim/total_tokens": 4394057728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.7305514812469482, "created_at": "2025-01-16T05:13:48.562917+00:00"} {"global_step": 8382, "acc_step": 0, "speed/wps": 12895.159501374983, "speed/FLOPS": 202536199775151.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05933455377817154, "optim/lr": 0.0029721146999201415, "optim/total_tokens": 4394582016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.022648811340332, "created_at": "2025-01-16T05:13:58.731429+00:00"} {"global_step": 8383, "acc_step": 0, "speed/wps": 12888.137485830357, "speed/FLOPS": 202425909371763.1, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049309078603982925, "optim/lr": 0.0029720982535209072, "optim/total_tokens": 4395106304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288006, "loss/out": 2.9858553409576416, "created_at": "2025-01-16T05:14:08.907175+00:00"} {"global_step": 8384, "acc_step": 0, "speed/wps": 12887.41634610841, "speed/FLOPS": 202414582881480.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07591699808835983, "optim/lr": 0.002972081802318694, "optim/total_tokens": 4395630592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0568909645080566, "created_at": "2025-01-16T05:14:19.082086+00:00"} {"global_step": 8385, "acc_step": 0, "speed/wps": 12891.925744190723, "speed/FLOPS": 202485409174925.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06249762326478958, "optim/lr": 0.002972065346313554, "optim/total_tokens": 4396154880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8742876052856445, "created_at": "2025-01-16T05:14:29.261744+00:00"} {"global_step": 8386, "acc_step": 0, "speed/wps": 12893.292937713244, "speed/FLOPS": 202506882827917.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06126194819808006, "optim/lr": 0.0029720488855055427, "optim/total_tokens": 4396679168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 3.0064520835876465, "created_at": "2025-01-16T05:14:39.459293+00:00"} {"global_step": 8387, "acc_step": 0, "speed/wps": 12893.43439206852, "speed/FLOPS": 202509104562945.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07564049959182739, "optim/lr": 0.0029720324198947135, "optim/total_tokens": 4397203456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 2.9326112270355225, "created_at": "2025-01-16T05:14:49.628628+00:00"} {"global_step": 8388, "acc_step": 0, "speed/wps": 12892.418974431725, "speed/FLOPS": 202493156033629.44, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0504060722887516, "optim/lr": 0.002972015949481119, "optim/total_tokens": 4397727744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.023984909057617, "created_at": "2025-01-16T05:14:59.799491+00:00"} {"global_step": 8389, "acc_step": 0, "speed/wps": 12884.044015048667, "speed/FLOPS": 202361615788118.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08320779353380203, "optim/lr": 0.0029719994742648144, "optim/total_tokens": 4398252032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.7769081592559814, "created_at": "2025-01-16T05:15:09.979464+00:00"} {"global_step": 8390, "acc_step": 0, "speed/wps": 12896.052600301737, "speed/FLOPS": 202550227121043.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05977492779493332, "optim/lr": 0.002971982994245852, "optim/total_tokens": 4398776320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 2.984036445617676, "created_at": "2025-01-16T05:15:20.148595+00:00"} {"global_step": 8391, "acc_step": 0, "speed/wps": 12888.482412094621, "speed/FLOPS": 202431326912721.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04934001713991165, "optim/lr": 0.0029719665094242865, "optim/total_tokens": 4399300608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.9487757682800293, "created_at": "2025-01-16T05:15:30.321948+00:00"} {"global_step": 8392, "acc_step": 0, "speed/wps": 12889.257780176189, "speed/FLOPS": 202443505133910.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05837618187069893, "optim/lr": 0.0029719500198001716, "optim/total_tokens": 4399824896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.935197353363037, "created_at": "2025-01-16T05:15:40.494607+00:00"} {"global_step": 8393, "acc_step": 0, "speed/wps": 12891.3572237208, "speed/FLOPS": 202476479779716.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0643341913819313, "optim/lr": 0.002971933525373561, "optim/total_tokens": 4400349184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407486, "loss/out": 2.960141658782959, "created_at": "2025-01-16T05:15:50.667453+00:00"} {"global_step": 8394, "acc_step": 0, "speed/wps": 12889.40732549715, "speed/FLOPS": 202445853948672.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06412218511104584, "optim/lr": 0.0029719170261445086, "optim/total_tokens": 4400873472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.0147385597229004, "created_at": "2025-01-16T05:16:00.839933+00:00"} {"global_step": 8395, "acc_step": 0, "speed/wps": 12891.551619660022, "speed/FLOPS": 202479533035069.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052820295095443726, "optim/lr": 0.002971900522113068, "optim/total_tokens": 4401397760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.847700357437134, "created_at": "2025-01-16T05:16:11.011043+00:00"} {"global_step": 8396, "acc_step": 0, "speed/wps": 12891.753010041191, "speed/FLOPS": 202482696147745.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0740148201584816, "optim/lr": 0.0029718840132792928, "optim/total_tokens": 4401922048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.958700656890869, "created_at": "2025-01-16T05:16:21.181744+00:00"} {"global_step": 8397, "acc_step": 0, "speed/wps": 12889.57137321184, "speed/FLOPS": 202448430543457.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06697135418653488, "optim/lr": 0.002971867499643238, "optim/total_tokens": 4402446336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.958385944366455, "created_at": "2025-01-16T05:16:31.357078+00:00"} {"global_step": 8398, "acc_step": 0, "speed/wps": 12889.10905399441, "speed/FLOPS": 202441169184854.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.12875863909721375, "optim/lr": 0.002971850981204956, "optim/total_tokens": 4402970624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 3.0508720874786377, "created_at": "2025-01-16T05:16:41.533161+00:00"} {"global_step": 8399, "acc_step": 0, "speed/wps": 12894.304081790362, "speed/FLOPS": 202522764235105.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09570871293544769, "optim/lr": 0.0029718344579645014, "optim/total_tokens": 4403494912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9637391567230225, "created_at": "2025-01-16T05:16:51.702218+00:00"} {"global_step": 8400, "acc_step": 0, "speed/wps": 12889.708883228785, "speed/FLOPS": 202450590327231.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04681511968374252, "optim/lr": 0.0029718179299219284, "optim/total_tokens": 4404019200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 2.9436757564544678, "created_at": "2025-01-16T05:17:01.876088+00:00"} {"global_step": 8401, "acc_step": 0, "speed/wps": 12895.668015190418, "speed/FLOPS": 202544186683394.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06722450256347656, "optim/lr": 0.0029718013970772904, "optim/total_tokens": 4404543488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.012921094894409, "created_at": "2025-01-16T05:17:12.048666+00:00"} {"global_step": 8402, "acc_step": 0, "speed/wps": 12893.210864780804, "speed/FLOPS": 202505593759733.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05544142797589302, "optim/lr": 0.002971784859430642, "optim/total_tokens": 4405067776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 3.0633115768432617, "created_at": "2025-01-16T05:17:22.219814+00:00"} {"global_step": 8403, "acc_step": 0, "speed/wps": 12888.962679043434, "speed/FLOPS": 202438870165108.75, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07010970264673233, "optim/lr": 0.002971768316982036, "optim/total_tokens": 4405592064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.0719072818756104, "created_at": "2025-01-16T05:17:32.398116+00:00"} {"global_step": 8404, "acc_step": 0, "speed/wps": 12896.502157089357, "speed/FLOPS": 202557288028148.5, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05554795265197754, "optim/lr": 0.002971751769731528, "optim/total_tokens": 4406116352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8432509899139404, "created_at": "2025-01-16T05:17:42.565458+00:00"} {"global_step": 8405, "acc_step": 0, "speed/wps": 12890.935283173463, "speed/FLOPS": 202469852623613.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06387735903263092, "optim/lr": 0.0029717352176791704, "optim/total_tokens": 4406640640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9533534049987793, "created_at": "2025-01-16T05:17:52.737107+00:00"} {"global_step": 8406, "acc_step": 0, "speed/wps": 12892.463149635498, "speed/FLOPS": 202493849865907.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05678956210613251, "optim/lr": 0.0029717186608250183, "optim/total_tokens": 4407164928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9787697792053223, "created_at": "2025-01-16T05:18:02.910079+00:00"} {"global_step": 8407, "acc_step": 0, "speed/wps": 12887.14853432057, "speed/FLOPS": 202410376529340.1, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04802286997437477, "optim/lr": 0.0029717020991691253, "optim/total_tokens": 4407689216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291378, "loss/out": 2.9383583068847656, "created_at": "2025-01-16T05:18:13.084518+00:00"} {"global_step": 8408, "acc_step": 0, "speed/wps": 12898.8262122284, "speed/FLOPS": 202593790507693.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05969424173235893, "optim/lr": 0.002971685532711546, "optim/total_tokens": 4408213504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 3.0637590885162354, "created_at": "2025-01-16T05:18:23.274173+00:00"} {"global_step": 8409, "acc_step": 0, "speed/wps": 12893.898303340415, "speed/FLOPS": 202516390926951.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05547111853957176, "optim/lr": 0.0029716689614523325, "optim/total_tokens": 4408737792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.1182007789611816, "created_at": "2025-01-16T05:18:33.443139+00:00"} {"global_step": 8410, "acc_step": 0, "speed/wps": 12892.863462541993, "speed/FLOPS": 202500137330191.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06868090480566025, "optim/lr": 0.002971652385391541, "optim/total_tokens": 4409262080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417722, "loss/out": 2.9281601905822754, "created_at": "2025-01-16T05:18:43.615779+00:00"} {"global_step": 8411, "acc_step": 0, "speed/wps": 12892.885989821018, "speed/FLOPS": 202500491152064.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06026829034090042, "optim/lr": 0.002971635804529225, "optim/total_tokens": 4409786368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9637296199798584, "created_at": "2025-01-16T05:18:53.787919+00:00"} {"global_step": 8412, "acc_step": 0, "speed/wps": 12896.114623996653, "speed/FLOPS": 202551201288401.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047950439155101776, "optim/lr": 0.0029716192188654384, "optim/total_tokens": 4410310656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.793288230895996, "created_at": "2025-01-16T05:19:03.960966+00:00"} {"global_step": 8413, "acc_step": 0, "speed/wps": 12879.441050195392, "speed/FLOPS": 202289319899961.1, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05791905149817467, "optim/lr": 0.0029716026284002353, "optim/total_tokens": 4410834944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.013953447341919, "created_at": "2025-01-16T05:19:14.145432+00:00"} {"global_step": 8414, "acc_step": 0, "speed/wps": 12881.90048892158, "speed/FLOPS": 202327948764778.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05292527377605438, "optim/lr": 0.00297158603313367, "optim/total_tokens": 4411359232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8958077430725098, "created_at": "2025-01-16T05:19:24.324692+00:00"} {"global_step": 8415, "acc_step": 0, "speed/wps": 12892.144080231721, "speed/FLOPS": 202488838442474.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053120553493499756, "optim/lr": 0.002971569433065796, "optim/total_tokens": 4411883520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.84804368019104, "created_at": "2025-01-16T05:19:34.497790+00:00"} {"global_step": 8416, "acc_step": 0, "speed/wps": 12890.780880313738, "speed/FLOPS": 202467427514529.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07006318122148514, "optim/lr": 0.002971552828196669, "optim/total_tokens": 4412407808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9329824447631836, "created_at": "2025-01-16T05:19:44.671402+00:00"} {"global_step": 8417, "acc_step": 0, "speed/wps": 12887.93929462666, "speed/FLOPS": 202422796506565.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04649433493614197, "optim/lr": 0.0029715362185263414, "optim/total_tokens": 4412932096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397147, "loss/out": 2.894390821456909, "created_at": "2025-01-16T05:19:54.847827+00:00"} {"global_step": 8418, "acc_step": 0, "speed/wps": 12889.635859700256, "speed/FLOPS": 202449443392371.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0718446671962738, "optim/lr": 0.0029715196040548683, "optim/total_tokens": 4413456384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 2.927840232849121, "created_at": "2025-01-16T05:20:05.022112+00:00"} {"global_step": 8419, "acc_step": 0, "speed/wps": 12881.647645645735, "speed/FLOPS": 202323977513688.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06169985979795456, "optim/lr": 0.0029715029847823043, "optim/total_tokens": 4413980672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.972450017929077, "created_at": "2025-01-16T05:20:15.201018+00:00"} {"global_step": 8420, "acc_step": 0, "speed/wps": 12891.313805917607, "speed/FLOPS": 202475797843454.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05308106541633606, "optim/lr": 0.0029714863607087027, "optim/total_tokens": 4414504960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.913539409637451, "created_at": "2025-01-16T05:20:25.376845+00:00"} {"global_step": 8421, "acc_step": 0, "speed/wps": 12886.81375898299, "speed/FLOPS": 202405118422633.88, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056654322892427444, "optim/lr": 0.0029714697318341176, "optim/total_tokens": 4415029248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.085404396057129, "created_at": "2025-01-16T05:20:35.551514+00:00"} {"global_step": 8422, "acc_step": 0, "speed/wps": 12893.67751937939, "speed/FLOPS": 202512923211454.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06381352990865707, "optim/lr": 0.0029714530981586047, "optim/total_tokens": 4415553536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 2.8335609436035156, "created_at": "2025-01-16T05:20:45.724077+00:00"} {"global_step": 8423, "acc_step": 0, "speed/wps": 12895.608056517403, "speed/FLOPS": 202543244950043.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04905370995402336, "optim/lr": 0.0029714364596822164, "optim/total_tokens": 4416077824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.0247011184692383, "created_at": "2025-01-16T05:20:55.897999+00:00"} {"global_step": 8424, "acc_step": 0, "speed/wps": 12888.43061939589, "speed/FLOPS": 202430513437217.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057880375534296036, "optim/lr": 0.002971419816405009, "optim/total_tokens": 4416602112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415331, "loss/out": 2.952240467071533, "created_at": "2025-01-16T05:21:06.072883+00:00"} {"global_step": 8425, "acc_step": 0, "speed/wps": 12889.05448702031, "speed/FLOPS": 202440312135543.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04797602444887161, "optim/lr": 0.002971403168327035, "optim/total_tokens": 4417126400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 2.9798808097839355, "created_at": "2025-01-16T05:21:16.246945+00:00"} {"global_step": 8426, "acc_step": 0, "speed/wps": 12878.807588535261, "speed/FLOPS": 202279370514121.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0643363818526268, "optim/lr": 0.00297138651544835, "optim/total_tokens": 4417650688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9959521293640137, "created_at": "2025-01-16T05:21:26.433621+00:00"} {"global_step": 8427, "acc_step": 0, "speed/wps": 12885.096501795342, "speed/FLOPS": 202378146538743.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05715765431523323, "optim/lr": 0.0029713698577690075, "optim/total_tokens": 4418174976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.0198841094970703, "created_at": "2025-01-16T05:21:36.611225+00:00"} {"global_step": 8428, "acc_step": 0, "speed/wps": 12882.927076186235, "speed/FLOPS": 202344072728449.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04449479281902313, "optim/lr": 0.0029713531952890618, "optim/total_tokens": 4418699264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.021381139755249, "created_at": "2025-01-16T05:21:46.789168+00:00"} {"global_step": 8429, "acc_step": 0, "speed/wps": 12890.396254182202, "speed/FLOPS": 202461386432601.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04862036183476448, "optim/lr": 0.0029713365280085685, "optim/total_tokens": 4419223552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 2.9438014030456543, "created_at": "2025-01-16T05:21:56.961405+00:00"} {"global_step": 8430, "acc_step": 0, "speed/wps": 12889.063390847672, "speed/FLOPS": 202440451982386.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05327359586954117, "optim/lr": 0.0029713198559275806, "optim/total_tokens": 4419747840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.9261157512664795, "created_at": "2025-01-16T05:22:07.134601+00:00"} {"global_step": 8431, "acc_step": 0, "speed/wps": 12880.185646551508, "speed/FLOPS": 202301014808916.53, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06472162902355194, "optim/lr": 0.0029713031790461527, "optim/total_tokens": 4420272128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8246572017669678, "created_at": "2025-01-16T05:22:17.314683+00:00"} {"global_step": 8432, "acc_step": 0, "speed/wps": 12890.49715250463, "speed/FLOPS": 202462971179404.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04542526602745056, "optim/lr": 0.00297128649736434, "optim/total_tokens": 4420796416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462981, "loss/out": 2.9802603721618652, "created_at": "2025-01-16T05:22:27.486475+00:00"} {"global_step": 8433, "acc_step": 0, "speed/wps": 12891.41311738472, "speed/FLOPS": 202477357666513.8, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04565617814660072, "optim/lr": 0.002971269810882197, "optim/total_tokens": 4421320704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.994760751724243, "created_at": "2025-01-16T05:22:37.657929+00:00"} {"global_step": 8434, "acc_step": 0, "speed/wps": 12889.994277554446, "speed/FLOPS": 202455072837289.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059361834079027176, "optim/lr": 0.0029712531195997764, "optim/total_tokens": 4421844992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.0298447608947754, "created_at": "2025-01-16T05:22:47.832076+00:00"} {"global_step": 8435, "acc_step": 0, "speed/wps": 12890.798432991345, "speed/FLOPS": 202467703203450.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0785861387848854, "optim/lr": 0.002971236423517134, "optim/total_tokens": 4422369280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.969801902770996, "created_at": "2025-01-16T05:22:58.003603+00:00"} {"global_step": 8436, "acc_step": 0, "speed/wps": 12894.013133409193, "speed/FLOPS": 202518194490974.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051747314631938934, "optim/lr": 0.002971219722634324, "optim/total_tokens": 4422893568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286610, "loss/out": 2.956836700439453, "created_at": "2025-01-16T05:23:08.172905+00:00"} {"global_step": 8437, "acc_step": 0, "speed/wps": 12888.018696896126, "speed/FLOPS": 202424043628317.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0692676305770874, "optim/lr": 0.0029712030169514016, "optim/total_tokens": 4423417856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.957932233810425, "created_at": "2025-01-16T05:23:18.348517+00:00"} {"global_step": 8438, "acc_step": 0, "speed/wps": 12893.230178118927, "speed/FLOPS": 202505897102247.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04657793417572975, "optim/lr": 0.0029711863064684207, "optim/total_tokens": 4423942144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9648501873016357, "created_at": "2025-01-16T05:23:28.520429+00:00"} {"global_step": 8439, "acc_step": 0, "speed/wps": 12891.79472911072, "speed/FLOPS": 202483351403060.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0439334362745285, "optim/lr": 0.0029711695911854353, "optim/total_tokens": 4424466432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 2.943948984146118, "created_at": "2025-01-16T05:23:38.694166+00:00"} {"global_step": 8440, "acc_step": 0, "speed/wps": 12890.947148571346, "speed/FLOPS": 202470038985991.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047520969063043594, "optim/lr": 0.002971152871102501, "optim/total_tokens": 4424990720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9835400581359863, "created_at": "2025-01-16T05:23:48.871437+00:00"} {"global_step": 8441, "acc_step": 0, "speed/wps": 12892.281151489498, "speed/FLOPS": 202490991334933.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05262003466486931, "optim/lr": 0.0029711361462196714, "optim/total_tokens": 4425515008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8951663970947266, "created_at": "2025-01-16T05:23:59.043053+00:00"} {"global_step": 8442, "acc_step": 0, "speed/wps": 12891.659497387676, "speed/FLOPS": 202481227403021.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04987238347530365, "optim/lr": 0.002971119416537002, "optim/total_tokens": 4426039296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.99545955657959, "created_at": "2025-01-16T05:24:09.213825+00:00"} {"global_step": 8443, "acc_step": 0, "speed/wps": 12886.194118676534, "speed/FLOPS": 202395386120146.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04374202340841293, "optim/lr": 0.002971102682054546, "optim/total_tokens": 4426563584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.934299945831299, "created_at": "2025-01-16T05:24:19.391664+00:00"} {"global_step": 8444, "acc_step": 0, "speed/wps": 12891.11065217641, "speed/FLOPS": 202472607034786.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0575154647231102, "optim/lr": 0.00297108594277236, "optim/total_tokens": 4427087872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.029670238494873, "created_at": "2025-01-16T05:24:29.565280+00:00"} {"global_step": 8445, "acc_step": 0, "speed/wps": 12888.017986404464, "speed/FLOPS": 202424032469070.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0730682760477066, "optim/lr": 0.002971069198690496, "optim/total_tokens": 4427612160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.857235908508301, "created_at": "2025-01-16T05:24:39.741457+00:00"} {"global_step": 8446, "acc_step": 0, "speed/wps": 12886.684739282462, "speed/FLOPS": 202403091990945.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06552115082740784, "optim/lr": 0.0029710524498090113, "optim/total_tokens": 4428136448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.865907669067383, "created_at": "2025-01-16T05:24:49.917117+00:00"} {"global_step": 8447, "acc_step": 0, "speed/wps": 12890.351674012061, "speed/FLOPS": 202460686239770.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05914739519357681, "optim/lr": 0.0029710356961279586, "optim/total_tokens": 4428660736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.879289150238037, "created_at": "2025-01-16T05:25:00.092148+00:00"} {"global_step": 8448, "acc_step": 0, "speed/wps": 12889.220265720314, "speed/FLOPS": 202442915917830.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05113808438181877, "optim/lr": 0.002971018937647394, "optim/total_tokens": 4429185024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9156713485717773, "created_at": "2025-01-16T05:25:10.268185+00:00"} {"global_step": 8449, "acc_step": 0, "speed/wps": 12884.729511489602, "speed/FLOPS": 202372382451694.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08824961632490158, "optim/lr": 0.0029710021743673716, "optim/total_tokens": 4429709312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.00860857963562, "created_at": "2025-01-16T05:25:20.444374+00:00"} {"global_step": 8450, "acc_step": 0, "speed/wps": 12889.675521583924, "speed/FLOPS": 202450066336755.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05302520468831062, "optim/lr": 0.002970985406287945, "optim/total_tokens": 4430233600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.8626999855041504, "created_at": "2025-01-16T05:25:30.617150+00:00"} {"global_step": 8451, "acc_step": 0, "speed/wps": 12892.889357167605, "speed/FLOPS": 202500544040870.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0706356018781662, "optim/lr": 0.0029709686334091706, "optim/total_tokens": 4430757888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.085696220397949, "created_at": "2025-01-16T05:25:40.790267+00:00"} {"global_step": 8452, "acc_step": 0, "speed/wps": 12887.884664384566, "speed/FLOPS": 202421938463543.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06200350448489189, "optim/lr": 0.002970951855731102, "optim/total_tokens": 4431282176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8264026641845703, "created_at": "2025-01-16T05:25:50.964054+00:00"} {"global_step": 8453, "acc_step": 0, "speed/wps": 12889.60877284107, "speed/FLOPS": 202449017956026.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07214956730604172, "optim/lr": 0.002970935073253795, "optim/total_tokens": 4431806464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 3.029076337814331, "created_at": "2025-01-16T05:26:01.136381+00:00"} {"global_step": 8454, "acc_step": 0, "speed/wps": 12890.313101238862, "speed/FLOPS": 202460080401362.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08516915887594223, "optim/lr": 0.002970918285977303, "optim/total_tokens": 4432330752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.942221164703369, "created_at": "2025-01-16T05:26:11.310628+00:00"} {"global_step": 8455, "acc_step": 0, "speed/wps": 12888.190989896422, "speed/FLOPS": 202426749726642.94, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06738700717687607, "optim/lr": 0.002970901493901682, "optim/total_tokens": 4432855040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9587948322296143, "created_at": "2025-01-16T05:26:21.484302+00:00"} {"global_step": 8456, "acc_step": 0, "speed/wps": 12883.491479093707, "speed/FLOPS": 202352937451681.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07614552229642868, "optim/lr": 0.002970884697026986, "optim/total_tokens": 4433379328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.0309038162231445, "created_at": "2025-01-16T05:26:31.661726+00:00"} {"global_step": 8457, "acc_step": 0, "speed/wps": 12892.782018693331, "speed/FLOPS": 202498858142633.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06254132837057114, "optim/lr": 0.0029708678953532706, "optim/total_tokens": 4433903616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.8636693954467773, "created_at": "2025-01-16T05:26:41.831613+00:00"} {"global_step": 8458, "acc_step": 0, "speed/wps": 12879.679999641581, "speed/FLOPS": 202293072929364.47, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05420660600066185, "optim/lr": 0.0029708510888805896, "optim/total_tokens": 4434427904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.981316328048706, "created_at": "2025-01-16T05:26:52.012119+00:00"} {"global_step": 8459, "acc_step": 0, "speed/wps": 12876.4884353323, "speed/FLOPS": 202242945026218.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06648200750350952, "optim/lr": 0.002970834277608998, "optim/total_tokens": 4434952192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9783143997192383, "created_at": "2025-01-16T05:27:02.194861+00:00"} {"global_step": 8460, "acc_step": 0, "speed/wps": 12889.84241063224, "speed/FLOPS": 202452687558588.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07862899452447891, "optim/lr": 0.002970817461538552, "optim/total_tokens": 4435476480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 3.027798652648926, "created_at": "2025-01-16T05:27:12.367088+00:00"} {"global_step": 8461, "acc_step": 0, "speed/wps": 12888.876306822554, "speed/FLOPS": 202437513570691.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10539036989212036, "optim/lr": 0.0029708006406693045, "optim/total_tokens": 4436000768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9289209842681885, "created_at": "2025-01-16T05:27:22.543337+00:00"} {"global_step": 8462, "acc_step": 0, "speed/wps": 12891.86396508159, "speed/FLOPS": 202484438849122.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06339680403470993, "optim/lr": 0.0029707838150013115, "optim/total_tokens": 4436525056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 2.998626708984375, "created_at": "2025-01-16T05:27:32.714212+00:00"} {"global_step": 8463, "acc_step": 0, "speed/wps": 12883.999365574471, "speed/FLOPS": 202360914506771.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06508883088827133, "optim/lr": 0.0029707669845346276, "optim/total_tokens": 4437049344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400071, "loss/out": 2.9459586143493652, "created_at": "2025-01-16T05:27:42.896613+00:00"} {"global_step": 8464, "acc_step": 0, "speed/wps": 12890.771327077098, "speed/FLOPS": 202467277467820.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07557383179664612, "optim/lr": 0.0029707501492693086, "optim/total_tokens": 4437573632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9721269607543945, "created_at": "2025-01-16T05:27:53.068597+00:00"} {"global_step": 8465, "acc_step": 0, "speed/wps": 12887.88931081091, "speed/FLOPS": 202422011442054.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04831601679325104, "optim/lr": 0.0029707333092054084, "optim/total_tokens": 4438097920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.8837456703186035, "created_at": "2025-01-16T05:28:03.242278+00:00"} {"global_step": 8466, "acc_step": 0, "speed/wps": 12886.362198973531, "speed/FLOPS": 202398026052176.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07282351702451706, "optim/lr": 0.0029707164643429816, "optim/total_tokens": 4438622208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.953700304031372, "created_at": "2025-01-16T05:28:13.417751+00:00"} {"global_step": 8467, "acc_step": 0, "speed/wps": 12885.715923284115, "speed/FLOPS": 202387875404398.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047465335577726364, "optim/lr": 0.0029706996146820846, "optim/total_tokens": 4439146496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 2.932830810546875, "created_at": "2025-01-16T05:28:23.593134+00:00"} {"global_step": 8468, "acc_step": 0, "speed/wps": 12890.48784290675, "speed/FLOPS": 202462824959376.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06673014909029007, "optim/lr": 0.0029706827602227707, "optim/total_tokens": 4439670784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.0349698066711426, "created_at": "2025-01-16T05:28:33.764986+00:00"} {"global_step": 8469, "acc_step": 0, "speed/wps": 12885.494949904636, "speed/FLOPS": 202384404713824.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05781634896993637, "optim/lr": 0.0029706659009650963, "optim/total_tokens": 4440195072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.9237782955169678, "created_at": "2025-01-16T05:28:43.940779+00:00"} {"global_step": 8470, "acc_step": 0, "speed/wps": 12888.318224498504, "speed/FLOPS": 202428748120907.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0502571240067482, "optim/lr": 0.002970649036909115, "optim/total_tokens": 4440719360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 2.8668875694274902, "created_at": "2025-01-16T05:28:54.114383+00:00"} {"global_step": 8471, "acc_step": 0, "speed/wps": 12887.503563794213, "speed/FLOPS": 202415952755085.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05567379668354988, "optim/lr": 0.0029706321680548838, "optim/total_tokens": 4441243648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.915055274963379, "created_at": "2025-01-16T05:29:04.290280+00:00"} {"global_step": 8472, "acc_step": 0, "speed/wps": 12886.406676038274, "speed/FLOPS": 202398724625596.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060878824442625046, "optim/lr": 0.002970615294402456, "optim/total_tokens": 4441767936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.9197213649749756, "created_at": "2025-01-16T05:29:14.470208+00:00"} {"global_step": 8473, "acc_step": 0, "speed/wps": 12899.24999247476, "speed/FLOPS": 202600446558797.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07805602252483368, "optim/lr": 0.0029705984159518876, "optim/total_tokens": 4442292224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.0242185592651367, "created_at": "2025-01-16T05:29:24.638304+00:00"} {"global_step": 8474, "acc_step": 0, "speed/wps": 12893.588820701843, "speed/FLOPS": 202511530076838.88, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057049866765737534, "optim/lr": 0.0029705815327032326, "optim/total_tokens": 4442816512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403804, "loss/out": 2.98463773727417, "created_at": "2025-01-16T05:29:34.807833+00:00"} {"global_step": 8475, "acc_step": 0, "speed/wps": 12885.324644852959, "speed/FLOPS": 202381729838964.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04736092686653137, "optim/lr": 0.002970564644656548, "optim/total_tokens": 4443340800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 2.926978826522827, "created_at": "2025-01-16T05:29:44.988232+00:00"} {"global_step": 8476, "acc_step": 0, "speed/wps": 12888.654158014651, "speed/FLOPS": 202434024418399.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06582701951265335, "optim/lr": 0.002970547751811886, "optim/total_tokens": 4443865088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9155972003936768, "created_at": "2025-01-16T05:29:55.166056+00:00"} {"global_step": 8477, "acc_step": 0, "speed/wps": 12888.249446780128, "speed/FLOPS": 202427667872329.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06330443173646927, "optim/lr": 0.002970530854169305, "optim/total_tokens": 4444389376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411033, "loss/out": 2.919699192047119, "created_at": "2025-01-16T05:30:05.342135+00:00"} {"global_step": 8478, "acc_step": 0, "speed/wps": 12883.39107134959, "speed/FLOPS": 202351360410085.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06359606981277466, "optim/lr": 0.0029705139517288578, "optim/total_tokens": 4444913664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 2.966695547103882, "created_at": "2025-01-16T05:30:15.520725+00:00"} {"global_step": 8479, "acc_step": 0, "speed/wps": 12886.068981328277, "speed/FLOPS": 202393420666134.88, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06337998807430267, "optim/lr": 0.0029704970444906003, "optim/total_tokens": 4445437952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.0028088092803955, "created_at": "2025-01-16T05:30:25.698189+00:00"} {"global_step": 8480, "acc_step": 0, "speed/wps": 12889.58949396087, "speed/FLOPS": 202448715154721.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050638772547245026, "optim/lr": 0.0029704801324545883, "optim/total_tokens": 4445962240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.010282278060913, "created_at": "2025-01-16T05:30:35.870614+00:00"} {"global_step": 8481, "acc_step": 0, "speed/wps": 12889.924478463901, "speed/FLOPS": 202453976546658.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08210162073373795, "optim/lr": 0.0029704632156208758, "optim/total_tokens": 4446486528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 2.8931052684783936, "created_at": "2025-01-16T05:30:46.045138+00:00"} {"global_step": 8482, "acc_step": 0, "speed/wps": 12888.820811690908, "speed/FLOPS": 202436641943391.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06053303927183151, "optim/lr": 0.0029704462939895183, "optim/total_tokens": 4447010816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 3.0090484619140625, "created_at": "2025-01-16T05:30:56.221630+00:00"} {"global_step": 8483, "acc_step": 0, "speed/wps": 12891.437045981898, "speed/FLOPS": 202477733498013.8, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06227214261889458, "optim/lr": 0.0029704293675605715, "optim/total_tokens": 4447535104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9653735160827637, "created_at": "2025-01-16T05:31:06.392715+00:00"} {"global_step": 8484, "acc_step": 0, "speed/wps": 12890.608422426021, "speed/FLOPS": 202464718826420.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05471407249569893, "optim/lr": 0.0029704124363340904, "optim/total_tokens": 4448059392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367183, "loss/out": 2.92031192779541, "created_at": "2025-01-16T05:31:16.564295+00:00"} {"global_step": 8485, "acc_step": 0, "speed/wps": 12887.1926946551, "speed/FLOPS": 202411070128076.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05638608708977699, "optim/lr": 0.0029703955003101304, "optim/total_tokens": 4448583680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9329800605773926, "created_at": "2025-01-16T05:31:26.740236+00:00"} {"global_step": 8486, "acc_step": 0, "speed/wps": 12873.318878906844, "speed/FLOPS": 202193162787129.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0026, "optim/grad_norm": 0.04743501916527748, "optim/lr": 0.0029703785594887466, "optim/total_tokens": 4449107968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 2.9351511001586914, "created_at": "2025-01-16T05:31:36.929123+00:00"} {"global_step": 8487, "acc_step": 0, "speed/wps": 12893.56688276547, "speed/FLOPS": 202511185511402.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052393533289432526, "optim/lr": 0.0029703616138699934, "optim/total_tokens": 4449632256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.112407922744751, "created_at": "2025-01-16T05:31:47.098893+00:00"} {"global_step": 8488, "acc_step": 0, "speed/wps": 12886.152403990607, "speed/FLOPS": 202394730933681.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05898596718907356, "optim/lr": 0.002970344663453928, "optim/total_tokens": 4450156544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9666552543640137, "created_at": "2025-01-16T05:31:57.276035+00:00"} {"global_step": 8489, "acc_step": 0, "speed/wps": 12886.96337975921, "speed/FLOPS": 202407468422524.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05090026557445526, "optim/lr": 0.002970327708240604, "optim/total_tokens": 4450680832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 2.9063992500305176, "created_at": "2025-01-16T05:32:07.453668+00:00"} {"global_step": 8490, "acc_step": 0, "speed/wps": 12880.397591971694, "speed/FLOPS": 202304343702984.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.04876101016998291, "optim/lr": 0.0029703107482300776, "optim/total_tokens": 4451205120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9575510025024414, "created_at": "2025-01-16T05:32:17.635788+00:00"} {"global_step": 8491, "acc_step": 0, "speed/wps": 12887.91717409672, "speed/FLOPS": 202422449073246.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048593513667583466, "optim/lr": 0.0029702937834224035, "optim/total_tokens": 4451729408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.886920928955078, "created_at": "2025-01-16T05:32:27.814418+00:00"} {"global_step": 8492, "acc_step": 0, "speed/wps": 12895.354706764152, "speed/FLOPS": 202539265744029.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04836579039692879, "optim/lr": 0.002970276813817638, "optim/total_tokens": 4452253696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420607, "loss/out": 2.890270709991455, "created_at": "2025-01-16T05:32:37.982461+00:00"} {"global_step": 8493, "acc_step": 0, "speed/wps": 12888.165058668608, "speed/FLOPS": 202426342441076.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046124693006277084, "optim/lr": 0.0029702598394158357, "optim/total_tokens": 4452777984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.059908866882324, "created_at": "2025-01-16T05:32:48.156600+00:00"} {"global_step": 8494, "acc_step": 0, "speed/wps": 12884.390189602409, "speed/FLOPS": 202367052935178.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04497222602367401, "optim/lr": 0.002970242860217052, "optim/total_tokens": 4453302272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8747172355651855, "created_at": "2025-01-16T05:32:58.333097+00:00"} {"global_step": 8495, "acc_step": 0, "speed/wps": 12893.778094612304, "speed/FLOPS": 202514502883692.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04734266921877861, "optim/lr": 0.002970225876221343, "optim/total_tokens": 4453826560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.903475284576416, "created_at": "2025-01-16T05:33:08.502594+00:00"} {"global_step": 8496, "acc_step": 0, "speed/wps": 12885.211475968013, "speed/FLOPS": 202379952366116.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06103719398379326, "optim/lr": 0.002970208887428763, "optim/total_tokens": 4454350848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.9699084758758545, "created_at": "2025-01-16T05:33:18.680827+00:00"} {"global_step": 8497, "acc_step": 0, "speed/wps": 12890.362680004373, "speed/FLOPS": 202460859104003.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04256707802414894, "optim/lr": 0.002970191893839368, "optim/total_tokens": 4454875136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0361499786376953, "created_at": "2025-01-16T05:33:28.853392+00:00"} {"global_step": 8498, "acc_step": 0, "speed/wps": 12880.457080389637, "speed/FLOPS": 202305278050333.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049763020128011703, "optim/lr": 0.002970174895453214, "optim/total_tokens": 4455399424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9539499282836914, "created_at": "2025-01-16T05:33:39.035566+00:00"} {"global_step": 8499, "acc_step": 0, "speed/wps": 12887.319819860239, "speed/FLOPS": 202413066804111.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05832333490252495, "optim/lr": 0.0029701578922703554, "optim/total_tokens": 4455923712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.894746780395508, "created_at": "2025-01-16T05:33:49.211498+00:00"} {"global_step": 8500, "acc_step": 0, "speed/wps": 12887.03643673214, "speed/FLOPS": 202408615882675.56, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06233450770378113, "optim/lr": 0.0029701408842908485, "optim/total_tokens": 4456448000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 2.9438986778259277, "created_at": "2025-01-16T05:33:59.390325+00:00"} {"global_step": 8501, "acc_step": 0, "speed/wps": 12888.940400369558, "speed/FLOPS": 202438520247921.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05747295543551445, "optim/lr": 0.0029701238715147482, "optim/total_tokens": 4456972288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.048698902130127, "created_at": "2025-01-16T05:34:09.565170+00:00"} {"global_step": 8502, "acc_step": 0, "speed/wps": 12883.770286969622, "speed/FLOPS": 202357316512495.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05057600513100624, "optim/lr": 0.0029701068539421107, "optim/total_tokens": 4457496576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.984776496887207, "created_at": "2025-01-16T05:34:19.742034+00:00"} {"global_step": 8503, "acc_step": 0, "speed/wps": 12886.252602283234, "speed/FLOPS": 202396304685554.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06283512711524963, "optim/lr": 0.002970089831572991, "optim/total_tokens": 4458020864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424656, "loss/out": 2.9857778549194336, "created_at": "2025-01-16T05:34:29.917999+00:00"} {"global_step": 8504, "acc_step": 0, "speed/wps": 12887.13371515155, "speed/FLOPS": 202410143773926.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0514582097530365, "optim/lr": 0.0029700728044074446, "optim/total_tokens": 4458545152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332063, "loss/out": 2.866462469100952, "created_at": "2025-01-16T05:34:40.094619+00:00"} {"global_step": 8505, "acc_step": 0, "speed/wps": 12885.876650222028, "speed/FLOPS": 202390399841820.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05627793073654175, "optim/lr": 0.0029700557724455278, "optim/total_tokens": 4459069440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.915933609008789, "created_at": "2025-01-16T05:34:50.271025+00:00"} {"global_step": 8506, "acc_step": 0, "speed/wps": 12888.44567406623, "speed/FLOPS": 202430749891502.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08897587656974792, "optim/lr": 0.0029700387356872953, "optim/total_tokens": 4459593728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9477288722991943, "created_at": "2025-01-16T05:35:00.449008+00:00"} {"global_step": 8507, "acc_step": 0, "speed/wps": 12874.162179638946, "speed/FLOPS": 202206407983943.84, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.06553371250629425, "optim/lr": 0.0029700216941328027, "optim/total_tokens": 4460118016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.923001766204834, "created_at": "2025-01-16T05:35:10.633717+00:00"} {"global_step": 8508, "acc_step": 0, "speed/wps": 12879.924976581966, "speed/FLOPS": 202296920628846.78, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06382346898317337, "optim/lr": 0.0029700046477821058, "optim/total_tokens": 4460642304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.932924747467041, "created_at": "2025-01-16T05:35:20.817021+00:00"} {"global_step": 8509, "acc_step": 0, "speed/wps": 12885.485589722608, "speed/FLOPS": 202384257699303.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07096216827630997, "optim/lr": 0.0029699875966352606, "optim/total_tokens": 4461166592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.888874053955078, "created_at": "2025-01-16T05:35:30.993288+00:00"} {"global_step": 8510, "acc_step": 0, "speed/wps": 12887.023069265575, "speed/FLOPS": 202408405928244.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05060141533613205, "optim/lr": 0.0029699705406923222, "optim/total_tokens": 4461690880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9432809352874756, "created_at": "2025-01-16T05:35:41.170741+00:00"} {"global_step": 8511, "acc_step": 0, "speed/wps": 12887.859863830829, "speed/FLOPS": 202421548936767.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08018246293067932, "optim/lr": 0.002969953479953347, "optim/total_tokens": 4462215168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.0284547805786133, "created_at": "2025-01-16T05:35:51.345015+00:00"} {"global_step": 8512, "acc_step": 0, "speed/wps": 12886.08654359378, "speed/FLOPS": 202393696505647.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07303143292665482, "optim/lr": 0.002969936414418389, "optim/total_tokens": 4462739456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 2.819075345993042, "created_at": "2025-01-16T05:36:01.522504+00:00"} {"global_step": 8513, "acc_step": 0, "speed/wps": 12886.216922347061, "speed/FLOPS": 202395744283127.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06687221676111221, "optim/lr": 0.002969919344087506, "optim/total_tokens": 4463263744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.9242911338806152, "created_at": "2025-01-16T05:36:11.700668+00:00"} {"global_step": 8514, "acc_step": 0, "speed/wps": 12890.572580224361, "speed/FLOPS": 202464155875391.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05990617722272873, "optim/lr": 0.002969902268960753, "optim/total_tokens": 4463788032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8632893562316895, "created_at": "2025-01-16T05:36:21.898196+00:00"} {"global_step": 8515, "acc_step": 0, "speed/wps": 12887.087215185264, "speed/FLOPS": 202409413428058.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058927398175001144, "optim/lr": 0.0029698851890381846, "optim/total_tokens": 4464312320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9428634643554688, "created_at": "2025-01-16T05:36:32.074062+00:00"} {"global_step": 8516, "acc_step": 0, "speed/wps": 12888.38645411787, "speed/FLOPS": 202429819760836.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05258708447217941, "optim/lr": 0.002969868104319858, "optim/total_tokens": 4464836608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411556, "loss/out": 3.085196018218994, "created_at": "2025-01-16T05:36:42.257850+00:00"} {"global_step": 8517, "acc_step": 0, "speed/wps": 12887.688265355446, "speed/FLOPS": 202418853746910.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05070782080292702, "optim/lr": 0.002969851014805828, "optim/total_tokens": 4465360896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.9238734245300293, "created_at": "2025-01-16T05:36:52.433689+00:00"} {"global_step": 8518, "acc_step": 0, "speed/wps": 12890.753896087446, "speed/FLOPS": 202467003690175.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0436871163547039, "optim/lr": 0.0029698339204961505, "optim/total_tokens": 4465885184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.921008348464966, "created_at": "2025-01-16T05:37:02.607400+00:00"} {"global_step": 8519, "acc_step": 0, "speed/wps": 12886.019933483785, "speed/FLOPS": 202392650302338.3, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051489561796188354, "optim/lr": 0.002969816821390881, "optim/total_tokens": 4466409472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.1097657680511475, "created_at": "2025-01-16T05:37:12.789459+00:00"} {"global_step": 8520, "acc_step": 0, "speed/wps": 12886.959089187918, "speed/FLOPS": 202407401033206.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04945512115955353, "optim/lr": 0.0029697997174900764, "optim/total_tokens": 4466933760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 2.9666335582733154, "created_at": "2025-01-16T05:37:22.963965+00:00"} {"global_step": 8521, "acc_step": 0, "speed/wps": 12889.872057923152, "speed/FLOPS": 202453153210032.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05121564120054245, "optim/lr": 0.002969782608793792, "optim/total_tokens": 4467458048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.8832898139953613, "created_at": "2025-01-16T05:37:33.138948+00:00"} {"global_step": 8522, "acc_step": 0, "speed/wps": 12886.807306756607, "speed/FLOPS": 202405017081552.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06172000616788864, "optim/lr": 0.002969765495302083, "optim/total_tokens": 4467982336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 2.9771931171417236, "created_at": "2025-01-16T05:37:43.317259+00:00"} {"global_step": 8523, "acc_step": 0, "speed/wps": 12888.818714181427, "speed/FLOPS": 202436608999122.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050035711377859116, "optim/lr": 0.0029697483770150055, "optim/total_tokens": 4468506624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.908053159713745, "created_at": "2025-01-16T05:37:53.493860+00:00"} {"global_step": 8524, "acc_step": 0, "speed/wps": 12888.38943670161, "speed/FLOPS": 202429866606412.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06112111732363701, "optim/lr": 0.002969731253932616, "optim/total_tokens": 4469030912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 2.979055643081665, "created_at": "2025-01-16T05:38:03.669574+00:00"} {"global_step": 8525, "acc_step": 0, "speed/wps": 12889.621211887088, "speed/FLOPS": 202449213328337.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053229015320539474, "optim/lr": 0.0029697141260549694, "optim/total_tokens": 4469555200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9730498790740967, "created_at": "2025-01-16T05:38:13.842208+00:00"} {"global_step": 8526, "acc_step": 0, "speed/wps": 12886.214866981096, "speed/FLOPS": 202395712000780.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05622575432062149, "optim/lr": 0.0029696969933821226, "optim/total_tokens": 4470079488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.9011802673339844, "created_at": "2025-01-16T05:38:24.018509+00:00"} {"global_step": 8527, "acc_step": 0, "speed/wps": 12889.793027116837, "speed/FLOPS": 202451911922619.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05861388519406319, "optim/lr": 0.0029696798559141306, "optim/total_tokens": 4470603776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.03267765045166, "created_at": "2025-01-16T05:38:34.192535+00:00"} {"global_step": 8528, "acc_step": 0, "speed/wps": 12887.668167291686, "speed/FLOPS": 202418538079202.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05336114764213562, "optim/lr": 0.00296966271365105, "optim/total_tokens": 4471128064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437181, "loss/out": 3.015075922012329, "created_at": "2025-01-16T05:38:44.371060+00:00"} {"global_step": 8529, "acc_step": 0, "speed/wps": 12883.332044082592, "speed/FLOPS": 202350433305745.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052812594920396805, "optim/lr": 0.002969645566592936, "optim/total_tokens": 4471652352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8587682247161865, "created_at": "2025-01-16T05:38:54.551520+00:00"} {"global_step": 8530, "acc_step": 0, "speed/wps": 12884.366611534511, "speed/FLOPS": 202366682609222.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07505592703819275, "optim/lr": 0.002969628414739845, "optim/total_tokens": 4472176640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.9850006103515625, "created_at": "2025-01-16T05:39:04.732274+00:00"} {"global_step": 8531, "acc_step": 0, "speed/wps": 12884.929884253597, "speed/FLOPS": 202375529581293.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0530431903898716, "optim/lr": 0.002969611258091833, "optim/total_tokens": 4472700928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.927051544189453, "created_at": "2025-01-16T05:39:14.909188+00:00"} {"global_step": 8532, "acc_step": 0, "speed/wps": 12890.505846984594, "speed/FLOPS": 202463107738159.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07120314985513687, "optim/lr": 0.002969594096648956, "optim/total_tokens": 4473225216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 2.952360153198242, "created_at": "2025-01-16T05:39:25.081106+00:00"} {"global_step": 8533, "acc_step": 0, "speed/wps": 12886.502975038595, "speed/FLOPS": 202400237133725.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057276539504528046, "optim/lr": 0.00296957693041127, "optim/total_tokens": 4473749504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 2.8454740047454834, "created_at": "2025-01-16T05:39:35.255782+00:00"} {"global_step": 8534, "acc_step": 0, "speed/wps": 12884.709368390813, "speed/FLOPS": 202372066076648.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07086385041475296, "optim/lr": 0.0029695597593788307, "optim/total_tokens": 4474273792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.013500213623047, "created_at": "2025-01-16T05:39:45.434901+00:00"} {"global_step": 8535, "acc_step": 0, "speed/wps": 12885.912631490995, "speed/FLOPS": 202390964977093.16, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05626513436436653, "optim/lr": 0.0029695425835516947, "optim/total_tokens": 4474798080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.947664499282837, "created_at": "2025-01-16T05:39:55.610199+00:00"} {"global_step": 8536, "acc_step": 0, "speed/wps": 12893.693412009308, "speed/FLOPS": 202513172827046.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06893625855445862, "optim/lr": 0.002969525402929917, "optim/total_tokens": 4475322368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0077764987945557, "created_at": "2025-01-16T05:40:05.779317+00:00"} {"global_step": 8537, "acc_step": 0, "speed/wps": 12884.436136945185, "speed/FLOPS": 202367774601335.1, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07122346758842468, "optim/lr": 0.002969508217513555, "optim/total_tokens": 4475846656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9624974727630615, "created_at": "2025-01-16T05:40:15.958967+00:00"} {"global_step": 8538, "acc_step": 0, "speed/wps": 12892.583640920564, "speed/FLOPS": 202495742347112.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051108330488204956, "optim/lr": 0.002969491027302664, "optim/total_tokens": 4476370944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.989436626434326, "created_at": "2025-01-16T05:40:26.131819+00:00"} {"global_step": 8539, "acc_step": 0, "speed/wps": 12885.517535204484, "speed/FLOPS": 202384759446994.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05413384735584259, "optim/lr": 0.0029694738322973, "optim/total_tokens": 4476895232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 2.910125970840454, "created_at": "2025-01-16T05:40:36.313382+00:00"} {"global_step": 8540, "acc_step": 0, "speed/wps": 12889.284557462588, "speed/FLOPS": 202443925707988.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05861432105302811, "optim/lr": 0.002969456632497519, "optim/total_tokens": 4477419520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.92301607131958, "created_at": "2025-01-16T05:40:46.487779+00:00"} {"global_step": 8541, "acc_step": 0, "speed/wps": 12876.345673793276, "speed/FLOPS": 202240702760072.97, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05637265741825104, "optim/lr": 0.0029694394279033776, "optim/total_tokens": 4477943808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8519468307495117, "created_at": "2025-01-16T05:40:56.670680+00:00"} {"global_step": 8542, "acc_step": 0, "speed/wps": 12877.60744531109, "speed/FLOPS": 202260520615616.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0454108826816082, "optim/lr": 0.0029694222185149316, "optim/total_tokens": 4478468096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.02803897857666, "created_at": "2025-01-16T05:41:06.857015+00:00"} {"global_step": 8543, "acc_step": 0, "speed/wps": 12878.254469304984, "speed/FLOPS": 202270683016544.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06119551509618759, "optim/lr": 0.0029694050043322373, "optim/total_tokens": 4478992384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.95333194732666, "created_at": "2025-01-16T05:41:17.038803+00:00"} {"global_step": 8544, "acc_step": 0, "speed/wps": 12877.720897794776, "speed/FLOPS": 202262302542772.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08536849170923233, "optim/lr": 0.002969387785355351, "optim/total_tokens": 4479516672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.8843507766723633, "created_at": "2025-01-16T05:41:27.223408+00:00"} {"global_step": 8545, "acc_step": 0, "speed/wps": 12877.882963108914, "speed/FLOPS": 202264848001231.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08627896010875702, "optim/lr": 0.0029693705615843288, "optim/total_tokens": 4480040960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 2.9207208156585693, "created_at": "2025-01-16T05:41:37.407236+00:00"} {"global_step": 8546, "acc_step": 0, "speed/wps": 12880.025507119854, "speed/FLOPS": 202298499599087.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05531825125217438, "optim/lr": 0.002969353333019227, "optim/total_tokens": 4480565248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9774985313415527, "created_at": "2025-01-16T05:41:47.591044+00:00"} {"global_step": 8547, "acc_step": 0, "speed/wps": 12891.495072962784, "speed/FLOPS": 202478644891486.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05779275670647621, "optim/lr": 0.002969336099660101, "optim/total_tokens": 4481089536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.8794162273406982, "created_at": "2025-01-16T05:41:57.765533+00:00"} {"global_step": 8548, "acc_step": 0, "speed/wps": 12883.82360027358, "speed/FLOPS": 202358153871193.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061492063105106354, "optim/lr": 0.002969318861507008, "optim/total_tokens": 4481613824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 2.9092392921447754, "created_at": "2025-01-16T05:42:07.946034+00:00"} {"global_step": 8549, "acc_step": 0, "speed/wps": 12882.671434367818, "speed/FLOPS": 202340057522402.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06554096192121506, "optim/lr": 0.0029693016185600037, "optim/total_tokens": 4482138112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 2.863957405090332, "created_at": "2025-01-16T05:42:18.126893+00:00"} {"global_step": 8550, "acc_step": 0, "speed/wps": 12885.172361798115, "speed/FLOPS": 202379338024329.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.059554215520620346, "optim/lr": 0.002969284370819145, "optim/total_tokens": 4482662400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 3.068413257598877, "created_at": "2025-01-16T05:42:28.307367+00:00"} {"global_step": 8551, "acc_step": 0, "speed/wps": 12884.712187665898, "speed/FLOPS": 202372110357237.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07120639085769653, "optim/lr": 0.0029692671182844875, "optim/total_tokens": 4483186688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.03924822807312, "created_at": "2025-01-16T05:42:38.515543+00:00"} {"global_step": 8552, "acc_step": 0, "speed/wps": 12885.756405787783, "speed/FLOPS": 202388511237747.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05757258087396622, "optim/lr": 0.0029692498609560874, "optim/total_tokens": 4483710976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9749417304992676, "created_at": "2025-01-16T05:42:48.694850+00:00"} {"global_step": 8553, "acc_step": 0, "speed/wps": 12888.975081430646, "speed/FLOPS": 202439064961642.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08860138058662415, "optim/lr": 0.0029692325988340015, "optim/total_tokens": 4484235264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9524986743927, "created_at": "2025-01-16T05:42:58.870110+00:00"} {"global_step": 8554, "acc_step": 0, "speed/wps": 12885.011293809854, "speed/FLOPS": 202376808230243.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0607878640294075, "optim/lr": 0.0029692153319182857, "optim/total_tokens": 4484759552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 2.942258358001709, "created_at": "2025-01-16T05:43:09.046967+00:00"} {"global_step": 8555, "acc_step": 0, "speed/wps": 12878.372253395763, "speed/FLOPS": 202272532977542.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.058071546256542206, "optim/lr": 0.0029691980602089972, "optim/total_tokens": 4485283840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.928624153137207, "created_at": "2025-01-16T05:43:19.230195+00:00"} {"global_step": 8556, "acc_step": 0, "speed/wps": 12885.91802451026, "speed/FLOPS": 202391049681871.5, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06908329576253891, "optim/lr": 0.002969180783706191, "optim/total_tokens": 4485808128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9392545223236084, "created_at": "2025-01-16T05:43:29.405631+00:00"} {"global_step": 8557, "acc_step": 0, "speed/wps": 12886.124252187188, "speed/FLOPS": 202394288770924.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07168855518102646, "optim/lr": 0.002969163502409925, "optim/total_tokens": 4486332416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 2.9656450748443604, "created_at": "2025-01-16T05:43:39.582192+00:00"} {"global_step": 8558, "acc_step": 0, "speed/wps": 12885.826808644697, "speed/FLOPS": 202389617011359.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053400639444589615, "optim/lr": 0.0029691462163202544, "optim/total_tokens": 4486856704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.891624927520752, "created_at": "2025-01-16T05:43:49.760972+00:00"} {"global_step": 8559, "acc_step": 0, "speed/wps": 12892.862988918401, "speed/FLOPS": 202500129891282.53, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047982655465602875, "optim/lr": 0.002969128925437236, "optim/total_tokens": 4487380992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.870335578918457, "created_at": "2025-01-16T05:43:59.932001+00:00"} {"global_step": 8560, "acc_step": 0, "speed/wps": 12882.966576506282, "speed/FLOPS": 202344693135254.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06677218526601791, "optim/lr": 0.0029691116297609265, "optim/total_tokens": 4487905280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9663989543914795, "created_at": "2025-01-16T05:44:10.111354+00:00"} {"global_step": 8561, "acc_step": 0, "speed/wps": 12886.25135867367, "speed/FLOPS": 202396285152957.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044416896998882294, "optim/lr": 0.0029690943292913813, "optim/total_tokens": 4488429568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.7893643379211426, "created_at": "2025-01-16T05:44:20.288675+00:00"} {"global_step": 8562, "acc_step": 0, "speed/wps": 12884.497704515552, "speed/FLOPS": 202368741604630.4, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0689552053809166, "optim/lr": 0.002969077024028658, "optim/total_tokens": 4488953856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 2.833517551422119, "created_at": "2025-01-16T05:44:30.468314+00:00"} {"global_step": 8563, "acc_step": 0, "speed/wps": 12887.22080038396, "speed/FLOPS": 202411511567169.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04709189012646675, "optim/lr": 0.0029690597139728128, "optim/total_tokens": 4489478144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9836246967315674, "created_at": "2025-01-16T05:44:40.649346+00:00"} {"global_step": 8564, "acc_step": 0, "speed/wps": 12886.462520882029, "speed/FLOPS": 202399601745606.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06644686311483383, "optim/lr": 0.0029690423991239018, "optim/total_tokens": 4490002432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.892731189727783, "created_at": "2025-01-16T05:44:50.824454+00:00"} {"global_step": 8565, "acc_step": 0, "speed/wps": 12888.556064646986, "speed/FLOPS": 202432483727265.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04392087832093239, "optim/lr": 0.0029690250794819815, "optim/total_tokens": 4490526720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.098086357116699, "created_at": "2025-01-16T05:45:01.005660+00:00"} {"global_step": 8566, "acc_step": 0, "speed/wps": 12884.366524247813, "speed/FLOPS": 202366681238265.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05275547504425049, "optim/lr": 0.002969007755047109, "optim/total_tokens": 4491051008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 2.9066879749298096, "created_at": "2025-01-16T05:45:11.182309+00:00"} {"global_step": 8567, "acc_step": 0, "speed/wps": 12887.804026938427, "speed/FLOPS": 202420671941644.78, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049200285226106644, "optim/lr": 0.00296899042581934, "optim/total_tokens": 4491575296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9005913734436035, "created_at": "2025-01-16T05:45:21.357878+00:00"} {"global_step": 8568, "acc_step": 0, "speed/wps": 12892.585165662736, "speed/FLOPS": 202495766295283.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048384107649326324, "optim/lr": 0.002968973091798732, "optim/total_tokens": 4492099584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8251566886901855, "created_at": "2025-01-16T05:45:31.532078+00:00"} {"global_step": 8569, "acc_step": 0, "speed/wps": 12887.647678660678, "speed/FLOPS": 202418216277097.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049696989357471466, "optim/lr": 0.002968955752985341, "optim/total_tokens": 4492623872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392005, "loss/out": 2.980132579803467, "created_at": "2025-01-16T05:45:41.707123+00:00"} {"global_step": 8570, "acc_step": 0, "speed/wps": 12886.87462239244, "speed/FLOPS": 202406074366114.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04930480569601059, "optim/lr": 0.002968938409379223, "optim/total_tokens": 4493148160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 2.957082748413086, "created_at": "2025-01-16T05:45:51.884205+00:00"} {"global_step": 8571, "acc_step": 0, "speed/wps": 12887.921097083312, "speed/FLOPS": 202422510689141.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058792468160390854, "optim/lr": 0.0029689210609804358, "optim/total_tokens": 4493672448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.9666314125061035, "created_at": "2025-01-16T05:46:02.059575+00:00"} {"global_step": 8572, "acc_step": 0, "speed/wps": 12891.568382073501, "speed/FLOPS": 202479796311807.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047513578087091446, "optim/lr": 0.002968903707789035, "optim/total_tokens": 4494196736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8755550384521484, "created_at": "2025-01-16T05:46:12.233361+00:00"} {"global_step": 8573, "acc_step": 0, "speed/wps": 12884.970024435574, "speed/FLOPS": 202376160038009.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0537799708545208, "optim/lr": 0.0029688863498050773, "optim/total_tokens": 4494721024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.0584828853607178, "created_at": "2025-01-16T05:46:22.411999+00:00"} {"global_step": 8574, "acc_step": 0, "speed/wps": 12892.01231907715, "speed/FLOPS": 202486768952482.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07302520424127579, "optim/lr": 0.0029688689870286207, "optim/total_tokens": 4495245312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301523, "loss/out": 2.9137258529663086, "created_at": "2025-01-16T05:46:32.584285+00:00"} {"global_step": 8575, "acc_step": 0, "speed/wps": 12883.377566382103, "speed/FLOPS": 202351148296013.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05209707096219063, "optim/lr": 0.0029688516194597203, "optim/total_tokens": 4495769600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9691905975341797, "created_at": "2025-01-16T05:46:42.765459+00:00"} {"global_step": 8576, "acc_step": 0, "speed/wps": 12877.627411851186, "speed/FLOPS": 202260834217565.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.043663740158081055, "optim/lr": 0.0029688342470984325, "optim/total_tokens": 4496293888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.894726276397705, "created_at": "2025-01-16T05:46:52.948302+00:00"} {"global_step": 8577, "acc_step": 0, "speed/wps": 12878.457306423556, "speed/FLOPS": 202273868852218.8, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0752052590250969, "optim/lr": 0.002968816869944816, "optim/total_tokens": 4496818176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.020235061645508, "created_at": "2025-01-16T05:47:03.129890+00:00"} {"global_step": 8578, "acc_step": 0, "speed/wps": 12886.108818302237, "speed/FLOPS": 202394046360552.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07291033118963242, "optim/lr": 0.0029687994879989254, "optim/total_tokens": 4497342464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284448, "loss/out": 2.9857072830200195, "created_at": "2025-01-16T05:47:13.306693+00:00"} {"global_step": 8579, "acc_step": 0, "speed/wps": 12885.792677798574, "speed/FLOPS": 202389080939520.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06612487137317657, "optim/lr": 0.0029687821012608184, "optim/total_tokens": 4497866752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.934145927429199, "created_at": "2025-01-16T05:47:23.482335+00:00"} {"global_step": 8580, "acc_step": 0, "speed/wps": 12878.091467238231, "speed/FLOPS": 202268122845097.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06079545617103577, "optim/lr": 0.0029687647097305517, "optim/total_tokens": 4498391040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.9187495708465576, "created_at": "2025-01-16T05:47:33.664026+00:00"} {"global_step": 8581, "acc_step": 0, "speed/wps": 12882.015559350339, "speed/FLOPS": 202329756103985.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07004973292350769, "optim/lr": 0.0029687473134081817, "optim/total_tokens": 4498915328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.9750821590423584, "created_at": "2025-01-16T05:47:43.845014+00:00"} {"global_step": 8582, "acc_step": 0, "speed/wps": 12877.05802152928, "speed/FLOPS": 202251891160135.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07183893024921417, "optim/lr": 0.0029687299122937655, "optim/total_tokens": 4499439616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 2.888698101043701, "created_at": "2025-01-16T05:47:54.027364+00:00"} {"global_step": 8583, "acc_step": 0, "speed/wps": 12880.443791404601, "speed/FLOPS": 202305069328562.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05403561517596245, "optim/lr": 0.0029687125063873596, "optim/total_tokens": 4499963904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.0439395904541016, "created_at": "2025-01-16T05:48:04.213087+00:00"} {"global_step": 8584, "acc_step": 0, "speed/wps": 12882.663957992921, "speed/FLOPS": 202339940095660.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06539836525917053, "optim/lr": 0.0029686950956890213, "optim/total_tokens": 4500488192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.926807403564453, "created_at": "2025-01-16T05:48:14.393948+00:00"} {"global_step": 8585, "acc_step": 0, "speed/wps": 12885.907501578806, "speed/FLOPS": 202390884404773.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06849799305200577, "optim/lr": 0.0029686776801988064, "optim/total_tokens": 4501012480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9263954162597656, "created_at": "2025-01-16T05:48:24.575200+00:00"} {"global_step": 8586, "acc_step": 0, "speed/wps": 12890.145022935505, "speed/FLOPS": 202457440500644.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06872832775115967, "optim/lr": 0.0029686602599167736, "optim/total_tokens": 4501536768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9893314838409424, "created_at": "2025-01-16T05:48:34.747084+00:00"} {"global_step": 8587, "acc_step": 0, "speed/wps": 12877.951895365235, "speed/FLOPS": 202265930677040.3, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.06765824556350708, "optim/lr": 0.002968642834842977, "optim/total_tokens": 4502061056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367183, "loss/out": 2.8687548637390137, "created_at": "2025-01-16T05:48:44.929823+00:00"} {"global_step": 8588, "acc_step": 0, "speed/wps": 12882.915561643835, "speed/FLOPS": 202343891876737.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04545621946454048, "optim/lr": 0.0029686254049774764, "optim/total_tokens": 4502585344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.002197742462158, "created_at": "2025-01-16T05:48:55.109344+00:00"} {"global_step": 8589, "acc_step": 0, "speed/wps": 12887.173179908461, "speed/FLOPS": 202410763622164.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05495823174715042, "optim/lr": 0.002968607970320326, "optim/total_tokens": 4503109632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.073423385620117, "created_at": "2025-01-16T05:49:05.284227+00:00"} {"global_step": 8590, "acc_step": 0, "speed/wps": 12882.04115454941, "speed/FLOPS": 202330158111758.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056834105402231216, "optim/lr": 0.0029685905308715847, "optim/total_tokens": 4503633920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 2.896482467651367, "created_at": "2025-01-16T05:49:15.465257+00:00"} {"global_step": 8591, "acc_step": 0, "speed/wps": 12889.389410662065, "speed/FLOPS": 202445572571570.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04329391568899155, "optim/lr": 0.002968573086631308, "optim/total_tokens": 4504158208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 2.8719747066497803, "created_at": "2025-01-16T05:49:25.643016+00:00"} {"global_step": 8592, "acc_step": 0, "speed/wps": 12892.502210055423, "speed/FLOPS": 202494463363477.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04822782054543495, "optim/lr": 0.0029685556375995538, "optim/total_tokens": 4504682496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8283097743988037, "created_at": "2025-01-16T05:49:35.815929+00:00"} {"global_step": 8593, "acc_step": 0, "speed/wps": 12889.214597255444, "speed/FLOPS": 202442826886800.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05045526847243309, "optim/lr": 0.0029685381837763784, "optim/total_tokens": 4505206784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.013944625854492, "created_at": "2025-01-16T05:49:45.990798+00:00"} {"global_step": 8594, "acc_step": 0, "speed/wps": 12888.993985467352, "speed/FLOPS": 202439361875515.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06591540575027466, "optim/lr": 0.002968520725161839, "optim/total_tokens": 4505731072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.926443576812744, "created_at": "2025-01-16T05:49:56.169811+00:00"} {"global_step": 8595, "acc_step": 0, "speed/wps": 12886.341437215497, "speed/FLOPS": 202397699960237.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.044033486396074295, "optim/lr": 0.0029685032617559927, "optim/total_tokens": 4506255360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9664268493652344, "created_at": "2025-01-16T05:50:06.348265+00:00"} {"global_step": 8596, "acc_step": 0, "speed/wps": 12891.682889594551, "speed/FLOPS": 202481594809775.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053485896438360214, "optim/lr": 0.002968485793558897, "optim/total_tokens": 4506779648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.190077304840088, "created_at": "2025-01-16T05:50:16.520845+00:00"} {"global_step": 8597, "acc_step": 0, "speed/wps": 12888.270602993925, "speed/FLOPS": 202428000159739.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054827336221933365, "optim/lr": 0.0029684683205706067, "optim/total_tokens": 4507303936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.888019561767578, "created_at": "2025-01-16T05:50:26.696510+00:00"} {"global_step": 8598, "acc_step": 0, "speed/wps": 12892.780423039134, "speed/FLOPS": 202498833080691.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059920769184827805, "optim/lr": 0.0029684508427911815, "optim/total_tokens": 4507828224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 2.7531304359436035, "created_at": "2025-01-16T05:50:36.869421+00:00"} {"global_step": 8599, "acc_step": 0, "speed/wps": 12888.431424356597, "speed/FLOPS": 202430526080231.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05245133489370346, "optim/lr": 0.002968433360220677, "optim/total_tokens": 4508352512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8439760208129883, "created_at": "2025-01-16T05:50:47.046546+00:00"} {"global_step": 8600, "acc_step": 0, "speed/wps": 12892.946057831756, "speed/FLOPS": 202501434602715.4, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06068296357989311, "optim/lr": 0.0029684158728591507, "optim/total_tokens": 4508876800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9028537273406982, "created_at": "2025-01-16T05:50:57.220060+00:00"} {"global_step": 8601, "acc_step": 0, "speed/wps": 12890.591667231605, "speed/FLOPS": 202464455663068.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05926719307899475, "optim/lr": 0.002968398380706659, "optim/total_tokens": 4509401088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.9343838691711426, "created_at": "2025-01-16T05:51:07.394806+00:00"} {"global_step": 8602, "acc_step": 0, "speed/wps": 12892.211433918872, "speed/FLOPS": 202489896324684.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05939299985766411, "optim/lr": 0.00296838088376326, "optim/total_tokens": 4509925376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.993238687515259, "created_at": "2025-01-16T05:51:17.568455+00:00"} {"global_step": 8603, "acc_step": 0, "speed/wps": 12889.255156045881, "speed/FLOPS": 202443463918337.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06792581081390381, "optim/lr": 0.00296836338202901, "optim/total_tokens": 4510449664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9079394340515137, "created_at": "2025-01-16T05:51:27.741275+00:00"} {"global_step": 8604, "acc_step": 0, "speed/wps": 12886.143734260335, "speed/FLOPS": 202394594763654.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05669106915593147, "optim/lr": 0.002968345875503966, "optim/total_tokens": 4510973952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.889714002609253, "created_at": "2025-01-16T05:51:37.916513+00:00"} {"global_step": 8605, "acc_step": 0, "speed/wps": 12887.310728447297, "speed/FLOPS": 202412924010978.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08113346248865128, "optim/lr": 0.0029683283641881854, "optim/total_tokens": 4511498240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421661, "loss/out": 2.9878904819488525, "created_at": "2025-01-16T05:51:48.091779+00:00"} {"global_step": 8606, "acc_step": 0, "speed/wps": 12888.998442649492, "speed/FLOPS": 202439431881686.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05271819606423378, "optim/lr": 0.0029683108480817256, "optim/total_tokens": 4512022528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.85023832321167, "created_at": "2025-01-16T05:51:58.268988+00:00"} {"global_step": 8607, "acc_step": 0, "speed/wps": 12888.718452371944, "speed/FLOPS": 202435034249630.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06390289217233658, "optim/lr": 0.0029682933271846434, "optim/total_tokens": 4512546816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.8826422691345215, "created_at": "2025-01-16T05:52:08.443038+00:00"} {"global_step": 8608, "acc_step": 0, "speed/wps": 12890.275283120382, "speed/FLOPS": 202459486415844.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06840937584638596, "optim/lr": 0.002968275801496996, "optim/total_tokens": 4513071104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.9130566120147705, "created_at": "2025-01-16T05:52:18.617294+00:00"} {"global_step": 8609, "acc_step": 0, "speed/wps": 12890.989895592875, "speed/FLOPS": 202470710386705.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06179405748844147, "optim/lr": 0.0029682582710188416, "optim/total_tokens": 4513595392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.960631847381592, "created_at": "2025-01-16T05:52:28.794627+00:00"} {"global_step": 8610, "acc_step": 0, "speed/wps": 12890.77145813735, "speed/FLOPS": 202467279526301.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06830467283725739, "optim/lr": 0.002968240735750235, "optim/total_tokens": 4514119680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.8431544303894043, "created_at": "2025-01-16T05:52:38.973752+00:00"} {"global_step": 8611, "acc_step": 0, "speed/wps": 12894.168994702188, "speed/FLOPS": 202520642506756.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05878383293747902, "optim/lr": 0.0029682231956912357, "optim/total_tokens": 4514643968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.936558723449707, "created_at": "2025-01-16T05:52:49.147311+00:00"} {"global_step": 8612, "acc_step": 0, "speed/wps": 12887.450032392338, "speed/FLOPS": 202415111970859.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05160842835903168, "optim/lr": 0.0029682056508419, "optim/total_tokens": 4515168256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.938126564025879, "created_at": "2025-01-16T05:52:59.322685+00:00"} {"global_step": 8613, "acc_step": 0, "speed/wps": 12889.971257453879, "speed/FLOPS": 202454711274976.78, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06285548955202103, "optim/lr": 0.0029681881012022853, "optim/total_tokens": 4515692544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 3.0117461681365967, "created_at": "2025-01-16T05:53:09.496674+00:00"} {"global_step": 8614, "acc_step": 0, "speed/wps": 12887.19392310548, "speed/FLOPS": 202411089422577.84, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04895858094096184, "optim/lr": 0.0029681705467724485, "optim/total_tokens": 4516216832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.0007195472717285, "created_at": "2025-01-16T05:53:19.671066+00:00"} {"global_step": 8615, "acc_step": 0, "speed/wps": 12889.141503780147, "speed/FLOPS": 202441678853329.78, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06829097121953964, "optim/lr": 0.0029681529875524477, "optim/total_tokens": 4516741120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.865081787109375, "created_at": "2025-01-16T05:53:29.843741+00:00"} {"global_step": 8616, "acc_step": 0, "speed/wps": 12893.222750904562, "speed/FLOPS": 202505780447639.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053956761956214905, "optim/lr": 0.0029681354235423394, "optim/total_tokens": 4517265408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9311845302581787, "created_at": "2025-01-16T05:53:40.019264+00:00"} {"global_step": 8617, "acc_step": 0, "speed/wps": 12888.219868326189, "speed/FLOPS": 202427203302065.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05738240107893944, "optim/lr": 0.0029681178547421813, "optim/total_tokens": 4517789696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.979508876800537, "created_at": "2025-01-16T05:53:50.193151+00:00"} {"global_step": 8618, "acc_step": 0, "speed/wps": 12891.086801541645, "speed/FLOPS": 202472232427793.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061402540653944016, "optim/lr": 0.0029681002811520305, "optim/total_tokens": 4518313984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.043382167816162, "created_at": "2025-01-16T05:54:00.366169+00:00"} {"global_step": 8619, "acc_step": 0, "speed/wps": 12891.070391071487, "speed/FLOPS": 202471974678809.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052621982991695404, "optim/lr": 0.002968082702771944, "optim/total_tokens": 4518838272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398201, "loss/out": 3.0300354957580566, "created_at": "2025-01-16T05:54:10.537431+00:00"} {"global_step": 8620, "acc_step": 0, "speed/wps": 12888.580810044481, "speed/FLOPS": 202432872387737.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06197555363178253, "optim/lr": 0.00296806511960198, "optim/total_tokens": 4519362560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.852726697921753, "created_at": "2025-01-16T05:54:20.713152+00:00"} {"global_step": 8621, "acc_step": 0, "speed/wps": 12887.72353212112, "speed/FLOPS": 202419407659927.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056785948574543, "optim/lr": 0.0029680475316421952, "optim/total_tokens": 4519886848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.0270774364471436, "created_at": "2025-01-16T05:54:30.886950+00:00"} {"global_step": 8622, "acc_step": 0, "speed/wps": 12894.04593748189, "speed/FLOPS": 202518709724013.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05436740070581436, "optim/lr": 0.002968029938892648, "optim/total_tokens": 4520411136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372174, "loss/out": 2.9873318672180176, "created_at": "2025-01-16T05:54:41.055808+00:00"} {"global_step": 8623, "acc_step": 0, "speed/wps": 12884.393206903032, "speed/FLOPS": 202367100326031.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05725201591849327, "optim/lr": 0.0029680123413533942, "optim/total_tokens": 4520935424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.9657139778137207, "created_at": "2025-01-16T05:54:51.234498+00:00"} {"global_step": 8624, "acc_step": 0, "speed/wps": 12889.251385704882, "speed/FLOPS": 202443404699951.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.047083817422389984, "optim/lr": 0.0029679947390244925, "optim/total_tokens": 4521459712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.0006041526794434, "created_at": "2025-01-16T05:55:01.410062+00:00"} {"global_step": 8625, "acc_step": 0, "speed/wps": 12887.817671007579, "speed/FLOPS": 202420886240499.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04990338534116745, "optim/lr": 0.0029679771319059997, "optim/total_tokens": 4521984000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.9746036529541016, "created_at": "2025-01-16T05:55:11.586299+00:00"} {"global_step": 8626, "acc_step": 0, "speed/wps": 12890.657271427963, "speed/FLOPS": 202465486067121.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057620055973529816, "optim/lr": 0.002967959519997973, "optim/total_tokens": 4522508288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.0016419887542725, "created_at": "2025-01-16T05:55:21.760878+00:00"} {"global_step": 8627, "acc_step": 0, "speed/wps": 12889.638710644553, "speed/FLOPS": 202449488170369.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055689409375190735, "optim/lr": 0.0029679419033004713, "optim/total_tokens": 4523032576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.905900478363037, "created_at": "2025-01-16T05:55:31.934341+00:00"} {"global_step": 8628, "acc_step": 0, "speed/wps": 12888.615791334445, "speed/FLOPS": 202433421816965.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05638333037495613, "optim/lr": 0.0029679242818135506, "optim/total_tokens": 4523556864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.0052170753479004, "created_at": "2025-01-16T05:55:42.107825+00:00"} {"global_step": 8629, "acc_step": 0, "speed/wps": 12887.796469073615, "speed/FLOPS": 202420553234992.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07232265919446945, "optim/lr": 0.0029679066555372688, "optim/total_tokens": 4524081152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.9259586334228516, "created_at": "2025-01-16T05:55:52.283338+00:00"} {"global_step": 8630, "acc_step": 0, "speed/wps": 12887.470863809376, "speed/FLOPS": 202415439156888.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06831733882427216, "optim/lr": 0.002967889024471683, "optim/total_tokens": 4524605440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.0086636543273926, "created_at": "2025-01-16T05:56:02.458483+00:00"} {"global_step": 8631, "acc_step": 0, "speed/wps": 12889.78855521861, "speed/FLOPS": 202451841685312.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05104123428463936, "optim/lr": 0.0029678713886168518, "optim/total_tokens": 4525129728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9221694469451904, "created_at": "2025-01-16T05:56:12.630888+00:00"} {"global_step": 8632, "acc_step": 0, "speed/wps": 12890.557907849141, "speed/FLOPS": 202463925425576.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06123223900794983, "optim/lr": 0.002967853747972832, "optim/total_tokens": 4525654016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.979642391204834, "created_at": "2025-01-16T05:56:22.802577+00:00"} {"global_step": 8633, "acc_step": 0, "speed/wps": 12885.457883238792, "speed/FLOPS": 202383822530902.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05484696105122566, "optim/lr": 0.002967836102539681, "optim/total_tokens": 4526178304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8564653396606445, "created_at": "2025-01-16T05:56:32.980815+00:00"} {"global_step": 8634, "acc_step": 0, "speed/wps": 12891.980148889812, "speed/FLOPS": 202486263675484.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04529125615954399, "optim/lr": 0.002967818452317457, "optim/total_tokens": 4526702592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.938542366027832, "created_at": "2025-01-16T05:56:43.151113+00:00"} {"global_step": 8635, "acc_step": 0, "speed/wps": 12891.906542225051, "speed/FLOPS": 202485107581668.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050948575139045715, "optim/lr": 0.002967800797306217, "optim/total_tokens": 4527226880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.052912712097168, "created_at": "2025-01-16T05:56:53.326066+00:00"} {"global_step": 8636, "acc_step": 0, "speed/wps": 12890.95290711693, "speed/FLOPS": 202470129431863.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057799264788627625, "optim/lr": 0.0029677831375060195, "optim/total_tokens": 4527751168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 2.8752994537353516, "created_at": "2025-01-16T05:57:03.497464+00:00"} {"global_step": 8637, "acc_step": 0, "speed/wps": 12884.803830113038, "speed/FLOPS": 202373549727799.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06412491202354431, "optim/lr": 0.0029677654729169207, "optim/total_tokens": 4528275456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300751, "loss/out": 2.9853930473327637, "created_at": "2025-01-16T05:57:13.676006+00:00"} {"global_step": 8638, "acc_step": 0, "speed/wps": 12888.799196870725, "speed/FLOPS": 202436302452938.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05840477719902992, "optim/lr": 0.002967747803538979, "optim/total_tokens": 4528799744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.0039007663726807, "created_at": "2025-01-16T05:57:23.854701+00:00"} {"global_step": 8639, "acc_step": 0, "speed/wps": 12892.331644142885, "speed/FLOPS": 202491784391438.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05062730237841606, "optim/lr": 0.0029677301293722525, "optim/total_tokens": 4529324032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.9021642208099365, "created_at": "2025-01-16T05:57:34.026859+00:00"} {"global_step": 8640, "acc_step": 0, "speed/wps": 12890.372775734291, "speed/FLOPS": 202461017671314.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05515306815505028, "optim/lr": 0.0029677124504167984, "optim/total_tokens": 4529848320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.882136344909668, "created_at": "2025-01-16T05:57:44.199332+00:00"} {"global_step": 8641, "acc_step": 0, "speed/wps": 12892.411737115684, "speed/FLOPS": 202493042361635.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06033585965633392, "optim/lr": 0.0029676947666726745, "optim/total_tokens": 4530372608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.9461421966552734, "created_at": "2025-01-16T05:57:54.373038+00:00"} {"global_step": 8642, "acc_step": 0, "speed/wps": 12889.820405165427, "speed/FLOPS": 202452341932493.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056249696761369705, "optim/lr": 0.0029676770781399383, "optim/total_tokens": 4530896896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 2.8843753337860107, "created_at": "2025-01-16T05:58:04.553827+00:00"} {"global_step": 8643, "acc_step": 0, "speed/wps": 12886.232191306144, "speed/FLOPS": 202395984103111.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07304949313402176, "optim/lr": 0.0029676593848186467, "optim/total_tokens": 4531421184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.922307014465332, "created_at": "2025-01-16T05:58:14.731924+00:00"} {"global_step": 8644, "acc_step": 0, "speed/wps": 12893.206078682108, "speed/FLOPS": 202505518587476.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05586270987987518, "optim/lr": 0.0029676416867088595, "optim/total_tokens": 4531945472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 2.9353322982788086, "created_at": "2025-01-16T05:58:24.901446+00:00"} {"global_step": 8645, "acc_step": 0, "speed/wps": 12889.641316042782, "speed/FLOPS": 202449529091728.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060565538704395294, "optim/lr": 0.0029676239838106327, "optim/total_tokens": 4532469760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.92502498626709, "created_at": "2025-01-16T05:58:35.074504+00:00"} {"global_step": 8646, "acc_step": 0, "speed/wps": 12889.283462004498, "speed/FLOPS": 202443908502314.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050736136734485626, "optim/lr": 0.002967606276124025, "optim/total_tokens": 4532994048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9734511375427246, "created_at": "2025-01-16T05:58:45.249890+00:00"} {"global_step": 8647, "acc_step": 0, "speed/wps": 12891.280490281943, "speed/FLOPS": 202475274575615.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06176679581403732, "optim/lr": 0.0029675885636490937, "optim/total_tokens": 4533518336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.942795753479004, "created_at": "2025-01-16T05:58:55.424802+00:00"} {"global_step": 8648, "acc_step": 0, "speed/wps": 12892.318318834128, "speed/FLOPS": 202491575099153.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0505693219602108, "optim/lr": 0.0029675708463858964, "optim/total_tokens": 4534042624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 2.922092914581299, "created_at": "2025-01-16T05:59:05.597973+00:00"} {"global_step": 8649, "acc_step": 0, "speed/wps": 12889.863382006479, "speed/FLOPS": 202453016942838.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06553568691015244, "optim/lr": 0.0029675531243344914, "optim/total_tokens": 4534566912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9236345291137695, "created_at": "2025-01-16T05:59:15.774197+00:00"} {"global_step": 8650, "acc_step": 0, "speed/wps": 12891.641057884162, "speed/FLOPS": 202480937785280.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05032304674386978, "optim/lr": 0.0029675353974949366, "optim/total_tokens": 4535091200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 3.042011260986328, "created_at": "2025-01-16T05:59:25.947366+00:00"} {"global_step": 8651, "acc_step": 0, "speed/wps": 12890.849119479508, "speed/FLOPS": 202468499304398.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07364847511053085, "optim/lr": 0.002967517665867289, "optim/total_tokens": 4535615488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425994, "loss/out": 2.861112117767334, "created_at": "2025-01-16T05:59:36.118683+00:00"} {"global_step": 8652, "acc_step": 0, "speed/wps": 12887.221598113118, "speed/FLOPS": 202411524096602.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06031686067581177, "optim/lr": 0.0029674999294516074, "optim/total_tokens": 4536139776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9811911582946777, "created_at": "2025-01-16T05:59:46.293017+00:00"} {"global_step": 8653, "acc_step": 0, "speed/wps": 12887.778376969152, "speed/FLOPS": 202420269073630.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05658693611621857, "optim/lr": 0.00296748218824795, "optim/total_tokens": 4536664064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.0075745582580566, "created_at": "2025-01-16T05:59:56.467790+00:00"} {"global_step": 8654, "acc_step": 0, "speed/wps": 12891.428379766807, "speed/FLOPS": 202477597383197.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08056757599115372, "optim/lr": 0.0029674644422563725, "optim/total_tokens": 4537188352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.876955509185791, "created_at": "2025-01-16T06:00:06.638677+00:00"} {"global_step": 8655, "acc_step": 0, "speed/wps": 12888.210079293523, "speed/FLOPS": 202427049551856.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061474841088056564, "optim/lr": 0.0029674466914769346, "optim/total_tokens": 4537712640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 2.858603000640869, "created_at": "2025-01-16T06:00:16.812530+00:00"} {"global_step": 8656, "acc_step": 0, "speed/wps": 12890.684897554378, "speed/FLOPS": 202465919973399.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06463025510311127, "optim/lr": 0.002967428935909695, "optim/total_tokens": 4538236928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9290127754211426, "created_at": "2025-01-16T06:00:26.984137+00:00"} {"global_step": 8657, "acc_step": 0, "speed/wps": 12886.48217978869, "speed/FLOPS": 202399910515749.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06901401281356812, "optim/lr": 0.0029674111755547093, "optim/total_tokens": 4538761216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 3.086472272872925, "created_at": "2025-01-16T06:00:37.159049+00:00"} {"global_step": 8658, "acc_step": 0, "speed/wps": 12887.592676786166, "speed/FLOPS": 202417352397077.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06150294840335846, "optim/lr": 0.002967393410412037, "optim/total_tokens": 4539285504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410019, "loss/out": 2.944716453552246, "created_at": "2025-01-16T06:00:47.335003+00:00"} {"global_step": 8659, "acc_step": 0, "speed/wps": 12893.929366771394, "speed/FLOPS": 202516878820820.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061529289931058884, "optim/lr": 0.002967375640481736, "optim/total_tokens": 4539809792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.888113021850586, "created_at": "2025-01-16T06:00:57.505343+00:00"} {"global_step": 8660, "acc_step": 0, "speed/wps": 12888.120757122797, "speed/FLOPS": 202425646624423.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050860270857810974, "optim/lr": 0.002967357865763864, "optim/total_tokens": 4540334080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.0478296279907227, "created_at": "2025-01-16T06:01:07.679967+00:00"} {"global_step": 8661, "acc_step": 0, "speed/wps": 12890.869658436208, "speed/FLOPS": 202468821896936.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.067144013941288, "optim/lr": 0.002967340086258479, "optim/total_tokens": 4540858368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.847313642501831, "created_at": "2025-01-16T06:01:17.851574+00:00"} {"global_step": 8662, "acc_step": 0, "speed/wps": 12892.523669589233, "speed/FLOPS": 202494800414944.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049364037811756134, "optim/lr": 0.002967322301965639, "optim/total_tokens": 4541382656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.958669900894165, "created_at": "2025-01-16T06:01:28.021625+00:00"} {"global_step": 8663, "acc_step": 0, "speed/wps": 12891.394170826055, "speed/FLOPS": 202477060084774.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058507904410362244, "optim/lr": 0.0029673045128854017, "optim/total_tokens": 4541906944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8925862312316895, "created_at": "2025-01-16T06:01:38.195071+00:00"} {"global_step": 8664, "acc_step": 0, "speed/wps": 12889.570464224467, "speed/FLOPS": 202448416266561.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04905780777335167, "optim/lr": 0.0029672867190178253, "optim/total_tokens": 4542431232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9334914684295654, "created_at": "2025-01-16T06:01:48.368808+00:00"} {"global_step": 8665, "acc_step": 0, "speed/wps": 12893.560030424056, "speed/FLOPS": 202511077885964.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05500456690788269, "optim/lr": 0.002967268920362969, "optim/total_tokens": 4542955520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 2.9311938285827637, "created_at": "2025-01-16T06:01:58.539035+00:00"} {"global_step": 8666, "acc_step": 0, "speed/wps": 12891.443863003264, "speed/FLOPS": 202477840568701.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06276111304759979, "optim/lr": 0.002967251116920889, "optim/total_tokens": 4543479808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.0250020027160645, "created_at": "2025-01-16T06:02:08.713943+00:00"} {"global_step": 8667, "acc_step": 0, "speed/wps": 12888.164446121182, "speed/FLOPS": 202426332820177.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06737572699785233, "optim/lr": 0.0029672333086916444, "optim/total_tokens": 4544004096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.880833148956299, "created_at": "2025-01-16T06:02:18.891838+00:00"} {"global_step": 8668, "acc_step": 0, "speed/wps": 12892.517863579531, "speed/FLOPS": 202494709223583.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0569663867354393, "optim/lr": 0.0029672154956752934, "optim/total_tokens": 4544528384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418281, "loss/out": 2.908844470977783, "created_at": "2025-01-16T06:02:29.062531+00:00"} {"global_step": 8669, "acc_step": 0, "speed/wps": 12885.422106245453, "speed/FLOPS": 202383260604059.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050808150321245193, "optim/lr": 0.0029671976778718936, "optim/total_tokens": 4545052672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8885769844055176, "created_at": "2025-01-16T06:02:39.241188+00:00"} {"global_step": 8670, "acc_step": 0, "speed/wps": 12890.834968359439, "speed/FLOPS": 202468277041611.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04664937034249306, "optim/lr": 0.002967179855281504, "optim/total_tokens": 4545576960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.8606717586517334, "created_at": "2025-01-16T06:02:49.414718+00:00"} {"global_step": 8671, "acc_step": 0, "speed/wps": 12885.816751889333, "speed/FLOPS": 202389459056196.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04860008880496025, "optim/lr": 0.0029671620279041818, "optim/total_tokens": 4546101248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.905785083770752, "created_at": "2025-01-16T06:02:59.594700+00:00"} {"global_step": 8672, "acc_step": 0, "speed/wps": 12888.326914859123, "speed/FLOPS": 202428884614962.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05029922351241112, "optim/lr": 0.0029671441957399855, "optim/total_tokens": 4546625536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 2.822709798812866, "created_at": "2025-01-16T06:03:09.768071+00:00"} {"global_step": 8673, "acc_step": 0, "speed/wps": 12887.872149617147, "speed/FLOPS": 202421741901924.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04504837468266487, "optim/lr": 0.0029671263587889734, "optim/total_tokens": 4547149824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 2.900550603866577, "created_at": "2025-01-16T06:03:19.943265+00:00"} {"global_step": 8674, "acc_step": 0, "speed/wps": 12891.676844621654, "speed/FLOPS": 202481499865169.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05506114289164543, "optim/lr": 0.0029671085170512035, "optim/total_tokens": 4547674112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.9542622566223145, "created_at": "2025-01-16T06:03:30.117035+00:00"} {"global_step": 8675, "acc_step": 0, "speed/wps": 12885.573706217274, "speed/FLOPS": 202385641689937.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048137787729501724, "optim/lr": 0.002967090670526734, "optim/total_tokens": 4548198400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 3.002450466156006, "created_at": "2025-01-16T06:03:40.294048+00:00"} {"global_step": 8676, "acc_step": 0, "speed/wps": 12885.390252076526, "speed/FLOPS": 202382760290564.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056343305855989456, "optim/lr": 0.0029670728192156236, "optim/total_tokens": 4548722688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.145728588104248, "created_at": "2025-01-16T06:03:50.470556+00:00"} {"global_step": 8677, "acc_step": 0, "speed/wps": 12892.235914640682, "speed/FLOPS": 202490280828060.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04737776517868042, "optim/lr": 0.0029670549631179302, "optim/total_tokens": 4549246976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 2.9259390830993652, "created_at": "2025-01-16T06:04:00.640892+00:00"} {"global_step": 8678, "acc_step": 0, "speed/wps": 12889.966013335592, "speed/FLOPS": 202454628908892.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05047459900379181, "optim/lr": 0.0029670371022337124, "optim/total_tokens": 4549771264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.967066764831543, "created_at": "2025-01-16T06:04:10.813419+00:00"} {"global_step": 8679, "acc_step": 0, "speed/wps": 12887.346756668641, "speed/FLOPS": 202413489883702.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060506947338581085, "optim/lr": 0.002967019236563028, "optim/total_tokens": 4550295552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350651, "loss/out": 3.0590171813964844, "created_at": "2025-01-16T06:04:20.990112+00:00"} {"global_step": 8680, "acc_step": 0, "speed/wps": 12890.182744659833, "speed/FLOPS": 202458032972160.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058534666895866394, "optim/lr": 0.002967001366105935, "optim/total_tokens": 4550819840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.8917622566223145, "created_at": "2025-01-16T06:04:31.162072+00:00"} {"global_step": 8681, "acc_step": 0, "speed/wps": 12890.539109027892, "speed/FLOPS": 202463630164256.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06388210505247116, "optim/lr": 0.0029669834908624924, "optim/total_tokens": 4551344128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.960002899169922, "created_at": "2025-01-16T06:04:41.334919+00:00"} {"global_step": 8682, "acc_step": 0, "speed/wps": 12889.370030917955, "speed/FLOPS": 202445268186058.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059327591210603714, "optim/lr": 0.0029669656108327583, "optim/total_tokens": 4551868416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 2.9506919384002686, "created_at": "2025-01-16T06:04:51.509030+00:00"} {"global_step": 8683, "acc_step": 0, "speed/wps": 12891.137522178895, "speed/FLOPS": 202473029065098.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07606718689203262, "optim/lr": 0.0029669477260167907, "optim/total_tokens": 4552392704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.9154255390167236, "created_at": "2025-01-16T06:05:01.680651+00:00"} {"global_step": 8684, "acc_step": 0, "speed/wps": 12891.134995288672, "speed/FLOPS": 202472989376815.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0759599506855011, "optim/lr": 0.0029669298364146486, "optim/total_tokens": 4552916992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.886561393737793, "created_at": "2025-01-16T06:05:11.851978+00:00"} {"global_step": 8685, "acc_step": 0, "speed/wps": 12891.529513887974, "speed/FLOPS": 202479185833542.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05890810117125511, "optim/lr": 0.0029669119420263896, "optim/total_tokens": 4553441280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.012702226638794, "created_at": "2025-01-16T06:05:22.024688+00:00"} {"global_step": 8686, "acc_step": 0, "speed/wps": 12889.425047964385, "speed/FLOPS": 202446132304373.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06292737275362015, "optim/lr": 0.002966894042852073, "optim/total_tokens": 4553965568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 2.8661773204803467, "created_at": "2025-01-16T06:05:32.197381+00:00"} {"global_step": 8687, "acc_step": 0, "speed/wps": 12879.128697112215, "speed/FLOPS": 202284413965571.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06791859865188599, "optim/lr": 0.0029668761388917564, "optim/total_tokens": 4554489856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.9137277603149414, "created_at": "2025-01-16T06:05:42.379177+00:00"} {"global_step": 8688, "acc_step": 0, "speed/wps": 12887.488223375018, "speed/FLOPS": 202415711812721.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049538686871528625, "optim/lr": 0.0029668582301454987, "optim/total_tokens": 4555014144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.0282020568847656, "created_at": "2025-01-16T06:05:52.553086+00:00"} {"global_step": 8689, "acc_step": 0, "speed/wps": 12888.88948337164, "speed/FLOPS": 202437720526501.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05151863023638725, "optim/lr": 0.002966840316613358, "optim/total_tokens": 4555538432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 2.7690186500549316, "created_at": "2025-01-16T06:06:02.728519+00:00"} {"global_step": 8690, "acc_step": 0, "speed/wps": 12891.73445926562, "speed/FLOPS": 202482404782323.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056200772523880005, "optim/lr": 0.0029668223982953932, "optim/total_tokens": 4556062720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.935790538787842, "created_at": "2025-01-16T06:06:12.899625+00:00"} {"global_step": 8691, "acc_step": 0, "speed/wps": 12891.558109702366, "speed/FLOPS": 202479634970102.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06574373692274094, "optim/lr": 0.002966804475191662, "optim/total_tokens": 4556587008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9511241912841797, "created_at": "2025-01-16T06:06:23.071359+00:00"} {"global_step": 8692, "acc_step": 0, "speed/wps": 12893.4998161267, "speed/FLOPS": 202510132137681.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05351463332772255, "optim/lr": 0.002966786547302224, "optim/total_tokens": 4557111296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9082202911376953, "created_at": "2025-01-16T06:06:33.242338+00:00"} {"global_step": 8693, "acc_step": 0, "speed/wps": 12892.473170707568, "speed/FLOPS": 202494007260615.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055469170212745667, "optim/lr": 0.002966768614627136, "optim/total_tokens": 4557635584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304572, "loss/out": 3.075765609741211, "created_at": "2025-01-16T06:06:43.414087+00:00"} {"global_step": 8694, "acc_step": 0, "speed/wps": 12892.886432737427, "speed/FLOPS": 202500498108675.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05444434657692909, "optim/lr": 0.002966750677166458, "optim/total_tokens": 4558159872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9564085006713867, "created_at": "2025-01-16T06:06:53.588069+00:00"} {"global_step": 8695, "acc_step": 0, "speed/wps": 12887.259889228773, "speed/FLOPS": 202412125511191.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056878797709941864, "optim/lr": 0.002966732734920248, "optim/total_tokens": 4558684160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.9637019634246826, "created_at": "2025-01-16T06:07:03.764538+00:00"} {"global_step": 8696, "acc_step": 0, "speed/wps": 12886.489496584525, "speed/FLOPS": 202400025436082.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05609147623181343, "optim/lr": 0.0029667147878885652, "optim/total_tokens": 4559208448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299229, "loss/out": 2.9528329372406006, "created_at": "2025-01-16T06:07:13.940316+00:00"} {"global_step": 8697, "acc_step": 0, "speed/wps": 12892.940495933524, "speed/FLOPS": 202501347245460.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051902737468481064, "optim/lr": 0.002966696836071467, "optim/total_tokens": 4559732736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.817732334136963, "created_at": "2025-01-16T06:07:24.110651+00:00"} {"global_step": 8698, "acc_step": 0, "speed/wps": 12890.981136698812, "speed/FLOPS": 202470572816237.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04491293057799339, "optim/lr": 0.0029666788794690123, "optim/total_tokens": 4560257024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8886666297912598, "created_at": "2025-01-16T06:07:34.284247+00:00"} {"global_step": 8699, "acc_step": 0, "speed/wps": 12884.91737404434, "speed/FLOPS": 202375333091266.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046349260956048965, "optim/lr": 0.00296666091808126, "optim/total_tokens": 4560781312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.811032772064209, "created_at": "2025-01-16T06:07:44.462935+00:00"} {"global_step": 8700, "acc_step": 0, "speed/wps": 12886.24861895781, "speed/FLOPS": 202396242121954.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05241923779249191, "optim/lr": 0.002966642951908269, "optim/total_tokens": 4561305600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 3.011680841445923, "created_at": "2025-01-16T06:07:54.638210+00:00"} {"global_step": 8701, "acc_step": 0, "speed/wps": 12893.817199620162, "speed/FLOPS": 202515117081576.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0532253198325634, "optim/lr": 0.002966624980950098, "optim/total_tokens": 4561829888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.0187485218048096, "created_at": "2025-01-16T06:08:04.807474+00:00"} {"global_step": 8702, "acc_step": 0, "speed/wps": 12891.64197779154, "speed/FLOPS": 202480952233690.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.044272907078266144, "optim/lr": 0.002966607005206804, "optim/total_tokens": 4562354176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8933396339416504, "created_at": "2025-01-16T06:08:14.979777+00:00"} {"global_step": 8703, "acc_step": 0, "speed/wps": 12887.662254719977, "speed/FLOPS": 202418445214138.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05492033436894417, "optim/lr": 0.0029665890246784476, "optim/total_tokens": 4562878464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.908005714416504, "created_at": "2025-01-16T06:08:25.153701+00:00"} {"global_step": 8704, "acc_step": 0, "speed/wps": 12889.92255649974, "speed/FLOPS": 202453946359569.88, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05251903086900711, "optim/lr": 0.0029665710393650857, "optim/total_tokens": 4563402752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.0081377029418945, "created_at": "2025-01-16T06:08:35.328655+00:00"} {"global_step": 8705, "acc_step": 0, "speed/wps": 12883.90271476513, "speed/FLOPS": 202359396472998.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055560313165187836, "optim/lr": 0.0029665530492667793, "optim/total_tokens": 4563927040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8392252922058105, "created_at": "2025-01-16T06:08:45.510626+00:00"} {"global_step": 8706, "acc_step": 0, "speed/wps": 12891.419492745965, "speed/FLOPS": 202477457800322.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049097515642642975, "optim/lr": 0.002966535054383585, "optim/total_tokens": 4564451328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.91355037689209, "created_at": "2025-01-16T06:08:55.683498+00:00"} {"global_step": 8707, "acc_step": 0, "speed/wps": 12893.377817041153, "speed/FLOPS": 202508215974398.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05351437255740166, "optim/lr": 0.002966517054715562, "optim/total_tokens": 4564975616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371642, "loss/out": 3.051158905029297, "created_at": "2025-01-16T06:09:05.854660+00:00"} {"global_step": 8708, "acc_step": 0, "speed/wps": 12888.68240821832, "speed/FLOPS": 202434468126667.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05412659794092178, "optim/lr": 0.00296649905026277, "optim/total_tokens": 4565499904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.0230371952056885, "created_at": "2025-01-16T06:09:16.030772+00:00"} {"global_step": 8709, "acc_step": 0, "speed/wps": 12891.856329237096, "speed/FLOPS": 202484318917692.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05970940366387367, "optim/lr": 0.0029664810410252665, "optim/total_tokens": 4566024192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.7204411029815674, "created_at": "2025-01-16T06:09:26.203765+00:00"} {"global_step": 8710, "acc_step": 0, "speed/wps": 12889.343038432447, "speed/FLOPS": 202444844231981.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08040741086006165, "optim/lr": 0.002966463027003111, "optim/total_tokens": 4566548480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402750, "loss/out": 2.833932876586914, "created_at": "2025-01-16T06:09:36.379085+00:00"} {"global_step": 8711, "acc_step": 0, "speed/wps": 12888.395255517547, "speed/FLOPS": 202429957998913.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044140059500932693, "optim/lr": 0.0029664450081963623, "optim/total_tokens": 4567072768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330191, "loss/out": 2.87926983833313, "created_at": "2025-01-16T06:09:46.553099+00:00"} {"global_step": 8712, "acc_step": 0, "speed/wps": 12888.256564938369, "speed/FLOPS": 202427779672786.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06273992359638214, "optim/lr": 0.0029664269846050788, "optim/total_tokens": 4567597056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.937222957611084, "created_at": "2025-01-16T06:09:56.728339+00:00"} {"global_step": 8713, "acc_step": 0, "speed/wps": 12885.832432836527, "speed/FLOPS": 202389705347020.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06154230237007141, "optim/lr": 0.0029664089562293195, "optim/total_tokens": 4568121344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.044386386871338, "created_at": "2025-01-16T06:10:06.905747+00:00"} {"global_step": 8714, "acc_step": 0, "speed/wps": 12888.472466863279, "speed/FLOPS": 202431170709197.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05247153341770172, "optim/lr": 0.0029663909230691426, "optim/total_tokens": 4568645632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 2.9869656562805176, "created_at": "2025-01-16T06:10:17.083234+00:00"} {"global_step": 8715, "acc_step": 0, "speed/wps": 12888.787557334252, "speed/FLOPS": 202436119638023.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0596502311527729, "optim/lr": 0.0029663728851246083, "optim/total_tokens": 4569169920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.997743606567383, "created_at": "2025-01-16T06:10:27.256777+00:00"} {"global_step": 8716, "acc_step": 0, "speed/wps": 12888.387509293756, "speed/FLOPS": 202429836333824.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04993180185556412, "optim/lr": 0.0029663548423957743, "optim/total_tokens": 4569694208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9305813312530518, "created_at": "2025-01-16T06:10:37.430495+00:00"} {"global_step": 8717, "acc_step": 0, "speed/wps": 12886.933232538246, "speed/FLOPS": 202406994918993.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050603557378053665, "optim/lr": 0.0029663367948827, "optim/total_tokens": 4570218496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9272446632385254, "created_at": "2025-01-16T06:10:47.607755+00:00"} {"global_step": 8718, "acc_step": 0, "speed/wps": 12886.696429300164, "speed/FLOPS": 202403275598737.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056348104029893875, "optim/lr": 0.0029663187425854436, "optim/total_tokens": 4570742784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.953951120376587, "created_at": "2025-01-16T06:10:57.785160+00:00"} {"global_step": 8719, "acc_step": 0, "speed/wps": 12891.963168112303, "speed/FLOPS": 202485996969039.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05849522352218628, "optim/lr": 0.0029663006855040652, "optim/total_tokens": 4571267072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 2.9433815479278564, "created_at": "2025-01-16T06:11:07.958100+00:00"} {"global_step": 8720, "acc_step": 0, "speed/wps": 12887.771386797842, "speed/FLOPS": 202420159283384.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04884350299835205, "optim/lr": 0.0029662826236386227, "optim/total_tokens": 4571791360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8601772785186768, "created_at": "2025-01-16T06:11:18.133939+00:00"} {"global_step": 8721, "acc_step": 0, "speed/wps": 12888.24121446466, "speed/FLOPS": 202427538572502.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05556129664182663, "optim/lr": 0.002966264556989175, "optim/total_tokens": 4572315648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 3.0480270385742188, "created_at": "2025-01-16T06:11:28.310147+00:00"} {"global_step": 8722, "acc_step": 0, "speed/wps": 12891.259384406914, "speed/FLOPS": 202474943078846.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05752317234873772, "optim/lr": 0.0029662464855557823, "optim/total_tokens": 4572839936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.9300010204315186, "created_at": "2025-01-16T06:11:38.485250+00:00"} {"global_step": 8723, "acc_step": 0, "speed/wps": 12891.51392653018, "speed/FLOPS": 202478941012669.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04813365638256073, "optim/lr": 0.0029662284093385023, "optim/total_tokens": 4573364224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9993443489074707, "created_at": "2025-01-16T06:11:48.656573+00:00"} {"global_step": 8724, "acc_step": 0, "speed/wps": 12893.719006384248, "speed/FLOPS": 202513574821874.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04758421331644058, "optim/lr": 0.002966210328337394, "optim/total_tokens": 4573888512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463218, "loss/out": 2.9943971633911133, "created_at": "2025-01-16T06:11:58.826824+00:00"} {"global_step": 8725, "acc_step": 0, "speed/wps": 12890.734272554608, "speed/FLOPS": 202466695475626.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046762142330408096, "optim/lr": 0.002966192242552517, "optim/total_tokens": 4574412800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.8368122577667236, "created_at": "2025-01-16T06:12:08.999590+00:00"} {"global_step": 8726, "acc_step": 0, "speed/wps": 12894.160712294432, "speed/FLOPS": 202520512420161.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05642450228333473, "optim/lr": 0.00296617415198393, "optim/total_tokens": 4574937088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.8478684425354004, "created_at": "2025-01-16T06:12:19.168335+00:00"} {"global_step": 8727, "acc_step": 0, "speed/wps": 12891.719812586844, "speed/FLOPS": 202482174736106.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06151886284351349, "optim/lr": 0.002966156056631692, "optim/total_tokens": 4575461376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9962918758392334, "created_at": "2025-01-16T06:12:29.340221+00:00"} {"global_step": 8728, "acc_step": 0, "speed/wps": 12888.608316320708, "speed/FLOPS": 202433304411602.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07176744937896729, "optim/lr": 0.0029661379564958623, "optim/total_tokens": 4575985664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 271085, "loss/out": 2.8119711875915527, "created_at": "2025-01-16T06:12:39.517847+00:00"} {"global_step": 8729, "acc_step": 0, "speed/wps": 12897.050518363072, "speed/FLOPS": 202565900795479.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05899522826075554, "optim/lr": 0.0029661198515765, "optim/total_tokens": 4576509952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.801208972930908, "created_at": "2025-01-16T06:12:49.684470+00:00"} {"global_step": 8730, "acc_step": 0, "speed/wps": 12896.140935827216, "speed/FLOPS": 202551614551856.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056807614862918854, "optim/lr": 0.0029661017418736637, "optim/total_tokens": 4577034240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.7892754077911377, "created_at": "2025-01-16T06:12:59.853532+00:00"} {"global_step": 8731, "acc_step": 0, "speed/wps": 12891.38335200588, "speed/FLOPS": 202476890160337.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051265403628349304, "optim/lr": 0.0029660836273874124, "optim/total_tokens": 4577558528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 2.9602832794189453, "created_at": "2025-01-16T06:13:10.026344+00:00"} {"global_step": 8732, "acc_step": 0, "speed/wps": 12887.847448247318, "speed/FLOPS": 202421353932968.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054155394434928894, "optim/lr": 0.002966065508117806, "optim/total_tokens": 4578082816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.852412223815918, "created_at": "2025-01-16T06:13:20.205215+00:00"} {"global_step": 8733, "acc_step": 0, "speed/wps": 12891.004407186614, "speed/FLOPS": 202470938311225.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06008337810635567, "optim/lr": 0.002966047384064903, "optim/total_tokens": 4578607104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.0718271732330322, "created_at": "2025-01-16T06:13:30.380542+00:00"} {"global_step": 8734, "acc_step": 0, "speed/wps": 12884.274770067854, "speed/FLOPS": 202365240112785.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05360783264040947, "optim/lr": 0.002966029255228763, "optim/total_tokens": 4579131392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8983538150787354, "created_at": "2025-01-16T06:13:40.557099+00:00"} {"global_step": 8735, "acc_step": 0, "speed/wps": 12890.235723801905, "speed/FLOPS": 202458865082387.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0511709526181221, "optim/lr": 0.002966011121609444, "optim/total_tokens": 4579655680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.9126648902893066, "created_at": "2025-01-16T06:13:50.731346+00:00"} {"global_step": 8736, "acc_step": 0, "speed/wps": 12891.473644012609, "speed/FLOPS": 202478308320377.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056393228471279144, "optim/lr": 0.002965992983207006, "optim/total_tokens": 4580179968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.847348928451538, "created_at": "2025-01-16T06:14:00.902337+00:00"} {"global_step": 8737, "acc_step": 0, "speed/wps": 12891.543500031095, "speed/FLOPS": 202479405505139.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053216561675071716, "optim/lr": 0.0029659748400215086, "optim/total_tokens": 4580704256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9876670837402344, "created_at": "2025-01-16T06:14:11.077065+00:00"} {"global_step": 8738, "acc_step": 0, "speed/wps": 12885.710870215293, "speed/FLOPS": 202387796039009.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055467311292886734, "optim/lr": 0.002965956692053011, "optim/total_tokens": 4581228544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 2.9172844886779785, "created_at": "2025-01-16T06:14:21.253484+00:00"} {"global_step": 8739, "acc_step": 0, "speed/wps": 12889.507948630478, "speed/FLOPS": 202447434373252.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0618714801967144, "optim/lr": 0.002965938539301571, "optim/total_tokens": 4581752832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9478378295898438, "created_at": "2025-01-16T06:14:31.427507+00:00"} {"global_step": 8740, "acc_step": 0, "speed/wps": 12892.316784707024, "speed/FLOPS": 202491551003579.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05235922709107399, "optim/lr": 0.0029659203817672497, "optim/total_tokens": 4582277120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.9385151863098145, "created_at": "2025-01-16T06:14:41.601455+00:00"} {"global_step": 8741, "acc_step": 0, "speed/wps": 12885.251720539434, "speed/FLOPS": 202380584462412.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05116764456033707, "optim/lr": 0.0029659022194501047, "optim/total_tokens": 4582801408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 3.0173299312591553, "created_at": "2025-01-16T06:14:51.777417+00:00"} {"global_step": 8742, "acc_step": 0, "speed/wps": 12886.62794500146, "speed/FLOPS": 202402199958718.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05955018848180771, "optim/lr": 0.0029658840523501965, "optim/total_tokens": 4583325696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 2.895571231842041, "created_at": "2025-01-16T06:15:01.953476+00:00"} {"global_step": 8743, "acc_step": 0, "speed/wps": 12885.601414378956, "speed/FLOPS": 202386076884691.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06713225692510605, "optim/lr": 0.0029658658804675836, "optim/total_tokens": 4583849984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.005697250366211, "created_at": "2025-01-16T06:15:12.132356+00:00"} {"global_step": 8744, "acc_step": 0, "speed/wps": 12887.757613007883, "speed/FLOPS": 202419942947086.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08330359309911728, "optim/lr": 0.0029658477038023256, "optim/total_tokens": 4584374272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.952393054962158, "created_at": "2025-01-16T06:15:22.308021+00:00"} {"global_step": 8745, "acc_step": 0, "speed/wps": 12888.338816765478, "speed/FLOPS": 202429071550756.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055546872317790985, "optim/lr": 0.0029658295223544823, "optim/total_tokens": 4584898560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9333817958831787, "created_at": "2025-01-16T06:15:32.484938+00:00"} {"global_step": 8746, "acc_step": 0, "speed/wps": 12889.401046563713, "speed/FLOPS": 202445755329394.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06054515391588211, "optim/lr": 0.002965811336124112, "optim/total_tokens": 4585422848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 2.997267007827759, "created_at": "2025-01-16T06:15:42.661113+00:00"} {"global_step": 8747, "acc_step": 0, "speed/wps": 12887.678885464453, "speed/FLOPS": 202418706422833.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04767119884490967, "optim/lr": 0.002965793145111274, "optim/total_tokens": 4585947136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.820047378540039, "created_at": "2025-01-16T06:15:52.834905+00:00"} {"global_step": 8748, "acc_step": 0, "speed/wps": 12885.725946847726, "speed/FLOPS": 202388032838239.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05484023317694664, "optim/lr": 0.0029657749493160294, "optim/total_tokens": 4586471424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9577174186706543, "created_at": "2025-01-16T06:16:03.011654+00:00"} {"global_step": 8749, "acc_step": 0, "speed/wps": 12887.015925328977, "speed/FLOPS": 202408293722903.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04484640806913376, "optim/lr": 0.0029657567487384358, "optim/total_tokens": 4586995712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 3.074148178100586, "created_at": "2025-01-16T06:16:13.188999+00:00"} {"global_step": 8750, "acc_step": 0, "speed/wps": 12887.31303198166, "speed/FLOPS": 202412960191150.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05195382609963417, "optim/lr": 0.002965738543378553, "optim/total_tokens": 4587520000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9181389808654785, "created_at": "2025-01-16T06:16:23.363255+00:00"} {"global_step": 8751, "acc_step": 0, "speed/wps": 12888.498762862584, "speed/FLOPS": 202431583724000.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04848114773631096, "optim/lr": 0.0029657203332364404, "optim/total_tokens": 4588044288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.7598981857299805, "created_at": "2025-01-16T06:16:33.536457+00:00"} {"global_step": 8752, "acc_step": 0, "speed/wps": 12890.547674870117, "speed/FLOPS": 202463764702579.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06867456436157227, "optim/lr": 0.0029657021183121576, "optim/total_tokens": 4588568576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.0099005699157715, "created_at": "2025-01-16T06:16:43.708065+00:00"} {"global_step": 8753, "acc_step": 0, "speed/wps": 12883.999857418796, "speed/FLOPS": 202360922231862.25, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055561572313308716, "optim/lr": 0.0029656838986057642, "optim/total_tokens": 4589092864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369627, "loss/out": 3.020937442779541, "created_at": "2025-01-16T06:16:53.885118+00:00"} {"global_step": 8754, "acc_step": 0, "speed/wps": 12892.755511514972, "speed/FLOPS": 202498441810972.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04942718520760536, "optim/lr": 0.0029656656741173194, "optim/total_tokens": 4589617152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9855051040649414, "created_at": "2025-01-16T06:17:04.057722+00:00"} {"global_step": 8755, "acc_step": 0, "speed/wps": 12886.895748132605, "speed/FLOPS": 202406406174893.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05335817113518715, "optim/lr": 0.0029656474448468825, "optim/total_tokens": 4590141440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9228739738464355, "created_at": "2025-01-16T06:17:14.232537+00:00"} {"global_step": 8756, "acc_step": 0, "speed/wps": 12884.02956040361, "speed/FLOPS": 202361388758054.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05058281496167183, "optim/lr": 0.002965629210794513, "optim/total_tokens": 4590665728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364786, "loss/out": 2.8696393966674805, "created_at": "2025-01-16T06:17:24.409389+00:00"} {"global_step": 8757, "acc_step": 0, "speed/wps": 12892.13772659227, "speed/FLOPS": 202488738649835.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061820678412914276, "optim/lr": 0.002965610971960271, "optim/total_tokens": 4591190016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.943478584289551, "created_at": "2025-01-16T06:17:34.585156+00:00"} {"global_step": 8758, "acc_step": 0, "speed/wps": 12889.303697317479, "speed/FLOPS": 202444226325710.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050314392894506454, "optim/lr": 0.002965592728344215, "optim/total_tokens": 4591714304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9616594314575195, "created_at": "2025-01-16T06:17:44.767297+00:00"} {"global_step": 8759, "acc_step": 0, "speed/wps": 12889.986870595838, "speed/FLOPS": 202454956500825.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05179768428206444, "optim/lr": 0.0029655744799464054, "optim/total_tokens": 4592238592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.9411520957946777, "created_at": "2025-01-16T06:17:54.942669+00:00"} {"global_step": 8760, "acc_step": 0, "speed/wps": 12889.682769994555, "speed/FLOPS": 202450180183004.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05594249442219734, "optim/lr": 0.0029655562267669015, "optim/total_tokens": 4592762880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477885, "loss/out": 2.8521642684936523, "created_at": "2025-01-16T06:18:05.117800+00:00"} {"global_step": 8761, "acc_step": 0, "speed/wps": 12892.008078285953, "speed/FLOPS": 202486702345029.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042548734694719315, "optim/lr": 0.0029655379688057625, "optim/total_tokens": 4593287168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.9510860443115234, "created_at": "2025-01-16T06:18:15.288903+00:00"} {"global_step": 8762, "acc_step": 0, "speed/wps": 12891.761428702832, "speed/FLOPS": 202482828374394.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05548976734280586, "optim/lr": 0.002965519706063048, "optim/total_tokens": 4593811456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9686501026153564, "created_at": "2025-01-16T06:18:25.461967+00:00"} {"global_step": 8763, "acc_step": 0, "speed/wps": 12885.585196017451, "speed/FLOPS": 202385822153037.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057125408202409744, "optim/lr": 0.0029655014385388187, "optim/total_tokens": 4594335744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 2.9402854442596436, "created_at": "2025-01-16T06:18:35.637832+00:00"} {"global_step": 8764, "acc_step": 0, "speed/wps": 12891.425360351097, "speed/FLOPS": 202477549959124.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06265810132026672, "optim/lr": 0.002965483166233133, "optim/total_tokens": 4594860032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 3.0788216590881348, "created_at": "2025-01-16T06:18:45.809149+00:00"} {"global_step": 8765, "acc_step": 0, "speed/wps": 12892.088852197796, "speed/FLOPS": 202487971010310.1, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06869178265333176, "optim/lr": 0.0029654648891460503, "optim/total_tokens": 4595384320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.851383686065674, "created_at": "2025-01-16T06:18:55.985148+00:00"} {"global_step": 8766, "acc_step": 0, "speed/wps": 12890.608947834486, "speed/FLOPS": 202464727078682.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05524729564785957, "optim/lr": 0.002965446607277631, "optim/total_tokens": 4595908608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.908511161804199, "created_at": "2025-01-16T06:19:06.158985+00:00"} {"global_step": 8767, "acc_step": 0, "speed/wps": 12890.625699531734, "speed/FLOPS": 202464990187107.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056060172617435455, "optim/lr": 0.0029654283206279348, "optim/total_tokens": 4596432896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.803799629211426, "created_at": "2025-01-16T06:19:16.330672+00:00"} {"global_step": 8768, "acc_step": 0, "speed/wps": 12889.828576929258, "speed/FLOPS": 202452470281274.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04727829247713089, "optim/lr": 0.002965410029197021, "optim/total_tokens": 4596957184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.977687358856201, "created_at": "2025-01-16T06:19:26.505727+00:00"} {"global_step": 8769, "acc_step": 0, "speed/wps": 12888.993634889332, "speed/FLOPS": 202439356369206.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.054355185478925705, "optim/lr": 0.002965391732984949, "optim/total_tokens": 4597481472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.8140671253204346, "created_at": "2025-01-16T06:19:36.679403+00:00"} {"global_step": 8770, "acc_step": 0, "speed/wps": 12889.510335584826, "speed/FLOPS": 202447471863650.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046413518488407135, "optim/lr": 0.0029653734319917795, "optim/total_tokens": 4598005760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.903388738632202, "created_at": "2025-01-16T06:19:46.852294+00:00"} {"global_step": 8771, "acc_step": 0, "speed/wps": 12886.845528150154, "speed/FLOPS": 202405617401059.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046190328896045685, "optim/lr": 0.0029653551262175705, "optim/total_tokens": 4598530048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 2.9409449100494385, "created_at": "2025-01-16T06:19:57.031102+00:00"} {"global_step": 8772, "acc_step": 0, "speed/wps": 12887.587806347134, "speed/FLOPS": 202417275900140.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.24263803660869598, "optim/lr": 0.002965336815662384, "optim/total_tokens": 4599054336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 2.864786148071289, "created_at": "2025-01-16T06:20:07.207073+00:00"} {"global_step": 8773, "acc_step": 0, "speed/wps": 12892.20752367277, "speed/FLOPS": 202489834908896.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06648297607898712, "optim/lr": 0.002965318500326278, "optim/total_tokens": 4599578624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 2.9005329608917236, "created_at": "2025-01-16T06:20:17.381309+00:00"} {"global_step": 8774, "acc_step": 0, "speed/wps": 12895.152304698207, "speed/FLOPS": 202536086741453.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057488951832056046, "optim/lr": 0.002965300180209313, "optim/total_tokens": 4600102912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 2.9953417778015137, "created_at": "2025-01-16T06:20:27.559693+00:00"} {"global_step": 8775, "acc_step": 0, "speed/wps": 12888.283903412932, "speed/FLOPS": 202428209061096.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3593139052391052, "optim/lr": 0.002965281855311548, "optim/total_tokens": 4600627200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.9941537380218506, "created_at": "2025-01-16T06:20:37.735704+00:00"} {"global_step": 8776, "acc_step": 0, "speed/wps": 12884.208826244205, "speed/FLOPS": 202364204374418.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6154184937477112, "optim/lr": 0.0029652635256330437, "optim/total_tokens": 4601151488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.005946159362793, "created_at": "2025-01-16T06:20:47.912451+00:00"} {"global_step": 8777, "acc_step": 0, "speed/wps": 12889.519759459215, "speed/FLOPS": 202447619878547.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12220197916030884, "optim/lr": 0.0029652451911738598, "optim/total_tokens": 4601675776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.927891731262207, "created_at": "2025-01-16T06:20:58.090261+00:00"} {"global_step": 8778, "acc_step": 0, "speed/wps": 12888.520729588006, "speed/FLOPS": 202431928741609.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1526498943567276, "optim/lr": 0.0029652268519340555, "optim/total_tokens": 4602200064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 2.9190735816955566, "created_at": "2025-01-16T06:21:08.263627+00:00"} {"global_step": 8779, "acc_step": 0, "speed/wps": 12886.98564923894, "speed/FLOPS": 202407818195305.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11412634700536728, "optim/lr": 0.0029652085079136916, "optim/total_tokens": 4602724352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9549999237060547, "created_at": "2025-01-16T06:21:18.459230+00:00"} {"global_step": 8780, "acc_step": 0, "speed/wps": 12885.322726619728, "speed/FLOPS": 202381699710475.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07424239069223404, "optim/lr": 0.0029651901591128266, "optim/total_tokens": 4603248640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.851752519607544, "created_at": "2025-01-16T06:21:28.637277+00:00"} {"global_step": 8781, "acc_step": 0, "speed/wps": 12884.779456643153, "speed/FLOPS": 202373166908962.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07454324513673782, "optim/lr": 0.002965171805531521, "optim/total_tokens": 4603772928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.009455919265747, "created_at": "2025-01-16T06:21:38.815871+00:00"} {"global_step": 8782, "acc_step": 0, "speed/wps": 12884.936522165159, "speed/FLOPS": 202375633838816.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06581209599971771, "optim/lr": 0.0029651534471698353, "optim/total_tokens": 4604297216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.962165594100952, "created_at": "2025-01-16T06:21:48.993420+00:00"} {"global_step": 8783, "acc_step": 0, "speed/wps": 12887.900582863947, "speed/FLOPS": 202422188485136.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.054519664496183395, "optim/lr": 0.0029651350840278297, "optim/total_tokens": 4604821504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 2.918029308319092, "created_at": "2025-01-16T06:21:59.169087+00:00"} {"global_step": 8784, "acc_step": 0, "speed/wps": 12889.183439674694, "speed/FLOPS": 202442337514179.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05660651624202728, "optim/lr": 0.002965116716105562, "optim/total_tokens": 4605345792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.8688251972198486, "created_at": "2025-01-16T06:22:09.343061+00:00"} {"global_step": 8785, "acc_step": 0, "speed/wps": 12891.771621096366, "speed/FLOPS": 202482988459940.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0455206036567688, "optim/lr": 0.002965098343403094, "optim/total_tokens": 4605870080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.9961280822753906, "created_at": "2025-01-16T06:22:19.515773+00:00"} {"global_step": 8786, "acc_step": 0, "speed/wps": 12890.300356253127, "speed/FLOPS": 202459880223848.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053690437227487564, "optim/lr": 0.0029650799659204845, "optim/total_tokens": 4606394368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9828662872314453, "created_at": "2025-01-16T06:22:29.693317+00:00"} {"global_step": 8787, "acc_step": 0, "speed/wps": 12892.787221402768, "speed/FLOPS": 202498939858334.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052146315574645996, "optim/lr": 0.002965061583657795, "optim/total_tokens": 4606918656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 3.0602636337280273, "created_at": "2025-01-16T06:22:39.863065+00:00"} {"global_step": 8788, "acc_step": 0, "speed/wps": 12884.811203984671, "speed/FLOPS": 202373665544586.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06880208104848862, "optim/lr": 0.002965043196615084, "optim/total_tokens": 4607442944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.0623385906219482, "created_at": "2025-01-16T06:22:50.044815+00:00"} {"global_step": 8789, "acc_step": 0, "speed/wps": 12890.93441413931, "speed/FLOPS": 202469838974237.44, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0628766417503357, "optim/lr": 0.002965024804792412, "optim/total_tokens": 4607967232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9172587394714355, "created_at": "2025-01-16T06:23:00.220383+00:00"} {"global_step": 8790, "acc_step": 0, "speed/wps": 12889.042518869783, "speed/FLOPS": 202440124159293.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05466127023100853, "optim/lr": 0.002965006408189839, "optim/total_tokens": 4608491520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372786, "loss/out": 2.8955602645874023, "created_at": "2025-01-16T06:23:10.396228+00:00"} {"global_step": 8791, "acc_step": 0, "speed/wps": 12884.27009911818, "speed/FLOPS": 202365166749102.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05363583192229271, "optim/lr": 0.0029649880068074257, "optim/total_tokens": 4609015808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.8726091384887695, "created_at": "2025-01-16T06:23:20.573856+00:00"} {"global_step": 8792, "acc_step": 0, "speed/wps": 12885.193503192162, "speed/FLOPS": 202379670078974.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04915667325258255, "optim/lr": 0.0029649696006452306, "optim/total_tokens": 4609540096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.9245591163635254, "created_at": "2025-01-16T06:23:30.750091+00:00"} {"global_step": 8793, "acc_step": 0, "speed/wps": 12884.262751815553, "speed/FLOPS": 202365051349618.1, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05211430415511131, "optim/lr": 0.002964951189703315, "optim/total_tokens": 4610064384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9720654487609863, "created_at": "2025-01-16T06:23:40.926801+00:00"} {"global_step": 8794, "acc_step": 0, "speed/wps": 12879.882621169074, "speed/FLOPS": 202296255378884.88, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04686683416366577, "optim/lr": 0.0029649327739817385, "optim/total_tokens": 4610588672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 3.0100626945495605, "created_at": "2025-01-16T06:23:51.109699+00:00"} {"global_step": 8795, "acc_step": 0, "speed/wps": 12882.300913783018, "speed/FLOPS": 202334237987469.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05075155198574066, "optim/lr": 0.0029649143534805616, "optim/total_tokens": 4611112960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307266, "loss/out": 2.8914670944213867, "created_at": "2025-01-16T06:24:01.290999+00:00"} {"global_step": 8796, "acc_step": 0, "speed/wps": 12885.03072179198, "speed/FLOPS": 202377113373399.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04242189973592758, "optim/lr": 0.0029648959281998438, "optim/total_tokens": 4611637248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.8227155208587646, "created_at": "2025-01-16T06:24:11.468952+00:00"} {"global_step": 8797, "acc_step": 0, "speed/wps": 12888.388750955088, "speed/FLOPS": 202429855835821.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04679718613624573, "optim/lr": 0.0029648774981396457, "optim/total_tokens": 4612161536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.0170342922210693, "created_at": "2025-01-16T06:24:21.644687+00:00"} {"global_step": 8798, "acc_step": 0, "speed/wps": 12886.119339202753, "speed/FLOPS": 202394211605752.47, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05076500400900841, "optim/lr": 0.002964859063300027, "optim/total_tokens": 4612685824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9969935417175293, "created_at": "2025-01-16T06:24:31.821567+00:00"} {"global_step": 8799, "acc_step": 0, "speed/wps": 12890.933258182376, "speed/FLOPS": 202469820818345.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05054514482617378, "optim/lr": 0.0029648406236810485, "optim/total_tokens": 4613210112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.9458260536193848, "created_at": "2025-01-16T06:24:41.993706+00:00"} {"global_step": 8800, "acc_step": 0, "speed/wps": 12886.894551599418, "speed/FLOPS": 202406387381695.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05649116635322571, "optim/lr": 0.00296482217928277, "optim/total_tokens": 4613734400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.856901168823242, "created_at": "2025-01-16T06:24:52.171825+00:00"} {"global_step": 8801, "acc_step": 0, "speed/wps": 12884.184208506364, "speed/FLOPS": 202363817719017.47, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047969210892915726, "optim/lr": 0.002964803730105251, "optim/total_tokens": 4614258688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 2.988220691680908, "created_at": "2025-01-16T06:25:02.352024+00:00"} {"global_step": 8802, "acc_step": 0, "speed/wps": 12894.491355317705, "speed/FLOPS": 202525705623193.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.049632780253887177, "optim/lr": 0.0029647852761485525, "optim/total_tokens": 4614782976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.9272494316101074, "created_at": "2025-01-16T06:25:12.522801+00:00"} {"global_step": 8803, "acc_step": 0, "speed/wps": 12884.743987758784, "speed/FLOPS": 202372609821394.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04886748641729355, "optim/lr": 0.0029647668174127345, "optim/total_tokens": 4615307264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9445290565490723, "created_at": "2025-01-16T06:25:22.702304+00:00"} {"global_step": 8804, "acc_step": 0, "speed/wps": 12887.612636537577, "speed/FLOPS": 202417665892401.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05066509544849396, "optim/lr": 0.0029647483538978573, "optim/total_tokens": 4615831552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.865290403366089, "created_at": "2025-01-16T06:25:32.879068+00:00"} {"global_step": 8805, "acc_step": 0, "speed/wps": 12886.115036214898, "speed/FLOPS": 202394144021415.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044960565865039825, "optim/lr": 0.0029647298856039814, "optim/total_tokens": 4616355840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9247827529907227, "created_at": "2025-01-16T06:25:43.056667+00:00"} {"global_step": 8806, "acc_step": 0, "speed/wps": 12887.436259289265, "speed/FLOPS": 202414895645349.0, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051166336983442307, "optim/lr": 0.002964711412531166, "optim/total_tokens": 4616880128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413993, "loss/out": 2.891035318374634, "created_at": "2025-01-16T06:25:53.232976+00:00"} {"global_step": 8807, "acc_step": 0, "speed/wps": 12887.308914650586, "speed/FLOPS": 202412895522808.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053075339645147324, "optim/lr": 0.002964692934679473, "optim/total_tokens": 4617404416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474863, "loss/out": 2.868422508239746, "created_at": "2025-01-16T06:26:03.412463+00:00"} {"global_step": 8808, "acc_step": 0, "speed/wps": 12888.778081414392, "speed/FLOPS": 202435970805680.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05177921801805496, "optim/lr": 0.0029646744520489612, "optim/total_tokens": 4617928704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.829246997833252, "created_at": "2025-01-16T06:26:13.590932+00:00"} {"global_step": 8809, "acc_step": 0, "speed/wps": 12887.723762253241, "speed/FLOPS": 202419411274468.72, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04572585970163345, "optim/lr": 0.002964655964639692, "optim/total_tokens": 4618452992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.9740686416625977, "created_at": "2025-01-16T06:26:23.764822+00:00"} {"global_step": 8810, "acc_step": 0, "speed/wps": 12883.9015683252, "speed/FLOPS": 202359378466583.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04549203813076019, "optim/lr": 0.002964637472451725, "optim/total_tokens": 4618977280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 2.854297637939453, "created_at": "2025-01-16T06:26:33.946209+00:00"} {"global_step": 8811, "acc_step": 0, "speed/wps": 12886.627038789118, "speed/FLOPS": 202402185725408.7, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04090006649494171, "optim/lr": 0.0029646189754851207, "optim/total_tokens": 4619501568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.904496669769287, "created_at": "2025-01-16T06:26:44.122173+00:00"} {"global_step": 8812, "acc_step": 0, "speed/wps": 12890.00261600082, "speed/FLOPS": 202455203804047.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.041151754558086395, "optim/lr": 0.0029646004737399396, "optim/total_tokens": 4620025856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.973637104034424, "created_at": "2025-01-16T06:26:54.296420+00:00"} {"global_step": 8813, "acc_step": 0, "speed/wps": 12890.698221848206, "speed/FLOPS": 202466129249743.44, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04343102127313614, "optim/lr": 0.002964581967216242, "optim/total_tokens": 4620550144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476257, "loss/out": 2.9687910079956055, "created_at": "2025-01-16T06:27:04.467952+00:00"} {"global_step": 8814, "acc_step": 0, "speed/wps": 12885.918419754686, "speed/FLOPS": 202391055889728.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051215942949056625, "optim/lr": 0.0029645634559140885, "optim/total_tokens": 4621074432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.8728954792022705, "created_at": "2025-01-16T06:27:14.643785+00:00"} {"global_step": 8815, "acc_step": 0, "speed/wps": 12889.289064424658, "speed/FLOPS": 202443996496021.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054885294288396835, "optim/lr": 0.0029645449398335386, "optim/total_tokens": 4621598720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.9932403564453125, "created_at": "2025-01-16T06:27:24.819899+00:00"} {"global_step": 8816, "acc_step": 0, "speed/wps": 12883.671510738792, "speed/FLOPS": 202355765096057.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04956132918596268, "optim/lr": 0.002964526418974654, "optim/total_tokens": 4622123008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.021881580352783, "created_at": "2025-01-16T06:27:34.997828+00:00"} {"global_step": 8817, "acc_step": 0, "speed/wps": 12888.866294882371, "speed/FLOPS": 202437356319413.56, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06170029938220978, "optim/lr": 0.002964507893337494, "optim/total_tokens": 4622647296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299938, "loss/out": 2.8668086528778076, "created_at": "2025-01-16T06:27:45.171538+00:00"} {"global_step": 8818, "acc_step": 0, "speed/wps": 12884.87146927527, "speed/FLOPS": 202374612093788.38, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04859544709324837, "optim/lr": 0.00296448936292212, "optim/total_tokens": 4623171584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 2.9946208000183105, "created_at": "2025-01-16T06:27:55.351571+00:00"} {"global_step": 8819, "acc_step": 0, "speed/wps": 12892.46205127498, "speed/FLOPS": 202493832614646.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04999031871557236, "optim/lr": 0.0029644708277285917, "optim/total_tokens": 4623695872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481042, "loss/out": 2.9083800315856934, "created_at": "2025-01-16T06:28:05.524898+00:00"} {"global_step": 8820, "acc_step": 0, "speed/wps": 12886.357927681176, "speed/FLOPS": 202397958965661.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06052462011575699, "optim/lr": 0.00296445228775697, "optim/total_tokens": 4624220160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 2.9995169639587402, "created_at": "2025-01-16T06:28:15.709950+00:00"} {"global_step": 8821, "acc_step": 0, "speed/wps": 12885.845895758286, "speed/FLOPS": 202389916800706.47, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05090362951159477, "optim/lr": 0.0029644337430073152, "optim/total_tokens": 4624744448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.966944694519043, "created_at": "2025-01-16T06:28:25.886912+00:00"} {"global_step": 8822, "acc_step": 0, "speed/wps": 12887.878333825849, "speed/FLOPS": 202421839033420.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04333354905247688, "optim/lr": 0.002964415193479688, "optim/total_tokens": 4625268736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.891432523727417, "created_at": "2025-01-16T06:28:36.062436+00:00"} {"global_step": 8823, "acc_step": 0, "speed/wps": 12885.989927539049, "speed/FLOPS": 202392179017743.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05185087025165558, "optim/lr": 0.0029643966391741487, "optim/total_tokens": 4625793024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.1090548038482666, "created_at": "2025-01-16T06:28:46.241033+00:00"} {"global_step": 8824, "acc_step": 0, "speed/wps": 12888.792785149015, "speed/FLOPS": 202436201748038.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05113212764263153, "optim/lr": 0.0029643780800907575, "optim/total_tokens": 4626317312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9764113426208496, "created_at": "2025-01-16T06:28:56.414589+00:00"} {"global_step": 8825, "acc_step": 0, "speed/wps": 12889.843318477415, "speed/FLOPS": 202452701817544.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05459906905889511, "optim/lr": 0.002964359516229576, "optim/total_tokens": 4626841600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 2.913268566131592, "created_at": "2025-01-16T06:29:06.589132+00:00"} {"global_step": 8826, "acc_step": 0, "speed/wps": 12889.687846247898, "speed/FLOPS": 202450259912539.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055996060371398926, "optim/lr": 0.0029643409475906637, "optim/total_tokens": 4627365888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.992917537689209, "created_at": "2025-01-16T06:29:16.762947+00:00"} {"global_step": 8827, "acc_step": 0, "speed/wps": 12891.33747908754, "speed/FLOPS": 202476169663118.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06199069321155548, "optim/lr": 0.002964322374174082, "optim/total_tokens": 4627890176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.720806360244751, "created_at": "2025-01-16T06:29:26.934331+00:00"} {"global_step": 8828, "acc_step": 0, "speed/wps": 12889.51009122299, "speed/FLOPS": 202447468025611.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05284586921334267, "optim/lr": 0.0029643037959798915, "optim/total_tokens": 4628414464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 2.8680665493011475, "created_at": "2025-01-16T06:29:37.110336+00:00"} {"global_step": 8829, "acc_step": 0, "speed/wps": 12885.357205356564, "speed/FLOPS": 202382241246416.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05756831169128418, "optim/lr": 0.002964285213008152, "optim/total_tokens": 4628938752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9161412715911865, "created_at": "2025-01-16T06:29:47.286467+00:00"} {"global_step": 8830, "acc_step": 0, "speed/wps": 12886.946380334273, "speed/FLOPS": 202407201423196.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06882777065038681, "optim/lr": 0.002964266625258924, "optim/total_tokens": 4629463040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8619415760040283, "created_at": "2025-01-16T06:29:57.462676+00:00"} {"global_step": 8831, "acc_step": 0, "speed/wps": 12888.74680672845, "speed/FLOPS": 202435479593762.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06331739574670792, "optim/lr": 0.002964248032732269, "optim/total_tokens": 4629987328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8802378177642822, "created_at": "2025-01-16T06:30:07.636595+00:00"} {"global_step": 8832, "acc_step": 0, "speed/wps": 12888.058141108484, "speed/FLOPS": 202424663153875.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0489344522356987, "optim/lr": 0.0029642294354282482, "optim/total_tokens": 4630511616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336107, "loss/out": 3.019744396209717, "created_at": "2025-01-16T06:30:17.810887+00:00"} {"global_step": 8833, "acc_step": 0, "speed/wps": 12892.271552332137, "speed/FLOPS": 202490840566976.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07959655672311783, "optim/lr": 0.002964210833346921, "optim/total_tokens": 4631035904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.93448543548584, "created_at": "2025-01-16T06:30:27.981328+00:00"} {"global_step": 8834, "acc_step": 0, "speed/wps": 12890.767434234638, "speed/FLOPS": 202467216325380.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05483875796198845, "optim/lr": 0.002964192226488348, "optim/total_tokens": 4631560192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 2.9838075637817383, "created_at": "2025-01-16T06:30:38.153045+00:00"} {"global_step": 8835, "acc_step": 0, "speed/wps": 12886.166607365436, "speed/FLOPS": 202394954017200.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04674513638019562, "optim/lr": 0.0029641736148525912, "optim/total_tokens": 4632084480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315344, "loss/out": 2.942625045776367, "created_at": "2025-01-16T06:30:48.331990+00:00"} {"global_step": 8836, "acc_step": 0, "speed/wps": 12894.226603271323, "speed/FLOPS": 202521547328497.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045579101890325546, "optim/lr": 0.0029641549984397103, "optim/total_tokens": 4632608768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.900393009185791, "created_at": "2025-01-16T06:30:58.505432+00:00"} {"global_step": 8837, "acc_step": 0, "speed/wps": 12889.44371250315, "speed/FLOPS": 202446425456602.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05268453434109688, "optim/lr": 0.0029641363772497666, "optim/total_tokens": 4633133056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.966660976409912, "created_at": "2025-01-16T06:31:08.680983+00:00"} {"global_step": 8838, "acc_step": 0, "speed/wps": 12888.024175474004, "speed/FLOPS": 202424129676911.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049507588148117065, "optim/lr": 0.00296411775128282, "optim/total_tokens": 4633657344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9993667602539062, "created_at": "2025-01-16T06:31:18.860187+00:00"} {"global_step": 8839, "acc_step": 0, "speed/wps": 12891.76004232127, "speed/FLOPS": 202482806599367.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.039464373141527176, "optim/lr": 0.0029640991205389317, "optim/total_tokens": 4634181632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309387, "loss/out": 2.8502495288848877, "created_at": "2025-01-16T06:31:29.031764+00:00"} {"global_step": 8840, "acc_step": 0, "speed/wps": 12892.467082476303, "speed/FLOPS": 202493911636576.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04772099852561951, "optim/lr": 0.0029640804850181637, "optim/total_tokens": 4634705920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.926223039627075, "created_at": "2025-01-16T06:31:39.207465+00:00"} {"global_step": 8841, "acc_step": 0, "speed/wps": 12893.390660246978, "speed/FLOPS": 202508417694594.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05058462545275688, "optim/lr": 0.0029640618447205748, "optim/total_tokens": 4635230208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8980507850646973, "created_at": "2025-01-16T06:31:49.378018+00:00"} {"global_step": 8842, "acc_step": 0, "speed/wps": 12890.935416598555, "speed/FLOPS": 202469854719237.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05026617273688316, "optim/lr": 0.002964043199646227, "optim/total_tokens": 4635754496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 2.9059736728668213, "created_at": "2025-01-16T06:31:59.552489+00:00"} {"global_step": 8843, "acc_step": 0, "speed/wps": 12892.32558557048, "speed/FLOPS": 202491689233233.84, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05505030229687691, "optim/lr": 0.002964024549795181, "optim/total_tokens": 4636278784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.0324418544769287, "created_at": "2025-01-16T06:32:09.723059+00:00"} {"global_step": 8844, "acc_step": 0, "speed/wps": 12887.369610442714, "speed/FLOPS": 202413848833628.8, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06831549108028412, "optim/lr": 0.002964005895167497, "optim/total_tokens": 4636803072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.979947090148926, "created_at": "2025-01-16T06:32:19.898633+00:00"} {"global_step": 8845, "acc_step": 0, "speed/wps": 12888.186069434678, "speed/FLOPS": 202426672444029.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06491529196500778, "optim/lr": 0.0029639872357632375, "optim/total_tokens": 4637327360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8632946014404297, "created_at": "2025-01-16T06:32:30.072202+00:00"} {"global_step": 8846, "acc_step": 0, "speed/wps": 12887.723824801975, "speed/FLOPS": 202419412256882.53, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053547151386737823, "optim/lr": 0.0029639685715824615, "optim/total_tokens": 4637851648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.949305295944214, "created_at": "2025-01-16T06:32:40.248680+00:00"} {"global_step": 8847, "acc_step": 0, "speed/wps": 12888.423617894388, "speed/FLOPS": 202430403469015.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05500837415456772, "optim/lr": 0.0029639499026252306, "optim/total_tokens": 4638375936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.810819149017334, "created_at": "2025-01-16T06:32:50.432565+00:00"} {"global_step": 8848, "acc_step": 0, "speed/wps": 12888.748318759706, "speed/FLOPS": 202435503342290.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05212937667965889, "optim/lr": 0.002963931228891606, "optim/total_tokens": 4638900224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9871082305908203, "created_at": "2025-01-16T06:33:00.606014+00:00"} {"global_step": 8849, "acc_step": 0, "speed/wps": 12890.532773499808, "speed/FLOPS": 202463530656082.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056335873901844025, "optim/lr": 0.0029639125503816486, "optim/total_tokens": 4639424512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 2.8943190574645996, "created_at": "2025-01-16T06:33:10.779042+00:00"} {"global_step": 8850, "acc_step": 0, "speed/wps": 12884.314421377061, "speed/FLOPS": 202365862891083.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051734793931245804, "optim/lr": 0.002963893867095419, "optim/total_tokens": 4639948800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8971478939056396, "created_at": "2025-01-16T06:33:20.959037+00:00"} {"global_step": 8851, "acc_step": 0, "speed/wps": 12891.594119164462, "speed/FLOPS": 202480200548187.16, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04793859273195267, "optim/lr": 0.0029638751790329785, "optim/total_tokens": 4640473088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.031054735183716, "created_at": "2025-01-16T06:33:31.130985+00:00"} {"global_step": 8852, "acc_step": 0, "speed/wps": 12890.940853968556, "speed/FLOPS": 202469940120605.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05354795977473259, "optim/lr": 0.0029638564861943872, "optim/total_tokens": 4640997376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8928399085998535, "created_at": "2025-01-16T06:33:41.302528+00:00"} {"global_step": 8853, "acc_step": 0, "speed/wps": 12883.821467827305, "speed/FLOPS": 202358120378194.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.044036801904439926, "optim/lr": 0.0029638377885797076, "optim/total_tokens": 4641521664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413463, "loss/out": 3.033686876296997, "created_at": "2025-01-16T06:33:51.479831+00:00"} {"global_step": 8854, "acc_step": 0, "speed/wps": 12888.439337054882, "speed/FLOPS": 202430650360031.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057381562888622284, "optim/lr": 0.0029638190861889994, "optim/total_tokens": 4642045952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.959300994873047, "created_at": "2025-01-16T06:34:01.653771+00:00"} {"global_step": 8855, "acc_step": 0, "speed/wps": 12887.023727725937, "speed/FLOPS": 202408416270269.16, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04883287101984024, "optim/lr": 0.0029638003790223243, "optim/total_tokens": 4642570240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8785853385925293, "created_at": "2025-01-16T06:34:11.828157+00:00"} {"global_step": 8856, "acc_step": 0, "speed/wps": 12890.502149113627, "speed/FLOPS": 202463049658014.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05059242993593216, "optim/lr": 0.002963781667079743, "optim/total_tokens": 4643094528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421973, "loss/out": 2.934511661529541, "created_at": "2025-01-16T06:34:22.001486+00:00"} {"global_step": 8857, "acc_step": 0, "speed/wps": 12890.62855327531, "speed/FLOPS": 202465035009071.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053245142102241516, "optim/lr": 0.0029637629503613165, "optim/total_tokens": 4643618816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.964972496032715, "created_at": "2025-01-16T06:34:32.176414+00:00"} {"global_step": 8858, "acc_step": 0, "speed/wps": 12889.600917732225, "speed/FLOPS": 202448894580748.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0621715746819973, "optim/lr": 0.0029637442288671066, "optim/total_tokens": 4644143104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9237046241760254, "created_at": "2025-01-16T06:34:42.352539+00:00"} {"global_step": 8859, "acc_step": 0, "speed/wps": 12889.672999988912, "speed/FLOPS": 202450026731640.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055615123361349106, "optim/lr": 0.0029637255025971735, "optim/total_tokens": 4644667392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.9757132530212402, "created_at": "2025-01-16T06:34:52.530172+00:00"} {"global_step": 8860, "acc_step": 0, "speed/wps": 12890.72413493343, "speed/FLOPS": 202466536250355.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051408614963293076, "optim/lr": 0.0029637067715515787, "optim/total_tokens": 4645191680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372547, "loss/out": 2.9172873497009277, "created_at": "2025-01-16T06:35:02.705223+00:00"} {"global_step": 8861, "acc_step": 0, "speed/wps": 12886.840777456131, "speed/FLOPS": 202405542784881.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0772135928273201, "optim/lr": 0.002963688035730383, "optim/total_tokens": 4645715968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9554219245910645, "created_at": "2025-01-16T06:35:12.881739+00:00"} {"global_step": 8862, "acc_step": 0, "speed/wps": 12889.154750516653, "speed/FLOPS": 202441886911529.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04656453803181648, "optim/lr": 0.002963669295133648, "optim/total_tokens": 4646240256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.873617172241211, "created_at": "2025-01-16T06:35:23.054895+00:00"} {"global_step": 8863, "acc_step": 0, "speed/wps": 12895.913027604096, "speed/FLOPS": 202548034940033.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052456364035606384, "optim/lr": 0.0029636505497614347, "optim/total_tokens": 4646764544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9360668659210205, "created_at": "2025-01-16T06:35:33.224380+00:00"} {"global_step": 8864, "acc_step": 0, "speed/wps": 12889.574325650927, "speed/FLOPS": 202448476915569.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04518814757466316, "optim/lr": 0.002963631799613804, "optim/total_tokens": 4647288832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 2.914052963256836, "created_at": "2025-01-16T06:35:43.398237+00:00"} {"global_step": 8865, "acc_step": 0, "speed/wps": 12890.82426146671, "speed/FLOPS": 202468108875148.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054961513727903366, "optim/lr": 0.0029636130446908174, "optim/total_tokens": 4647813120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.956494092941284, "created_at": "2025-01-16T06:35:53.570090+00:00"} {"global_step": 8866, "acc_step": 0, "speed/wps": 12890.89792189512, "speed/FLOPS": 202469265813396.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05507230386137962, "optim/lr": 0.0029635942849925356, "optim/total_tokens": 4648337408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9394326210021973, "created_at": "2025-01-16T06:36:03.741498+00:00"} {"global_step": 8867, "acc_step": 0, "speed/wps": 12896.920882322449, "speed/FLOPS": 202563864683322.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04895513504743576, "optim/lr": 0.0029635755205190207, "optim/total_tokens": 4648861696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 2.7396669387817383, "created_at": "2025-01-16T06:36:13.909039+00:00"} {"global_step": 8868, "acc_step": 0, "speed/wps": 12888.765384317412, "speed/FLOPS": 202435771380325.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07000225782394409, "optim/lr": 0.0029635567512703324, "optim/total_tokens": 4649385984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.9874823093414307, "created_at": "2025-01-16T06:36:24.082788+00:00"} {"global_step": 8869, "acc_step": 0, "speed/wps": 12890.564272919057, "speed/FLOPS": 202464025397746.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06367877870798111, "optim/lr": 0.002963537977246534, "optim/total_tokens": 4649910272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.996242046356201, "created_at": "2025-01-16T06:36:34.255712+00:00"} {"global_step": 8870, "acc_step": 0, "speed/wps": 12893.088266415802, "speed/FLOPS": 202503668183942.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057124339044094086, "optim/lr": 0.0029635191984476853, "optim/total_tokens": 4650434560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 2.7972002029418945, "created_at": "2025-01-16T06:36:44.427387+00:00"} {"global_step": 8871, "acc_step": 0, "speed/wps": 12891.5960888621, "speed/FLOPS": 202480231484995.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08047215640544891, "optim/lr": 0.0029635004148738475, "optim/total_tokens": 4650958848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.948049545288086, "created_at": "2025-01-16T06:36:54.599888+00:00"} {"global_step": 8872, "acc_step": 0, "speed/wps": 12893.778741950402, "speed/FLOPS": 202514513051026.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05227028205990791, "optim/lr": 0.0029634816265250827, "optim/total_tokens": 4651483136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.9413259029388428, "created_at": "2025-01-16T06:37:04.769585+00:00"} {"global_step": 8873, "acc_step": 0, "speed/wps": 12894.422458536023, "speed/FLOPS": 202524623504563.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06643617898225784, "optim/lr": 0.0029634628334014525, "optim/total_tokens": 4652007424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.915987491607666, "created_at": "2025-01-16T06:37:14.938694+00:00"} {"global_step": 8874, "acc_step": 0, "speed/wps": 12890.004140132563, "speed/FLOPS": 202455227742631.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06952786445617676, "optim/lr": 0.0029634440355030163, "optim/total_tokens": 4652531712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9609999656677246, "created_at": "2025-01-16T06:37:25.113972+00:00"} {"global_step": 8875, "acc_step": 0, "speed/wps": 12889.467587750916, "speed/FLOPS": 202446800450177.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06675991415977478, "optim/lr": 0.002963425232829838, "optim/total_tokens": 4653056000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.7947702407836914, "created_at": "2025-01-16T06:37:35.286730+00:00"} {"global_step": 8876, "acc_step": 0, "speed/wps": 12891.608119618282, "speed/FLOPS": 202480420444553.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0938972607254982, "optim/lr": 0.002963406425381976, "optim/total_tokens": 4653580288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9468061923980713, "created_at": "2025-01-16T06:37:45.461726+00:00"} {"global_step": 8877, "acc_step": 0, "speed/wps": 12889.33820916692, "speed/FLOPS": 202444768381730.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05338328331708908, "optim/lr": 0.0029633876131594946, "optim/total_tokens": 4654104576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 3.016937255859375, "created_at": "2025-01-16T06:37:55.634504+00:00"} {"global_step": 8878, "acc_step": 0, "speed/wps": 12892.564886975653, "speed/FLOPS": 202495447790636.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07285675406455994, "optim/lr": 0.0029633687961624533, "optim/total_tokens": 4654628864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9971563816070557, "created_at": "2025-01-16T06:38:05.807300+00:00"} {"global_step": 8879, "acc_step": 0, "speed/wps": 12885.907856708289, "speed/FLOPS": 202390889982569.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06544865667819977, "optim/lr": 0.002963349974390914, "optim/total_tokens": 4655153152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.0196008682250977, "created_at": "2025-01-16T06:38:15.984007+00:00"} {"global_step": 8880, "acc_step": 0, "speed/wps": 12880.22129781128, "speed/FLOPS": 202301574760941.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05998262017965317, "optim/lr": 0.0029633311478449386, "optim/total_tokens": 4655677440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387212, "loss/out": 2.829813003540039, "created_at": "2025-01-16T06:38:26.164251+00:00"} {"global_step": 8881, "acc_step": 0, "speed/wps": 12892.372454031884, "speed/FLOPS": 202492425366824.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06370069086551666, "optim/lr": 0.0029633123165245874, "optim/total_tokens": 4656201728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.966888427734375, "created_at": "2025-01-16T06:38:36.336139+00:00"} {"global_step": 8882, "acc_step": 0, "speed/wps": 12890.356378888475, "speed/FLOPS": 202460760136319.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06565870344638824, "optim/lr": 0.0029632934804299225, "optim/total_tokens": 4656726016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.0008554458618164, "created_at": "2025-01-16T06:38:46.510971+00:00"} {"global_step": 8883, "acc_step": 0, "speed/wps": 12891.82920234089, "speed/FLOPS": 202483892852511.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050859492272138596, "optim/lr": 0.002963274639561006, "optim/total_tokens": 4657250304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.948115587234497, "created_at": "2025-01-16T06:38:56.682279+00:00"} {"global_step": 8884, "acc_step": 0, "speed/wps": 12888.955244940982, "speed/FLOPS": 202438753402314.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06645183265209198, "optim/lr": 0.0029632557939178986, "optim/total_tokens": 4657774592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 2.9752564430236816, "created_at": "2025-01-16T06:39:06.855230+00:00"} {"global_step": 8885, "acc_step": 0, "speed/wps": 12883.949493698441, "speed/FLOPS": 202360131200427.1, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056518737226724625, "optim/lr": 0.002963236943500662, "optim/total_tokens": 4658298880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.977599620819092, "created_at": "2025-01-16T06:39:17.035311+00:00"} {"global_step": 8886, "acc_step": 0, "speed/wps": 12893.362946925217, "speed/FLOPS": 202507982418793.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.068950355052948, "optim/lr": 0.0029632180883093573, "optim/total_tokens": 4658823168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 2.8903753757476807, "created_at": "2025-01-16T06:39:27.208824+00:00"} {"global_step": 8887, "acc_step": 0, "speed/wps": 12887.670159392917, "speed/FLOPS": 202418569367889.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06379524618387222, "optim/lr": 0.0029631992283440465, "optim/total_tokens": 4659347456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285262, "loss/out": 2.94743275642395, "created_at": "2025-01-16T06:39:37.386771+00:00"} {"global_step": 8888, "acc_step": 0, "speed/wps": 12890.39469808007, "speed/FLOPS": 202461361991879.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06811729073524475, "optim/lr": 0.0029631803636047914, "optim/total_tokens": 4659871744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9397809505462646, "created_at": "2025-01-16T06:39:47.565026+00:00"} {"global_step": 8889, "acc_step": 0, "speed/wps": 12890.268869929987, "speed/FLOPS": 202459385687876.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09776589274406433, "optim/lr": 0.0029631614940916528, "optim/total_tokens": 4660396032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.9109535217285156, "created_at": "2025-01-16T06:39:57.739142+00:00"} {"global_step": 8890, "acc_step": 0, "speed/wps": 12888.922773665789, "speed/FLOPS": 202438243396317.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06889012455940247, "optim/lr": 0.002963142619804692, "optim/total_tokens": 4660920320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9547901153564453, "created_at": "2025-01-16T06:40:07.912466+00:00"} {"global_step": 8891, "acc_step": 0, "speed/wps": 12885.60591640609, "speed/FLOPS": 202386147595214.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05717708542943001, "optim/lr": 0.002963123740743972, "optim/total_tokens": 4661444608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415854, "loss/out": 2.9538848400115967, "created_at": "2025-01-16T06:40:18.087925+00:00"} {"global_step": 8892, "acc_step": 0, "speed/wps": 12885.993340838002, "speed/FLOPS": 202392232628294.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06995096802711487, "optim/lr": 0.0029631048569095536, "optim/total_tokens": 4661968896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.955953598022461, "created_at": "2025-01-16T06:40:28.263871+00:00"} {"global_step": 8893, "acc_step": 0, "speed/wps": 12892.234518709138, "speed/FLOPS": 202490258903037.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05898429825901985, "optim/lr": 0.002963085968301498, "optim/total_tokens": 4662493184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 2.8784804344177246, "created_at": "2025-01-16T06:40:38.437280+00:00"} {"global_step": 8894, "acc_step": 0, "speed/wps": 12885.69789483048, "speed/FLOPS": 202387592242760.7, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054203350096940994, "optim/lr": 0.0029630670749198674, "optim/total_tokens": 4663017472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 2.875518560409546, "created_at": "2025-01-16T06:40:48.613507+00:00"} {"global_step": 8895, "acc_step": 0, "speed/wps": 12893.353174883432, "speed/FLOPS": 202507828935449.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06089111790060997, "optim/lr": 0.0029630481767647233, "optim/total_tokens": 4663541760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.90934419631958, "created_at": "2025-01-16T06:40:58.784073+00:00"} {"global_step": 8896, "acc_step": 0, "speed/wps": 12890.683932916718, "speed/FLOPS": 202465904822439.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04773994907736778, "optim/lr": 0.0029630292738361272, "optim/total_tokens": 4664066048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.848149299621582, "created_at": "2025-01-16T06:41:08.958502+00:00"} {"global_step": 8897, "acc_step": 0, "speed/wps": 12887.703320651906, "speed/FLOPS": 202419090211030.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0670221596956253, "optim/lr": 0.0029630103661341415, "optim/total_tokens": 4664590336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.8552145957946777, "created_at": "2025-01-16T06:41:19.135962+00:00"} {"global_step": 8898, "acc_step": 0, "speed/wps": 12887.570208125608, "speed/FLOPS": 202416999495888.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05772201716899872, "optim/lr": 0.0029629914536588265, "optim/total_tokens": 4665114624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421131, "loss/out": 2.984632968902588, "created_at": "2025-01-16T06:41:29.313901+00:00"} {"global_step": 8899, "acc_step": 0, "speed/wps": 12886.88888990003, "speed/FLOPS": 202406298456926.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05914361774921417, "optim/lr": 0.002962972536410245, "optim/total_tokens": 4665638912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 2.8695435523986816, "created_at": "2025-01-16T06:41:39.489410+00:00"} {"global_step": 8900, "acc_step": 0, "speed/wps": 12891.752266072463, "speed/FLOPS": 202482684462693.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04218306392431259, "optim/lr": 0.0029629536143884585, "optim/total_tokens": 4666163200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9349565505981445, "created_at": "2025-01-16T06:41:49.662542+00:00"} {"global_step": 8901, "acc_step": 0, "speed/wps": 12888.544597875643, "speed/FLOPS": 202432303625864.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05369050055742264, "optim/lr": 0.0029629346875935285, "optim/total_tokens": 4666687488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410747, "loss/out": 2.7891640663146973, "created_at": "2025-01-16T06:41:59.835960+00:00"} {"global_step": 8902, "acc_step": 0, "speed/wps": 12885.122765145643, "speed/FLOPS": 202378559040749.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04447892680764198, "optim/lr": 0.002962915756025517, "optim/total_tokens": 4667211776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.979871988296509, "created_at": "2025-01-16T06:42:10.013317+00:00"} {"global_step": 8903, "acc_step": 0, "speed/wps": 12885.380404815774, "speed/FLOPS": 202382605625802.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046494025737047195, "optim/lr": 0.0029628968196844857, "optim/total_tokens": 4667736064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9276843070983887, "created_at": "2025-01-16T06:42:20.189274+00:00"} {"global_step": 8904, "acc_step": 0, "speed/wps": 12887.967893399054, "speed/FLOPS": 202423245689585.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04927438125014305, "optim/lr": 0.002962877878570496, "optim/total_tokens": 4668260352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9013538360595703, "created_at": "2025-01-16T06:42:30.362882+00:00"} {"global_step": 8905, "acc_step": 0, "speed/wps": 12894.2202487604, "speed/FLOPS": 202521447522171.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05459664762020111, "optim/lr": 0.00296285893268361, "optim/total_tokens": 4668784640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.7617127895355225, "created_at": "2025-01-16T06:42:40.531633+00:00"} {"global_step": 8906, "acc_step": 0, "speed/wps": 12896.102401664382, "speed/FLOPS": 202551009319877.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048473749309778214, "optim/lr": 0.0029628399820238903, "optim/total_tokens": 4669308928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 2.873641014099121, "created_at": "2025-01-16T06:42:50.701418+00:00"} {"global_step": 8907, "acc_step": 0, "speed/wps": 12896.825299748842, "speed/FLOPS": 202562363427660.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04823296144604683, "optim/lr": 0.0029628210265913975, "optim/total_tokens": 4669833216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.7927262783050537, "created_at": "2025-01-16T06:43:00.872429+00:00"} {"global_step": 8908, "acc_step": 0, "speed/wps": 12891.048836009575, "speed/FLOPS": 202471636126942.1, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04250919446349144, "optim/lr": 0.0029628020663861943, "optim/total_tokens": 4670357504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 2.8455843925476074, "created_at": "2025-01-16T06:43:11.044901+00:00"} {"global_step": 8909, "acc_step": 0, "speed/wps": 12890.213178629074, "speed/FLOPS": 202458510979467.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05286562815308571, "optim/lr": 0.002962783101408342, "optim/total_tokens": 4670881792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.9171559810638428, "created_at": "2025-01-16T06:43:21.216966+00:00"} {"global_step": 8910, "acc_step": 0, "speed/wps": 12889.84883048162, "speed/FLOPS": 202452788391145.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05449657514691353, "optim/lr": 0.0029627641316579026, "optim/total_tokens": 4671406080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.954894542694092, "created_at": "2025-01-16T06:43:31.390281+00:00"} {"global_step": 8911, "acc_step": 0, "speed/wps": 12894.150263315658, "speed/FLOPS": 202520348304591.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05424945056438446, "optim/lr": 0.002962745157134938, "optim/total_tokens": 4671930368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.958644390106201, "created_at": "2025-01-16T06:43:41.560722+00:00"} {"global_step": 8912, "acc_step": 0, "speed/wps": 12890.402394767949, "speed/FLOPS": 202461482878939.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0516979843378067, "optim/lr": 0.00296272617783951, "optim/total_tokens": 4672454656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 2.9941246509552, "created_at": "2025-01-16T06:43:51.732554+00:00"} {"global_step": 8913, "acc_step": 0, "speed/wps": 12891.006942291006, "speed/FLOPS": 202470978128523.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.050156429409980774, "optim/lr": 0.002962707193771681, "optim/total_tokens": 4672978944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.930562973022461, "created_at": "2025-01-16T06:44:01.903975+00:00"} {"global_step": 8914, "acc_step": 0, "speed/wps": 12886.845995431722, "speed/FLOPS": 202405624740358.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0583893246948719, "optim/lr": 0.002962688204931513, "optim/total_tokens": 4673503232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.945359706878662, "created_at": "2025-01-16T06:44:12.081630+00:00"} {"global_step": 8915, "acc_step": 0, "speed/wps": 12882.517149219735, "speed/FLOPS": 202337634262141.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05566902831196785, "optim/lr": 0.0029626692113190675, "optim/total_tokens": 4674027520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 2.907801866531372, "created_at": "2025-01-16T06:44:22.259897+00:00"} {"global_step": 8916, "acc_step": 0, "speed/wps": 12893.000746346388, "speed/FLOPS": 202502293560987.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059951506555080414, "optim/lr": 0.0029626502129344065, "optim/total_tokens": 4674551808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.95807147026062, "created_at": "2025-01-16T06:44:32.429844+00:00"} {"global_step": 8917, "acc_step": 0, "speed/wps": 12889.436969606179, "speed/FLOPS": 202446319550140.2, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05652555078268051, "optim/lr": 0.0029626312097775914, "optim/total_tokens": 4675076096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.954892873764038, "created_at": "2025-01-16T06:44:42.603224+00:00"} {"global_step": 8918, "acc_step": 0, "speed/wps": 12893.357075193751, "speed/FLOPS": 202507890195182.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06045735999941826, "optim/lr": 0.0029626122018486852, "optim/total_tokens": 4675600384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.8277854919433594, "created_at": "2025-01-16T06:44:52.772829+00:00"} {"global_step": 8919, "acc_step": 0, "speed/wps": 12894.710159876427, "speed/FLOPS": 202529142249457.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04416941851377487, "optim/lr": 0.00296259318914775, "optim/total_tokens": 4676124672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481042, "loss/out": 2.9836654663085938, "created_at": "2025-01-16T06:45:02.970553+00:00"} {"global_step": 8920, "acc_step": 0, "speed/wps": 12897.114518314149, "speed/FLOPS": 202566906002656.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05524752289056778, "optim/lr": 0.002962574171674847, "optim/total_tokens": 4676648960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.8975608348846436, "created_at": "2025-01-16T06:45:13.137683+00:00"} {"global_step": 8921, "acc_step": 0, "speed/wps": 12892.792659153962, "speed/FLOPS": 202499025265688.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06665419787168503, "optim/lr": 0.0029625551494300386, "optim/total_tokens": 4677173248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9202966690063477, "created_at": "2025-01-16T06:45:23.307633+00:00"} {"global_step": 8922, "acc_step": 0, "speed/wps": 12890.161382644677, "speed/FLOPS": 202457697452357.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06090148165822029, "optim/lr": 0.002962536122413387, "optim/total_tokens": 4677697536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.9397006034851074, "created_at": "2025-01-16T06:45:33.481080+00:00"} {"global_step": 8923, "acc_step": 0, "speed/wps": 12890.305586475813, "speed/FLOPS": 202459962371682.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05486943945288658, "optim/lr": 0.0029625170906249546, "optim/total_tokens": 4678221824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.8552186489105225, "created_at": "2025-01-16T06:45:43.653877+00:00"} {"global_step": 8924, "acc_step": 0, "speed/wps": 12891.5537109739, "speed/FLOPS": 202479565882027.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05544622614979744, "optim/lr": 0.002962498054064803, "optim/total_tokens": 4678746112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 2.9066357612609863, "created_at": "2025-01-16T06:45:53.824722+00:00"} {"global_step": 8925, "acc_step": 0, "speed/wps": 12892.644753719662, "speed/FLOPS": 202496702207599.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04956948757171631, "optim/lr": 0.002962479012732994, "optim/total_tokens": 4679270400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459256, "loss/out": 2.879049301147461, "created_at": "2025-01-16T06:46:03.997223+00:00"} {"global_step": 8926, "acc_step": 0, "speed/wps": 12890.895582837398, "speed/FLOPS": 202469229075281.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07148692011833191, "optim/lr": 0.0029624599666295904, "optim/total_tokens": 4679794688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.860804796218872, "created_at": "2025-01-16T06:46:14.170177+00:00"} {"global_step": 8927, "acc_step": 0, "speed/wps": 12891.642329694527, "speed/FLOPS": 202480957760810.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04504155367612839, "optim/lr": 0.0029624409157546537, "optim/total_tokens": 4680318976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497719, "loss/out": 2.9175233840942383, "created_at": "2025-01-16T06:46:24.346200+00:00"} {"global_step": 8928, "acc_step": 0, "speed/wps": 12897.671251668826, "speed/FLOPS": 202575650265023.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05478662624955177, "optim/lr": 0.0029624218601082466, "optim/total_tokens": 4680843264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310208, "loss/out": 2.8388776779174805, "created_at": "2025-01-16T06:46:34.518391+00:00"} {"global_step": 8929, "acc_step": 0, "speed/wps": 12893.727885944447, "speed/FLOPS": 202513714287569.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05848578363656998, "optim/lr": 0.0029624027996904313, "optim/total_tokens": 4681367552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8980703353881836, "created_at": "2025-01-16T06:46:44.687541+00:00"} {"global_step": 8930, "acc_step": 0, "speed/wps": 12891.637921461532, "speed/FLOPS": 202480888523453.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04473773390054703, "optim/lr": 0.0029623837345012696, "optim/total_tokens": 4681891840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.906205892562866, "created_at": "2025-01-16T06:46:54.863067+00:00"} {"global_step": 8931, "acc_step": 0, "speed/wps": 12894.788004551712, "speed/FLOPS": 202530364907052.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05823967978358269, "optim/lr": 0.002962364664540824, "optim/total_tokens": 4682416128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499005, "loss/out": 2.878575325012207, "created_at": "2025-01-16T06:47:05.032278+00:00"} {"global_step": 8932, "acc_step": 0, "speed/wps": 12894.496449536082, "speed/FLOPS": 202525785634893.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048296067863702774, "optim/lr": 0.0029623455898091567, "optim/total_tokens": 4682940416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.916370391845703, "created_at": "2025-01-16T06:47:15.203339+00:00"} {"global_step": 8933, "acc_step": 0, "speed/wps": 12891.398414755888, "speed/FLOPS": 202477126741524.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05834287032485008, "optim/lr": 0.0029623265103063292, "optim/total_tokens": 4683464704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.085569381713867, "created_at": "2025-01-16T06:47:25.377197+00:00"} {"global_step": 8934, "acc_step": 0, "speed/wps": 12894.692149959616, "speed/FLOPS": 202528859378965.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05973435938358307, "optim/lr": 0.002962307426032405, "optim/total_tokens": 4683988992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457645, "loss/out": 2.96224308013916, "created_at": "2025-01-16T06:47:35.546397+00:00"} {"global_step": 8935, "acc_step": 0, "speed/wps": 12896.980067266833, "speed/FLOPS": 202564794264203.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05101090669631958, "optim/lr": 0.002962288336987446, "optim/total_tokens": 4684513280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9183733463287354, "created_at": "2025-01-16T06:47:45.715779+00:00"} {"global_step": 8936, "acc_step": 0, "speed/wps": 12895.341779299122, "speed/FLOPS": 202539062700426.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06539598852396011, "optim/lr": 0.002962269243171514, "optim/total_tokens": 4685037568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.026711940765381, "created_at": "2025-01-16T06:47:55.887490+00:00"} {"global_step": 8937, "acc_step": 0, "speed/wps": 12888.985008551343, "speed/FLOPS": 202439220880714.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058081455528736115, "optim/lr": 0.0029622501445846715, "optim/total_tokens": 4685561856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8856263160705566, "created_at": "2025-01-16T06:48:06.062887+00:00"} {"global_step": 8938, "acc_step": 0, "speed/wps": 12889.974675223499, "speed/FLOPS": 202454764955745.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05581129714846611, "optim/lr": 0.0029622310412269807, "optim/total_tokens": 4686086144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411033, "loss/out": 2.970315456390381, "created_at": "2025-01-16T06:48:16.234938+00:00"} {"global_step": 8939, "acc_step": 0, "speed/wps": 12900.327499487803, "speed/FLOPS": 202617370287087.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09140150994062424, "optim/lr": 0.002962211933098504, "optim/total_tokens": 4686610432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.834824562072754, "created_at": "2025-01-16T06:48:26.399145+00:00"} {"global_step": 8940, "acc_step": 0, "speed/wps": 12893.712788218736, "speed/FLOPS": 202513477157040.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05378328263759613, "optim/lr": 0.002962192820199304, "optim/total_tokens": 4687134720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.934544086456299, "created_at": "2025-01-16T06:48:36.571144+00:00"} {"global_step": 8941, "acc_step": 0, "speed/wps": 12891.83103040031, "speed/FLOPS": 202483921564696.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07054741680622101, "optim/lr": 0.0029621737025294433, "optim/total_tokens": 4687659008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.924499034881592, "created_at": "2025-01-16T06:48:46.741887+00:00"} {"global_step": 8942, "acc_step": 0, "speed/wps": 12892.920011795453, "speed/FLOPS": 202501025513924.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07237383723258972, "optim/lr": 0.002962154580088983, "optim/total_tokens": 4688183296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.9680376052856445, "created_at": "2025-01-16T06:48:56.912613+00:00"} {"global_step": 8943, "acc_step": 0, "speed/wps": 12892.557795946022, "speed/FLOPS": 202495336416272.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057602446526288986, "optim/lr": 0.002962135452877987, "optim/total_tokens": 4688707584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.89013934135437, "created_at": "2025-01-16T06:49:07.082671+00:00"} {"global_step": 8944, "acc_step": 0, "speed/wps": 12891.080781924124, "speed/FLOPS": 202472137881428.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06872685998678207, "optim/lr": 0.0029621163208965167, "optim/total_tokens": 4689231872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 2.9077625274658203, "created_at": "2025-01-16T06:49:17.255671+00:00"} {"global_step": 8945, "acc_step": 0, "speed/wps": 12891.188613986276, "speed/FLOPS": 202473831532141.25, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061096761375665665, "optim/lr": 0.002962097184144635, "optim/total_tokens": 4689756160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.9538414478302, "created_at": "2025-01-16T06:49:27.428002+00:00"} {"global_step": 8946, "acc_step": 0, "speed/wps": 12893.854391178666, "speed/FLOPS": 202515701226109.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05276221036911011, "optim/lr": 0.0029620780426224042, "optim/total_tokens": 4690280448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.9503188133239746, "created_at": "2025-01-16T06:49:37.598157+00:00"} {"global_step": 8947, "acc_step": 0, "speed/wps": 12894.651542767488, "speed/FLOPS": 202528221587212.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054350681602954865, "optim/lr": 0.0029620588963298867, "optim/total_tokens": 4690804736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.962932586669922, "created_at": "2025-01-16T06:49:47.770023+00:00"} {"global_step": 8948, "acc_step": 0, "speed/wps": 12893.680665063323, "speed/FLOPS": 202512972618743.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056583598256111145, "optim/lr": 0.0029620397452671454, "optim/total_tokens": 4691329024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.973078727722168, "created_at": "2025-01-16T06:49:57.939305+00:00"} {"global_step": 8949, "acc_step": 0, "speed/wps": 12893.59321018806, "speed/FLOPS": 202511599019751.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061227843165397644, "optim/lr": 0.002962020589434242, "optim/total_tokens": 4691853312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8791592121124268, "created_at": "2025-01-16T06:50:08.112044+00:00"} {"global_step": 8950, "acc_step": 0, "speed/wps": 12888.011162383878, "speed/FLOPS": 202423925288449.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048868145793676376, "optim/lr": 0.0029620014288312397, "optim/total_tokens": 4692377600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.9828834533691406, "created_at": "2025-01-16T06:50:18.286913+00:00"} {"global_step": 8951, "acc_step": 0, "speed/wps": 12886.605030139937, "speed/FLOPS": 202401840049330.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05534759908914566, "optim/lr": 0.0029619822634582007, "optim/total_tokens": 4692901888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9840774536132812, "created_at": "2025-01-16T06:50:28.463271+00:00"} {"global_step": 8952, "acc_step": 0, "speed/wps": 12890.093854396806, "speed/FLOPS": 202456636828431.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049882110208272934, "optim/lr": 0.002961963093315187, "optim/total_tokens": 4693426176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.78269100189209, "created_at": "2025-01-16T06:50:38.635535+00:00"} {"global_step": 8953, "acc_step": 0, "speed/wps": 12891.919221893017, "speed/FLOPS": 202485306733277.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04364238306879997, "optim/lr": 0.002961943918402262, "optim/total_tokens": 4693950464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404138, "loss/out": 2.9451992511749268, "created_at": "2025-01-16T06:50:48.808165+00:00"} {"global_step": 8954, "acc_step": 0, "speed/wps": 12894.39175423333, "speed/FLOPS": 202524141251297.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04520375654101372, "optim/lr": 0.0029619247387194883, "optim/total_tokens": 4694474752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.898512363433838, "created_at": "2025-01-16T06:50:58.978391+00:00"} {"global_step": 8955, "acc_step": 0, "speed/wps": 12892.704811215035, "speed/FLOPS": 202497645493092.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04808935523033142, "optim/lr": 0.0029619055542669283, "optim/total_tokens": 4694999040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9348182678222656, "created_at": "2025-01-16T06:51:09.151345+00:00"} {"global_step": 8956, "acc_step": 0, "speed/wps": 12889.48104764528, "speed/FLOPS": 202447011856313.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04535705968737602, "optim/lr": 0.0029618863650446438, "optim/total_tokens": 4695523328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.7994179725646973, "created_at": "2025-01-16T06:51:19.324388+00:00"} {"global_step": 8957, "acc_step": 0, "speed/wps": 12893.124161586646, "speed/FLOPS": 202504231966926.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05331028252840042, "optim/lr": 0.0029618671710526985, "optim/total_tokens": 4696047616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.964250326156616, "created_at": "2025-01-16T06:51:29.501106+00:00"} {"global_step": 8958, "acc_step": 0, "speed/wps": 12896.54529903987, "speed/FLOPS": 202557965631764.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07254163175821304, "optim/lr": 0.002961847972291154, "optim/total_tokens": 4696571904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.023242473602295, "created_at": "2025-01-16T06:51:39.670551+00:00"} {"global_step": 8959, "acc_step": 0, "speed/wps": 12888.089930574151, "speed/FLOPS": 202425162451116.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05696956068277359, "optim/lr": 0.002961828768760074, "optim/total_tokens": 4697096192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9068491458892822, "created_at": "2025-01-16T06:51:49.847651+00:00"} {"global_step": 8960, "acc_step": 0, "speed/wps": 12894.070225526866, "speed/FLOPS": 202519091201135.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06502574682235718, "optim/lr": 0.0029618095604595205, "optim/total_tokens": 4697620480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293649, "loss/out": 2.9127914905548096, "created_at": "2025-01-16T06:52:00.016462+00:00"} {"global_step": 8961, "acc_step": 0, "speed/wps": 12892.254287348247, "speed/FLOPS": 202490569396679.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05972755700349808, "optim/lr": 0.0029617903473895565, "optim/total_tokens": 4698144768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.926596164703369, "created_at": "2025-01-16T06:52:10.192156+00:00"} {"global_step": 8962, "acc_step": 0, "speed/wps": 12893.83510195868, "speed/FLOPS": 202515398262403.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06803809106349945, "optim/lr": 0.002961771129550245, "optim/total_tokens": 4698669056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8681929111480713, "created_at": "2025-01-16T06:52:20.363123+00:00"} {"global_step": 8963, "acc_step": 0, "speed/wps": 12890.522052940942, "speed/FLOPS": 202463362274973.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05833030864596367, "optim/lr": 0.002961751906941647, "optim/total_tokens": 4699193344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8898684978485107, "created_at": "2025-01-16T06:52:30.540642+00:00"} {"global_step": 8964, "acc_step": 0, "speed/wps": 12898.802857929486, "speed/FLOPS": 202593423696336.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05971573665738106, "optim/lr": 0.0029617326795638274, "optim/total_tokens": 4699717632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283388, "loss/out": 2.9894251823425293, "created_at": "2025-01-16T06:52:40.705628+00:00"} {"global_step": 8965, "acc_step": 0, "speed/wps": 12892.371093501251, "speed/FLOPS": 202492403997821.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07628698647022247, "optim/lr": 0.0029617134474168474, "optim/total_tokens": 4700241920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.91841197013855, "created_at": "2025-01-16T06:52:50.875909+00:00"} {"global_step": 8966, "acc_step": 0, "speed/wps": 12894.567432311733, "speed/FLOPS": 202526900516925.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05047399923205376, "optim/lr": 0.002961694210500771, "optim/total_tokens": 4700766208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.9447977542877197, "created_at": "2025-01-16T06:53:01.048036+00:00"} {"global_step": 8967, "acc_step": 0, "speed/wps": 12889.460817677857, "speed/FLOPS": 202446694116876.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0723402351140976, "optim/lr": 0.0029616749688156593, "optim/total_tokens": 4701290496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.026813507080078, "created_at": "2025-01-16T06:53:11.220677+00:00"} {"global_step": 8968, "acc_step": 0, "speed/wps": 12892.151444792646, "speed/FLOPS": 202488954113023.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09218185395002365, "optim/lr": 0.002961655722361577, "optim/total_tokens": 4701814784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.9866747856140137, "created_at": "2025-01-16T06:53:21.391104+00:00"} {"global_step": 8969, "acc_step": 0, "speed/wps": 12888.697615741145, "speed/FLOPS": 202434706981711.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06290610134601593, "optim/lr": 0.0029616364711385857, "optim/total_tokens": 4702339072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.972212314605713, "created_at": "2025-01-16T06:53:31.566946+00:00"} {"global_step": 8970, "acc_step": 0, "speed/wps": 12895.39157871659, "speed/FLOPS": 202539844868709.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05784877389669418, "optim/lr": 0.0029616172151467482, "optim/total_tokens": 4702863360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 2.8841018676757812, "created_at": "2025-01-16T06:53:41.735306+00:00"} {"global_step": 8971, "acc_step": 0, "speed/wps": 12894.48485995883, "speed/FLOPS": 202525603604656.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060861144214868546, "optim/lr": 0.002961597954386128, "optim/total_tokens": 4703387648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 2.966423988342285, "created_at": "2025-01-16T06:53:51.907750+00:00"} {"global_step": 8972, "acc_step": 0, "speed/wps": 12892.529769684394, "speed/FLOPS": 202494896225321.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04932277649641037, "optim/lr": 0.002961578688856787, "optim/total_tokens": 4703911936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9170312881469727, "created_at": "2025-01-16T06:54:02.080308+00:00"} {"global_step": 8973, "acc_step": 0, "speed/wps": 12894.585389897604, "speed/FLOPS": 202527182565487.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06028853356838226, "optim/lr": 0.0029615594185587886, "optim/total_tokens": 4704436224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 2.9798197746276855, "created_at": "2025-01-16T06:54:12.251471+00:00"} {"global_step": 8974, "acc_step": 0, "speed/wps": 12889.961542497907, "speed/FLOPS": 202454558688243.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06032593920826912, "optim/lr": 0.0029615401434921963, "optim/total_tokens": 4704960512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8951892852783203, "created_at": "2025-01-16T06:54:22.433362+00:00"} {"global_step": 8975, "acc_step": 0, "speed/wps": 12892.517571861932, "speed/FLOPS": 202494704641757.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05789963901042938, "optim/lr": 0.002961520863657072, "optim/total_tokens": 4705484800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9210386276245117, "created_at": "2025-01-16T06:54:32.605764+00:00"} {"global_step": 8976, "acc_step": 0, "speed/wps": 12891.218808278833, "speed/FLOPS": 202474305774997.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07112113386392593, "optim/lr": 0.002961501579053479, "optim/total_tokens": 4706009088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.815506935119629, "created_at": "2025-01-16T06:54:42.780446+00:00"} {"global_step": 8977, "acc_step": 0, "speed/wps": 12893.72066724026, "speed/FLOPS": 202513600907900.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06279770284891129, "optim/lr": 0.0029614822896814803, "optim/total_tokens": 4706533376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348373, "loss/out": 2.8430099487304688, "created_at": "2025-01-16T06:54:52.949951+00:00"} {"global_step": 8978, "acc_step": 0, "speed/wps": 12890.18988384361, "speed/FLOPS": 202458145102851.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.042644158005714417, "optim/lr": 0.002961462995541138, "optim/total_tokens": 4707057664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.946389675140381, "created_at": "2025-01-16T06:55:03.125069+00:00"} {"global_step": 8979, "acc_step": 0, "speed/wps": 12893.170438186235, "speed/FLOPS": 202504958804513.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0655430406332016, "optim/lr": 0.0029614436966325162, "optim/total_tokens": 4707581952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.976611614227295, "created_at": "2025-01-16T06:55:13.300086+00:00"} {"global_step": 8980, "acc_step": 0, "speed/wps": 12886.601817109997, "speed/FLOPS": 202401789584279.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06542421877384186, "optim/lr": 0.002961424392955678, "optim/total_tokens": 4708106240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348781, "loss/out": 2.8791065216064453, "created_at": "2025-01-16T06:55:23.475978+00:00"} {"global_step": 8981, "acc_step": 0, "speed/wps": 12891.421393901399, "speed/FLOPS": 202477487660581.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05587557703256607, "optim/lr": 0.0029614050845106856, "optim/total_tokens": 4708630528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.009173631668091, "created_at": "2025-01-16T06:55:33.648311+00:00"} {"global_step": 8982, "acc_step": 0, "speed/wps": 12894.901008541088, "speed/FLOPS": 202532139789989.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05774970352649689, "optim/lr": 0.0029613857712976017, "optim/total_tokens": 4709154816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.941188335418701, "created_at": "2025-01-16T06:55:43.821254+00:00"} {"global_step": 8983, "acc_step": 0, "speed/wps": 12889.050634168163, "speed/FLOPS": 202440251621205.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06783056259155273, "optim/lr": 0.0029613664533164902, "optim/total_tokens": 4709679104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8771464824676514, "created_at": "2025-01-16T06:55:53.996961+00:00"} {"global_step": 8984, "acc_step": 0, "speed/wps": 12889.97572947888, "speed/FLOPS": 202454781514274.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0743168443441391, "optim/lr": 0.0029613471305674137, "optim/total_tokens": 4710203392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 3.0956716537475586, "created_at": "2025-01-16T06:56:04.170407+00:00"} {"global_step": 8985, "acc_step": 0, "speed/wps": 12887.62435420203, "speed/FLOPS": 202417849934423.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05496032536029816, "optim/lr": 0.0029613278030504353, "optim/total_tokens": 4710727680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.871636152267456, "created_at": "2025-01-16T06:56:14.346350+00:00"} {"global_step": 8986, "acc_step": 0, "speed/wps": 12889.035725635142, "speed/FLOPS": 202440017462208.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06840778887271881, "optim/lr": 0.002961308470765618, "optim/total_tokens": 4711251968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.975083589553833, "created_at": "2025-01-16T06:56:24.520417+00:00"} {"global_step": 8987, "acc_step": 0, "speed/wps": 12891.835413966093, "speed/FLOPS": 202483990414620.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054504115134477615, "optim/lr": 0.002961289133713025, "optim/total_tokens": 4711776256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 3.056642532348633, "created_at": "2025-01-16T06:56:34.694402+00:00"} {"global_step": 8988, "acc_step": 0, "speed/wps": 12885.859502646625, "speed/FLOPS": 202390130515585.66, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08048352599143982, "optim/lr": 0.0029612697918927192, "optim/total_tokens": 4712300544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 2.9151861667633057, "created_at": "2025-01-16T06:56:44.870076+00:00"} {"global_step": 8989, "acc_step": 0, "speed/wps": 12888.246722738233, "speed/FLOPS": 202427625087508.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041626010090112686, "optim/lr": 0.002961250445304764, "optim/total_tokens": 4712824832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 2.9489905834198, "created_at": "2025-01-16T06:56:55.068855+00:00"} {"global_step": 8990, "acc_step": 0, "speed/wps": 12887.757098453367, "speed/FLOPS": 202419934865300.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.072090744972229, "optim/lr": 0.0029612310939492216, "optim/total_tokens": 4713349120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.0722930431365967, "created_at": "2025-01-16T06:57:05.242760+00:00"} {"global_step": 8991, "acc_step": 0, "speed/wps": 12887.18675773826, "speed/FLOPS": 202410976880639.4, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051307741552591324, "optim/lr": 0.0029612117378261566, "optim/total_tokens": 4713873408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.913008689880371, "created_at": "2025-01-16T06:57:15.417425+00:00"} {"global_step": 8992, "acc_step": 0, "speed/wps": 12889.116724358264, "speed/FLOPS": 202441289658458.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07191355526447296, "optim/lr": 0.0029611923769356312, "optim/total_tokens": 4714397696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295655, "loss/out": 3.018160820007324, "created_at": "2025-01-16T06:57:25.590682+00:00"} {"global_step": 8993, "acc_step": 0, "speed/wps": 12892.361407995912, "speed/FLOPS": 202492251873650.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06144387274980545, "optim/lr": 0.0029611730112777093, "optim/total_tokens": 4714921984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 2.9622650146484375, "created_at": "2025-01-16T06:57:35.764339+00:00"} {"global_step": 8994, "acc_step": 0, "speed/wps": 12893.419917526824, "speed/FLOPS": 202508877220377.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05881331115961075, "optim/lr": 0.0029611536408524534, "optim/total_tokens": 4715446272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 3.0538368225097656, "created_at": "2025-01-16T06:57:45.934154+00:00"} {"global_step": 8995, "acc_step": 0, "speed/wps": 12894.377285811715, "speed/FLOPS": 202523914004854.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07002244889736176, "optim/lr": 0.0029611342656599266, "optim/total_tokens": 4715970560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 2.979689121246338, "created_at": "2025-01-16T06:57:56.102707+00:00"} {"global_step": 8996, "acc_step": 0, "speed/wps": 12891.436911365578, "speed/FLOPS": 202477731383679.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06384816765785217, "optim/lr": 0.0029611148857001924, "optim/total_tokens": 4716494848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.0075278282165527, "created_at": "2025-01-16T06:58:06.274404+00:00"} {"global_step": 8997, "acc_step": 0, "speed/wps": 12888.296793125517, "speed/FLOPS": 202428411511744.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07199454307556152, "optim/lr": 0.002961095500973314, "optim/total_tokens": 4717019136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.0502829551696777, "created_at": "2025-01-16T06:58:16.447747+00:00"} {"global_step": 8998, "acc_step": 0, "speed/wps": 12887.635267026428, "speed/FLOPS": 202418021335326.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05305732414126396, "optim/lr": 0.0029610761114793552, "optim/total_tokens": 4717543424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.820042133331299, "created_at": "2025-01-16T06:58:26.622921+00:00"} {"global_step": 8999, "acc_step": 0, "speed/wps": 12898.601052173355, "speed/FLOPS": 202590254059626.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050499554723501205, "optim/lr": 0.002961056717218378, "optim/total_tokens": 4718067712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.978966474533081, "created_at": "2025-01-16T06:58:36.791672+00:00"} {"global_step": 9000, "acc_step": 0, "speed/wps": 12889.620430387999, "speed/FLOPS": 202449201053820.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05873442813754082, "optim/lr": 0.002961037318190447, "optim/total_tokens": 4718592000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.002434730529785, "created_at": "2025-01-16T06:58:46.964290+00:00"} {"global_step": 9001, "acc_step": 0, "speed/wps": 12583.58231033152, "speed/FLOPS": 197642451837886.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.2444, "optim/grad_norm": 0.06571227312088013, "optim/lr": 0.002961017914395625, "optim/total_tokens": 4719116288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.897346019744873, "created_at": "2025-01-16T06:58:57.386280+00:00"} {"global_step": 9002, "acc_step": 0, "speed/wps": 12911.00474415052, "speed/FLOPS": 202785071086587.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07751605659723282, "optim/lr": 0.002960998505833975, "optim/total_tokens": 4719640576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332595, "loss/out": 2.8677525520324707, "created_at": "2025-01-16T06:59:07.545341+00:00"} {"global_step": 9003, "acc_step": 0, "speed/wps": 12908.528280882718, "speed/FLOPS": 202746174827949.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04613577201962471, "optim/lr": 0.0029609790925055604, "optim/total_tokens": 4720164864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.067570686340332, "created_at": "2025-01-16T06:59:17.710132+00:00"} {"global_step": 9004, "acc_step": 0, "speed/wps": 12905.826795252675, "speed/FLOPS": 202703744283899.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08552984148263931, "optim/lr": 0.0029609596744104446, "optim/total_tokens": 4720689152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.937920570373535, "created_at": "2025-01-16T06:59:27.869814+00:00"} {"global_step": 9005, "acc_step": 0, "speed/wps": 12904.527180599456, "speed/FLOPS": 202683332049912.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053579043596982956, "optim/lr": 0.002960940251548691, "optim/total_tokens": 4721213440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.8131141662597656, "created_at": "2025-01-16T06:59:38.031316+00:00"} {"global_step": 9006, "acc_step": 0, "speed/wps": 12912.489373471357, "speed/FLOPS": 202808389230165.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06382699310779572, "optim/lr": 0.002960920823920364, "optim/total_tokens": 4721737728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.865048885345459, "created_at": "2025-01-16T06:59:48.187919+00:00"} {"global_step": 9007, "acc_step": 0, "speed/wps": 12898.879704590476, "speed/FLOPS": 202594630678746.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05024532228708267, "optim/lr": 0.0029609013915255253, "optim/total_tokens": 4722262016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0238986015319824, "created_at": "2025-01-16T06:59:58.354010+00:00"} {"global_step": 9008, "acc_step": 0, "speed/wps": 12909.916643314687, "speed/FLOPS": 202767980967753.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05182957649230957, "optim/lr": 0.0029608819543642385, "optim/total_tokens": 4722786304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.947249412536621, "created_at": "2025-01-16T07:00:08.511610+00:00"} {"global_step": 9009, "acc_step": 0, "speed/wps": 12903.642632463152, "speed/FLOPS": 202669439005935.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0589931346476078, "optim/lr": 0.0029608625124365674, "optim/total_tokens": 4723310592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 2.9489340782165527, "created_at": "2025-01-16T07:00:18.675525+00:00"} {"global_step": 9010, "acc_step": 0, "speed/wps": 12907.65072507542, "speed/FLOPS": 202732391608107.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07483009994029999, "optim/lr": 0.0029608430657425766, "optim/total_tokens": 4723834880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.015838146209717, "created_at": "2025-01-16T07:00:28.837109+00:00"} {"global_step": 9011, "acc_step": 0, "speed/wps": 12901.54583966978, "speed/FLOPS": 202636505993820.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05345975607633591, "optim/lr": 0.002960823614282328, "optim/total_tokens": 4724359168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 2.948291063308716, "created_at": "2025-01-16T07:00:39.000089+00:00"} {"global_step": 9012, "acc_step": 0, "speed/wps": 12895.81848563897, "speed/FLOPS": 202546550028557.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06674102693796158, "optim/lr": 0.0029608041580558857, "optim/total_tokens": 4724883456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 2.9284050464630127, "created_at": "2025-01-16T07:00:49.167520+00:00"} {"global_step": 9013, "acc_step": 0, "speed/wps": 12897.491363984826, "speed/FLOPS": 202572824881754.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0675903782248497, "optim/lr": 0.002960784697063313, "optim/total_tokens": 4725407744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9542155265808105, "created_at": "2025-01-16T07:00:59.340272+00:00"} {"global_step": 9014, "acc_step": 0, "speed/wps": 12906.977569559795, "speed/FLOPS": 202721818775721.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05451840162277222, "optim/lr": 0.002960765231304673, "optim/total_tokens": 4725932032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.9301512241363525, "created_at": "2025-01-16T07:01:09.498940+00:00"} {"global_step": 9015, "acc_step": 0, "speed/wps": 12905.111401940943, "speed/FLOPS": 202692508048884.8, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04803458973765373, "optim/lr": 0.00296074576078003, "optim/total_tokens": 4726456320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.932523250579834, "created_at": "2025-01-16T07:01:19.662436+00:00"} {"global_step": 9016, "acc_step": 0, "speed/wps": 12906.62835624098, "speed/FLOPS": 202716333900680.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06652702391147614, "optim/lr": 0.002960726285489447, "optim/total_tokens": 4726980608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.882993698120117, "created_at": "2025-01-16T07:01:29.822148+00:00"} {"global_step": 9017, "acc_step": 0, "speed/wps": 12906.52475989272, "speed/FLOPS": 202714706777674.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06009536609053612, "optim/lr": 0.002960706805432988, "optim/total_tokens": 4727504896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.036465883255005, "created_at": "2025-01-16T07:01:39.984780+00:00"} {"global_step": 9018, "acc_step": 0, "speed/wps": 12901.727883047912, "speed/FLOPS": 202639365235227.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07529187202453613, "optim/lr": 0.0029606873206107164, "optim/total_tokens": 4728029184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 2.947230339050293, "created_at": "2025-01-16T07:01:50.149156+00:00"} {"global_step": 9019, "acc_step": 0, "speed/wps": 12903.851131560994, "speed/FLOPS": 202672713770774.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054319705814123154, "optim/lr": 0.0029606678310226953, "optim/total_tokens": 4728553472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 2.913184642791748, "created_at": "2025-01-16T07:02:00.313333+00:00"} {"global_step": 9020, "acc_step": 0, "speed/wps": 12906.636161099212, "speed/FLOPS": 202716456486704.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05872425436973572, "optim/lr": 0.0029606483366689884, "optim/total_tokens": 4729077760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.784627676010132, "created_at": "2025-01-16T07:02:10.472283+00:00"} {"global_step": 9021, "acc_step": 0, "speed/wps": 12903.77740229381, "speed/FLOPS": 202671555751319.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06601019203662872, "optim/lr": 0.00296062883754966, "optim/total_tokens": 4729602048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.0489773750305176, "created_at": "2025-01-16T07:02:20.634196+00:00"} {"global_step": 9022, "acc_step": 0, "speed/wps": 12908.79029581591, "speed/FLOPS": 202750290132521.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052163030952215195, "optim/lr": 0.0029606093336647732, "optim/total_tokens": 4730126336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 2.9868059158325195, "created_at": "2025-01-16T07:02:30.791509+00:00"} {"global_step": 9023, "acc_step": 0, "speed/wps": 12906.648691029362, "speed/FLOPS": 202716653286475.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05457894131541252, "optim/lr": 0.0029605898250143915, "optim/total_tokens": 4730650624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.9718472957611084, "created_at": "2025-01-16T07:02:40.951550+00:00"} {"global_step": 9024, "acc_step": 0, "speed/wps": 12906.549793305216, "speed/FLOPS": 202715099961817.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07216965407133102, "optim/lr": 0.002960570311598579, "optim/total_tokens": 4731174912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 2.9883081912994385, "created_at": "2025-01-16T07:02:51.110504+00:00"} {"global_step": 9025, "acc_step": 0, "speed/wps": 12902.74944961542, "speed/FLOPS": 202655410341949.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1313178837299347, "optim/lr": 0.0029605507934173983, "optim/total_tokens": 4731699200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9354965686798096, "created_at": "2025-01-16T07:03:01.275231+00:00"} {"global_step": 9026, "acc_step": 0, "speed/wps": 12910.666180653574, "speed/FLOPS": 202779753481632.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06336916238069534, "optim/lr": 0.002960531270470915, "optim/total_tokens": 4732223488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379498, "loss/out": 2.977440357208252, "created_at": "2025-01-16T07:03:11.433848+00:00"} {"global_step": 9027, "acc_step": 0, "speed/wps": 12906.772138063432, "speed/FLOPS": 202718592191779.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0643426775932312, "optim/lr": 0.002960511742759191, "optim/total_tokens": 4732747776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8511013984680176, "created_at": "2025-01-16T07:03:21.594409+00:00"} {"global_step": 9028, "acc_step": 0, "speed/wps": 12901.71598833378, "speed/FLOPS": 202639178412397.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04720853641629219, "optim/lr": 0.002960492210282291, "optim/total_tokens": 4733272064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.776249885559082, "created_at": "2025-01-16T07:03:31.757646+00:00"} {"global_step": 9029, "acc_step": 0, "speed/wps": 12903.684072410442, "speed/FLOPS": 202670089877252.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05420723557472229, "optim/lr": 0.0029604726730402785, "optim/total_tokens": 4733796352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 2.949032783508301, "created_at": "2025-01-16T07:03:41.919116+00:00"} {"global_step": 9030, "acc_step": 0, "speed/wps": 12905.41257662771, "speed/FLOPS": 202697238411197.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052830904722213745, "optim/lr": 0.002960453131033217, "optim/total_tokens": 4734320640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.982614278793335, "created_at": "2025-01-16T07:03:52.088219+00:00"} {"global_step": 9031, "acc_step": 0, "speed/wps": 12908.156504711167, "speed/FLOPS": 202740335572300.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05472110956907272, "optim/lr": 0.0029604335842611704, "optim/total_tokens": 4734844928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.0041284561157227, "created_at": "2025-01-16T07:04:02.246830+00:00"} {"global_step": 9032, "acc_step": 0, "speed/wps": 12904.955026318377, "speed/FLOPS": 202690051954849.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055288027971982956, "optim/lr": 0.002960414032724202, "optim/total_tokens": 4735369216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9379372596740723, "created_at": "2025-01-16T07:04:12.407179+00:00"} {"global_step": 9033, "acc_step": 0, "speed/wps": 12909.058575669045, "speed/FLOPS": 202754503836269.75, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6749265789985657, "optim/lr": 0.0029603944764223767, "optim/total_tokens": 4735893504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9466185569763184, "created_at": "2025-01-16T07:04:22.565459+00:00"} {"global_step": 9034, "acc_step": 0, "speed/wps": 12913.24954734759, "speed/FLOPS": 202820328805479.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3767417669296265, "optim/lr": 0.0029603749153557578, "optim/total_tokens": 4736417792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.955718517303467, "created_at": "2025-01-16T07:04:32.720559+00:00"} {"global_step": 9035, "acc_step": 0, "speed/wps": 12905.522353908718, "speed/FLOPS": 202698962614245.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2518032193183899, "optim/lr": 0.002960355349524409, "optim/total_tokens": 4736942080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9301562309265137, "created_at": "2025-01-16T07:04:42.882588+00:00"} {"global_step": 9036, "acc_step": 0, "speed/wps": 12892.31506043835, "speed/FLOPS": 202491523921570.7, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33592331409454346, "optim/lr": 0.002960335778928394, "optim/total_tokens": 4737466368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.9437131881713867, "created_at": "2025-01-16T07:04:53.056674+00:00"} {"global_step": 9037, "acc_step": 0, "speed/wps": 12883.724921118172, "speed/FLOPS": 202356603979458.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13737666606903076, "optim/lr": 0.0029603162035677765, "optim/total_tokens": 4737990656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 2.910313606262207, "created_at": "2025-01-16T07:05:03.234975+00:00"} {"global_step": 9038, "acc_step": 0, "speed/wps": 12886.032175647333, "speed/FLOPS": 202392842582339.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11467073112726212, "optim/lr": 0.002960296623442621, "optim/total_tokens": 4738514944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.8761446475982666, "created_at": "2025-01-16T07:05:13.413403+00:00"} {"global_step": 9039, "acc_step": 0, "speed/wps": 12893.430566127661, "speed/FLOPS": 202509044471286.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10381574183702469, "optim/lr": 0.0029602770385529914, "optim/total_tokens": 4739039232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8761024475097656, "created_at": "2025-01-16T07:05:23.584059+00:00"} {"global_step": 9040, "acc_step": 0, "speed/wps": 12906.4427728917, "speed/FLOPS": 202713419059161.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09789066761732101, "optim/lr": 0.0029602574488989506, "optim/total_tokens": 4739563520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.850965738296509, "created_at": "2025-01-16T07:05:33.744662+00:00"} {"global_step": 9041, "acc_step": 0, "speed/wps": 12904.437595433848, "speed/FLOPS": 202681924991783.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08215391635894775, "optim/lr": 0.0029602378544805637, "optim/total_tokens": 4740087808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9273979663848877, "created_at": "2025-01-16T07:05:43.906155+00:00"} {"global_step": 9042, "acc_step": 0, "speed/wps": 12906.834138260498, "speed/FLOPS": 202719565990071.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1599777787923813, "optim/lr": 0.0029602182552978936, "optim/total_tokens": 4740612096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 3.004697322845459, "created_at": "2025-01-16T07:05:54.067423+00:00"} {"global_step": 9043, "acc_step": 0, "speed/wps": 12900.352083173364, "speed/FLOPS": 202617756407650.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06414467096328735, "optim/lr": 0.0029601986513510055, "optim/total_tokens": 4741136384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.012287139892578, "created_at": "2025-01-16T07:06:04.232104+00:00"} {"global_step": 9044, "acc_step": 0, "speed/wps": 12902.811322451156, "speed/FLOPS": 202656382139857.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.062053993344306946, "optim/lr": 0.0029601790426399617, "optim/total_tokens": 4741660672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.88153076171875, "created_at": "2025-01-16T07:06:14.395164+00:00"} {"global_step": 9045, "acc_step": 0, "speed/wps": 12905.209742739602, "speed/FLOPS": 202694052626262.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056563280522823334, "optim/lr": 0.0029601594291648278, "optim/total_tokens": 4742184960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.000387191772461, "created_at": "2025-01-16T07:06:24.558451+00:00"} {"global_step": 9046, "acc_step": 0, "speed/wps": 12904.81898469206, "speed/FLOPS": 202687915234167.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06748069077730179, "optim/lr": 0.002960139810925667, "optim/total_tokens": 4742709248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.9829862117767334, "created_at": "2025-01-16T07:06:34.726015+00:00"} {"global_step": 9047, "acc_step": 0, "speed/wps": 12902.02803402754, "speed/FLOPS": 202644079518813.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04811030998826027, "optim/lr": 0.002960120187922543, "optim/total_tokens": 4743233536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378926, "loss/out": 3.034785509109497, "created_at": "2025-01-16T07:06:44.897936+00:00"} {"global_step": 9048, "acc_step": 0, "speed/wps": 12902.9873045554, "speed/FLOPS": 202659146180629.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05477633327245712, "optim/lr": 0.0029601005601555203, "optim/total_tokens": 4743757824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.814800500869751, "created_at": "2025-01-16T07:06:55.062654+00:00"} {"global_step": 9049, "acc_step": 0, "speed/wps": 12909.44830216045, "speed/FLOPS": 202760625026357.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0622502900660038, "optim/lr": 0.002960080927624663, "optim/total_tokens": 4744282112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.93841814994812, "created_at": "2025-01-16T07:07:05.221574+00:00"} {"global_step": 9050, "acc_step": 0, "speed/wps": 12909.162643708012, "speed/FLOPS": 202756138367826.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05468803271651268, "optim/lr": 0.002960061290330035, "optim/total_tokens": 4744806400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277427, "loss/out": 2.8853201866149902, "created_at": "2025-01-16T07:07:15.378667+00:00"} {"global_step": 9051, "acc_step": 0, "speed/wps": 12908.80516128138, "speed/FLOPS": 202750523615083.62, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05241664871573448, "optim/lr": 0.0029600416482717003, "optim/total_tokens": 4745330688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.7071292400360107, "created_at": "2025-01-16T07:07:25.539081+00:00"} {"global_step": 9052, "acc_step": 0, "speed/wps": 12909.350750510675, "speed/FLOPS": 202759092843647.88, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05474413186311722, "optim/lr": 0.0029600220014497237, "optim/total_tokens": 4745854976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.9940643310546875, "created_at": "2025-01-16T07:07:35.700393+00:00"} {"global_step": 9053, "acc_step": 0, "speed/wps": 12907.485400030162, "speed/FLOPS": 202729794951091.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05147068202495575, "optim/lr": 0.002960002349864168, "optim/total_tokens": 4746379264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 2.9601082801818848, "created_at": "2025-01-16T07:07:45.864535+00:00"} {"global_step": 9054, "acc_step": 0, "speed/wps": 12902.98519296882, "speed/FLOPS": 202659113015260.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05301103740930557, "optim/lr": 0.0029599826935150977, "optim/total_tokens": 4746903552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.9784352779388428, "created_at": "2025-01-16T07:07:56.026301+00:00"} {"global_step": 9055, "acc_step": 0, "speed/wps": 12908.639054823228, "speed/FLOPS": 202747914684904.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04895555600523949, "optim/lr": 0.002959963032402578, "optim/total_tokens": 4747427840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8404622077941895, "created_at": "2025-01-16T07:08:06.188434+00:00"} {"global_step": 9056, "acc_step": 0, "speed/wps": 12909.936809646279, "speed/FLOPS": 202768297707702.16, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058307945728302, "optim/lr": 0.002959943366526672, "optim/total_tokens": 4747952128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.9669907093048096, "created_at": "2025-01-16T07:08:16.347031+00:00"} {"global_step": 9057, "acc_step": 0, "speed/wps": 12905.577351534042, "speed/FLOPS": 202699826427526.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04674830287694931, "optim/lr": 0.0029599236958874437, "optim/total_tokens": 4748476416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 2.980842113494873, "created_at": "2025-01-16T07:08:26.508184+00:00"} {"global_step": 9058, "acc_step": 0, "speed/wps": 12902.271933837334, "speed/FLOPS": 202647910300479.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0675022080540657, "optim/lr": 0.0029599040204849577, "optim/total_tokens": 4749000704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9326932430267334, "created_at": "2025-01-16T07:08:36.675791+00:00"} {"global_step": 9059, "acc_step": 0, "speed/wps": 12900.323411675772, "speed/FLOPS": 202617306082382.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046227917075157166, "optim/lr": 0.0029598843403192787, "optim/total_tokens": 4749524992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.006915807723999, "created_at": "2025-01-16T07:08:46.843189+00:00"} {"global_step": 9060, "acc_step": 0, "speed/wps": 12901.87680005437, "speed/FLOPS": 202641704181447.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04660459980368614, "optim/lr": 0.00295986465539047, "optim/total_tokens": 4750049280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 2.9274792671203613, "created_at": "2025-01-16T07:08:57.006773+00:00"} {"global_step": 9061, "acc_step": 0, "speed/wps": 12904.760735940037, "speed/FLOPS": 202687000357475.0, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041543666273355484, "optim/lr": 0.002959844965698597, "optim/total_tokens": 4750573568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9233808517456055, "created_at": "2025-01-16T07:09:07.171572+00:00"} {"global_step": 9062, "acc_step": 0, "speed/wps": 12899.562825741628, "speed/FLOPS": 202605360035131.88, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04609156772494316, "optim/lr": 0.002959825271243722, "optim/total_tokens": 4751097856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.912781238555908, "created_at": "2025-01-16T07:09:17.337536+00:00"} {"global_step": 9063, "acc_step": 0, "speed/wps": 12901.469068862143, "speed/FLOPS": 202635300202792.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04393923282623291, "optim/lr": 0.0029598055720259117, "optim/total_tokens": 4751622144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.950336456298828, "created_at": "2025-01-16T07:09:27.502360+00:00"} {"global_step": 9064, "acc_step": 0, "speed/wps": 12903.676512472153, "speed/FLOPS": 202669971138032.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04835599660873413, "optim/lr": 0.0029597858680452285, "optim/total_tokens": 4752146432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9226908683776855, "created_at": "2025-01-16T07:09:37.664105+00:00"} {"global_step": 9065, "acc_step": 0, "speed/wps": 12903.565760983232, "speed/FLOPS": 202668231633710.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0497007854282856, "optim/lr": 0.0029597661593017367, "optim/total_tokens": 4752670720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408061, "loss/out": 3.0045158863067627, "created_at": "2025-01-16T07:09:47.825401+00:00"} {"global_step": 9066, "acc_step": 0, "speed/wps": 12903.339148300644, "speed/FLOPS": 202664672370132.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04727331921458244, "optim/lr": 0.002959746445795502, "optim/total_tokens": 4753195008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.997629165649414, "created_at": "2025-01-16T07:09:57.988866+00:00"} {"global_step": 9067, "acc_step": 0, "speed/wps": 12904.71326855399, "speed/FLOPS": 202686254816950.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06511477380990982, "optim/lr": 0.0029597267275265878, "optim/total_tokens": 4753719296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9095983505249023, "created_at": "2025-01-16T07:10:08.149313+00:00"} {"global_step": 9068, "acc_step": 0, "speed/wps": 12898.294191425002, "speed/FLOPS": 202585434389903.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07059002667665482, "optim/lr": 0.0029597070044950586, "optim/total_tokens": 4754243584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 2.9431028366088867, "created_at": "2025-01-16T07:10:18.315178+00:00"} {"global_step": 9069, "acc_step": 0, "speed/wps": 12906.405225799273, "speed/FLOPS": 202712829330480.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04272100701928139, "optim/lr": 0.002959687276700978, "optim/total_tokens": 4754767872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.884878158569336, "created_at": "2025-01-16T07:10:28.475406+00:00"} {"global_step": 9070, "acc_step": 0, "speed/wps": 12904.16054859101, "speed/FLOPS": 202677573590414.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.062322165817022324, "optim/lr": 0.002959667544144412, "optim/total_tokens": 4755292160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.914088726043701, "created_at": "2025-01-16T07:10:38.639473+00:00"} {"global_step": 9071, "acc_step": 0, "speed/wps": 12904.24974773855, "speed/FLOPS": 202678974585599.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05089838057756424, "optim/lr": 0.0029596478068254235, "optim/total_tokens": 4755816448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.990678548812866, "created_at": "2025-01-16T07:10:48.800695+00:00"} {"global_step": 9072, "acc_step": 0, "speed/wps": 12905.787352265927, "speed/FLOPS": 202703124777591.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062221284955739975, "optim/lr": 0.002959628064744078, "optim/total_tokens": 4756340736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.9339070320129395, "created_at": "2025-01-16T07:10:58.965370+00:00"} {"global_step": 9073, "acc_step": 0, "speed/wps": 12906.433300583458, "speed/FLOPS": 202713270283544.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046307701617479324, "optim/lr": 0.0029596083179004385, "optim/total_tokens": 4756865024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.930819272994995, "created_at": "2025-01-16T07:11:09.125660+00:00"} {"global_step": 9074, "acc_step": 0, "speed/wps": 12909.400291104412, "speed/FLOPS": 202759870946748.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06202074512839317, "optim/lr": 0.0029595885662945707, "optim/total_tokens": 4757389312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.911750555038452, "created_at": "2025-01-16T07:11:19.282432+00:00"} {"global_step": 9075, "acc_step": 0, "speed/wps": 12905.052755313338, "speed/FLOPS": 202691586923009.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04718758165836334, "optim/lr": 0.0029595688099265387, "optim/total_tokens": 4757913600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 3.008390188217163, "created_at": "2025-01-16T07:11:29.443238+00:00"} {"global_step": 9076, "acc_step": 0, "speed/wps": 12908.563016525335, "speed/FLOPS": 202746720398948.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06418247520923615, "optim/lr": 0.0029595490487964067, "optim/total_tokens": 4758437888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.0254733562469482, "created_at": "2025-01-16T07:11:39.602294+00:00"} {"global_step": 9077, "acc_step": 0, "speed/wps": 12903.235236832468, "speed/FLOPS": 202663040297735.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06470798701047897, "optim/lr": 0.0029595292829042393, "optim/total_tokens": 4758962176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.980912685394287, "created_at": "2025-01-16T07:11:49.764406+00:00"} {"global_step": 9078, "acc_step": 0, "speed/wps": 12905.01907161053, "speed/FLOPS": 202691057874170.5, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057867199182510376, "optim/lr": 0.0029595095122501016, "optim/total_tokens": 4759486464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8554275035858154, "created_at": "2025-01-16T07:11:59.924791+00:00"} {"global_step": 9079, "acc_step": 0, "speed/wps": 12904.418375109984, "speed/FLOPS": 202681623110185.97, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05877881869673729, "optim/lr": 0.002959489736834057, "optim/total_tokens": 4760010752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8927364349365234, "created_at": "2025-01-16T07:12:10.089396+00:00"} {"global_step": 9080, "acc_step": 0, "speed/wps": 12904.772509657085, "speed/FLOPS": 202687185279880.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05908314511179924, "optim/lr": 0.0029594699566561707, "optim/total_tokens": 4760535040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.934915781021118, "created_at": "2025-01-16T07:12:20.252016+00:00"} {"global_step": 9081, "acc_step": 0, "speed/wps": 12908.294170037445, "speed/FLOPS": 202742497795422.06, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05674188584089279, "optim/lr": 0.002959450171716507, "optim/total_tokens": 4761059328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493882, "loss/out": 2.960969924926758, "created_at": "2025-01-16T07:12:30.440963+00:00"} {"global_step": 9082, "acc_step": 0, "speed/wps": 12906.555076965542, "speed/FLOPS": 202715182948963.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0731552317738533, "optim/lr": 0.002959430382015131, "optim/total_tokens": 4761583616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9820351600646973, "created_at": "2025-01-16T07:12:40.601376+00:00"} {"global_step": 9083, "acc_step": 0, "speed/wps": 12907.38474460827, "speed/FLOPS": 202728214019375.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04395713657140732, "optim/lr": 0.002959410587552107, "optim/total_tokens": 4762107904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.9253311157226562, "created_at": "2025-01-16T07:12:50.759746+00:00"} {"global_step": 9084, "acc_step": 0, "speed/wps": 12906.763263017141, "speed/FLOPS": 202718452796981.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05572027713060379, "optim/lr": 0.002959390788327499, "optim/total_tokens": 4762632192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 2.95347261428833, "created_at": "2025-01-16T07:13:00.923416+00:00"} {"global_step": 9085, "acc_step": 0, "speed/wps": 12907.22455906643, "speed/FLOPS": 202725698085322.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050766877830028534, "optim/lr": 0.0029593709843413717, "optim/total_tokens": 4763156480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 2.929116725921631, "created_at": "2025-01-16T07:13:11.082207+00:00"} {"global_step": 9086, "acc_step": 0, "speed/wps": 12903.483199932762, "speed/FLOPS": 202666934898962.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04626630246639252, "optim/lr": 0.0029593511755937898, "optim/total_tokens": 4763680768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.8446264266967773, "created_at": "2025-01-16T07:13:21.245082+00:00"} {"global_step": 9087, "acc_step": 0, "speed/wps": 12904.088545573175, "speed/FLOPS": 202676442684077.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05282606557011604, "optim/lr": 0.002959331362084819, "optim/total_tokens": 4764205056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8715176582336426, "created_at": "2025-01-16T07:13:31.406693+00:00"} {"global_step": 9088, "acc_step": 0, "speed/wps": 12908.258904257691, "speed/FLOPS": 202741943897890.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05817164108157158, "optim/lr": 0.002959311543814523, "optim/total_tokens": 4764729344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8997297286987305, "created_at": "2025-01-16T07:13:41.567330+00:00"} {"global_step": 9089, "acc_step": 0, "speed/wps": 12909.802369415189, "speed/FLOPS": 202766186139132.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05219621583819389, "optim/lr": 0.0029592917207829665, "optim/total_tokens": 4765253632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9115991592407227, "created_at": "2025-01-16T07:13:51.725110+00:00"} {"global_step": 9090, "acc_step": 0, "speed/wps": 12907.184954834109, "speed/FLOPS": 202725076046432.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07229728251695633, "optim/lr": 0.0029592718929902138, "optim/total_tokens": 4765777920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376202, "loss/out": 2.9753079414367676, "created_at": "2025-01-16T07:14:01.884045+00:00"} {"global_step": 9091, "acc_step": 0, "speed/wps": 12903.935948826924, "speed/FLOPS": 202674045942487.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.06940092891454697, "optim/lr": 0.0029592520604363303, "optim/total_tokens": 4766302208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9013888835906982, "created_at": "2025-01-16T07:14:12.046686+00:00"} {"global_step": 9092, "acc_step": 0, "speed/wps": 12899.580156417052, "speed/FLOPS": 202605632237204.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05391756072640419, "optim/lr": 0.0029592322231213803, "optim/total_tokens": 4766826496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.9471123218536377, "created_at": "2025-01-16T07:14:22.213757+00:00"} {"global_step": 9093, "acc_step": 0, "speed/wps": 12909.569288763549, "speed/FLOPS": 202762525287212.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05039013549685478, "optim/lr": 0.0029592123810454284, "optim/total_tokens": 4767350784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.88908314704895, "created_at": "2025-01-16T07:14:32.373535+00:00"} {"global_step": 9094, "acc_step": 0, "speed/wps": 12902.71619195456, "speed/FLOPS": 202654887984685.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05216682702302933, "optim/lr": 0.0029591925342085405, "optim/total_tokens": 4767875072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.913046360015869, "created_at": "2025-01-16T07:14:42.539117+00:00"} {"global_step": 9095, "acc_step": 0, "speed/wps": 12900.690303419102, "speed/FLOPS": 202623068621373.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06867683678865433, "optim/lr": 0.00295917268261078, "optim/total_tokens": 4768399360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.9688992500305176, "created_at": "2025-01-16T07:14:52.702717+00:00"} {"global_step": 9096, "acc_step": 0, "speed/wps": 12906.032646394387, "speed/FLOPS": 202706977458948.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06963089853525162, "optim/lr": 0.002959152826252212, "optim/total_tokens": 4768923648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.0426087379455566, "created_at": "2025-01-16T07:15:02.864255+00:00"} {"global_step": 9097, "acc_step": 0, "speed/wps": 12908.556461990693, "speed/FLOPS": 202746617450974.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046486996114254, "optim/lr": 0.0029591329651329013, "optim/total_tokens": 4769447936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.826707363128662, "created_at": "2025-01-16T07:15:13.024015+00:00"} {"global_step": 9098, "acc_step": 0, "speed/wps": 12896.671796930188, "speed/FLOPS": 202559952454958.3, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06351495534181595, "optim/lr": 0.002959113099252913, "optim/total_tokens": 4769972224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.930138111114502, "created_at": "2025-01-16T07:15:23.191997+00:00"} {"global_step": 9099, "acc_step": 0, "speed/wps": 12905.53524380566, "speed/FLOPS": 202699165067789.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08095979690551758, "optim/lr": 0.0029590932286123117, "optim/total_tokens": 4770496512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9414353370666504, "created_at": "2025-01-16T07:15:33.352983+00:00"} {"global_step": 9100, "acc_step": 0, "speed/wps": 12902.570882296548, "speed/FLOPS": 202652605696829.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04929884895682335, "optim/lr": 0.002959073353211162, "optim/total_tokens": 4771020800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 2.989959239959717, "created_at": "2025-01-16T07:15:43.515230+00:00"} {"global_step": 9101, "acc_step": 0, "speed/wps": 12906.272061612617, "speed/FLOPS": 202710737803941.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1018194630742073, "optim/lr": 0.0029590534730495294, "optim/total_tokens": 4771545088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9618165493011475, "created_at": "2025-01-16T07:15:53.676397+00:00"} {"global_step": 9102, "acc_step": 0, "speed/wps": 12907.559420274949, "speed/FLOPS": 202730957540750.1, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06649357080459595, "optim/lr": 0.0029590335881274784, "optim/total_tokens": 4772069376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474921, "loss/out": 2.918555736541748, "created_at": "2025-01-16T07:16:03.837009+00:00"} {"global_step": 9103, "acc_step": 0, "speed/wps": 12905.770657000872, "speed/FLOPS": 202702862555512.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06209414824843407, "optim/lr": 0.002959013698445074, "optim/total_tokens": 4772593664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 3.0439248085021973, "created_at": "2025-01-16T07:16:14.003539+00:00"} {"global_step": 9104, "acc_step": 0, "speed/wps": 12900.546897118029, "speed/FLOPS": 202620816228354.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05138132721185684, "optim/lr": 0.0029589938040023803, "optim/total_tokens": 4773117952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.820936679840088, "created_at": "2025-01-16T07:16:24.169776+00:00"} {"global_step": 9105, "acc_step": 0, "speed/wps": 12899.857146136323, "speed/FLOPS": 202609982741369.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06069347634911537, "optim/lr": 0.002958973904799463, "optim/total_tokens": 4773642240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 2.9404191970825195, "created_at": "2025-01-16T07:16:34.336750+00:00"} {"global_step": 9106, "acc_step": 0, "speed/wps": 12905.877943205356, "speed/FLOPS": 202704547632780.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05013302341103554, "optim/lr": 0.002958954000836387, "optim/total_tokens": 4774166528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 2.915292978286743, "created_at": "2025-01-16T07:16:44.503427+00:00"} {"global_step": 9107, "acc_step": 0, "speed/wps": 12907.603177236204, "speed/FLOPS": 202731644803954.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04411743953824043, "optim/lr": 0.0029589340921132175, "optim/total_tokens": 4774690816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.7963340282440186, "created_at": "2025-01-16T07:16:54.662924+00:00"} {"global_step": 9108, "acc_step": 0, "speed/wps": 12903.01692362251, "speed/FLOPS": 202659611388779.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051491737365722656, "optim/lr": 0.0029589141786300183, "optim/total_tokens": 4775215104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.896183729171753, "created_at": "2025-01-16T07:17:04.825399+00:00"} {"global_step": 9109, "acc_step": 0, "speed/wps": 12899.624889117878, "speed/FLOPS": 202606334825739.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05337122455239296, "optim/lr": 0.002958894260386856, "optim/total_tokens": 4775739392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.973822593688965, "created_at": "2025-01-16T07:17:14.994022+00:00"} {"global_step": 9110, "acc_step": 0, "speed/wps": 12911.885300601092, "speed/FLOPS": 202798901435654.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05582909658551216, "optim/lr": 0.0029588743373837943, "optim/total_tokens": 4776263680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 2.9630730152130127, "created_at": "2025-01-16T07:17:25.150596+00:00"} {"global_step": 9111, "acc_step": 0, "speed/wps": 12905.861341201542, "speed/FLOPS": 202704286875497.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04812522977590561, "optim/lr": 0.002958854409620898, "optim/total_tokens": 4776787968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.072608470916748, "created_at": "2025-01-16T07:17:35.310375+00:00"} {"global_step": 9112, "acc_step": 0, "speed/wps": 12905.080599378185, "speed/FLOPS": 202692024252309.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05590600147843361, "optim/lr": 0.0029588344770982333, "optim/total_tokens": 4777312256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418530, "loss/out": 2.875047206878662, "created_at": "2025-01-16T07:17:45.470527+00:00"} {"global_step": 9113, "acc_step": 0, "speed/wps": 12906.655356072022, "speed/FLOPS": 202716757970128.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.054434049874544144, "optim/lr": 0.002958814539815865, "optim/total_tokens": 4777836544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.924281358718872, "created_at": "2025-01-16T07:17:55.634156+00:00"} {"global_step": 9114, "acc_step": 0, "speed/wps": 12917.235007989662, "speed/FLOPS": 202882925941460.47, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061724405735731125, "optim/lr": 0.002958794597773857, "optim/total_tokens": 4778360832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481042, "loss/out": 2.870723247528076, "created_at": "2025-01-16T07:18:05.787712+00:00"} {"global_step": 9115, "acc_step": 0, "speed/wps": 12905.75504583691, "speed/FLOPS": 202702617360730.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07620049268007278, "optim/lr": 0.002958774650972276, "optim/total_tokens": 4778885120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.931155204772949, "created_at": "2025-01-16T07:18:15.947390+00:00"} {"global_step": 9116, "acc_step": 0, "speed/wps": 12907.5339887418, "speed/FLOPS": 202730558103575.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04571561887860298, "optim/lr": 0.0029587546994111857, "optim/total_tokens": 4779409408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.944760322570801, "created_at": "2025-01-16T07:18:26.108115+00:00"} {"global_step": 9117, "acc_step": 0, "speed/wps": 12906.010672009834, "speed/FLOPS": 202706632321043.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049060650169849396, "optim/lr": 0.0029587347430906526, "optim/total_tokens": 4779933696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336107, "loss/out": 2.887512683868408, "created_at": "2025-01-16T07:18:36.268271+00:00"} {"global_step": 9118, "acc_step": 0, "speed/wps": 12906.371680616538, "speed/FLOPS": 202712302457289.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045580290257930756, "optim/lr": 0.00295871478201074, "optim/total_tokens": 4780457984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9617538452148438, "created_at": "2025-01-16T07:18:46.431392+00:00"} {"global_step": 9119, "acc_step": 0, "speed/wps": 12905.724822465714, "speed/FLOPS": 202702142661154.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04586990922689438, "optim/lr": 0.002958694816171515, "optim/total_tokens": 4780982272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 2.8679068088531494, "created_at": "2025-01-16T07:18:56.592194+00:00"} {"global_step": 9120, "acc_step": 0, "speed/wps": 12910.78766028947, "speed/FLOPS": 202781661486244.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04776575043797493, "optim/lr": 0.0029586748455730405, "optim/total_tokens": 4781506560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.952434539794922, "created_at": "2025-01-16T07:19:06.750879+00:00"} {"global_step": 9121, "acc_step": 0, "speed/wps": 12905.04226974063, "speed/FLOPS": 202691422232681.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05264337360858917, "optim/lr": 0.002958654870215384, "optim/total_tokens": 4782030848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.9045772552490234, "created_at": "2025-01-16T07:19:16.913817+00:00"} {"global_step": 9122, "acc_step": 0, "speed/wps": 12908.319101391835, "speed/FLOPS": 202742889376602.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05388651043176651, "optim/lr": 0.002958634890098609, "optim/total_tokens": 4782555136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.9007184505462646, "created_at": "2025-01-16T07:19:27.071407+00:00"} {"global_step": 9123, "acc_step": 0, "speed/wps": 12903.894603095212, "speed/FLOPS": 202673396550955.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049518145620822906, "optim/lr": 0.0029586149052227818, "optim/total_tokens": 4783079424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.014134645462036, "created_at": "2025-01-16T07:19:37.237424+00:00"} {"global_step": 9124, "acc_step": 0, "speed/wps": 12904.658309058026, "speed/FLOPS": 202685391602543.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05482497811317444, "optim/lr": 0.0029585949155879663, "optim/total_tokens": 4783603712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415331, "loss/out": 2.907109498977661, "created_at": "2025-01-16T07:19:47.398143+00:00"} {"global_step": 9125, "acc_step": 0, "speed/wps": 12905.792451863192, "speed/FLOPS": 202703204873774.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04916377365589142, "optim/lr": 0.002958574921194229, "optim/total_tokens": 4784128000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.035780906677246, "created_at": "2025-01-16T07:19:57.557817+00:00"} {"global_step": 9126, "acc_step": 0, "speed/wps": 12907.276455574127, "speed/FLOPS": 202726513191289.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054978542029857635, "optim/lr": 0.0029585549220416344, "optim/total_tokens": 4784652288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389901, "loss/out": 2.9889168739318848, "created_at": "2025-01-16T07:20:07.717053+00:00"} {"global_step": 9127, "acc_step": 0, "speed/wps": 12907.026511926344, "speed/FLOPS": 202722587482842.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042390767484903336, "optim/lr": 0.0029585349181302477, "optim/total_tokens": 4785176576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.871476888656616, "created_at": "2025-01-16T07:20:17.877511+00:00"} {"global_step": 9128, "acc_step": 0, "speed/wps": 12889.561067427465, "speed/FLOPS": 202448268676951.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07685848325490952, "optim/lr": 0.002958514909460135, "optim/total_tokens": 4785700864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.957907199859619, "created_at": "2025-01-16T07:20:28.052225+00:00"} {"global_step": 9129, "acc_step": 0, "speed/wps": 12902.580095798347, "speed/FLOPS": 202652750407536.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07068763673305511, "optim/lr": 0.002958494896031361, "optim/total_tokens": 4786225152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 3.008960008621216, "created_at": "2025-01-16T07:20:38.220464+00:00"} {"global_step": 9130, "acc_step": 0, "speed/wps": 12901.566405630554, "speed/FLOPS": 202636829010494.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054434336721897125, "optim/lr": 0.0029584748778439906, "optim/total_tokens": 4786749440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.007685422897339, "created_at": "2025-01-16T07:20:48.384372+00:00"} {"global_step": 9131, "acc_step": 0, "speed/wps": 12900.86840360496, "speed/FLOPS": 202625865929527.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058401159942150116, "optim/lr": 0.00295845485489809, "optim/total_tokens": 4787273728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.0299041271209717, "created_at": "2025-01-16T07:20:58.552115+00:00"} {"global_step": 9132, "acc_step": 0, "speed/wps": 12902.09266887373, "speed/FLOPS": 202645094697890.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04746736213564873, "optim/lr": 0.002958434827193724, "optim/total_tokens": 4787798016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 3.0161736011505127, "created_at": "2025-01-16T07:21:08.716259+00:00"} {"global_step": 9133, "acc_step": 0, "speed/wps": 12912.873132094519, "speed/FLOPS": 202814416686681.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05070753023028374, "optim/lr": 0.002958414794730958, "optim/total_tokens": 4788322304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8532590866088867, "created_at": "2025-01-16T07:21:18.870783+00:00"} {"global_step": 9134, "acc_step": 0, "speed/wps": 12899.906409337198, "speed/FLOPS": 202610756487633.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05242058262228966, "optim/lr": 0.002958394757509857, "optim/total_tokens": 4788846592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475206, "loss/out": 2.9274680614471436, "created_at": "2025-01-16T07:21:29.037631+00:00"} {"global_step": 9135, "acc_step": 0, "speed/wps": 12903.069011065689, "speed/FLOPS": 202660429493649.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05371252819895744, "optim/lr": 0.002958374715530487, "optim/total_tokens": 4789370880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.884047031402588, "created_at": "2025-01-16T07:21:39.202534+00:00"} {"global_step": 9136, "acc_step": 0, "speed/wps": 12907.239386597474, "speed/FLOPS": 202725930972072.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047657959163188934, "optim/lr": 0.002958354668792913, "optim/total_tokens": 4789895168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.00506329536438, "created_at": "2025-01-16T07:21:49.361801+00:00"} {"global_step": 9137, "acc_step": 0, "speed/wps": 12905.844345139662, "speed/FLOPS": 202704019928990.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07251395285129547, "optim/lr": 0.0029583346172972007, "optim/total_tokens": 4790419456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405960, "loss/out": 3.0383172035217285, "created_at": "2025-01-16T07:21:59.527905+00:00"} {"global_step": 9138, "acc_step": 0, "speed/wps": 12904.198139417027, "speed/FLOPS": 202678164005992.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0539734922349453, "optim/lr": 0.0029583145610434154, "optim/total_tokens": 4790943744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.018552303314209, "created_at": "2025-01-16T07:22:09.690159+00:00"} {"global_step": 9139, "acc_step": 0, "speed/wps": 12905.563086402288, "speed/FLOPS": 202699602374029.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05721350386738777, "optim/lr": 0.002958294500031622, "optim/total_tokens": 4791468032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.856501579284668, "created_at": "2025-01-16T07:22:19.851219+00:00"} {"global_step": 9140, "acc_step": 0, "speed/wps": 12910.952279929796, "speed/FLOPS": 202784247063907.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056882623583078384, "optim/lr": 0.002958274434261887, "optim/total_tokens": 4791992320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.042485237121582, "created_at": "2025-01-16T07:22:30.007126+00:00"} {"global_step": 9141, "acc_step": 0, "speed/wps": 12905.616874894822, "speed/FLOPS": 202700447196218.12, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052379459142684937, "optim/lr": 0.0029582543637342753, "optim/total_tokens": 4792516608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.861222505569458, "created_at": "2025-01-16T07:22:40.167340+00:00"} {"global_step": 9142, "acc_step": 0, "speed/wps": 12914.192963248659, "speed/FLOPS": 202835146448595.47, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051646746695041656, "optim/lr": 0.0029582342884488517, "optim/total_tokens": 4793040896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.7673323154449463, "created_at": "2025-01-16T07:22:50.321710+00:00"} {"global_step": 9143, "acc_step": 0, "speed/wps": 12903.31732623297, "speed/FLOPS": 202664329624572.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05186544358730316, "optim/lr": 0.002958214208405683, "optim/total_tokens": 4793565184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.745393991470337, "created_at": "2025-01-16T07:23:00.484202+00:00"} {"global_step": 9144, "acc_step": 0, "speed/wps": 12908.381256158336, "speed/FLOPS": 202743865602619.53, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06081920489668846, "optim/lr": 0.002958194123604834, "optim/total_tokens": 4794089472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 2.976986885070801, "created_at": "2025-01-16T07:23:10.645724+00:00"} {"global_step": 9145, "acc_step": 0, "speed/wps": 12907.765143197501, "speed/FLOPS": 202734188701941.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05864950641989708, "optim/lr": 0.0029581740340463712, "optim/total_tokens": 4794613760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.93951153755188, "created_at": "2025-01-16T07:23:20.804799+00:00"} {"global_step": 9146, "acc_step": 0, "speed/wps": 12909.101330404035, "speed/FLOPS": 202755175358133.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0544586107134819, "optim/lr": 0.0029581539397303585, "optim/total_tokens": 4795138048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.883415699005127, "created_at": "2025-01-16T07:23:30.966161+00:00"} {"global_step": 9147, "acc_step": 0, "speed/wps": 12911.7209355398, "speed/FLOPS": 202796319856505.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05528583750128746, "optim/lr": 0.002958133840656862, "optim/total_tokens": 4795662336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9179916381835938, "created_at": "2025-01-16T07:23:41.124261+00:00"} {"global_step": 9148, "acc_step": 0, "speed/wps": 12904.30452509081, "speed/FLOPS": 202679834939188.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.044335514307022095, "optim/lr": 0.002958113736825948, "optim/total_tokens": 4796186624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.876131772994995, "created_at": "2025-01-16T07:23:51.286147+00:00"} {"global_step": 9149, "acc_step": 0, "speed/wps": 12906.159025116267, "speed/FLOPS": 202708962410438.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06027428060770035, "optim/lr": 0.0029580936282376812, "optim/total_tokens": 4796710912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.9299535751342773, "created_at": "2025-01-16T07:24:01.446329+00:00"} {"global_step": 9150, "acc_step": 0, "speed/wps": 12909.32818342026, "speed/FLOPS": 202758738396482.47, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05353975668549538, "optim/lr": 0.002958073514892128, "optim/total_tokens": 4797235200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.97546124458313, "created_at": "2025-01-16T07:24:11.606445+00:00"} {"global_step": 9151, "acc_step": 0, "speed/wps": 12906.669107552776, "speed/FLOPS": 202716973956030.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05209069326519966, "optim/lr": 0.002958053396789353, "optim/total_tokens": 4797759488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 2.9713525772094727, "created_at": "2025-01-16T07:24:21.772011+00:00"} {"global_step": 9152, "acc_step": 0, "speed/wps": 12909.46271679632, "speed/FLOPS": 202760851428021.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04937514662742615, "optim/lr": 0.002958033273929423, "optim/total_tokens": 4798283776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.9281697273254395, "created_at": "2025-01-16T07:24:31.938214+00:00"} {"global_step": 9153, "acc_step": 0, "speed/wps": 12905.622195661583, "speed/FLOPS": 202700530766171.6, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05419205129146576, "optim/lr": 0.0029580131463124034, "optim/total_tokens": 4798808064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377626, "loss/out": 3.1047959327697754, "created_at": "2025-01-16T07:24:42.098084+00:00"} {"global_step": 9154, "acc_step": 0, "speed/wps": 12903.891382620966, "speed/FLOPS": 202673345968982.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05940139666199684, "optim/lr": 0.0029579930139383595, "optim/total_tokens": 4799332352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9707512855529785, "created_at": "2025-01-16T07:24:52.259450+00:00"} {"global_step": 9155, "acc_step": 0, "speed/wps": 12905.541342018721, "speed/FLOPS": 202699260848605.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06092563271522522, "optim/lr": 0.002957972876807357, "optim/total_tokens": 4799856640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.913238525390625, "created_at": "2025-01-16T07:25:02.422147+00:00"} {"global_step": 9156, "acc_step": 0, "speed/wps": 12907.43829850952, "speed/FLOPS": 202729055156985.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04623449221253395, "optim/lr": 0.0029579527349194615, "optim/total_tokens": 4800380928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286322, "loss/out": 2.997978687286377, "created_at": "2025-01-16T07:25:12.580338+00:00"} {"global_step": 9157, "acc_step": 0, "speed/wps": 12903.303009312047, "speed/FLOPS": 202664104757656.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07517296820878983, "optim/lr": 0.002957932588274739, "optim/total_tokens": 4800905216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9095613956451416, "created_at": "2025-01-16T07:25:22.743073+00:00"} {"global_step": 9158, "acc_step": 0, "speed/wps": 12911.26101658826, "speed/FLOPS": 202789096197377.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06481770426034927, "optim/lr": 0.002957912436873255, "optim/total_tokens": 4801429504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 2.917401075363159, "created_at": "2025-01-16T07:25:32.900831+00:00"} {"global_step": 9159, "acc_step": 0, "speed/wps": 12905.539444980352, "speed/FLOPS": 202699231053011.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06070536747574806, "optim/lr": 0.002957892280715076, "optim/total_tokens": 4801953792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.843285322189331, "created_at": "2025-01-16T07:25:43.065077+00:00"} {"global_step": 9160, "acc_step": 0, "speed/wps": 12895.910778895875, "speed/FLOPS": 202547999620981.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06336034089326859, "optim/lr": 0.0029578721198002668, "optim/total_tokens": 4802478080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.821503162384033, "created_at": "2025-01-16T07:25:53.235411+00:00"} {"global_step": 9161, "acc_step": 0, "speed/wps": 12907.732592467155, "speed/FLOPS": 202733677447992.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.062360573559999466, "optim/lr": 0.0029578519541288936, "optim/total_tokens": 4803002368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 2.9259731769561768, "created_at": "2025-01-16T07:26:03.393776+00:00"} {"global_step": 9162, "acc_step": 0, "speed/wps": 12908.311802407612, "speed/FLOPS": 202742774736024.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06136789545416832, "optim/lr": 0.0029578317837010222, "optim/total_tokens": 4803526656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.002635955810547, "created_at": "2025-01-16T07:26:13.557274+00:00"} {"global_step": 9163, "acc_step": 0, "speed/wps": 12910.214251426669, "speed/FLOPS": 202772655312108.16, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048142652958631516, "optim/lr": 0.0029578116085167178, "optim/total_tokens": 4804050944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9511914253234863, "created_at": "2025-01-16T07:26:23.713485+00:00"} {"global_step": 9164, "acc_step": 0, "speed/wps": 12914.376818953811, "speed/FLOPS": 202838034155089.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049280617386102676, "optim/lr": 0.002957791428576047, "optim/total_tokens": 4804575232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9613473415374756, "created_at": "2025-01-16T07:26:33.870133+00:00"} {"global_step": 9165, "acc_step": 0, "speed/wps": 12913.587589911804, "speed/FLOPS": 202825638228470.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04183272644877434, "optim/lr": 0.002957771243879076, "optim/total_tokens": 4805099520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 2.9562599658966064, "created_at": "2025-01-16T07:26:44.024288+00:00"} {"global_step": 9166, "acc_step": 0, "speed/wps": 12904.394412496325, "speed/FLOPS": 202681246744410.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06228748708963394, "optim/lr": 0.0029577510544258694, "optim/total_tokens": 4805623808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8914408683776855, "created_at": "2025-01-16T07:26:54.196002+00:00"} {"global_step": 9167, "acc_step": 0, "speed/wps": 12916.870892358527, "speed/FLOPS": 202877207005126.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04902152717113495, "optim/lr": 0.0029577308602164944, "optim/total_tokens": 4806148096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.869994640350342, "created_at": "2025-01-16T07:27:04.350064+00:00"} {"global_step": 9168, "acc_step": 0, "speed/wps": 12907.2767112638, "speed/FLOPS": 202726517207247.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054954394698143005, "optim/lr": 0.0029577106612510154, "optim/total_tokens": 4806672384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 2.988602638244629, "created_at": "2025-01-16T07:27:14.509681+00:00"} {"global_step": 9169, "acc_step": 0, "speed/wps": 12909.135110018278, "speed/FLOPS": 202755705913393.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06719059497117996, "optim/lr": 0.0029576904575294998, "optim/total_tokens": 4807196672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0276803970336914, "created_at": "2025-01-16T07:27:24.667606+00:00"} {"global_step": 9170, "acc_step": 0, "speed/wps": 12907.946012516999, "speed/FLOPS": 202737029503144.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0006827116012573, "optim/lr": 0.0029576702490520124, "optim/total_tokens": 4807720960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.0008411407470703, "created_at": "2025-01-16T07:27:34.826388+00:00"} {"global_step": 9171, "acc_step": 0, "speed/wps": 12910.120867489002, "speed/FLOPS": 202771188589044.12, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06692631542682648, "optim/lr": 0.0029576500358186194, "optim/total_tokens": 4808245248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.888749837875366, "created_at": "2025-01-16T07:27:44.988361+00:00"} {"global_step": 9172, "acc_step": 0, "speed/wps": 12909.918114131973, "speed/FLOPS": 202768004068959.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06366381049156189, "optim/lr": 0.002957629817829388, "optim/total_tokens": 4808769536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.0357882976531982, "created_at": "2025-01-16T07:27:55.148745+00:00"} {"global_step": 9173, "acc_step": 0, "speed/wps": 12909.358794303766, "speed/FLOPS": 202759219182471.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0503198504447937, "optim/lr": 0.002957609595084382, "optim/total_tokens": 4809293824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9228320121765137, "created_at": "2025-01-16T07:28:05.305851+00:00"} {"global_step": 9174, "acc_step": 0, "speed/wps": 12914.330134030786, "speed/FLOPS": 202837300904221.2, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05078032240271568, "optim/lr": 0.002957589367583669, "optim/total_tokens": 4809818112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.758997917175293, "created_at": "2025-01-16T07:28:15.458862+00:00"} {"global_step": 9175, "acc_step": 0, "speed/wps": 12911.437230445681, "speed/FLOPS": 202791863878151.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05978381261229515, "optim/lr": 0.002957569135327314, "optim/total_tokens": 4810342400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.0058083534240723, "created_at": "2025-01-16T07:28:25.615502+00:00"} {"global_step": 9176, "acc_step": 0, "speed/wps": 12912.460678799505, "speed/FLOPS": 202807938540913.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06332337856292725, "optim/lr": 0.002957548898315384, "optim/total_tokens": 4810866688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.813349723815918, "created_at": "2025-01-16T07:28:35.771897+00:00"} {"global_step": 9177, "acc_step": 0, "speed/wps": 12911.256530966188, "speed/FLOPS": 202789025744518.47, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051518023014068604, "optim/lr": 0.0029575286565479445, "optim/total_tokens": 4811390976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371737, "loss/out": 2.968108654022217, "created_at": "2025-01-16T07:28:45.929651+00:00"} {"global_step": 9178, "acc_step": 0, "speed/wps": 12904.394336770252, "speed/FLOPS": 202681245555028.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04650363698601723, "optim/lr": 0.002957508410025061, "optim/total_tokens": 4811915264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9933624267578125, "created_at": "2025-01-16T07:28:56.090557+00:00"} {"global_step": 9179, "acc_step": 0, "speed/wps": 12912.471971365603, "speed/FLOPS": 202808115906181.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053823988884687424, "optim/lr": 0.0029574881587468003, "optim/total_tokens": 4812439552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 2.8636412620544434, "created_at": "2025-01-16T07:29:06.247438+00:00"} {"global_step": 9180, "acc_step": 0, "speed/wps": 12913.828163561078, "speed/FLOPS": 202829416768214.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061851684004068375, "optim/lr": 0.0029574679027132285, "optim/total_tokens": 4812963840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8528528213500977, "created_at": "2025-01-16T07:29:16.406213+00:00"} {"global_step": 9181, "acc_step": 0, "speed/wps": 12914.720803445547, "speed/FLOPS": 202843436904216.03, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05591820925474167, "optim/lr": 0.002957447641924412, "optim/total_tokens": 4813488128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9325568675994873, "created_at": "2025-01-16T07:29:26.561902+00:00"} {"global_step": 9182, "acc_step": 0, "speed/wps": 12911.307755131575, "speed/FLOPS": 202789830290425.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056224919855594635, "optim/lr": 0.0029574273763804154, "optim/total_tokens": 4814012416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 2.8269927501678467, "created_at": "2025-01-16T07:29:36.717109+00:00"} {"global_step": 9183, "acc_step": 0, "speed/wps": 12910.41303949361, "speed/FLOPS": 202775777551862.88, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05630750581622124, "optim/lr": 0.002957407106081306, "optim/total_tokens": 4814536704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.849538803100586, "created_at": "2025-01-16T07:29:46.876512+00:00"} {"global_step": 9184, "acc_step": 0, "speed/wps": 12908.54218435238, "speed/FLOPS": 202746393201045.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059429045766592026, "optim/lr": 0.00295738683102715, "optim/total_tokens": 4815060992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.9082627296447754, "created_at": "2025-01-16T07:29:57.036854+00:00"} {"global_step": 9185, "acc_step": 0, "speed/wps": 12909.089051472814, "speed/FLOPS": 202754982500646.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04787140339612961, "optim/lr": 0.0029573665512180134, "optim/total_tokens": 4815585280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8255112171173096, "created_at": "2025-01-16T07:30:07.196717+00:00"} {"global_step": 9186, "acc_step": 0, "speed/wps": 12910.114650073174, "speed/FLOPS": 202771090935984.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06202070787549019, "optim/lr": 0.002957346266653962, "optim/total_tokens": 4816109568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 2.9962644577026367, "created_at": "2025-01-16T07:30:17.359955+00:00"} {"global_step": 9187, "acc_step": 0, "speed/wps": 12912.804817130422, "speed/FLOPS": 202813343706295.9, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07563655078411102, "optim/lr": 0.002957325977335062, "optim/total_tokens": 4816633856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 2.8967409133911133, "created_at": "2025-01-16T07:30:27.518038+00:00"} {"global_step": 9188, "acc_step": 0, "speed/wps": 12906.236106214721, "speed/FLOPS": 202710173075010.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05270874872803688, "optim/lr": 0.00295730568326138, "optim/total_tokens": 4817158144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.97613263130188, "created_at": "2025-01-16T07:30:37.677928+00:00"} {"global_step": 9189, "acc_step": 0, "speed/wps": 12903.628854304561, "speed/FLOPS": 202669222601021.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07376785576343536, "optim/lr": 0.0029572853844329824, "optim/total_tokens": 4817682432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.9494245052337646, "created_at": "2025-01-16T07:30:47.841778+00:00"} {"global_step": 9190, "acc_step": 0, "speed/wps": 12909.693162324971, "speed/FLOPS": 202764470891714.53, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05058010667562485, "optim/lr": 0.002957265080849935, "optim/total_tokens": 4818206720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9507734775543213, "created_at": "2025-01-16T07:30:58.000864+00:00"} {"global_step": 9191, "acc_step": 0, "speed/wps": 12909.427735930483, "speed/FLOPS": 202760302005455.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06729846447706223, "optim/lr": 0.002957244772512304, "optim/total_tokens": 4818731008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 2.9468536376953125, "created_at": "2025-01-16T07:31:08.157709+00:00"} {"global_step": 9192, "acc_step": 0, "speed/wps": 12905.950679847463, "speed/FLOPS": 202705690061695.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07095581293106079, "optim/lr": 0.0029572244594201556, "optim/total_tokens": 4819255296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.91965913772583, "created_at": "2025-01-16T07:31:18.317310+00:00"} {"global_step": 9193, "acc_step": 0, "speed/wps": 12909.391555723883, "speed/FLOPS": 202759733745593.53, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046318721026182175, "optim/lr": 0.0029572041415735564, "optim/total_tokens": 4819779584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 2.908407211303711, "created_at": "2025-01-16T07:31:28.484538+00:00"} {"global_step": 9194, "acc_step": 0, "speed/wps": 12907.502871563021, "speed/FLOPS": 202730069365523.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058641575276851654, "optim/lr": 0.0029571838189725724, "optim/total_tokens": 4820303872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.95957612991333, "created_at": "2025-01-16T07:31:38.644021+00:00"} {"global_step": 9195, "acc_step": 0, "speed/wps": 12899.549879197048, "speed/FLOPS": 202605156691858.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058485135436058044, "optim/lr": 0.00295716349161727, "optim/total_tokens": 4820828160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.000736951828003, "created_at": "2025-01-16T07:31:48.813414+00:00"} {"global_step": 9196, "acc_step": 0, "speed/wps": 12917.048108225923, "speed/FLOPS": 202879990423843.75, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05363990366458893, "optim/lr": 0.002957143159507716, "optim/total_tokens": 4821352448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 2.9448189735412598, "created_at": "2025-01-16T07:31:58.968589+00:00"} {"global_step": 9197, "acc_step": 0, "speed/wps": 12907.788735950146, "speed/FLOPS": 202734559258541.8, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04670430347323418, "optim/lr": 0.002957122822643976, "optim/total_tokens": 4821876736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9097366333007812, "created_at": "2025-01-16T07:32:09.126642+00:00"} {"global_step": 9198, "acc_step": 0, "speed/wps": 12906.1213968272, "speed/FLOPS": 202708371406451.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05812491104006767, "optim/lr": 0.002957102481026116, "optim/total_tokens": 4822401024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 2.8639228343963623, "created_at": "2025-01-16T07:32:19.290069+00:00"} {"global_step": 9199, "acc_step": 0, "speed/wps": 12907.647210323366, "speed/FLOPS": 202732336404096.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05051209032535553, "optim/lr": 0.0029570821346542042, "optim/total_tokens": 4822925312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.03364634513855, "created_at": "2025-01-16T07:32:29.449004+00:00"} {"global_step": 9200, "acc_step": 0, "speed/wps": 12908.312397928643, "speed/FLOPS": 202742784089501.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05581264570355415, "optim/lr": 0.002957061783528305, "optim/total_tokens": 4823449600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.9222207069396973, "created_at": "2025-01-16T07:32:39.609990+00:00"} {"global_step": 9201, "acc_step": 0, "speed/wps": 12906.943010532794, "speed/FLOPS": 202721275978713.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05352548882365227, "optim/lr": 0.002957041427648486, "optim/total_tokens": 4823973888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.900064706802368, "created_at": "2025-01-16T07:32:49.769770+00:00"} {"global_step": 9202, "acc_step": 0, "speed/wps": 12907.325555279644, "speed/FLOPS": 202727284369634.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05634453892707825, "optim/lr": 0.0029570210670148136, "optim/total_tokens": 4824498176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9590694904327393, "created_at": "2025-01-16T07:32:59.930992+00:00"} {"global_step": 9203, "acc_step": 0, "speed/wps": 12907.679887292083, "speed/FLOPS": 202732849640793.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05894066393375397, "optim/lr": 0.002957000701627353, "optim/total_tokens": 4825022464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 3.065070629119873, "created_at": "2025-01-16T07:33:10.091507+00:00"} {"global_step": 9204, "acc_step": 0, "speed/wps": 12905.437911046118, "speed/FLOPS": 202697636323052.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04413096606731415, "optim/lr": 0.002956980331486172, "optim/total_tokens": 4825546752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.0138304233551025, "created_at": "2025-01-16T07:33:20.256175+00:00"} {"global_step": 9205, "acc_step": 0, "speed/wps": 12911.152932559922, "speed/FLOPS": 202787398589188.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05851918086409569, "optim/lr": 0.0029569599565913364, "optim/total_tokens": 4826071040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 2.9527761936187744, "created_at": "2025-01-16T07:33:30.415171+00:00"} {"global_step": 9206, "acc_step": 0, "speed/wps": 12911.065394481775, "speed/FLOPS": 202786023683381.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057725127786397934, "optim/lr": 0.0029569395769429127, "optim/total_tokens": 4826595328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.905484676361084, "created_at": "2025-01-16T07:33:40.574984+00:00"} {"global_step": 9207, "acc_step": 0, "speed/wps": 12904.667099564082, "speed/FLOPS": 202685529669520.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052236609160900116, "optim/lr": 0.0029569191925409677, "optim/total_tokens": 4827119616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.995604991912842, "created_at": "2025-01-16T07:33:50.740979+00:00"} {"global_step": 9208, "acc_step": 0, "speed/wps": 12908.083107450979, "speed/FLOPS": 202739182767471.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058281559497117996, "optim/lr": 0.0029568988033855676, "optim/total_tokens": 4827643904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.0486466884613037, "created_at": "2025-01-16T07:34:00.901278+00:00"} {"global_step": 9209, "acc_step": 0, "speed/wps": 12907.564698842938, "speed/FLOPS": 202731040447913.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050240255892276764, "optim/lr": 0.0029568784094767798, "optim/total_tokens": 4828168192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8816752433776855, "created_at": "2025-01-16T07:34:11.062433+00:00"} {"global_step": 9210, "acc_step": 0, "speed/wps": 12908.657435240695, "speed/FLOPS": 202748203374617.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07617590576410294, "optim/lr": 0.002956858010814669, "optim/total_tokens": 4828692480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 2.9490153789520264, "created_at": "2025-01-16T07:34:21.222309+00:00"} {"global_step": 9211, "acc_step": 0, "speed/wps": 12905.682447497376, "speed/FLOPS": 202701477104047.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06284715980291367, "optim/lr": 0.0029568376073993035, "optim/total_tokens": 4829216768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.875143051147461, "created_at": "2025-01-16T07:34:31.385393+00:00"} {"global_step": 9212, "acc_step": 0, "speed/wps": 12906.980640061829, "speed/FLOPS": 202721867002175.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0548209510743618, "optim/lr": 0.002956817199230749, "optim/total_tokens": 4829741056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385710, "loss/out": 2.8841519355773926, "created_at": "2025-01-16T07:34:41.547421+00:00"} {"global_step": 9213, "acc_step": 0, "speed/wps": 12908.402368409199, "speed/FLOPS": 202744197199530.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06399017572402954, "optim/lr": 0.0029567967863090724, "optim/total_tokens": 4830265344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8494036197662354, "created_at": "2025-01-16T07:34:51.707729+00:00"} {"global_step": 9214, "acc_step": 0, "speed/wps": 12907.764475512848, "speed/FLOPS": 202734178215036.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06439205259084702, "optim/lr": 0.00295677636863434, "optim/total_tokens": 4830789632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.7662363052368164, "created_at": "2025-01-16T07:35:01.871642+00:00"} {"global_step": 9215, "acc_step": 0, "speed/wps": 12903.364449735413, "speed/FLOPS": 202665069763933.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.054991479963064194, "optim/lr": 0.002956755946206619, "optim/total_tokens": 4831313920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.938204050064087, "created_at": "2025-01-16T07:35:12.035260+00:00"} {"global_step": 9216, "acc_step": 0, "speed/wps": 12920.391575655094, "speed/FLOPS": 202932504174224.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049822431057691574, "optim/lr": 0.0029567355190259757, "optim/total_tokens": 4831838208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.0321247577667236, "created_at": "2025-01-16T07:35:22.183291+00:00"} {"global_step": 9217, "acc_step": 0, "speed/wps": 12912.615344364558, "speed/FLOPS": 202810367776145.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11281290650367737, "optim/lr": 0.002956715087092476, "optim/total_tokens": 4832362496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423864, "loss/out": 3.0142059326171875, "created_at": "2025-01-16T07:35:32.341549+00:00"} {"global_step": 9218, "acc_step": 0, "speed/wps": 12912.181331501648, "speed/FLOPS": 202803551007736.97, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05559544637799263, "optim/lr": 0.002956694650406188, "optim/total_tokens": 4832886784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.896853446960449, "created_at": "2025-01-16T07:35:42.496836+00:00"} {"global_step": 9219, "acc_step": 0, "speed/wps": 12907.683249348753, "speed/FLOPS": 202732902446513.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07842686027288437, "optim/lr": 0.0029566742089671776, "optim/total_tokens": 4833411072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288730, "loss/out": 2.8747119903564453, "created_at": "2025-01-16T07:35:52.659384+00:00"} {"global_step": 9220, "acc_step": 0, "speed/wps": 12913.665583289, "speed/FLOPS": 202826863221633.0, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05726013705134392, "optim/lr": 0.002956653762775511, "optim/total_tokens": 4833935360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.937500476837158, "created_at": "2025-01-16T07:36:02.813049+00:00"} {"global_step": 9221, "acc_step": 0, "speed/wps": 12908.957860188104, "speed/FLOPS": 202752921961244.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09826728701591492, "optim/lr": 0.0029566333118312554, "optim/total_tokens": 4834459648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8648312091827393, "created_at": "2025-01-16T07:36:12.973937+00:00"} {"global_step": 9222, "acc_step": 0, "speed/wps": 12908.2303503291, "speed/FLOPS": 202741495419205.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08232387155294418, "optim/lr": 0.0029566128561344784, "optim/total_tokens": 4834983936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 2.899933338165283, "created_at": "2025-01-16T07:36:23.137652+00:00"} {"global_step": 9223, "acc_step": 0, "speed/wps": 12909.265365903826, "speed/FLOPS": 202757751761066.56, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053343240171670914, "optim/lr": 0.0029565923956852455, "optim/total_tokens": 4835508224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.819072723388672, "created_at": "2025-01-16T07:36:33.297374+00:00"} {"global_step": 9224, "acc_step": 0, "speed/wps": 12900.89131245692, "speed/FLOPS": 202626225744527.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06759858131408691, "optim/lr": 0.002956571930483624, "optim/total_tokens": 4836032512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435320, "loss/out": 2.8565423488616943, "created_at": "2025-01-16T07:36:43.461385+00:00"} {"global_step": 9225, "acc_step": 0, "speed/wps": 12907.925465279799, "speed/FLOPS": 202736706780549.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04442595690488815, "optim/lr": 0.00295655146052968, "optim/total_tokens": 4836556800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.947711944580078, "created_at": "2025-01-16T07:36:53.624055+00:00"} {"global_step": 9226, "acc_step": 0, "speed/wps": 12909.20598171071, "speed/FLOPS": 202756819050711.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05670233815908432, "optim/lr": 0.002956530985823481, "optim/total_tokens": 4837081088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457106, "loss/out": 2.976315975189209, "created_at": "2025-01-16T07:37:03.787725+00:00"} {"global_step": 9227, "acc_step": 0, "speed/wps": 12912.28031436473, "speed/FLOPS": 202805105669618.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052199531346559525, "optim/lr": 0.002956510506365094, "optim/total_tokens": 4837605376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.9013724327087402, "created_at": "2025-01-16T07:37:13.942315+00:00"} {"global_step": 9228, "acc_step": 0, "speed/wps": 12905.851049590172, "speed/FLOPS": 202704125231598.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07569237053394318, "optim/lr": 0.0029564900221545854, "optim/total_tokens": 4838129664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.9038572311401367, "created_at": "2025-01-16T07:37:24.102975+00:00"} {"global_step": 9229, "acc_step": 0, "speed/wps": 12910.498434545736, "speed/FLOPS": 202777118798501.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05717719346284866, "optim/lr": 0.0029564695331920215, "optim/total_tokens": 4838653952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 3.041081666946411, "created_at": "2025-01-16T07:37:34.263173+00:00"} {"global_step": 9230, "acc_step": 0, "speed/wps": 12911.246828909148, "speed/FLOPS": 202788873360380.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05083364620804787, "optim/lr": 0.0029564490394774707, "optim/total_tokens": 4839178240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.9749040603637695, "created_at": "2025-01-16T07:37:44.421429+00:00"} {"global_step": 9231, "acc_step": 0, "speed/wps": 12909.025356520082, "speed/FLOPS": 202753982083887.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060104336589574814, "optim/lr": 0.0029564285410109982, "optim/total_tokens": 4839702528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436610, "loss/out": 2.924525737762451, "created_at": "2025-01-16T07:37:54.579032+00:00"} {"global_step": 9232, "acc_step": 0, "speed/wps": 12911.285921500952, "speed/FLOPS": 202789487363253.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06015665456652641, "optim/lr": 0.0029564080377926715, "optim/total_tokens": 4840226816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.860591411590576, "created_at": "2025-01-16T07:38:04.743239+00:00"} {"global_step": 9233, "acc_step": 0, "speed/wps": 12911.691252807714, "speed/FLOPS": 202795853648410.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04508076608181, "optim/lr": 0.002956387529822558, "optim/total_tokens": 4840751104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.849519729614258, "created_at": "2025-01-16T07:38:14.901281+00:00"} {"global_step": 9234, "acc_step": 0, "speed/wps": 12911.159495653937, "speed/FLOPS": 202787501671599.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05263156443834305, "optim/lr": 0.002956367017100724, "optim/total_tokens": 4841275392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0154857635498047, "created_at": "2025-01-16T07:38:25.059445+00:00"} {"global_step": 9235, "acc_step": 0, "speed/wps": 12906.29806812208, "speed/FLOPS": 202711146271909.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04596773535013199, "optim/lr": 0.002956346499627237, "optim/total_tokens": 4841799680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.884751796722412, "created_at": "2025-01-16T07:38:35.221569+00:00"} {"global_step": 9236, "acc_step": 0, "speed/wps": 12909.018120676805, "speed/FLOPS": 202753868435025.62, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0482226125895977, "optim/lr": 0.0029563259774021628, "optim/total_tokens": 4842323968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 2.943437099456787, "created_at": "2025-01-16T07:38:45.383321+00:00"} {"global_step": 9237, "acc_step": 0, "speed/wps": 12905.859248793486, "speed/FLOPS": 202704254011353.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05354535952210426, "optim/lr": 0.0029563054504255695, "optim/total_tokens": 4842848256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9567861557006836, "created_at": "2025-01-16T07:38:55.546532+00:00"} {"global_step": 9238, "acc_step": 0, "speed/wps": 12911.693253527124, "speed/FLOPS": 202795885072457.66, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04489963874220848, "optim/lr": 0.0029562849186975234, "optim/total_tokens": 4843372544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374420, "loss/out": 2.844562530517578, "created_at": "2025-01-16T07:39:05.702304+00:00"} {"global_step": 9239, "acc_step": 0, "speed/wps": 12908.166797631959, "speed/FLOPS": 202740497236765.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07930316030979156, "optim/lr": 0.002956264382218092, "optim/total_tokens": 4843896832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.898709297180176, "created_at": "2025-01-16T07:39:15.860413+00:00"} {"global_step": 9240, "acc_step": 0, "speed/wps": 12911.175180303044, "speed/FLOPS": 202787748020566.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07172121107578278, "optim/lr": 0.002956243840987342, "optim/total_tokens": 4844421120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.8037781715393066, "created_at": "2025-01-16T07:39:26.017098+00:00"} {"global_step": 9241, "acc_step": 0, "speed/wps": 12909.381260849766, "speed/FLOPS": 202759572050448.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04718070849776268, "optim/lr": 0.002956223295005341, "optim/total_tokens": 4844945408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8629870414733887, "created_at": "2025-01-16T07:39:36.176181+00:00"} {"global_step": 9242, "acc_step": 0, "speed/wps": 12903.364066433374, "speed/FLOPS": 202665063743647.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07938943058252335, "optim/lr": 0.0029562027442721545, "optim/total_tokens": 4845469696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.893698215484619, "created_at": "2025-01-16T07:39:46.339280+00:00"} {"global_step": 9243, "acc_step": 0, "speed/wps": 12906.98118811227, "speed/FLOPS": 202721875610060.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05890696868300438, "optim/lr": 0.0029561821887878516, "optim/total_tokens": 4845993984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 2.832460641860962, "created_at": "2025-01-16T07:39:56.498720+00:00"} {"global_step": 9244, "acc_step": 0, "speed/wps": 12908.336418901297, "speed/FLOPS": 202743161371884.88, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06447435170412064, "optim/lr": 0.002956161628552498, "optim/total_tokens": 4846518272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.0266642570495605, "created_at": "2025-01-16T07:40:06.662557+00:00"} {"global_step": 9245, "acc_step": 0, "speed/wps": 12911.277893073344, "speed/FLOPS": 202789361265765.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07324075698852539, "optim/lr": 0.002956141063566161, "optim/total_tokens": 4847042560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 2.8145828247070312, "created_at": "2025-01-16T07:40:16.819190+00:00"} {"global_step": 9246, "acc_step": 0, "speed/wps": 12910.822629575336, "speed/FLOPS": 202782210726932.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06658667325973511, "optim/lr": 0.002956120493828908, "optim/total_tokens": 4847566848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 2.9346067905426025, "created_at": "2025-01-16T07:40:26.976841+00:00"} {"global_step": 9247, "acc_step": 0, "speed/wps": 12906.352500690438, "speed/FLOPS": 202712001210194.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11977110058069229, "optim/lr": 0.0029560999193408055, "optim/total_tokens": 4848091136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472483, "loss/out": 2.8708255290985107, "created_at": "2025-01-16T07:40:37.135991+00:00"} {"global_step": 9248, "acc_step": 0, "speed/wps": 12910.484788600175, "speed/FLOPS": 202776904470175.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06620022654533386, "optim/lr": 0.0029560793401019214, "optim/total_tokens": 4848615424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.767716884613037, "created_at": "2025-01-16T07:40:47.297825+00:00"} {"global_step": 9249, "acc_step": 0, "speed/wps": 12913.214126523764, "speed/FLOPS": 202819772472768.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0743255764245987, "optim/lr": 0.0029560587561123226, "optim/total_tokens": 4849139712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.791018009185791, "created_at": "2025-01-16T07:40:57.451591+00:00"} {"global_step": 9250, "acc_step": 0, "speed/wps": 12910.934768262581, "speed/FLOPS": 202783972019110.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06294170022010803, "optim/lr": 0.002956038167372076, "optim/total_tokens": 4849664000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.8891162872314453, "created_at": "2025-01-16T07:41:07.611003+00:00"} {"global_step": 9251, "acc_step": 0, "speed/wps": 12904.96569516463, "speed/FLOPS": 202690219523740.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07109906524419785, "optim/lr": 0.002956017573881249, "optim/total_tokens": 4850188288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.9714553356170654, "created_at": "2025-01-16T07:41:17.775678+00:00"} {"global_step": 9252, "acc_step": 0, "speed/wps": 12906.34474588553, "speed/FLOPS": 202711879410326.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05236582085490227, "optim/lr": 0.0029559969756399085, "optim/total_tokens": 4850712576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 2.881387948989868, "created_at": "2025-01-16T07:41:27.945554+00:00"} {"global_step": 9253, "acc_step": 0, "speed/wps": 12913.278802405006, "speed/FLOPS": 202820788296357.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05416573956608772, "optim/lr": 0.0029559763726481223, "optim/total_tokens": 4851236864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 2.9279167652130127, "created_at": "2025-01-16T07:41:38.102190+00:00"} {"global_step": 9254, "acc_step": 0, "speed/wps": 12904.738995443402, "speed/FLOPS": 202686658893100.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05328677222132683, "optim/lr": 0.002955955764905957, "optim/total_tokens": 4851761152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.942809820175171, "created_at": "2025-01-16T07:41:48.267216+00:00"} {"global_step": 9255, "acc_step": 0, "speed/wps": 12909.98775570088, "speed/FLOPS": 202769097885496.9, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05058743432164192, "optim/lr": 0.0029559351524134805, "optim/total_tokens": 4852285440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 2.834357261657715, "created_at": "2025-01-16T07:41:58.424654+00:00"} {"global_step": 9256, "acc_step": 0, "speed/wps": 12910.472787724313, "speed/FLOPS": 202776715979928.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048366717994213104, "optim/lr": 0.002955914535170759, "optim/total_tokens": 4852809728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 2.8435754776000977, "created_at": "2025-01-16T07:42:08.582713+00:00"} {"global_step": 9257, "acc_step": 0, "speed/wps": 12904.338160101643, "speed/FLOPS": 202680363223252.97, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05689140781760216, "optim/lr": 0.002955893913177861, "optim/total_tokens": 4853334016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376285, "loss/out": 2.880753993988037, "created_at": "2025-01-16T07:42:18.746273+00:00"} {"global_step": 9258, "acc_step": 0, "speed/wps": 12909.93794414493, "speed/FLOPS": 202768315526562.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052183598279953, "optim/lr": 0.0029558732864348523, "optim/total_tokens": 4853858304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9072957038879395, "created_at": "2025-01-16T07:42:28.908611+00:00"} {"global_step": 9259, "acc_step": 0, "speed/wps": 12902.67711930161, "speed/FLOPS": 202654274294979.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053548336029052734, "optim/lr": 0.002955852654941802, "optim/total_tokens": 4854382592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409401, "loss/out": 2.9085044860839844, "created_at": "2025-01-16T07:42:39.073049+00:00"} {"global_step": 9260, "acc_step": 0, "speed/wps": 12909.573156260856, "speed/FLOPS": 202762586031571.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05708232522010803, "optim/lr": 0.0029558320186987763, "optim/total_tokens": 4854906880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.7674922943115234, "created_at": "2025-01-16T07:42:49.232265+00:00"} {"global_step": 9261, "acc_step": 0, "speed/wps": 12907.446432254304, "speed/FLOPS": 202729182908624.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05428396910429001, "optim/lr": 0.0029558113777058428, "optim/total_tokens": 4855431168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.911942720413208, "created_at": "2025-01-16T07:42:59.391015+00:00"} {"global_step": 9262, "acc_step": 0, "speed/wps": 12909.608456458196, "speed/FLOPS": 202763140469678.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06491702795028687, "optim/lr": 0.0029557907319630685, "optim/total_tokens": 4855955456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.9015860557556152, "created_at": "2025-01-16T07:43:09.551637+00:00"} {"global_step": 9263, "acc_step": 0, "speed/wps": 12906.075738513486, "speed/FLOPS": 202707654279893.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05167662352323532, "optim/lr": 0.0029557700814705212, "optim/total_tokens": 4856479744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.9252545833587646, "created_at": "2025-01-16T07:43:19.711526+00:00"} {"global_step": 9264, "acc_step": 0, "speed/wps": 12905.118289049618, "speed/FLOPS": 202692616220390.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07093082368373871, "optim/lr": 0.002955749426228268, "optim/total_tokens": 4857004032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.992370128631592, "created_at": "2025-01-16T07:43:29.874425+00:00"} {"global_step": 9265, "acc_step": 0, "speed/wps": 12909.395220627248, "speed/FLOPS": 202759791307936.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06274693459272385, "optim/lr": 0.0029557287662363763, "optim/total_tokens": 4857528320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.985689640045166, "created_at": "2025-01-16T07:43:40.031185+00:00"} {"global_step": 9266, "acc_step": 0, "speed/wps": 12909.226386195684, "speed/FLOPS": 202757139531186.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05334389954805374, "optim/lr": 0.002955708101494914, "optim/total_tokens": 4858052608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426755, "loss/out": 2.9689221382141113, "created_at": "2025-01-16T07:43:50.191793+00:00"} {"global_step": 9267, "acc_step": 0, "speed/wps": 12911.414085127197, "speed/FLOPS": 202791500349119.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2670132517814636, "optim/lr": 0.0029556874320039476, "optim/total_tokens": 4858576896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.932523250579834, "created_at": "2025-01-16T07:44:00.350251+00:00"} {"global_step": 9268, "acc_step": 0, "speed/wps": 12907.12811110547, "speed/FLOPS": 202724183237562.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05351603031158447, "optim/lr": 0.002955666757763546, "optim/total_tokens": 4859101184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.9317870140075684, "created_at": "2025-01-16T07:44:10.508902+00:00"} {"global_step": 9269, "acc_step": 0, "speed/wps": 12907.449231893763, "speed/FLOPS": 202729226880808.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11219019442796707, "optim/lr": 0.002955646078773775, "optim/total_tokens": 4859625472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294309, "loss/out": 2.958115816116333, "created_at": "2025-01-16T07:44:20.668361+00:00"} {"global_step": 9270, "acc_step": 0, "speed/wps": 12911.059865501675, "speed/FLOPS": 202785936843151.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07583380490541458, "optim/lr": 0.002955625395034703, "optim/total_tokens": 4860149760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8457701206207275, "created_at": "2025-01-16T07:44:30.823686+00:00"} {"global_step": 9271, "acc_step": 0, "speed/wps": 12908.457198781669, "speed/FLOPS": 202745058385875.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07601092010736465, "optim/lr": 0.002955604706546397, "optim/total_tokens": 4860674048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.9684104919433594, "created_at": "2025-01-16T07:44:40.981361+00:00"} {"global_step": 9272, "acc_step": 0, "speed/wps": 12912.32740502848, "speed/FLOPS": 202805845293201.94, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06784332543611526, "optim/lr": 0.002955584013308925, "optim/total_tokens": 4861198336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.0166425704956055, "created_at": "2025-01-16T07:44:51.138599+00:00"} {"global_step": 9273, "acc_step": 0, "speed/wps": 12913.40117839524, "speed/FLOPS": 202822710379446.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3950744867324829, "optim/lr": 0.002955563315322354, "optim/total_tokens": 4861722624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329132, "loss/out": 2.9042105674743652, "created_at": "2025-01-16T07:45:01.298961+00:00"} {"global_step": 9274, "acc_step": 0, "speed/wps": 12909.713833609869, "speed/FLOPS": 202764795562649.16, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.058966126292943954, "optim/lr": 0.002955542612586752, "optim/total_tokens": 4862246912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.0252394676208496, "created_at": "2025-01-16T07:45:11.456434+00:00"} {"global_step": 9275, "acc_step": 0, "speed/wps": 12910.447485644858, "speed/FLOPS": 202776318576002.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055230915546417236, "optim/lr": 0.0029555219051021872, "optim/total_tokens": 4862771200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 2.914132595062256, "created_at": "2025-01-16T07:45:21.613855+00:00"} {"global_step": 9276, "acc_step": 0, "speed/wps": 12908.938879784033, "speed/FLOPS": 202752623847916.25, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06238126754760742, "optim/lr": 0.0029555011928687255, "optim/total_tokens": 4863295488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.994936466217041, "created_at": "2025-01-16T07:45:31.771625+00:00"} {"global_step": 9277, "acc_step": 0, "speed/wps": 12909.53398404376, "speed/FLOPS": 202761970778074.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.046819448471069336, "optim/lr": 0.0029554804758864356, "optim/total_tokens": 4863819776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.935776710510254, "created_at": "2025-01-16T07:45:41.929064+00:00"} {"global_step": 9278, "acc_step": 0, "speed/wps": 12906.506069524989, "speed/FLOPS": 202714413219766.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057030439376831055, "optim/lr": 0.0029554597541553845, "optim/total_tokens": 4864344064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.0134544372558594, "created_at": "2025-01-16T07:45:52.090785+00:00"} {"global_step": 9279, "acc_step": 0, "speed/wps": 12908.802717446233, "speed/FLOPS": 202750485231294.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04706402122974396, "optim/lr": 0.0029554390276756402, "optim/total_tokens": 4864868352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.947347640991211, "created_at": "2025-01-16T07:46:02.249468+00:00"} {"global_step": 9280, "acc_step": 0, "speed/wps": 12909.90339413189, "speed/FLOPS": 202767772871130.97, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.061123400926589966, "optim/lr": 0.0029554182964472704, "optim/total_tokens": 4865392640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.853954792022705, "created_at": "2025-01-16T07:46:12.406835+00:00"} {"global_step": 9281, "acc_step": 0, "speed/wps": 12911.007546518225, "speed/FLOPS": 202785115101622.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06064010411500931, "optim/lr": 0.0029553975604703422, "optim/total_tokens": 4865916928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.103346347808838, "created_at": "2025-01-16T07:46:22.564487+00:00"} {"global_step": 9282, "acc_step": 0, "speed/wps": 12916.90882875671, "speed/FLOPS": 202877802848389.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059578604996204376, "optim/lr": 0.0029553768197449243, "optim/total_tokens": 4866441216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 2.923701763153076, "created_at": "2025-01-16T07:46:32.717386+00:00"} {"global_step": 9283, "acc_step": 0, "speed/wps": 12906.552340445585, "speed/FLOPS": 202715139968157.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057751450687646866, "optim/lr": 0.0029553560742710833, "optim/total_tokens": 4866965504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.911130905151367, "created_at": "2025-01-16T07:46:42.879071+00:00"} {"global_step": 9284, "acc_step": 0, "speed/wps": 12909.577614661575, "speed/FLOPS": 202762656056881.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05574002489447594, "optim/lr": 0.0029553353240488875, "optim/total_tokens": 4867489792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 2.9499258995056152, "created_at": "2025-01-16T07:46:53.037794+00:00"} {"global_step": 9285, "acc_step": 0, "speed/wps": 12913.380159982433, "speed/FLOPS": 202822380256390.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05925900489091873, "optim/lr": 0.0029553145690784043, "optim/total_tokens": 4868014080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455507, "loss/out": 2.9006428718566895, "created_at": "2025-01-16T07:47:03.196128+00:00"} {"global_step": 9286, "acc_step": 0, "speed/wps": 12906.353608520472, "speed/FLOPS": 202712018610187.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051011793315410614, "optim/lr": 0.002955293809359701, "optim/total_tokens": 4868538368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.899416446685791, "created_at": "2025-01-16T07:47:13.356519+00:00"} {"global_step": 9287, "acc_step": 0, "speed/wps": 12905.717281438316, "speed/FLOPS": 202702024218956.72, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0517580471932888, "optim/lr": 0.0029552730448928464, "optim/total_tokens": 4869062656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.8466362953186035, "created_at": "2025-01-16T07:47:23.516321+00:00"} {"global_step": 9288, "acc_step": 0, "speed/wps": 12909.65051971965, "speed/FLOPS": 202763801131000.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056024037301540375, "optim/lr": 0.002955252275677908, "optim/total_tokens": 4869586944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.9244580268859863, "created_at": "2025-01-16T07:47:33.672831+00:00"} {"global_step": 9289, "acc_step": 0, "speed/wps": 12914.545667255506, "speed/FLOPS": 202840686149689.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06329832226037979, "optim/lr": 0.0029552315017149526, "optim/total_tokens": 4870111232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 2.8981990814208984, "created_at": "2025-01-16T07:47:43.826779+00:00"} {"global_step": 9290, "acc_step": 0, "speed/wps": 12908.977105877517, "speed/FLOPS": 202753224241242.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0628642737865448, "optim/lr": 0.0029552107230040488, "optim/total_tokens": 4870635520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8353662490844727, "created_at": "2025-01-16T07:47:53.987147+00:00"} {"global_step": 9291, "acc_step": 0, "speed/wps": 12909.075364695524, "speed/FLOPS": 202754767531001.16, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04532691091299057, "optim/lr": 0.002955189939545264, "optim/total_tokens": 4871159808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 3.06422758102417, "created_at": "2025-01-16T07:48:04.146277+00:00"} {"global_step": 9292, "acc_step": 0, "speed/wps": 12908.727082153191, "speed/FLOPS": 202749297275083.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06024416536092758, "optim/lr": 0.0029551691513386665, "optim/total_tokens": 4871684096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8565728664398193, "created_at": "2025-01-16T07:48:14.308594+00:00"} {"global_step": 9293, "acc_step": 0, "speed/wps": 12909.54321933435, "speed/FLOPS": 202762115831003.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05951392650604248, "optim/lr": 0.002955148358384324, "optim/total_tokens": 4872208384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.996056079864502, "created_at": "2025-01-16T07:48:24.468251+00:00"} {"global_step": 9294, "acc_step": 0, "speed/wps": 12905.924847529051, "speed/FLOPS": 202705284329637.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0657961368560791, "optim/lr": 0.0029551275606823036, "optim/total_tokens": 4872732672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328606, "loss/out": 2.8991165161132812, "created_at": "2025-01-16T07:48:34.629117+00:00"} {"global_step": 9295, "acc_step": 0, "speed/wps": 12907.872551354512, "speed/FLOPS": 202735875694641.8, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05585974082350731, "optim/lr": 0.002955106758232675, "optim/total_tokens": 4873256960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.974602222442627, "created_at": "2025-01-16T07:48:44.793682+00:00"} {"global_step": 9296, "acc_step": 0, "speed/wps": 12912.46413101495, "speed/FLOPS": 202807992762700.66, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06411123275756836, "optim/lr": 0.0029550859510355035, "optim/total_tokens": 4873781248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433975, "loss/out": 2.8262364864349365, "created_at": "2025-01-16T07:48:54.948012+00:00"} {"global_step": 9297, "acc_step": 0, "speed/wps": 12902.80171707, "speed/FLOPS": 202656231274147.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06397078186273575, "optim/lr": 0.002955065139090859, "optim/total_tokens": 4874305536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8996353149414062, "created_at": "2025-01-16T07:49:05.111003+00:00"} {"global_step": 9298, "acc_step": 0, "speed/wps": 12909.224633718815, "speed/FLOPS": 202757112006129.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04711069166660309, "optim/lr": 0.0029550443223988086, "optim/total_tokens": 4874829824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 3.017810344696045, "created_at": "2025-01-16T07:49:15.269193+00:00"} {"global_step": 9299, "acc_step": 0, "speed/wps": 12910.303877733864, "speed/FLOPS": 202774063016422.1, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06735268980264664, "optim/lr": 0.0029550235009594204, "optim/total_tokens": 4875354112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9629955291748047, "created_at": "2025-01-16T07:49:25.427792+00:00"} {"global_step": 9300, "acc_step": 0, "speed/wps": 12906.77006784949, "speed/FLOPS": 202718559676224.34, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04981505870819092, "optim/lr": 0.002955002674772762, "optim/total_tokens": 4875878400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.93326735496521, "created_at": "2025-01-16T07:49:35.587303+00:00"} {"global_step": 9301, "acc_step": 0, "speed/wps": 12909.051102973724, "speed/FLOPS": 202754386467321.2, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04844950512051582, "optim/lr": 0.0029549818438389017, "optim/total_tokens": 4876402688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 2.8349602222442627, "created_at": "2025-01-16T07:49:45.746851+00:00"} {"global_step": 9302, "acc_step": 0, "speed/wps": 12902.569938353981, "speed/FLOPS": 202652590870914.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05818374827504158, "optim/lr": 0.0029549610081579074, "optim/total_tokens": 4876926976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8866381645202637, "created_at": "2025-01-16T07:49:55.908980+00:00"} {"global_step": 9303, "acc_step": 0, "speed/wps": 12907.131120125598, "speed/FLOPS": 202724230498358.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05512344464659691, "optim/lr": 0.0029549401677298475, "optim/total_tokens": 4877451264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.9729270935058594, "created_at": "2025-01-16T07:50:06.067549+00:00"} {"global_step": 9304, "acc_step": 0, "speed/wps": 12905.850377369577, "speed/FLOPS": 202704114673450.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04609034210443497, "optim/lr": 0.0029549193225547887, "optim/total_tokens": 4877975552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8378005027770996, "created_at": "2025-01-16T07:50:16.232719+00:00"} {"global_step": 9305, "acc_step": 0, "speed/wps": 12910.40307460038, "speed/FLOPS": 202775621039521.5, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.044188857078552246, "optim/lr": 0.0029548984726328006, "optim/total_tokens": 4878499840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 3.025473117828369, "created_at": "2025-01-16T07:50:26.390984+00:00"} {"global_step": 9306, "acc_step": 0, "speed/wps": 12910.268000533397, "speed/FLOPS": 202773499515688.3, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0538189522922039, "optim/lr": 0.0029548776179639507, "optim/total_tokens": 4879024128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.021388530731201, "created_at": "2025-01-16T07:50:36.547196+00:00"} {"global_step": 9307, "acc_step": 0, "speed/wps": 12894.198591018134, "speed/FLOPS": 202521107357568.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05527538061141968, "optim/lr": 0.002954856758548306, "optim/total_tokens": 4879548416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.0001349449157715, "created_at": "2025-01-16T07:50:46.716833+00:00"} {"global_step": 9308, "acc_step": 0, "speed/wps": 12895.939109368079, "speed/FLOPS": 202548444589977.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05127173289656639, "optim/lr": 0.0029548358943859354, "optim/total_tokens": 4880072704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.1052358150482178, "created_at": "2025-01-16T07:50:56.887503+00:00"} {"global_step": 9309, "acc_step": 0, "speed/wps": 12904.12571477094, "speed/FLOPS": 202677026477403.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05273881182074547, "optim/lr": 0.002954815025476908, "optim/total_tokens": 4880596992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.907754898071289, "created_at": "2025-01-16T07:51:07.049349+00:00"} {"global_step": 9310, "acc_step": 0, "speed/wps": 12901.34370326487, "speed/FLOPS": 202633331163816.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0578891895711422, "optim/lr": 0.00295479415182129, "optim/total_tokens": 4881121280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 2.9308342933654785, "created_at": "2025-01-16T07:51:17.213413+00:00"} {"global_step": 9311, "acc_step": 0, "speed/wps": 12908.481325666526, "speed/FLOPS": 202745437331755.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05127276852726936, "optim/lr": 0.0029547732734191505, "optim/total_tokens": 4881645568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 2.882449150085449, "created_at": "2025-01-16T07:51:27.373582+00:00"} {"global_step": 9312, "acc_step": 0, "speed/wps": 12906.646172258152, "speed/FLOPS": 202716613725712.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08042466640472412, "optim/lr": 0.0029547523902705574, "optim/total_tokens": 4882169856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.9149534702301025, "created_at": "2025-01-16T07:51:37.534231+00:00"} {"global_step": 9313, "acc_step": 0, "speed/wps": 12906.87564013797, "speed/FLOPS": 202720217834086.47, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050828419625759125, "optim/lr": 0.002954731502375579, "optim/total_tokens": 4882694144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 2.975325584411621, "created_at": "2025-01-16T07:51:47.693721+00:00"} {"global_step": 9314, "acc_step": 0, "speed/wps": 12909.733086290276, "speed/FLOPS": 202765097952450.9, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08371911942958832, "optim/lr": 0.0029547106097342836, "optim/total_tokens": 4883218432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.9984588623046875, "created_at": "2025-01-16T07:51:57.850451+00:00"} {"global_step": 9315, "acc_step": 0, "speed/wps": 12908.114658367858, "speed/FLOPS": 202739678317976.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051829684525728226, "optim/lr": 0.0029546897123467392, "optim/total_tokens": 4883742720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9152355194091797, "created_at": "2025-01-16T07:52:08.010615+00:00"} {"global_step": 9316, "acc_step": 0, "speed/wps": 12902.913345020715, "speed/FLOPS": 202657984544505.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07017520070075989, "optim/lr": 0.0029546688102130136, "optim/total_tokens": 4884267008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 2.952739953994751, "created_at": "2025-01-16T07:52:18.172698+00:00"} {"global_step": 9317, "acc_step": 0, "speed/wps": 12907.501563475213, "speed/FLOPS": 202730048820206.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07454870641231537, "optim/lr": 0.002954647903333175, "optim/total_tokens": 4884791296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.0366063117980957, "created_at": "2025-01-16T07:52:28.337134+00:00"} {"global_step": 9318, "acc_step": 0, "speed/wps": 12904.90206337849, "speed/FLOPS": 202689220099100.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049766235053539276, "optim/lr": 0.002954626991707293, "optim/total_tokens": 4885315584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 3.001622438430786, "created_at": "2025-01-16T07:52:38.499098+00:00"} {"global_step": 9319, "acc_step": 0, "speed/wps": 12910.907773027999, "speed/FLOPS": 202783548021855.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05385275185108185, "optim/lr": 0.002954606075335434, "optim/total_tokens": 4885839872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 2.9729692935943604, "created_at": "2025-01-16T07:52:48.657117+00:00"} {"global_step": 9320, "acc_step": 0, "speed/wps": 12904.877786574985, "speed/FLOPS": 202688838798541.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06683936715126038, "optim/lr": 0.002954585154217667, "optim/total_tokens": 4886364160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444883, "loss/out": 3.0681633949279785, "created_at": "2025-01-16T07:52:58.818081+00:00"} {"global_step": 9321, "acc_step": 0, "speed/wps": 12905.191108248417, "speed/FLOPS": 202693759945971.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059318773448467255, "optim/lr": 0.0029545642283540603, "optim/total_tokens": 4886888448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.787464141845703, "created_at": "2025-01-16T07:53:08.978451+00:00"} {"global_step": 9322, "acc_step": 0, "speed/wps": 12910.118559346694, "speed/FLOPS": 202771152336497.44, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0674109235405922, "optim/lr": 0.002954543297744682, "optim/total_tokens": 4887412736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.858288288116455, "created_at": "2025-01-16T07:53:19.135443+00:00"} {"global_step": 9323, "acc_step": 0, "speed/wps": 12910.859991550022, "speed/FLOPS": 202782797548085.3, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06921114772558212, "optim/lr": 0.0029545223623896003, "optim/total_tokens": 4887937024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 2.855797290802002, "created_at": "2025-01-16T07:53:29.298644+00:00"} {"global_step": 9324, "acc_step": 0, "speed/wps": 12904.58040432791, "speed/FLOPS": 202684168001704.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05843126401305199, "optim/lr": 0.002954501422288885, "optim/total_tokens": 4888461312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 2.9223411083221436, "created_at": "2025-01-16T07:53:39.462887+00:00"} {"global_step": 9325, "acc_step": 0, "speed/wps": 12907.716492370264, "speed/FLOPS": 202733424573846.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07187388092279434, "optim/lr": 0.0029544804774426017, "optim/total_tokens": 4888985600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 2.7887935638427734, "created_at": "2025-01-16T07:53:49.622147+00:00"} {"global_step": 9326, "acc_step": 0, "speed/wps": 12897.952239810984, "speed/FLOPS": 202580063569914.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05360298603773117, "optim/lr": 0.0029544595278508208, "optim/total_tokens": 4889509888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.796621799468994, "created_at": "2025-01-16T07:53:59.798437+00:00"} {"global_step": 9327, "acc_step": 0, "speed/wps": 12908.786505749309, "speed/FLOPS": 202750230604317.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07237774133682251, "optim/lr": 0.0029544385735136096, "optim/total_tokens": 4890034176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9687113761901855, "created_at": "2025-01-16T07:54:09.955884+00:00"} {"global_step": 9328, "acc_step": 0, "speed/wps": 12908.802783751826, "speed/FLOPS": 202750486272714.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07487519085407257, "optim/lr": 0.0029544176144310377, "optim/total_tokens": 4890558464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.986362934112549, "created_at": "2025-01-16T07:54:20.114356+00:00"} {"global_step": 9329, "acc_step": 0, "speed/wps": 12903.3182915793, "speed/FLOPS": 202664344786663.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05107186734676361, "optim/lr": 0.0029543966506031727, "optim/total_tokens": 4891082752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9926395416259766, "created_at": "2025-01-16T07:54:30.280101+00:00"} {"global_step": 9330, "acc_step": 0, "speed/wps": 12897.998209463793, "speed/FLOPS": 202580785586480.4, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06175431236624718, "optim/lr": 0.002954375682030082, "optim/total_tokens": 4891607040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.872422456741333, "created_at": "2025-01-16T07:54:40.447528+00:00"} {"global_step": 9331, "acc_step": 0, "speed/wps": 12898.604919012434, "speed/FLOPS": 202590314793648.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05605950951576233, "optim/lr": 0.0029543547087118355, "optim/total_tokens": 4892131328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 2.9811413288116455, "created_at": "2025-01-16T07:54:50.620625+00:00"} {"global_step": 9332, "acc_step": 0, "speed/wps": 12905.608075994112, "speed/FLOPS": 202700308997391.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05765501409769058, "optim/lr": 0.0029543337306485013, "optim/total_tokens": 4892655616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.997291088104248, "created_at": "2025-01-16T07:55:00.783876+00:00"} {"global_step": 9333, "acc_step": 0, "speed/wps": 12904.615435840798, "speed/FLOPS": 202684718219753.22, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06810490041971207, "optim/lr": 0.0029543127478401476, "optim/total_tokens": 4893179904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419023, "loss/out": 2.9106123447418213, "created_at": "2025-01-16T07:55:10.948238+00:00"} {"global_step": 9334, "acc_step": 0, "speed/wps": 12900.45717209499, "speed/FLOPS": 202619406973572.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05411010980606079, "optim/lr": 0.0029542917602868425, "optim/total_tokens": 4893704192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.0698013305664062, "created_at": "2025-01-16T07:55:21.112746+00:00"} {"global_step": 9335, "acc_step": 0, "speed/wps": 12910.622483459274, "speed/FLOPS": 202779067157149.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08012723922729492, "optim/lr": 0.0029542707679886555, "optim/total_tokens": 4894228480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.0131301879882812, "created_at": "2025-01-16T07:55:31.271884+00:00"} {"global_step": 9336, "acc_step": 0, "speed/wps": 12906.979699025007, "speed/FLOPS": 202721852221899.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05245792493224144, "optim/lr": 0.0029542497709456545, "optim/total_tokens": 4894752768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 2.8751749992370605, "created_at": "2025-01-16T07:55:41.433404+00:00"} {"global_step": 9337, "acc_step": 0, "speed/wps": 12904.263753213538, "speed/FLOPS": 202679194560829.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053664758801460266, "optim/lr": 0.0029542287691579077, "optim/total_tokens": 4895277056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.0768985748291016, "created_at": "2025-01-16T07:55:51.597208+00:00"} {"global_step": 9338, "acc_step": 0, "speed/wps": 12908.49646983784, "speed/FLOPS": 202745675191776.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06258631497621536, "optim/lr": 0.002954207762625484, "optim/total_tokens": 4895801344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 2.8736376762390137, "created_at": "2025-01-16T07:56:01.761056+00:00"} {"global_step": 9339, "acc_step": 0, "speed/wps": 12904.521368506392, "speed/FLOPS": 202683240763004.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0517544224858284, "optim/lr": 0.0029541867513484515, "optim/total_tokens": 4896325632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.805919647216797, "created_at": "2025-01-16T07:56:11.924749+00:00"} {"global_step": 9340, "acc_step": 0, "speed/wps": 12900.448282081325, "speed/FLOPS": 202619267343691.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05046587064862251, "optim/lr": 0.002954165735326879, "optim/total_tokens": 4896849920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.998054265975952, "created_at": "2025-01-16T07:56:22.092510+00:00"} {"global_step": 9341, "acc_step": 0, "speed/wps": 12905.174254771966, "speed/FLOPS": 202693495238964.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05432162061333656, "optim/lr": 0.002954144714560836, "optim/total_tokens": 4897374208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.7547478675842285, "created_at": "2025-01-16T07:56:32.253391+00:00"} {"global_step": 9342, "acc_step": 0, "speed/wps": 12910.246595580466, "speed/FLOPS": 202773163321489.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04707678407430649, "optim/lr": 0.0029541236890503896, "optim/total_tokens": 4897898496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9356136322021484, "created_at": "2025-01-16T07:56:42.414717+00:00"} {"global_step": 9343, "acc_step": 0, "speed/wps": 12906.105000131873, "speed/FLOPS": 202708113873820.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06293658912181854, "optim/lr": 0.0029541026587956098, "optim/total_tokens": 4898422784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.904783248901367, "created_at": "2025-01-16T07:56:52.574611+00:00"} {"global_step": 9344, "acc_step": 0, "speed/wps": 12897.8895518699, "speed/FLOPS": 202579078969657.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0526747927069664, "optim/lr": 0.0029540816237965638, "optim/total_tokens": 4898947072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 2.921748399734497, "created_at": "2025-01-16T07:57:02.743868+00:00"} {"global_step": 9345, "acc_step": 0, "speed/wps": 12908.218152377467, "speed/FLOPS": 202741303833612.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045166727155447006, "optim/lr": 0.0029540605840533214, "optim/total_tokens": 4899471360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.9464218616485596, "created_at": "2025-01-16T07:57:12.903770+00:00"} {"global_step": 9346, "acc_step": 0, "speed/wps": 12905.771913848557, "speed/FLOPS": 202702882296032.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0476306714117527, "optim/lr": 0.00295403953956595, "optim/total_tokens": 4899995648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9271390438079834, "created_at": "2025-01-16T07:57:23.068519+00:00"} {"global_step": 9347, "acc_step": 0, "speed/wps": 12904.481260237344, "speed/FLOPS": 202682610807524.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.040547046810388565, "optim/lr": 0.0029540184903345196, "optim/total_tokens": 4900519936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.836303949356079, "created_at": "2025-01-16T07:57:33.231525+00:00"} {"global_step": 9348, "acc_step": 0, "speed/wps": 12903.801943523622, "speed/FLOPS": 202671941205056.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04288327321410179, "optim/lr": 0.0029539974363590985, "optim/total_tokens": 4901044224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.9995970726013184, "created_at": "2025-01-16T07:57:43.394516+00:00"} {"global_step": 9349, "acc_step": 0, "speed/wps": 12903.813067145105, "speed/FLOPS": 202672115916817.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05316096916794777, "optim/lr": 0.002953976377639755, "optim/total_tokens": 4901568512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 2.806514263153076, "created_at": "2025-01-16T07:57:53.557329+00:00"} {"global_step": 9350, "acc_step": 0, "speed/wps": 12905.462651489237, "speed/FLOPS": 202698024905708.25, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05263824015855789, "optim/lr": 0.0029539553141765583, "optim/total_tokens": 4902092800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.938821792602539, "created_at": "2025-01-16T07:58:03.719764+00:00"} {"global_step": 9351, "acc_step": 0, "speed/wps": 12906.240496031523, "speed/FLOPS": 202710242023115.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05160447210073471, "optim/lr": 0.0029539342459695764, "optim/total_tokens": 4902617088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 3.029776096343994, "created_at": "2025-01-16T07:58:13.880799+00:00"} {"global_step": 9352, "acc_step": 0, "speed/wps": 12913.585141895737, "speed/FLOPS": 202825599779013.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048790354281663895, "optim/lr": 0.002953913173018878, "optim/total_tokens": 4903141376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.980612277984619, "created_at": "2025-01-16T07:58:24.034508+00:00"} {"global_step": 9353, "acc_step": 0, "speed/wps": 12901.51726694254, "speed/FLOPS": 202636057219875.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050283659249544144, "optim/lr": 0.002953892095324533, "optim/total_tokens": 4903665664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 2.912191867828369, "created_at": "2025-01-16T07:58:34.197693+00:00"} {"global_step": 9354, "acc_step": 0, "speed/wps": 12904.690104738374, "speed/FLOPS": 202685890997395.47, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045903805643320084, "optim/lr": 0.002953871012886609, "optim/total_tokens": 4904189952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.903923988342285, "created_at": "2025-01-16T07:58:44.358231+00:00"} {"global_step": 9355, "acc_step": 0, "speed/wps": 12902.810854011053, "speed/FLOPS": 202656374782361.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046475980430841446, "optim/lr": 0.002953849925705176, "optim/total_tokens": 4904714240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.9352645874023438, "created_at": "2025-01-16T07:58:54.524067+00:00"} {"global_step": 9356, "acc_step": 0, "speed/wps": 12908.347783579131, "speed/FLOPS": 202743339869766.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058896422386169434, "optim/lr": 0.0029538288337803017, "optim/total_tokens": 4905238528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 2.936328887939453, "created_at": "2025-01-16T07:59:04.682923+00:00"} {"global_step": 9357, "acc_step": 0, "speed/wps": 12906.548956489338, "speed/FLOPS": 202715086818474.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05909417197108269, "optim/lr": 0.0029538077371120547, "optim/total_tokens": 4905762816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8764538764953613, "created_at": "2025-01-16T07:59:14.845344+00:00"} {"global_step": 9358, "acc_step": 0, "speed/wps": 12894.508102930176, "speed/FLOPS": 202525968667461.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06803280115127563, "optim/lr": 0.0029537866357005047, "optim/total_tokens": 4906287104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 2.961838722229004, "created_at": "2025-01-16T07:59:25.016250+00:00"} {"global_step": 9359, "acc_step": 0, "speed/wps": 12911.867749680483, "speed/FLOPS": 202798625774329.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06917653232812881, "optim/lr": 0.0029537655295457203, "optim/total_tokens": 4906811392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8098220825195312, "created_at": "2025-01-16T07:59:35.171554+00:00"} {"global_step": 9360, "acc_step": 0, "speed/wps": 12898.788826475658, "speed/FLOPS": 202593203313073.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06195986643433571, "optim/lr": 0.0029537444186477713, "optim/total_tokens": 4907335680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.965726375579834, "created_at": "2025-01-16T07:59:45.338660+00:00"} {"global_step": 9361, "acc_step": 0, "speed/wps": 12902.363375577877, "speed/FLOPS": 202649346518668.94, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0627175122499466, "optim/lr": 0.002953723303006725, "optim/total_tokens": 4907859968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 3.1051011085510254, "created_at": "2025-01-16T07:59:55.506573+00:00"} {"global_step": 9362, "acc_step": 0, "speed/wps": 12905.690318668714, "speed/FLOPS": 202701600731609.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04900972172617912, "optim/lr": 0.00295370218262265, "optim/total_tokens": 4908384256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.93403959274292, "created_at": "2025-01-16T08:00:05.668109+00:00"} {"global_step": 9363, "acc_step": 0, "speed/wps": 12900.909340877133, "speed/FLOPS": 202626508905640.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05043088272213936, "optim/lr": 0.0029536810574956173, "optim/total_tokens": 4908908544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.901026964187622, "created_at": "2025-01-16T08:00:15.834921+00:00"} {"global_step": 9364, "acc_step": 0, "speed/wps": 12909.242633336653, "speed/FLOPS": 202757394714861.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04561663791537285, "optim/lr": 0.002953659927625694, "optim/total_tokens": 4909432832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 3.028141975402832, "created_at": "2025-01-16T08:00:25.992334+00:00"} {"global_step": 9365, "acc_step": 0, "speed/wps": 12906.230119782924, "speed/FLOPS": 202710079049872.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05361402779817581, "optim/lr": 0.0029536387930129498, "optim/total_tokens": 4909957120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 2.8794615268707275, "created_at": "2025-01-16T08:00:36.157172+00:00"} {"global_step": 9366, "acc_step": 0, "speed/wps": 12904.011608439627, "speed/FLOPS": 202675234280672.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05026007071137428, "optim/lr": 0.002953617653657453, "optim/total_tokens": 4910481408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292430, "loss/out": 2.8644747734069824, "created_at": "2025-01-16T08:00:46.321364+00:00"} {"global_step": 9367, "acc_step": 0, "speed/wps": 12905.29978949639, "speed/FLOPS": 202695466934314.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049878183752298355, "optim/lr": 0.002953596509559274, "optim/total_tokens": 4911005696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.898818016052246, "created_at": "2025-01-16T08:00:56.484682+00:00"} {"global_step": 9368, "acc_step": 0, "speed/wps": 12904.648325831253, "speed/FLOPS": 202685234802248.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049177683889865875, "optim/lr": 0.0029535753607184806, "optim/total_tokens": 4911529984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.9795448780059814, "created_at": "2025-01-16T08:01:06.647341+00:00"} {"global_step": 9369, "acc_step": 0, "speed/wps": 12902.705016993144, "speed/FLOPS": 202654712466560.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05188143253326416, "optim/lr": 0.002953554207135142, "optim/total_tokens": 4912054272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 2.789137125015259, "created_at": "2025-01-16T08:01:16.812621+00:00"} {"global_step": 9370, "acc_step": 0, "speed/wps": 12904.292380702504, "speed/FLOPS": 202679644194882.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06611161679029465, "optim/lr": 0.002953533048809327, "optim/total_tokens": 4912578560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8858397006988525, "created_at": "2025-01-16T08:01:26.973764+00:00"} {"global_step": 9371, "acc_step": 0, "speed/wps": 12912.100591116789, "speed/FLOPS": 202802282869043.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05395220220088959, "optim/lr": 0.0029535118857411057, "optim/total_tokens": 4913102848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.9898159503936768, "created_at": "2025-01-16T08:01:37.129583+00:00"} {"global_step": 9372, "acc_step": 0, "speed/wps": 12909.82897397276, "speed/FLOPS": 202766604000267.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06197471544146538, "optim/lr": 0.002953490717930546, "optim/total_tokens": 4913627136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.930629014968872, "created_at": "2025-01-16T08:01:47.287793+00:00"} {"global_step": 9373, "acc_step": 0, "speed/wps": 12908.265289188188, "speed/FLOPS": 202742044181998.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05666809529066086, "optim/lr": 0.0029534695453777176, "optim/total_tokens": 4914151424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413178, "loss/out": 2.877176523208618, "created_at": "2025-01-16T08:01:57.446686+00:00"} {"global_step": 9374, "acc_step": 0, "speed/wps": 12907.882983517906, "speed/FLOPS": 202736039546103.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07762519270181656, "optim/lr": 0.002953448368082689, "optim/total_tokens": 4914675712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.958796977996826, "created_at": "2025-01-16T08:02:07.605449+00:00"} {"global_step": 9375, "acc_step": 0, "speed/wps": 12904.068780852209, "speed/FLOPS": 202676132251974.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05485661327838898, "optim/lr": 0.0029534271860455295, "optim/total_tokens": 4915200000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507980, "loss/out": 3.085026264190674, "created_at": "2025-01-16T08:02:17.768485+00:00"} {"global_step": 9376, "acc_step": 0, "speed/wps": 12905.562397643615, "speed/FLOPS": 202699591556128.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05349813029170036, "optim/lr": 0.0029534059992663082, "optim/total_tokens": 4915724288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 3.019343614578247, "created_at": "2025-01-16T08:02:27.931732+00:00"} {"global_step": 9377, "acc_step": 0, "speed/wps": 12911.026880045914, "speed/FLOPS": 202785418761242.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06088823825120926, "optim/lr": 0.002953384807745095, "optim/total_tokens": 4916248576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.902355194091797, "created_at": "2025-01-16T08:02:38.090363+00:00"} {"global_step": 9378, "acc_step": 0, "speed/wps": 12907.22648500998, "speed/FLOPS": 202725728334912.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06192830204963684, "optim/lr": 0.0029533636114819576, "optim/total_tokens": 4916772864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 2.9313700199127197, "created_at": "2025-01-16T08:02:48.254220+00:00"} {"global_step": 9379, "acc_step": 0, "speed/wps": 12906.799240290882, "speed/FLOPS": 202719017869503.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05424075201153755, "optim/lr": 0.0029533424104769665, "optim/total_tokens": 4917297152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8961341381073, "created_at": "2025-01-16T08:02:58.416742+00:00"} {"global_step": 9380, "acc_step": 0, "speed/wps": 12907.214655914026, "speed/FLOPS": 202725542542705.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045911967754364014, "optim/lr": 0.0029533212047301903, "optim/total_tokens": 4917821440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9335196018218994, "created_at": "2025-01-16T08:03:08.576812+00:00"} {"global_step": 9381, "acc_step": 0, "speed/wps": 12903.09617936911, "speed/FLOPS": 202660856209188.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05194113403558731, "optim/lr": 0.002953299994241698, "optim/total_tokens": 4918345728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 2.8192250728607178, "created_at": "2025-01-16T08:03:18.738534+00:00"} {"global_step": 9382, "acc_step": 0, "speed/wps": 12902.089267148081, "speed/FLOPS": 202645041269114.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05021299049258232, "optim/lr": 0.00295327877901156, "optim/total_tokens": 4918870016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9610207080841064, "created_at": "2025-01-16T08:03:28.902998+00:00"} {"global_step": 9383, "acc_step": 0, "speed/wps": 12909.852197688258, "speed/FLOPS": 202766968760632.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056932345032691956, "optim/lr": 0.002953257559039843, "optim/total_tokens": 4919394304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.888317108154297, "created_at": "2025-01-16T08:03:39.059331+00:00"} {"global_step": 9384, "acc_step": 0, "speed/wps": 12905.494338708286, "speed/FLOPS": 202698522597026.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060697074979543686, "optim/lr": 0.0029532363343266186, "optim/total_tokens": 4919918592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9036455154418945, "created_at": "2025-01-16T08:03:49.219091+00:00"} {"global_step": 9385, "acc_step": 0, "speed/wps": 12905.32409037209, "speed/FLOPS": 202695848612960.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057391729205846786, "optim/lr": 0.0029532151048719552, "optim/total_tokens": 4920442880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483664, "loss/out": 2.947657585144043, "created_at": "2025-01-16T08:03:59.380544+00:00"} {"global_step": 9386, "acc_step": 0, "speed/wps": 12904.911623371167, "speed/FLOPS": 202689370251921.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06741510331630707, "optim/lr": 0.0029531938706759218, "optim/total_tokens": 4920967168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0690417289733887, "created_at": "2025-01-16T08:04:09.544350+00:00"} {"global_step": 9387, "acc_step": 0, "speed/wps": 12884.051668769818, "speed/FLOPS": 202361736000326.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0517912395298481, "optim/lr": 0.0029531726317385877, "optim/total_tokens": 4921491456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.9180750846862793, "created_at": "2025-01-16T08:04:19.721458+00:00"} {"global_step": 9388, "acc_step": 0, "speed/wps": 12891.313935808068, "speed/FLOPS": 202475799883562.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060332655906677246, "optim/lr": 0.0029531513880600236, "optim/total_tokens": 4922015744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8871660232543945, "created_at": "2025-01-16T08:04:29.893735+00:00"} {"global_step": 9389, "acc_step": 0, "speed/wps": 12896.208964684787, "speed/FLOPS": 202552683038546.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05055452510714531, "optim/lr": 0.0029531301396402967, "optim/total_tokens": 4922540032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 2.912623882293701, "created_at": "2025-01-16T08:04:40.062999+00:00"} {"global_step": 9390, "acc_step": 0, "speed/wps": 12895.979737668556, "speed/FLOPS": 202549082713265.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.047532957047224045, "optim/lr": 0.0029531088864794774, "optim/total_tokens": 4923064320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.8503057956695557, "created_at": "2025-01-16T08:04:50.235102+00:00"} {"global_step": 9391, "acc_step": 0, "speed/wps": 12902.893111904697, "speed/FLOPS": 202657666755615.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054937925189733505, "optim/lr": 0.002953087628577635, "optim/total_tokens": 4923588608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8175599575042725, "created_at": "2025-01-16T08:05:00.397552+00:00"} {"global_step": 9392, "acc_step": 0, "speed/wps": 12900.306050623869, "speed/FLOPS": 202617033403205.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05473083257675171, "optim/lr": 0.0029530663659348387, "optim/total_tokens": 4924112896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9452924728393555, "created_at": "2025-01-16T08:05:10.564117+00:00"} {"global_step": 9393, "acc_step": 0, "speed/wps": 12896.493453336874, "speed/FLOPS": 202557151323755.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061525583267211914, "optim/lr": 0.0029530450985511586, "optim/total_tokens": 4924637184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.856318712234497, "created_at": "2025-01-16T08:05:20.733867+00:00"} {"global_step": 9394, "acc_step": 0, "speed/wps": 12902.730683827604, "speed/FLOPS": 202655115599465.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04729040339589119, "optim/lr": 0.002953023826426663, "optim/total_tokens": 4925161472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9132041931152344, "created_at": "2025-01-16T08:05:30.895924+00:00"} {"global_step": 9395, "acc_step": 0, "speed/wps": 12903.080335712963, "speed/FLOPS": 202660607362796.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054454900324344635, "optim/lr": 0.002953002549561422, "optim/total_tokens": 4925685760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.9896697998046875, "created_at": "2025-01-16T08:05:41.060788+00:00"} {"global_step": 9396, "acc_step": 0, "speed/wps": 12904.59715922624, "speed/FLOPS": 202684431160406.84, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050819139927625656, "optim/lr": 0.0029529812679555045, "optim/total_tokens": 4926210048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8556487560272217, "created_at": "2025-01-16T08:05:51.221420+00:00"} {"global_step": 9397, "acc_step": 0, "speed/wps": 12903.495255251986, "speed/FLOPS": 202667124244316.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04631862789392471, "optim/lr": 0.0029529599816089805, "optim/total_tokens": 4926734336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418737, "loss/out": 2.975159168243408, "created_at": "2025-01-16T08:06:01.388270+00:00"} {"global_step": 9398, "acc_step": 0, "speed/wps": 12904.893089139061, "speed/FLOPS": 202689079146338.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06884090602397919, "optim/lr": 0.0029529386905219187, "optim/total_tokens": 4927258624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8988637924194336, "created_at": "2025-01-16T08:06:11.549392+00:00"} {"global_step": 9399, "acc_step": 0, "speed/wps": 12901.421228309622, "speed/FLOPS": 202634548801174.97, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05927455797791481, "optim/lr": 0.0029529173946943894, "optim/total_tokens": 4927782912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.027986526489258, "created_at": "2025-01-16T08:06:21.715580+00:00"} {"global_step": 9400, "acc_step": 0, "speed/wps": 12898.633253048023, "speed/FLOPS": 202590759818612.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07860194891691208, "optim/lr": 0.002952896094126462, "optim/total_tokens": 4928307200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 2.9732038974761963, "created_at": "2025-01-16T08:06:31.881976+00:00"} {"global_step": 9401, "acc_step": 0, "speed/wps": 12905.214431250906, "speed/FLOPS": 202694126265774.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07183575630187988, "optim/lr": 0.002952874788818205, "optim/total_tokens": 4928831488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.7414302825927734, "created_at": "2025-01-16T08:06:42.044727+00:00"} {"global_step": 9402, "acc_step": 0, "speed/wps": 12903.198342488527, "speed/FLOPS": 202662460821365.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06073528528213501, "optim/lr": 0.002952853478769689, "optim/total_tokens": 4929355776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8739335536956787, "created_at": "2025-01-16T08:06:52.206475+00:00"} {"global_step": 9403, "acc_step": 0, "speed/wps": 12905.495103198688, "speed/FLOPS": 202698534604398.47, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05339460074901581, "optim/lr": 0.002952832163980983, "optim/total_tokens": 4929880064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9663033485412598, "created_at": "2025-01-16T08:07:02.366440+00:00"} {"global_step": 9404, "acc_step": 0, "speed/wps": 12899.98784271571, "speed/FLOPS": 202612035510744.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05690719187259674, "optim/lr": 0.0029528108444521572, "optim/total_tokens": 4930404352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.899712562561035, "created_at": "2025-01-16T08:07:12.532995+00:00"} {"global_step": 9405, "acc_step": 0, "speed/wps": 12906.542261966222, "speed/FLOPS": 202714981671789.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0490456186234951, "optim/lr": 0.00295278952018328, "optim/total_tokens": 4930928640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9131064414978027, "created_at": "2025-01-16T08:07:22.694037+00:00"} {"global_step": 9406, "acc_step": 0, "speed/wps": 12901.863769619928, "speed/FLOPS": 202641499520568.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05675281211733818, "optim/lr": 0.002952768191174422, "optim/total_tokens": 4931452928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.8514840602874756, "created_at": "2025-01-16T08:07:32.859179+00:00"} {"global_step": 9407, "acc_step": 0, "speed/wps": 12911.44727155976, "speed/FLOPS": 202792021587645.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06751453131437302, "optim/lr": 0.002952746857425652, "optim/total_tokens": 4931977216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 2.969419002532959, "created_at": "2025-01-16T08:07:43.014835+00:00"} {"global_step": 9408, "acc_step": 0, "speed/wps": 12905.430024798336, "speed/FLOPS": 202697512458693.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049068909138441086, "optim/lr": 0.0029527255189370407, "optim/total_tokens": 4932501504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8510706424713135, "created_at": "2025-01-16T08:07:53.174772+00:00"} {"global_step": 9409, "acc_step": 0, "speed/wps": 12904.117197131314, "speed/FLOPS": 202676892696169.03, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08301316201686859, "optim/lr": 0.002952704175708657, "optim/total_tokens": 4933025792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 2.9066061973571777, "created_at": "2025-01-16T08:08:03.340925+00:00"} {"global_step": 9410, "acc_step": 0, "speed/wps": 12902.491331370855, "speed/FLOPS": 202651356240227.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061151184141635895, "optim/lr": 0.00295268282774057, "optim/total_tokens": 4933550080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.8651435375213623, "created_at": "2025-01-16T08:08:13.503486+00:00"} {"global_step": 9411, "acc_step": 0, "speed/wps": 12904.593414218707, "speed/FLOPS": 202684372339917.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06732960790395737, "optim/lr": 0.0029526614750328504, "optim/total_tokens": 4934074368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.947969913482666, "created_at": "2025-01-16T08:08:23.665387+00:00"} {"global_step": 9412, "acc_step": 0, "speed/wps": 12900.319576915497, "speed/FLOPS": 202617245852202.6, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05536440759897232, "optim/lr": 0.002952640117585567, "optim/total_tokens": 4934598656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287670, "loss/out": 3.0112719535827637, "created_at": "2025-01-16T08:08:33.830514+00:00"} {"global_step": 9413, "acc_step": 0, "speed/wps": 12903.96391585193, "speed/FLOPS": 202674485203045.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04965665563941002, "optim/lr": 0.00295261875539879, "optim/total_tokens": 4935122944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.9544453620910645, "created_at": "2025-01-16T08:08:43.994125+00:00"} {"global_step": 9414, "acc_step": 0, "speed/wps": 12905.2345462101, "speed/FLOPS": 202694442198849.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05958249047398567, "optim/lr": 0.002952597388472589, "optim/total_tokens": 4935647232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.856963634490967, "created_at": "2025-01-16T08:08:54.156395+00:00"} {"global_step": 9415, "acc_step": 0, "speed/wps": 12904.648976627834, "speed/FLOPS": 202685245023903.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051976174116134644, "optim/lr": 0.0029525760168070342, "optim/total_tokens": 4936171520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 3.03757381439209, "created_at": "2025-01-16T08:09:04.320154+00:00"} {"global_step": 9416, "acc_step": 0, "speed/wps": 12905.95334273493, "speed/FLOPS": 202705731886002.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07015545666217804, "optim/lr": 0.002952554640402194, "optim/total_tokens": 4936695808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.831068992614746, "created_at": "2025-01-16T08:09:14.479929+00:00"} {"global_step": 9417, "acc_step": 0, "speed/wps": 12903.94827238134, "speed/FLOPS": 202674239500843.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06686335802078247, "optim/lr": 0.00295253325925814, "optim/total_tokens": 4937220096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 3.0111196041107178, "created_at": "2025-01-16T08:09:24.643005+00:00"} {"global_step": 9418, "acc_step": 0, "speed/wps": 12903.467968146031, "speed/FLOPS": 202666695662821.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05687951296567917, "optim/lr": 0.00295251187337494, "optim/total_tokens": 4937744384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8355982303619385, "created_at": "2025-01-16T08:09:34.808630+00:00"} {"global_step": 9419, "acc_step": 0, "speed/wps": 12904.227627976625, "speed/FLOPS": 202678627164342.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05653614550828934, "optim/lr": 0.002952490482752665, "optim/total_tokens": 4938268672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.9209439754486084, "created_at": "2025-01-16T08:09:44.971218+00:00"} {"global_step": 9420, "acc_step": 0, "speed/wps": 12907.563844138433, "speed/FLOPS": 202731027023605.12, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0568002387881279, "optim/lr": 0.0029524690873913846, "optim/total_tokens": 4938792960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.07100772857666, "created_at": "2025-01-16T08:09:55.129435+00:00"} {"global_step": 9421, "acc_step": 0, "speed/wps": 12905.182694521698, "speed/FLOPS": 202693627796831.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06634578108787537, "optim/lr": 0.002952447687291169, "optim/total_tokens": 4939317248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 2.8994317054748535, "created_at": "2025-01-16T08:10:05.291902+00:00"} {"global_step": 9422, "acc_step": 0, "speed/wps": 12910.240851760424, "speed/FLOPS": 202773073106902.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05189547315239906, "optim/lr": 0.0029524262824520866, "optim/total_tokens": 4939841536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414236, "loss/out": 2.878188133239746, "created_at": "2025-01-16T08:10:15.449607+00:00"} {"global_step": 9423, "acc_step": 0, "speed/wps": 12902.88524532796, "speed/FLOPS": 202657543200218.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05925276502966881, "optim/lr": 0.0029524048728742083, "optim/total_tokens": 4940365824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9657177925109863, "created_at": "2025-01-16T08:10:25.613259+00:00"} {"global_step": 9424, "acc_step": 0, "speed/wps": 12904.0125206481, "speed/FLOPS": 202675248608159.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050562817603349686, "optim/lr": 0.002952383458557604, "optim/total_tokens": 4940890112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 2.817673683166504, "created_at": "2025-01-16T08:10:35.774202+00:00"} {"global_step": 9425, "acc_step": 0, "speed/wps": 12907.688633378364, "speed/FLOPS": 202732987010096.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05486733466386795, "optim/lr": 0.0029523620395023437, "optim/total_tokens": 4941414400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9385952949523926, "created_at": "2025-01-16T08:10:45.933964+00:00"} {"global_step": 9426, "acc_step": 0, "speed/wps": 12903.955296675913, "speed/FLOPS": 202674349827042.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05179983004927635, "optim/lr": 0.0029523406157084967, "optim/total_tokens": 4941938688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.871487617492676, "created_at": "2025-01-16T08:10:56.098385+00:00"} {"global_step": 9427, "acc_step": 0, "speed/wps": 12904.655621856316, "speed/FLOPS": 202685349396348.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04734942317008972, "optim/lr": 0.0029523191871761332, "optim/total_tokens": 4942462976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 2.886061668395996, "created_at": "2025-01-16T08:11:06.260303+00:00"} {"global_step": 9428, "acc_step": 0, "speed/wps": 12903.154462780996, "speed/FLOPS": 202661771630261.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061512768268585205, "optim/lr": 0.0029522977539053227, "optim/total_tokens": 4942987264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.925654888153076, "created_at": "2025-01-16T08:11:16.422008+00:00"} {"global_step": 9429, "acc_step": 0, "speed/wps": 12906.969023297368, "speed/FLOPS": 202721684544926.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057330530136823654, "optim/lr": 0.0029522763158961363, "optim/total_tokens": 4943511552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.875690221786499, "created_at": "2025-01-16T08:11:26.582169+00:00"} {"global_step": 9430, "acc_step": 0, "speed/wps": 12899.256000807643, "speed/FLOPS": 202600540927922.0, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0569591298699379, "optim/lr": 0.0029522548731486424, "optim/total_tokens": 4944035840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 2.8914167881011963, "created_at": "2025-01-16T08:11:36.746927+00:00"} {"global_step": 9431, "acc_step": 0, "speed/wps": 12901.987825343627, "speed/FLOPS": 202643447986179.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050603412091732025, "optim/lr": 0.002952233425662912, "optim/total_tokens": 4944560128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.997041702270508, "created_at": "2025-01-16T08:11:46.912516+00:00"} {"global_step": 9432, "acc_step": 0, "speed/wps": 12905.484816905091, "speed/FLOPS": 202698373044022.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048750825226306915, "optim/lr": 0.0029522119734390144, "optim/total_tokens": 4945084416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 2.926326274871826, "created_at": "2025-01-16T08:11:57.074314+00:00"} {"global_step": 9433, "acc_step": 0, "speed/wps": 12905.551564497091, "speed/FLOPS": 202699421406675.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0661950334906578, "optim/lr": 0.002952190516477021, "optim/total_tokens": 4945608704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9529647827148438, "created_at": "2025-01-16T08:12:07.240152+00:00"} {"global_step": 9434, "acc_step": 0, "speed/wps": 12903.449121004009, "speed/FLOPS": 202666399642556.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06822362542152405, "optim/lr": 0.002952169054777, "optim/total_tokens": 4946132992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9293558597564697, "created_at": "2025-01-16T08:12:17.406959+00:00"} {"global_step": 9435, "acc_step": 0, "speed/wps": 12911.776331904744, "speed/FLOPS": 202797189932539.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054350029677152634, "optim/lr": 0.0029521475883390225, "optim/total_tokens": 4946657280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 2.921160936355591, "created_at": "2025-01-16T08:12:27.564661+00:00"} {"global_step": 9436, "acc_step": 0, "speed/wps": 12911.132498295136, "speed/FLOPS": 202787077640980.84, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06370344012975693, "optim/lr": 0.002952126117163158, "optim/total_tokens": 4947181568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.787543773651123, "created_at": "2025-01-16T08:12:37.719920+00:00"} {"global_step": 9437, "acc_step": 0, "speed/wps": 12898.331276419209, "speed/FLOPS": 202586016860697.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04935583099722862, "optim/lr": 0.0029521046412494764, "optim/total_tokens": 4947705856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 2.874155044555664, "created_at": "2025-01-16T08:12:47.886503+00:00"} {"global_step": 9438, "acc_step": 0, "speed/wps": 12910.242341600266, "speed/FLOPS": 202773096506884.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058305200189352036, "optim/lr": 0.0029520831605980486, "optim/total_tokens": 4948230144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 2.880683660507202, "created_at": "2025-01-16T08:12:58.050531+00:00"} {"global_step": 9439, "acc_step": 0, "speed/wps": 12905.70628708114, "speed/FLOPS": 202701851537469.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045866113156080246, "optim/lr": 0.0029520616752089445, "optim/total_tokens": 4948754432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.0310423374176025, "created_at": "2025-01-16T08:13:08.212367+00:00"} {"global_step": 9440, "acc_step": 0, "speed/wps": 12901.680881491453, "speed/FLOPS": 202638627011196.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04515581950545311, "optim/lr": 0.002952040185082234, "optim/total_tokens": 4949278720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 3.0069453716278076, "created_at": "2025-01-16T08:13:18.377203+00:00"} {"global_step": 9441, "acc_step": 0, "speed/wps": 12904.658575293395, "speed/FLOPS": 202685395784136.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04198233410716057, "optim/lr": 0.0029520186902179868, "optim/total_tokens": 4949803008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9483206272125244, "created_at": "2025-01-16T08:13:28.537633+00:00"} {"global_step": 9442, "acc_step": 0, "speed/wps": 12910.896011801486, "speed/FLOPS": 202783363295631.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05075450614094734, "optim/lr": 0.0029519971906162737, "optim/total_tokens": 4950327296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473062, "loss/out": 2.923198938369751, "created_at": "2025-01-16T08:13:38.697907+00:00"} {"global_step": 9443, "acc_step": 0, "speed/wps": 12901.52512357246, "speed/FLOPS": 202636180619044.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04787425696849823, "optim/lr": 0.0029519756862771643, "optim/total_tokens": 4950851584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.860745906829834, "created_at": "2025-01-16T08:13:48.862324+00:00"} {"global_step": 9444, "acc_step": 0, "speed/wps": 12906.103565689307, "speed/FLOPS": 202708091343928.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055643320083618164, "optim/lr": 0.0029519541772007295, "optim/total_tokens": 4951375872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.970428228378296, "created_at": "2025-01-16T08:13:59.021706+00:00"} {"global_step": 9445, "acc_step": 0, "speed/wps": 12905.546562131549, "speed/FLOPS": 202699342837651.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05446808040142059, "optim/lr": 0.0029519326633870384, "optim/total_tokens": 4951900160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.0303664207458496, "created_at": "2025-01-16T08:14:09.183735+00:00"} {"global_step": 9446, "acc_step": 0, "speed/wps": 12905.316135626857, "speed/FLOPS": 202695723672755.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044341377913951874, "optim/lr": 0.002951911144836162, "optim/total_tokens": 4952424448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.026228189468384, "created_at": "2025-01-16T08:14:19.344742+00:00"} {"global_step": 9447, "acc_step": 0, "speed/wps": 12901.617187477765, "speed/FLOPS": 202637626609186.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05301913619041443, "optim/lr": 0.00295188962154817, "optim/total_tokens": 4952948736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395218, "loss/out": 2.8790533542633057, "created_at": "2025-01-16T08:14:29.508938+00:00"} {"global_step": 9448, "acc_step": 0, "speed/wps": 12903.374923117517, "speed/FLOPS": 202665234262790.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05187034606933594, "optim/lr": 0.002951868093523134, "optim/total_tokens": 4953473024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9141879081726074, "created_at": "2025-01-16T08:14:39.674984+00:00"} {"global_step": 9449, "acc_step": 0, "speed/wps": 12904.170250649715, "speed/FLOPS": 202677725974578.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05025891587138176, "optim/lr": 0.0029518465607611217, "optim/total_tokens": 4953997312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 2.8184814453125, "created_at": "2025-01-16T08:14:49.836941+00:00"} {"global_step": 9450, "acc_step": 0, "speed/wps": 12903.198228920475, "speed/FLOPS": 202662459037623.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0557958222925663, "optim/lr": 0.0029518250232622055, "optim/total_tokens": 4954521600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 2.9544572830200195, "created_at": "2025-01-16T08:15:00.000880+00:00"} {"global_step": 9451, "acc_step": 0, "speed/wps": 12910.357298347682, "speed/FLOPS": 202774902060569.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05996391922235489, "optim/lr": 0.0029518034810264546, "optim/total_tokens": 4955045888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.889230251312256, "created_at": "2025-01-16T08:15:10.156991+00:00"} {"global_step": 9452, "acc_step": 0, "speed/wps": 12903.423310529186, "speed/FLOPS": 202665994253582.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06363478302955627, "optim/lr": 0.00295178193405394, "optim/total_tokens": 4955570176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 2.992213726043701, "created_at": "2025-01-16T08:15:20.319422+00:00"} {"global_step": 9453, "acc_step": 0, "speed/wps": 12903.792385174982, "speed/FLOPS": 202671791078056.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049991484731435776, "optim/lr": 0.002951760382344732, "optim/total_tokens": 4956094464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8791069984436035, "created_at": "2025-01-16T08:15:30.480543+00:00"} {"global_step": 9454, "acc_step": 0, "speed/wps": 12899.914156407058, "speed/FLOPS": 202610878166011.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06846857815980911, "optim/lr": 0.0029517388258989, "optim/total_tokens": 4956618752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.966754913330078, "created_at": "2025-01-16T08:15:40.644711+00:00"} {"global_step": 9455, "acc_step": 0, "speed/wps": 12904.525237712714, "speed/FLOPS": 202683301534206.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07970911264419556, "optim/lr": 0.0029517172647165156, "optim/total_tokens": 4957143040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.9097084999084473, "created_at": "2025-01-16T08:15:50.805442+00:00"} {"global_step": 9456, "acc_step": 0, "speed/wps": 12904.702260524788, "speed/FLOPS": 202686081920724.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04893196374177933, "optim/lr": 0.0029516956987976476, "optim/total_tokens": 4957667328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.0971486568450928, "created_at": "2025-01-16T08:16:00.968166+00:00"} {"global_step": 9457, "acc_step": 0, "speed/wps": 12903.127533195624, "speed/FLOPS": 202661348664118.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09361416846513748, "optim/lr": 0.002951674128142368, "optim/total_tokens": 4958191616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 2.89821195602417, "created_at": "2025-01-16T08:16:11.130904+00:00"} {"global_step": 9458, "acc_step": 0, "speed/wps": 12901.349280694232, "speed/FLOPS": 202633418765008.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12332981079816818, "optim/lr": 0.0029516525527507457, "optim/total_tokens": 4958715904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.905257225036621, "created_at": "2025-01-16T08:16:21.295057+00:00"} {"global_step": 9459, "acc_step": 0, "speed/wps": 12902.98546031812, "speed/FLOPS": 202659117214348.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10432109236717224, "optim/lr": 0.0029516309726228523, "optim/total_tokens": 4959240192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 3.016003131866455, "created_at": "2025-01-16T08:16:31.457217+00:00"} {"global_step": 9460, "acc_step": 0, "speed/wps": 12906.792420040196, "speed/FLOPS": 202718910748094.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09506803750991821, "optim/lr": 0.0029516093877587574, "optim/total_tokens": 4959764480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 2.96126651763916, "created_at": "2025-01-16T08:16:41.616194+00:00"} {"global_step": 9461, "acc_step": 0, "speed/wps": 12905.723918299378, "speed/FLOPS": 202702128459979.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06550541520118713, "optim/lr": 0.0029515877981585318, "optim/total_tokens": 4960288768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.969940662384033, "created_at": "2025-01-16T08:16:51.777758+00:00"} {"global_step": 9462, "acc_step": 0, "speed/wps": 12907.459227741127, "speed/FLOPS": 202729383879327.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.10128627717494965, "optim/lr": 0.002951566203822246, "optim/total_tokens": 4960813056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409163, "loss/out": 2.985997438430786, "created_at": "2025-01-16T08:17:01.937317+00:00"} {"global_step": 9463, "acc_step": 0, "speed/wps": 12902.662039646459, "speed/FLOPS": 202654037448273.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06162078678607941, "optim/lr": 0.0029515446047499705, "optim/total_tokens": 4961337344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.035754680633545, "created_at": "2025-01-16T08:17:12.104082+00:00"} {"global_step": 9464, "acc_step": 0, "speed/wps": 12905.664808108279, "speed/FLOPS": 202701200053202.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09250285476446152, "optim/lr": 0.002951523000941775, "optim/total_tokens": 4961861632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469813, "loss/out": 3.032660722732544, "created_at": "2025-01-16T08:17:22.272478+00:00"} {"global_step": 9465, "acc_step": 0, "speed/wps": 12909.200671013588, "speed/FLOPS": 202756735638915.28, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055102985352277756, "optim/lr": 0.0029515013923977307, "optim/total_tokens": 4962385920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 2.9200572967529297, "created_at": "2025-01-16T08:17:32.429322+00:00"} {"global_step": 9466, "acc_step": 0, "speed/wps": 12906.66717467374, "speed/FLOPS": 202716943597509.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09259039908647537, "optim/lr": 0.002951479779117908, "optim/total_tokens": 4962910208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.885634660720825, "created_at": "2025-01-16T08:17:42.588248+00:00"} {"global_step": 9467, "acc_step": 0, "speed/wps": 12904.13603675963, "speed/FLOPS": 202677188598420.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06041080132126808, "optim/lr": 0.002951458161102378, "optim/total_tokens": 4963434496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 2.934880018234253, "created_at": "2025-01-16T08:17:52.752070+00:00"} {"global_step": 9468, "acc_step": 0, "speed/wps": 12905.953431497865, "speed/FLOPS": 202705733280146.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07293188571929932, "optim/lr": 0.00295143653835121, "optim/total_tokens": 4963958784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0031042098999023, "created_at": "2025-01-16T08:18:02.911554+00:00"} {"global_step": 9469, "acc_step": 0, "speed/wps": 12903.902159739662, "speed/FLOPS": 202673515238440.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0801423117518425, "optim/lr": 0.0029514149108644757, "optim/total_tokens": 4964483072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 2.9970152378082275, "created_at": "2025-01-16T08:18:13.073503+00:00"} {"global_step": 9470, "acc_step": 0, "speed/wps": 12904.260594081186, "speed/FLOPS": 202679144942314.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059717509895563126, "optim/lr": 0.0029513932786422443, "optim/total_tokens": 4965007360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.025202989578247, "created_at": "2025-01-16T08:18:23.236784+00:00"} {"global_step": 9471, "acc_step": 0, "speed/wps": 12903.58688824496, "speed/FLOPS": 202668563466388.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08283281326293945, "optim/lr": 0.002951371641684588, "optim/total_tokens": 4965531648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9218554496765137, "created_at": "2025-01-16T08:18:33.402508+00:00"} {"global_step": 9472, "acc_step": 0, "speed/wps": 12900.65931383153, "speed/FLOPS": 202622581887316.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04953480884432793, "optim/lr": 0.0029513499999915763, "optim/total_tokens": 4966055936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287430, "loss/out": 3.0780081748962402, "created_at": "2025-01-16T08:18:43.566277+00:00"} {"global_step": 9473, "acc_step": 0, "speed/wps": 12905.107306363136, "speed/FLOPS": 202692443722207.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06166845187544823, "optim/lr": 0.0029513283535632802, "optim/total_tokens": 4966580224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.969972848892212, "created_at": "2025-01-16T08:18:53.727121+00:00"} {"global_step": 9474, "acc_step": 0, "speed/wps": 12909.612788176353, "speed/FLOPS": 202763208505264.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04386010393500328, "optim/lr": 0.00295130670239977, "optim/total_tokens": 4967104512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487384, "loss/out": 2.940298080444336, "created_at": "2025-01-16T08:19:03.884407+00:00"} {"global_step": 9475, "acc_step": 0, "speed/wps": 12909.897218373548, "speed/FLOPS": 202767675872359.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05417224019765854, "optim/lr": 0.0029512850465011173, "optim/total_tokens": 4967628800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 2.8103909492492676, "created_at": "2025-01-16T08:19:14.043887+00:00"} {"global_step": 9476, "acc_step": 0, "speed/wps": 12907.958193380078, "speed/FLOPS": 202737220820337.8, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046242065727710724, "optim/lr": 0.002951263385867391, "optim/total_tokens": 4968153088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.905109405517578, "created_at": "2025-01-16T08:19:24.206904+00:00"} {"global_step": 9477, "acc_step": 0, "speed/wps": 12901.304430776958, "speed/FLOPS": 202632714335427.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0028, "optim/grad_norm": 0.0508892722427845, "optim/lr": 0.0029512417204986636, "optim/total_tokens": 4968677376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.843393087387085, "created_at": "2025-01-16T08:19:34.370362+00:00"} {"global_step": 9478, "acc_step": 0, "speed/wps": 12899.34929383976, "speed/FLOPS": 202602006223189.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05044865980744362, "optim/lr": 0.0029512200503950046, "optim/total_tokens": 4969201664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9642279148101807, "created_at": "2025-01-16T08:19:44.535122+00:00"} {"global_step": 9479, "acc_step": 0, "speed/wps": 12890.832201898795, "speed/FLOPS": 202468233590545.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04663332179188728, "optim/lr": 0.0029511983755564856, "optim/total_tokens": 4969725952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410264, "loss/out": 2.722165822982788, "created_at": "2025-01-16T08:19:54.707719+00:00"} {"global_step": 9480, "acc_step": 0, "speed/wps": 12897.194137688039, "speed/FLOPS": 202568156534331.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05344056338071823, "optim/lr": 0.0029511766959831767, "optim/total_tokens": 4970250240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.05336856842041, "created_at": "2025-01-16T08:20:04.876039+00:00"} {"global_step": 9481, "acc_step": 0, "speed/wps": 12892.125908603353, "speed/FLOPS": 202488553032080.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04967794939875603, "optim/lr": 0.0029511550116751483, "optim/total_tokens": 4970774528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.985245943069458, "created_at": "2025-01-16T08:20:15.049511+00:00"} {"global_step": 9482, "acc_step": 0, "speed/wps": 12897.335219070907, "speed/FLOPS": 202570372411315.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05617421120405197, "optim/lr": 0.002951133322632472, "optim/total_tokens": 4971298816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.9887709617614746, "created_at": "2025-01-16T08:20:25.217198+00:00"} {"global_step": 9483, "acc_step": 0, "speed/wps": 12889.590745297863, "speed/FLOPS": 202448734808688.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06125592067837715, "optim/lr": 0.002951111628855218, "optim/total_tokens": 4971823104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318033, "loss/out": 2.854210138320923, "created_at": "2025-01-16T08:20:35.389893+00:00"} {"global_step": 9484, "acc_step": 0, "speed/wps": 12891.826698797824, "speed/FLOPS": 202483853530927.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04591574892401695, "optim/lr": 0.0029510899303434577, "optim/total_tokens": 4972347392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.923229217529297, "created_at": "2025-01-16T08:20:45.560672+00:00"} {"global_step": 9485, "acc_step": 0, "speed/wps": 12894.389504874232, "speed/FLOPS": 202524105922021.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05625506117939949, "optim/lr": 0.0029510682270972605, "optim/total_tokens": 4972871680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440909, "loss/out": 2.879410743713379, "created_at": "2025-01-16T08:20:55.735956+00:00"} {"global_step": 9486, "acc_step": 0, "speed/wps": 12891.262747361448, "speed/FLOPS": 202474995898668.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.053581416606903076, "optim/lr": 0.002951046519116699, "optim/total_tokens": 4973395968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 3.0291037559509277, "created_at": "2025-01-16T08:21:05.910220+00:00"} {"global_step": 9487, "acc_step": 0, "speed/wps": 12892.237123975869, "speed/FLOPS": 202490299822331.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05052732676267624, "optim/lr": 0.002951024806401843, "optim/total_tokens": 4973920256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.076754570007324, "created_at": "2025-01-16T08:21:16.080693+00:00"} {"global_step": 9488, "acc_step": 0, "speed/wps": 12897.86355086756, "speed/FLOPS": 202578670588186.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061962299048900604, "optim/lr": 0.0029510030889527633, "optim/total_tokens": 4974444544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9525954723358154, "created_at": "2025-01-16T08:21:26.250038+00:00"} {"global_step": 9489, "acc_step": 0, "speed/wps": 12895.940080700762, "speed/FLOPS": 202548459846091.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0442839041352272, "optim/lr": 0.002950981366769531, "optim/total_tokens": 4974968832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 2.9095544815063477, "created_at": "2025-01-16T08:21:36.419431+00:00"} {"global_step": 9490, "acc_step": 0, "speed/wps": 12897.05049117998, "speed/FLOPS": 202565900368531.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05703410133719444, "optim/lr": 0.0029509596398522167, "optim/total_tokens": 4975493120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7623543739318848, "created_at": "2025-01-16T08:21:46.586413+00:00"} {"global_step": 9491, "acc_step": 0, "speed/wps": 12899.86932119683, "speed/FLOPS": 202610173967424.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05320580676198006, "optim/lr": 0.0029509379082008924, "optim/total_tokens": 4976017408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.0305190086364746, "created_at": "2025-01-16T08:21:56.754612+00:00"} {"global_step": 9492, "acc_step": 0, "speed/wps": 12912.084381797125, "speed/FLOPS": 202802028279404.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04877492040395737, "optim/lr": 0.0029509161718156277, "optim/total_tokens": 4976541696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9871840476989746, "created_at": "2025-01-16T08:22:06.911983+00:00"} {"global_step": 9493, "acc_step": 0, "speed/wps": 12899.94591807188, "speed/FLOPS": 202611377026602.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05823522061109543, "optim/lr": 0.002950894430696494, "optim/total_tokens": 4977065984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.983368158340454, "created_at": "2025-01-16T08:22:17.076320+00:00"} {"global_step": 9494, "acc_step": 0, "speed/wps": 12897.754654413944, "speed/FLOPS": 202576960219742.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052085548639297485, "optim/lr": 0.002950872684843562, "optim/total_tokens": 4977590272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491104, "loss/out": 2.938797950744629, "created_at": "2025-01-16T08:22:27.242698+00:00"} {"global_step": 9495, "acc_step": 0, "speed/wps": 12891.28579096825, "speed/FLOPS": 202475357830178.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05534566566348076, "optim/lr": 0.002950850934256903, "optim/total_tokens": 4978114560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 2.9594480991363525, "created_at": "2025-01-16T08:22:37.413790+00:00"} {"global_step": 9496, "acc_step": 0, "speed/wps": 12901.275524429597, "speed/FLOPS": 202632260321521.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.045491497963666916, "optim/lr": 0.0029508291789365876, "optim/total_tokens": 4978638848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.919849395751953, "created_at": "2025-01-16T08:22:47.580070+00:00"} {"global_step": 9497, "acc_step": 0, "speed/wps": 12892.541375825238, "speed/FLOPS": 202495078515711.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04642592743039131, "optim/lr": 0.0029508074188826874, "optim/total_tokens": 4979163136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303758, "loss/out": 2.9396474361419678, "created_at": "2025-01-16T08:22:57.751546+00:00"} {"global_step": 9498, "acc_step": 0, "speed/wps": 12894.871394822821, "speed/FLOPS": 202531674665850.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04960665479302406, "optim/lr": 0.0029507856540952728, "optim/total_tokens": 4979687424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.0121469497680664, "created_at": "2025-01-16T08:23:07.920042+00:00"} {"global_step": 9499, "acc_step": 0, "speed/wps": 12892.363318877291, "speed/FLOPS": 202492281886668.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04976833239197731, "optim/lr": 0.002950763884574415, "optim/total_tokens": 4980211712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.0891261100769043, "created_at": "2025-01-16T08:23:18.095036+00:00"} {"global_step": 9500, "acc_step": 0, "speed/wps": 12893.819799620183, "speed/FLOPS": 202515157918149.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.04577389359474182, "optim/lr": 0.002950742110320185, "optim/total_tokens": 4980736000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 2.9645402431488037, "created_at": "2025-01-16T08:23:28.264700+00:00"} {"global_step": 9501, "acc_step": 0, "speed/wps": 12895.1436204905, "speed/FLOPS": 202535950344038.53, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05573992058634758, "optim/lr": 0.002950720331332654, "optim/total_tokens": 4981260288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.830228328704834, "created_at": "2025-01-16T08:23:38.432681+00:00"} {"global_step": 9502, "acc_step": 0, "speed/wps": 12895.912760547797, "speed/FLOPS": 202548030745547.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05787378177046776, "optim/lr": 0.0029506985476118923, "optim/total_tokens": 4981784576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8274965286254883, "created_at": "2025-01-16T08:23:48.602501+00:00"} {"global_step": 9503, "acc_step": 0, "speed/wps": 12894.299297242742, "speed/FLOPS": 202522689087210.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05179537832736969, "optim/lr": 0.0029506767591579723, "optim/total_tokens": 4982308864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 2.8339622020721436, "created_at": "2025-01-16T08:23:58.771464+00:00"} {"global_step": 9504, "acc_step": 0, "speed/wps": 12897.570393481314, "speed/FLOPS": 202574066148594.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05911402776837349, "optim/lr": 0.0029506549659709643, "optim/total_tokens": 4982833152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.000471830368042, "created_at": "2025-01-16T08:24:08.945688+00:00"} {"global_step": 9505, "acc_step": 0, "speed/wps": 12906.183879577815, "speed/FLOPS": 202709352783910.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062256548553705215, "optim/lr": 0.0029506331680509393, "optim/total_tokens": 4983357440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8559579849243164, "created_at": "2025-01-16T08:24:19.109200+00:00"} {"global_step": 9506, "acc_step": 0, "speed/wps": 12907.705438966701, "speed/FLOPS": 202733250964954.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06985436379909515, "optim/lr": 0.002950611365397969, "optim/total_tokens": 4983881728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.9560132026672363, "created_at": "2025-01-16T08:24:29.271770+00:00"} {"global_step": 9507, "acc_step": 0, "speed/wps": 12902.63503049656, "speed/FLOPS": 202653613232459.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045736588537693024, "optim/lr": 0.0029505895580121235, "optim/total_tokens": 4984406016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.001690149307251, "created_at": "2025-01-16T08:24:39.434320+00:00"} {"global_step": 9508, "acc_step": 0, "speed/wps": 12910.236930569927, "speed/FLOPS": 202773011519217.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052138976752758026, "optim/lr": 0.0029505677458934753, "optim/total_tokens": 4984930304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.915764808654785, "created_at": "2025-01-16T08:24:49.594651+00:00"} {"global_step": 9509, "acc_step": 0, "speed/wps": 12907.084961058887, "speed/FLOPS": 202723505506786.53, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04681707173585892, "optim/lr": 0.0029505459290420947, "optim/total_tokens": 4985454592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9365649223327637, "created_at": "2025-01-16T08:24:59.753434+00:00"} {"global_step": 9510, "acc_step": 0, "speed/wps": 12897.02189699533, "speed/FLOPS": 202565451257569.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07003647834062576, "optim/lr": 0.002950524107458053, "optim/total_tokens": 4985978880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.976243257522583, "created_at": "2025-01-16T08:25:09.921030+00:00"} {"global_step": 9511, "acc_step": 0, "speed/wps": 12910.660453041348, "speed/FLOPS": 202779663521611.94, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05637476220726967, "optim/lr": 0.002950502281141421, "optim/total_tokens": 4986503168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 2.919090747833252, "created_at": "2025-01-16T08:25:20.080173+00:00"} {"global_step": 9512, "acc_step": 0, "speed/wps": 12904.2618955962, "speed/FLOPS": 202679165384396.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0775899589061737, "optim/lr": 0.0029504804500922704, "optim/total_tokens": 4987027456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 2.963348865509033, "created_at": "2025-01-16T08:25:30.243634+00:00"} {"global_step": 9513, "acc_step": 0, "speed/wps": 12899.431761799608, "speed/FLOPS": 202603301495821.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07228341698646545, "optim/lr": 0.002950458614310673, "optim/total_tokens": 4987551744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9567136764526367, "created_at": "2025-01-16T08:25:40.409095+00:00"} {"global_step": 9514, "acc_step": 0, "speed/wps": 12899.536820357172, "speed/FLOPS": 202604951584833.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061419520527124405, "optim/lr": 0.0029504367737966993, "optim/total_tokens": 4988076032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.982276201248169, "created_at": "2025-01-16T08:25:50.577134+00:00"} {"global_step": 9515, "acc_step": 0, "speed/wps": 12913.537966224318, "speed/FLOPS": 202824858820267.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08831769227981567, "optim/lr": 0.0029504149285504203, "optim/total_tokens": 4988600320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8259334564208984, "created_at": "2025-01-16T08:26:00.731557+00:00"} {"global_step": 9516, "acc_step": 0, "speed/wps": 12884.583320924017, "speed/FLOPS": 202370086327974.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0456027127802372, "optim/lr": 0.002950393078571908, "optim/total_tokens": 4989124608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.8929696083068848, "created_at": "2025-01-16T08:26:10.908669+00:00"} {"global_step": 9517, "acc_step": 0, "speed/wps": 12898.320049946684, "speed/FLOPS": 202585840533519.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08047901093959808, "optim/lr": 0.002950371223861233, "optim/total_tokens": 4989648896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9413697719573975, "created_at": "2025-01-16T08:26:21.075755+00:00"} {"global_step": 9518, "acc_step": 0, "speed/wps": 12902.554874326832, "speed/FLOPS": 202652354269668.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.055453959852457047, "optim/lr": 0.0029503493644184673, "optim/total_tokens": 4990173184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.0016894340515137, "created_at": "2025-01-16T08:26:31.239608+00:00"} {"global_step": 9519, "acc_step": 0, "speed/wps": 12892.95096303076, "speed/FLOPS": 202501511645606.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08431284129619598, "optim/lr": 0.002950327500243682, "optim/total_tokens": 4990697472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396559, "loss/out": 2.7883365154266357, "created_at": "2025-01-16T08:26:41.410026+00:00"} {"global_step": 9520, "acc_step": 0, "speed/wps": 12896.794787275949, "speed/FLOPS": 202561884187344.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08417267352342606, "optim/lr": 0.002950305631336948, "optim/total_tokens": 4991221760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.008936643600464, "created_at": "2025-01-16T08:26:51.580955+00:00"} {"global_step": 9521, "acc_step": 0, "speed/wps": 12907.511983201031, "speed/FLOPS": 202730212476318.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06268798559904099, "optim/lr": 0.002950283757698337, "optim/total_tokens": 4991746048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.969982862472534, "created_at": "2025-01-16T08:27:01.740322+00:00"} {"global_step": 9522, "acc_step": 0, "speed/wps": 12902.60922209555, "speed/FLOPS": 202653207876057.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06081651151180267, "optim/lr": 0.00295026187932792, "optim/total_tokens": 4992270336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 2.958733081817627, "created_at": "2025-01-16T08:27:11.903956+00:00"} {"global_step": 9523, "acc_step": 0, "speed/wps": 12907.8567492399, "speed/FLOPS": 202735627500717.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04793575406074524, "optim/lr": 0.0029502399962257693, "optim/total_tokens": 4992794624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.9379661083221436, "created_at": "2025-01-16T08:27:22.063480+00:00"} {"global_step": 9524, "acc_step": 0, "speed/wps": 12900.160658082768, "speed/FLOPS": 202614749813557.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053619589656591415, "optim/lr": 0.0029502181083919557, "optim/total_tokens": 4993318912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388560, "loss/out": 2.967799663543701, "created_at": "2025-01-16T08:27:32.230851+00:00"} {"global_step": 9525, "acc_step": 0, "speed/wps": 12905.484014546477, "speed/FLOPS": 202698360441878.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044966962188482285, "optim/lr": 0.00295019621582655, "optim/total_tokens": 4993843200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.990586757659912, "created_at": "2025-01-16T08:27:42.394094+00:00"} {"global_step": 9526, "acc_step": 0, "speed/wps": 12893.026870663878, "speed/FLOPS": 202502703879292.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04904433712363243, "optim/lr": 0.002950174318529625, "optim/total_tokens": 4994367488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.0713858604431152, "created_at": "2025-01-16T08:27:52.568541+00:00"} {"global_step": 9527, "acc_step": 0, "speed/wps": 12904.697604334062, "speed/FLOPS": 202686008788851.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053221751004457474, "optim/lr": 0.002950152416501251, "optim/total_tokens": 4994891776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.836108684539795, "created_at": "2025-01-16T08:28:02.730786+00:00"} {"global_step": 9528, "acc_step": 0, "speed/wps": 12904.973134731013, "speed/FLOPS": 202690336372353.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1344895362854004, "optim/lr": 0.0029501305097415, "optim/total_tokens": 4995416064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 2.871990203857422, "created_at": "2025-01-16T08:28:12.891420+00:00"} {"global_step": 9529, "acc_step": 0, "speed/wps": 12897.946698415466, "speed/FLOPS": 202579976534682.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10019663721323013, "optim/lr": 0.002950108598250443, "optim/total_tokens": 4995940352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.91873836517334, "created_at": "2025-01-16T08:28:23.063190+00:00"} {"global_step": 9530, "acc_step": 0, "speed/wps": 12901.803720201773, "speed/FLOPS": 202640556361940.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07025055587291718, "optim/lr": 0.002950086682028152, "optim/total_tokens": 4996464640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283719, "loss/out": 2.8986012935638428, "created_at": "2025-01-16T08:28:33.229443+00:00"} {"global_step": 9531, "acc_step": 0, "speed/wps": 12902.229197123079, "speed/FLOPS": 202647239061655.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055130038410425186, "optim/lr": 0.002950064761074699, "optim/total_tokens": 4996988928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.9781875610351562, "created_at": "2025-01-16T08:28:43.394366+00:00"} {"global_step": 9532, "acc_step": 0, "speed/wps": 12901.272712093249, "speed/FLOPS": 202632216149914.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06742248684167862, "optim/lr": 0.002950042835390154, "optim/total_tokens": 4997513216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.7990949153900146, "created_at": "2025-01-16T08:28:53.558415+00:00"} {"global_step": 9533, "acc_step": 0, "speed/wps": 12909.111367885222, "speed/FLOPS": 202755333010568.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048960134387016296, "optim/lr": 0.0029500209049745894, "optim/total_tokens": 4998037504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.921318292617798, "created_at": "2025-01-16T08:29:03.715956+00:00"} {"global_step": 9534, "acc_step": 0, "speed/wps": 12893.125165567644, "speed/FLOPS": 202504247735827.38, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06350870430469513, "optim/lr": 0.0029499989698280773, "optim/total_tokens": 4998561792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.891664981842041, "created_at": "2025-01-16T08:29:13.885860+00:00"} {"global_step": 9535, "acc_step": 0, "speed/wps": 12905.304525408594, "speed/FLOPS": 202695541318323.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041470352560281754, "optim/lr": 0.002949977029950688, "optim/total_tokens": 4999086080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410747, "loss/out": 2.8226137161254883, "created_at": "2025-01-16T08:29:24.048870+00:00"} {"global_step": 9536, "acc_step": 0, "speed/wps": 12904.01253839535, "speed/FLOPS": 202675248886904.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05764703452587128, "optim/lr": 0.0029499550853424946, "optim/total_tokens": 4999610368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9542462825775146, "created_at": "2025-01-16T08:29:34.210177+00:00"} {"global_step": 9537, "acc_step": 0, "speed/wps": 12903.183177622179, "speed/FLOPS": 202662222636300.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055487196892499924, "optim/lr": 0.002949933136003567, "optim/total_tokens": 5000134656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 2.924903392791748, "created_at": "2025-01-16T08:29:44.381739+00:00"} {"global_step": 9538, "acc_step": 0, "speed/wps": 12912.337300627229, "speed/FLOPS": 202806000717178.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05306844413280487, "optim/lr": 0.002949911181933978, "optim/total_tokens": 5000658944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 2.805429697036743, "created_at": "2025-01-16T08:29:54.536263+00:00"} {"global_step": 9539, "acc_step": 0, "speed/wps": 12902.244213076741, "speed/FLOPS": 202647474907842.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046880725771188736, "optim/lr": 0.002949889223133799, "optim/total_tokens": 5001183232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7177138328552246, "created_at": "2025-01-16T08:30:04.702582+00:00"} {"global_step": 9540, "acc_step": 0, "speed/wps": 12908.420482945976, "speed/FLOPS": 202744481713222.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04976905137300491, "optim/lr": 0.002949867259603102, "optim/total_tokens": 5001707520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.9590561389923096, "created_at": "2025-01-16T08:30:14.866500+00:00"} {"global_step": 9541, "acc_step": 0, "speed/wps": 12902.866364399988, "speed/FLOPS": 202657246649299.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04201444610953331, "optim/lr": 0.0029498452913419575, "optim/total_tokens": 5002231808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.917574644088745, "created_at": "2025-01-16T08:30:25.028364+00:00"} {"global_step": 9542, "acc_step": 0, "speed/wps": 12905.374288803894, "speed/FLOPS": 202696637048311.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06762263923883438, "optim/lr": 0.0029498233183504383, "optim/total_tokens": 5002756096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.879408359527588, "created_at": "2025-01-16T08:30:35.188559+00:00"} {"global_step": 9543, "acc_step": 0, "speed/wps": 12903.451160576573, "speed/FLOPS": 202666431676846.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05546465143561363, "optim/lr": 0.0029498013406286158, "optim/total_tokens": 5003280384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.755092144012451, "created_at": "2025-01-16T08:30:45.351230+00:00"} {"global_step": 9544, "acc_step": 0, "speed/wps": 12905.314862300504, "speed/FLOPS": 202695703673415.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0438244491815567, "optim/lr": 0.0029497793581765614, "optim/total_tokens": 5003804672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.861482620239258, "created_at": "2025-01-16T08:30:55.511156+00:00"} {"global_step": 9545, "acc_step": 0, "speed/wps": 12903.199951369488, "speed/FLOPS": 202662486091051.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.044100724160671234, "optim/lr": 0.0029497573709943476, "optim/total_tokens": 5004328960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 2.903402805328369, "created_at": "2025-01-16T08:31:05.675923+00:00"} {"global_step": 9546, "acc_step": 0, "speed/wps": 12901.337852645995, "speed/FLOPS": 202633239271807.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04824066907167435, "optim/lr": 0.0029497353790820446, "optim/total_tokens": 5004853248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9032018184661865, "created_at": "2025-01-16T08:31:15.839582+00:00"} {"global_step": 9547, "acc_step": 0, "speed/wps": 12901.858787854178, "speed/FLOPS": 202641421275092.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04928039386868477, "optim/lr": 0.002949713382439726, "optim/total_tokens": 5005377536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.947183847427368, "created_at": "2025-01-16T08:31:26.018616+00:00"} {"global_step": 9548, "acc_step": 0, "speed/wps": 12905.75416770266, "speed/FLOPS": 202702603568425.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05736025422811508, "optim/lr": 0.002949691381067462, "optim/total_tokens": 5005901824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 2.991290330886841, "created_at": "2025-01-16T08:31:36.178281+00:00"} {"global_step": 9549, "acc_step": 0, "speed/wps": 12906.592326656866, "speed/FLOPS": 202715768006552.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05513988435268402, "optim/lr": 0.002949669374965325, "optim/total_tokens": 5006426112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9683477878570557, "created_at": "2025-01-16T08:31:46.338766+00:00"} {"global_step": 9550, "acc_step": 0, "speed/wps": 12901.376222962805, "speed/FLOPS": 202633841930359.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048965319991111755, "optim/lr": 0.002949647364133387, "optim/total_tokens": 5006950400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 3.016906976699829, "created_at": "2025-01-16T08:31:56.501835+00:00"} {"global_step": 9551, "acc_step": 0, "speed/wps": 12894.134454615916, "speed/FLOPS": 202520100007238.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05363837629556656, "optim/lr": 0.00294962534857172, "optim/total_tokens": 5007474688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.896312952041626, "created_at": "2025-01-16T08:32:06.671686+00:00"} {"global_step": 9552, "acc_step": 0, "speed/wps": 12898.69550963164, "speed/FLOPS": 202591737643807.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04711560159921646, "optim/lr": 0.0029496033282803958, "optim/total_tokens": 5007998976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.0783495903015137, "created_at": "2025-01-16T08:32:16.843881+00:00"} {"global_step": 9553, "acc_step": 0, "speed/wps": 12906.382109170476, "speed/FLOPS": 202712466252059.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04535273090004921, "optim/lr": 0.002949581303259485, "optim/total_tokens": 5008523264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 3.0077433586120605, "created_at": "2025-01-16T08:32:27.005238+00:00"} {"global_step": 9554, "acc_step": 0, "speed/wps": 12901.076012430873, "speed/FLOPS": 202629126711422.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04266154766082764, "optim/lr": 0.0029495592735090614, "optim/total_tokens": 5009047552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503532, "loss/out": 2.903665065765381, "created_at": "2025-01-16T08:32:37.169050+00:00"} {"global_step": 9555, "acc_step": 0, "speed/wps": 12905.302921919689, "speed/FLOPS": 202695516133327.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045328106731176376, "optim/lr": 0.002949537239029195, "optim/total_tokens": 5009571840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.878864288330078, "created_at": "2025-01-16T08:32:47.332399+00:00"} {"global_step": 9556, "acc_step": 0, "speed/wps": 12898.475425495444, "speed/FLOPS": 202588280920021.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047640420496463776, "optim/lr": 0.0029495151998199594, "optim/total_tokens": 5010096128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.0918407440185547, "created_at": "2025-01-16T08:32:57.497878+00:00"} {"global_step": 9557, "acc_step": 0, "speed/wps": 12908.039453785685, "speed/FLOPS": 202738497126671.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06251634657382965, "optim/lr": 0.002949493155881425, "optim/total_tokens": 5010620416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.0137603282928467, "created_at": "2025-01-16T08:33:07.659183+00:00"} {"global_step": 9558, "acc_step": 0, "speed/wps": 12901.822894889743, "speed/FLOPS": 202640857526763.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05712471529841423, "optim/lr": 0.0029494711072136645, "optim/total_tokens": 5011144704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 2.9062860012054443, "created_at": "2025-01-16T08:33:17.821878+00:00"} {"global_step": 9559, "acc_step": 0, "speed/wps": 12904.01586068101, "speed/FLOPS": 202675301067966.1, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05484883859753609, "optim/lr": 0.0029494490538167504, "optim/total_tokens": 5011668992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9001259803771973, "created_at": "2025-01-16T08:33:27.985159+00:00"} {"global_step": 9560, "acc_step": 0, "speed/wps": 12907.299098354444, "speed/FLOPS": 202726868827268.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06020580977201462, "optim/lr": 0.002949426995690753, "optim/total_tokens": 5012193280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.902482509613037, "created_at": "2025-01-16T08:33:38.147735+00:00"} {"global_step": 9561, "acc_step": 0, "speed/wps": 12904.43319737759, "speed/FLOPS": 202681855914266.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051267657428979874, "optim/lr": 0.0029494049328357462, "optim/total_tokens": 5012717568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352528, "loss/out": 2.760739803314209, "created_at": "2025-01-16T08:33:48.310439+00:00"} {"global_step": 9562, "acc_step": 0, "speed/wps": 12914.916949955916, "speed/FLOPS": 202846517654694.94, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05057623237371445, "optim/lr": 0.0029493828652518007, "optim/total_tokens": 5013241856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8282411098480225, "created_at": "2025-01-16T08:33:58.463433+00:00"} {"global_step": 9563, "acc_step": 0, "speed/wps": 12905.233579393647, "speed/FLOPS": 202694427013668.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06799153238534927, "optim/lr": 0.002949360792938989, "optim/total_tokens": 5013766144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 2.958369731903076, "created_at": "2025-01-16T08:34:08.623429+00:00"} {"global_step": 9564, "acc_step": 0, "speed/wps": 12907.748587242892, "speed/FLOPS": 202733928667924.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05664179101586342, "optim/lr": 0.0029493387158973827, "optim/total_tokens": 5014290432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.91933012008667, "created_at": "2025-01-16T08:34:18.783391+00:00"} {"global_step": 9565, "acc_step": 0, "speed/wps": 12908.372963770034, "speed/FLOPS": 202743735359266.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0721588060259819, "optim/lr": 0.002949316634127054, "optim/total_tokens": 5014814720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473062, "loss/out": 2.988433361053467, "created_at": "2025-01-16T08:34:28.943969+00:00"} {"global_step": 9566, "acc_step": 0, "speed/wps": 12906.066768559238, "speed/FLOPS": 202707513394436.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09268852323293686, "optim/lr": 0.0029492945476280756, "optim/total_tokens": 5015339008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8107306957244873, "created_at": "2025-01-16T08:34:39.104558+00:00"} {"global_step": 9567, "acc_step": 0, "speed/wps": 12905.669620029432, "speed/FLOPS": 202701275631036.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.059087060391902924, "optim/lr": 0.0029492724564005183, "optim/total_tokens": 5015863296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.886373281478882, "created_at": "2025-01-16T08:34:49.265687+00:00"} {"global_step": 9568, "acc_step": 0, "speed/wps": 12909.909130546273, "speed/FLOPS": 202767862969401.5, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09255284070968628, "optim/lr": 0.002949250360444455, "optim/total_tokens": 5016387584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 2.802337646484375, "created_at": "2025-01-16T08:34:59.422203+00:00"} {"global_step": 9569, "acc_step": 0, "speed/wps": 12904.07243681818, "speed/FLOPS": 202676189673944.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07537536323070526, "optim/lr": 0.0029492282597599583, "optim/total_tokens": 5016911872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8619017601013184, "created_at": "2025-01-16T08:35:09.587671+00:00"} {"global_step": 9570, "acc_step": 0, "speed/wps": 12901.105631002121, "speed/FLOPS": 202629591911784.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07027819752693176, "optim/lr": 0.002949206154347099, "optim/total_tokens": 5017436160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 2.8022403717041016, "created_at": "2025-01-16T08:35:19.754261+00:00"} {"global_step": 9571, "acc_step": 0, "speed/wps": 12908.569415964514, "speed/FLOPS": 202746820910934.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06760668009519577, "optim/lr": 0.00294918404420595, "optim/total_tokens": 5017960448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9043030738830566, "created_at": "2025-01-16T08:35:29.915141+00:00"} {"global_step": 9572, "acc_step": 0, "speed/wps": 12904.086793307828, "speed/FLOPS": 202676415162341.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06626240164041519, "optim/lr": 0.0029491619293365834, "optim/total_tokens": 5018484736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.931077003479004, "created_at": "2025-01-16T08:35:40.076045+00:00"} {"global_step": 9573, "acc_step": 0, "speed/wps": 12909.416144364175, "speed/FLOPS": 202760119943977.66, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07413709908723831, "optim/lr": 0.0029491398097390717, "optim/total_tokens": 5019009024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.018233060836792, "created_at": "2025-01-16T08:35:50.236939+00:00"} {"global_step": 9574, "acc_step": 0, "speed/wps": 12908.45044071271, "speed/FLOPS": 202744952241115.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05759473517537117, "optim/lr": 0.002949117685413486, "optim/total_tokens": 5019533312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9339981079101562, "created_at": "2025-01-16T08:36:00.395630+00:00"} {"global_step": 9575, "acc_step": 0, "speed/wps": 12913.165200016148, "speed/FLOPS": 202819004014734.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057937465608119965, "optim/lr": 0.0029490955563599, "optim/total_tokens": 5020057600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420607, "loss/out": 2.9401354789733887, "created_at": "2025-01-16T08:36:10.549586+00:00"} {"global_step": 9576, "acc_step": 0, "speed/wps": 12908.586597986903, "speed/FLOPS": 202747090778206.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05414780601859093, "optim/lr": 0.0029490734225783846, "optim/total_tokens": 5020581888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9327406883239746, "created_at": "2025-01-16T08:36:20.707086+00:00"} {"global_step": 9577, "acc_step": 0, "speed/wps": 12905.552914793256, "speed/FLOPS": 202699442614932.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04460800066590309, "optim/lr": 0.0029490512840690125, "optim/total_tokens": 5021106176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 2.8597359657287598, "created_at": "2025-01-16T08:36:30.870987+00:00"} {"global_step": 9578, "acc_step": 0, "speed/wps": 12895.803133689868, "speed/FLOPS": 202546308905100.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04005718231201172, "optim/lr": 0.002949029140831856, "optim/total_tokens": 5021630464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8677573204040527, "created_at": "2025-01-16T08:36:41.038475+00:00"} {"global_step": 9579, "acc_step": 0, "speed/wps": 12903.606927146337, "speed/FLOPS": 202668878204870.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05164359137415886, "optim/lr": 0.0029490069928669867, "optim/total_tokens": 5022154752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.971275806427002, "created_at": "2025-01-16T08:36:51.209681+00:00"} {"global_step": 9580, "acc_step": 0, "speed/wps": 12909.867831792324, "speed/FLOPS": 202767214315720.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0461769700050354, "optim/lr": 0.002948984840174478, "optim/total_tokens": 5022679040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379944, "loss/out": 2.8901896476745605, "created_at": "2025-01-16T08:37:01.371980+00:00"} {"global_step": 9581, "acc_step": 0, "speed/wps": 12907.979098252423, "speed/FLOPS": 202737549160084.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06437084823846817, "optim/lr": 0.0029489626827544015, "optim/total_tokens": 5023203328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 2.8360915184020996, "created_at": "2025-01-16T08:37:11.533871+00:00"} {"global_step": 9582, "acc_step": 0, "speed/wps": 12908.771591781875, "speed/FLOPS": 202749996359964.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05271391570568085, "optim/lr": 0.0029489405206068295, "optim/total_tokens": 5023727616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.919290542602539, "created_at": "2025-01-16T08:37:21.694254+00:00"} {"global_step": 9583, "acc_step": 0, "speed/wps": 12908.815750060337, "speed/FLOPS": 202750689926407.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06716218590736389, "optim/lr": 0.002948918353731834, "optim/total_tokens": 5024251904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.959588050842285, "created_at": "2025-01-16T08:37:31.860018+00:00"} {"global_step": 9584, "acc_step": 0, "speed/wps": 12904.98529223397, "speed/FLOPS": 202690527322643.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08766050636768341, "optim/lr": 0.0029488961821294885, "optim/total_tokens": 5024776192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.940039873123169, "created_at": "2025-01-16T08:37:42.022853+00:00"} {"global_step": 9585, "acc_step": 0, "speed/wps": 12899.12183495603, "speed/FLOPS": 202598433668859.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04154304042458534, "optim/lr": 0.002948874005799864, "optim/total_tokens": 5025300480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384815, "loss/out": 2.9408631324768066, "created_at": "2025-01-16T08:37:52.188621+00:00"} {"global_step": 9586, "acc_step": 0, "speed/wps": 12906.505207859032, "speed/FLOPS": 202714399686118.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07982035726308823, "optim/lr": 0.0029488518247430336, "optim/total_tokens": 5025824768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.954237699508667, "created_at": "2025-01-16T08:38:02.348606+00:00"} {"global_step": 9587, "acc_step": 0, "speed/wps": 12909.361571090252, "speed/FLOPS": 202759262795719.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06952980160713196, "optim/lr": 0.0029488296389590693, "optim/total_tokens": 5026349056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.0261006355285645, "created_at": "2025-01-16T08:38:12.505535+00:00"} {"global_step": 9588, "acc_step": 0, "speed/wps": 12904.501500700926, "speed/FLOPS": 202682928711818.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10469648241996765, "optim/lr": 0.0029488074484480434, "optim/total_tokens": 5026873344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.8683581352233887, "created_at": "2025-01-16T08:38:22.668603+00:00"} {"global_step": 9589, "acc_step": 0, "speed/wps": 12902.309042891435, "speed/FLOPS": 202648493149167.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.05859960988163948, "optim/lr": 0.0029487852532100287, "optim/total_tokens": 5027397632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8764138221740723, "created_at": "2025-01-16T08:38:32.831819+00:00"} {"global_step": 9590, "acc_step": 0, "speed/wps": 12906.298200681811, "speed/FLOPS": 202711148353942.53, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07419035583734512, "optim/lr": 0.0029487630532450976, "optim/total_tokens": 5027921920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.0214223861694336, "created_at": "2025-01-16T08:38:42.991429+00:00"} {"global_step": 9591, "acc_step": 0, "speed/wps": 12908.5034647497, "speed/FLOPS": 202745785056479.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05016028508543968, "optim/lr": 0.0029487408485533227, "optim/total_tokens": 5028446208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.901543140411377, "created_at": "2025-01-16T08:38:53.149739+00:00"} {"global_step": 9592, "acc_step": 0, "speed/wps": 12903.151388184187, "speed/FLOPS": 202661723339493.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055223166942596436, "optim/lr": 0.002948718639134776, "optim/total_tokens": 5028970496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.916447162628174, "created_at": "2025-01-16T08:39:03.314058+00:00"} {"global_step": 9593, "acc_step": 0, "speed/wps": 12906.821091892341, "speed/FLOPS": 202719361078931.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046791139990091324, "optim/lr": 0.00294869642498953, "optim/total_tokens": 5029494784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.87245512008667, "created_at": "2025-01-16T08:39:13.472925+00:00"} {"global_step": 9594, "acc_step": 0, "speed/wps": 12908.71476959756, "speed/FLOPS": 202749103889478.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05354221165180206, "optim/lr": 0.0029486742061176573, "optim/total_tokens": 5030019072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.023027181625366, "created_at": "2025-01-16T08:39:23.630664+00:00"} {"global_step": 9595, "acc_step": 0, "speed/wps": 12906.298293000196, "speed/FLOPS": 202711149803929.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05570722371339798, "optim/lr": 0.0029486519825192305, "optim/total_tokens": 5030543360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 2.9042296409606934, "created_at": "2025-01-16T08:39:33.797090+00:00"} {"global_step": 9596, "acc_step": 0, "speed/wps": 12901.346389087441, "speed/FLOPS": 202633373348350.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055126819759607315, "optim/lr": 0.0029486297541943215, "optim/total_tokens": 5031067648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.936610698699951, "created_at": "2025-01-16T08:39:43.960386+00:00"} {"global_step": 9597, "acc_step": 0, "speed/wps": 12902.416337790315, "speed/FLOPS": 202650178362992.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04550488293170929, "optim/lr": 0.002948607521143004, "optim/total_tokens": 5031591936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415820, "loss/out": 2.8191051483154297, "created_at": "2025-01-16T08:39:54.126099+00:00"} {"global_step": 9598, "acc_step": 0, "speed/wps": 12906.041251776445, "speed/FLOPS": 202707112618298.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05183397978544235, "optim/lr": 0.0029485852833653495, "optim/total_tokens": 5032116224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.901644706726074, "created_at": "2025-01-16T08:40:04.285532+00:00"} {"global_step": 9599, "acc_step": 0, "speed/wps": 12896.948394661935, "speed/FLOPS": 202564296802420.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042835742235183716, "optim/lr": 0.002948563040861431, "optim/total_tokens": 5032640512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.9591503143310547, "created_at": "2025-01-16T08:40:14.452869+00:00"} {"global_step": 9600, "acc_step": 0, "speed/wps": 12906.73810212603, "speed/FLOPS": 202718057610611.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05610137805342674, "optim/lr": 0.0029485407936313207, "optim/total_tokens": 5033164800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9744222164154053, "created_at": "2025-01-16T08:40:24.620857+00:00"} {"global_step": 9601, "acc_step": 0, "speed/wps": 12910.996671064213, "speed/FLOPS": 202784944287673.03, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059588152915239334, "optim/lr": 0.0029485185416750917, "optim/total_tokens": 5033689088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.082592487335205, "created_at": "2025-01-16T08:40:34.780383+00:00"} {"global_step": 9602, "acc_step": 0, "speed/wps": 12905.564676938064, "speed/FLOPS": 202699627355579.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06032456085085869, "optim/lr": 0.0029484962849928163, "optim/total_tokens": 5034213376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8311853408813477, "created_at": "2025-01-16T08:40:44.941199+00:00"} {"global_step": 9603, "acc_step": 0, "speed/wps": 12903.542856828717, "speed/FLOPS": 202667871892490.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04270846024155617, "optim/lr": 0.002948474023584567, "optim/total_tokens": 5034737664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.0020885467529297, "created_at": "2025-01-16T08:40:55.102729+00:00"} {"global_step": 9604, "acc_step": 0, "speed/wps": 12909.684383917238, "speed/FLOPS": 202764333014758.28, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06823945790529251, "optim/lr": 0.002948451757450417, "optim/total_tokens": 5035261952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.871514320373535, "created_at": "2025-01-16T08:41:05.261565+00:00"} {"global_step": 9605, "acc_step": 0, "speed/wps": 12902.783762616333, "speed/FLOPS": 202655949274780.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052387580275535583, "optim/lr": 0.002948429486590438, "optim/total_tokens": 5035786240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 2.8568344116210938, "created_at": "2025-01-16T08:41:15.423623+00:00"} {"global_step": 9606, "acc_step": 0, "speed/wps": 12906.907310622511, "speed/FLOPS": 202720715262566.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05356134474277496, "optim/lr": 0.0029484072110047043, "optim/total_tokens": 5036310528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 2.929077386856079, "created_at": "2025-01-16T08:41:25.582453+00:00"} {"global_step": 9607, "acc_step": 0, "speed/wps": 12911.033354178951, "speed/FLOPS": 202785520446397.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05385042726993561, "optim/lr": 0.0029483849306932864, "optim/total_tokens": 5036834816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372547, "loss/out": 2.963764190673828, "created_at": "2025-01-16T08:41:35.742624+00:00"} {"global_step": 9608, "acc_step": 0, "speed/wps": 12905.241789646165, "speed/FLOPS": 202694555966966.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053224582225084305, "optim/lr": 0.0029483626456562584, "optim/total_tokens": 5037359104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.932774305343628, "created_at": "2025-01-16T08:41:45.903993+00:00"} {"global_step": 9609, "acc_step": 0, "speed/wps": 12907.313959201896, "speed/FLOPS": 202727102237298.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04509562999010086, "optim/lr": 0.0029483403558936932, "optim/total_tokens": 5037883392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.9802894592285156, "created_at": "2025-01-16T08:41:56.066417+00:00"} {"global_step": 9610, "acc_step": 0, "speed/wps": 12910.069843406853, "speed/FLOPS": 202770387185719.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06029792129993439, "optim/lr": 0.0029483180614056624, "optim/total_tokens": 5038407680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9481253623962402, "created_at": "2025-01-16T08:42:06.223095+00:00"} {"global_step": 9611, "acc_step": 0, "speed/wps": 12910.108012248216, "speed/FLOPS": 202770986679822.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05963466688990593, "optim/lr": 0.0029482957621922395, "optim/total_tokens": 5038931968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.918722152709961, "created_at": "2025-01-16T08:42:16.380475+00:00"} {"global_step": 9612, "acc_step": 0, "speed/wps": 12913.87950885554, "speed/FLOPS": 202830223216620.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047143176198005676, "optim/lr": 0.0029482734582534976, "optim/total_tokens": 5039456256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.012040138244629, "created_at": "2025-01-16T08:42:26.533950+00:00"} {"global_step": 9613, "acc_step": 0, "speed/wps": 12905.175541087568, "speed/FLOPS": 202693515442319.0, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05459636077284813, "optim/lr": 0.002948251149589508, "optim/total_tokens": 5039980544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.995119094848633, "created_at": "2025-01-16T08:42:36.694316+00:00"} {"global_step": 9614, "acc_step": 0, "speed/wps": 12906.687430280532, "speed/FLOPS": 202717261739647.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.042124971747398376, "optim/lr": 0.0029482288362003457, "optim/total_tokens": 5040504832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394401, "loss/out": 2.9302258491516113, "created_at": "2025-01-16T08:42:46.856502+00:00"} {"global_step": 9615, "acc_step": 0, "speed/wps": 12906.808599504657, "speed/FLOPS": 202719164868816.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05711023509502411, "optim/lr": 0.0029482065180860815, "optim/total_tokens": 5041029120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.932849884033203, "created_at": "2025-01-16T08:42:57.016278+00:00"} {"global_step": 9616, "acc_step": 0, "speed/wps": 12904.48565241018, "speed/FLOPS": 202682679792634.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05321662873029709, "optim/lr": 0.002948184195246789, "optim/total_tokens": 5041553408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 2.919921875, "created_at": "2025-01-16T08:43:07.179859+00:00"} {"global_step": 9617, "acc_step": 0, "speed/wps": 12905.33539649369, "speed/FLOPS": 202696026191136.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05556798726320267, "optim/lr": 0.002948161867682541, "optim/total_tokens": 5042077696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9794511795043945, "created_at": "2025-01-16T08:43:17.341707+00:00"} {"global_step": 9618, "acc_step": 0, "speed/wps": 12908.581067577654, "speed/FLOPS": 202747003915529.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07772016525268555, "optim/lr": 0.002948139535393411, "optim/total_tokens": 5042601984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.813605785369873, "created_at": "2025-01-16T08:43:27.501376+00:00"} {"global_step": 9619, "acc_step": 0, "speed/wps": 12906.45695119838, "speed/FLOPS": 202713641748951.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052472446113824844, "optim/lr": 0.0029481171983794705, "optim/total_tokens": 5043126272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 2.979584217071533, "created_at": "2025-01-16T08:43:37.662762+00:00"} {"global_step": 9620, "acc_step": 0, "speed/wps": 12905.335179932876, "speed/FLOPS": 202696022789751.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05269298329949379, "optim/lr": 0.0029480948566407935, "optim/total_tokens": 5043650560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.021329402923584, "created_at": "2025-01-16T08:43:47.826446+00:00"} {"global_step": 9621, "acc_step": 0, "speed/wps": 12912.519673553119, "speed/FLOPS": 202808865134586.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05621141940355301, "optim/lr": 0.0029480725101774523, "optim/total_tokens": 5044174848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.7901666164398193, "created_at": "2025-01-16T08:43:57.983376+00:00"} {"global_step": 9622, "acc_step": 0, "speed/wps": 12909.004102511608, "speed/FLOPS": 202753648260478.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05416141077876091, "optim/lr": 0.0029480501589895206, "optim/total_tokens": 5044699136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9795660972595215, "created_at": "2025-01-16T08:44:08.140969+00:00"} {"global_step": 9623, "acc_step": 0, "speed/wps": 12907.30443352969, "speed/FLOPS": 202726952623527.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04784135892987251, "optim/lr": 0.00294802780307707, "optim/total_tokens": 5045223424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.798943519592285, "created_at": "2025-01-16T08:44:18.307293+00:00"} {"global_step": 9624, "acc_step": 0, "speed/wps": 12910.489316322188, "speed/FLOPS": 202776975584271.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054042138159275055, "optim/lr": 0.002948005442440175, "optim/total_tokens": 5045747712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.8821334838867188, "created_at": "2025-01-16T08:44:28.465876+00:00"} {"global_step": 9625, "acc_step": 0, "speed/wps": 12909.625237441578, "speed/FLOPS": 202763404038082.25, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04681653529405594, "optim/lr": 0.002947983077078907, "optim/total_tokens": 5046272000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.8853495121002197, "created_at": "2025-01-16T08:44:38.622893+00:00"} {"global_step": 9626, "acc_step": 0, "speed/wps": 12904.608919628958, "speed/FLOPS": 202684615873692.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05287392437458038, "optim/lr": 0.00294796070699334, "optim/total_tokens": 5046796288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366026, "loss/out": 2.832033157348633, "created_at": "2025-01-16T08:44:48.783430+00:00"} {"global_step": 9627, "acc_step": 0, "speed/wps": 12907.375389293971, "speed/FLOPS": 202728067081308.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042958639562129974, "optim/lr": 0.002947938332183547, "optim/total_tokens": 5047320576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.0027363300323486, "created_at": "2025-01-16T08:44:58.945120+00:00"} {"global_step": 9628, "acc_step": 0, "speed/wps": 12907.910301138685, "speed/FLOPS": 202736468606875.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050576068460941315, "optim/lr": 0.0029479159526496, "optim/total_tokens": 5047844864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 2.8007869720458984, "created_at": "2025-01-16T08:45:09.103318+00:00"} {"global_step": 9629, "acc_step": 0, "speed/wps": 12907.250107754626, "speed/FLOPS": 202726099362578.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04493757337331772, "optim/lr": 0.0029478935683915734, "optim/total_tokens": 5048369152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.009345054626465, "created_at": "2025-01-16T08:45:19.264187+00:00"} {"global_step": 9630, "acc_step": 0, "speed/wps": 12906.291326516519, "speed/FLOPS": 202711040385730.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055303074419498444, "optim/lr": 0.0029478711794095392, "optim/total_tokens": 5048893440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9424288272857666, "created_at": "2025-01-16T08:45:29.423701+00:00"} {"global_step": 9631, "acc_step": 0, "speed/wps": 12909.138905025948, "speed/FLOPS": 202755765519204.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05261462181806564, "optim/lr": 0.0029478487857035702, "optim/total_tokens": 5049417728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8266212940216064, "created_at": "2025-01-16T08:45:39.583415+00:00"} {"global_step": 9632, "acc_step": 0, "speed/wps": 12908.020731801385, "speed/FLOPS": 202738203072181.25, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047189630568027496, "optim/lr": 0.002947826387273741, "optim/total_tokens": 5049942016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 3.15556001663208, "created_at": "2025-01-16T08:45:49.742796+00:00"} {"global_step": 9633, "acc_step": 0, "speed/wps": 12910.521587240592, "speed/FLOPS": 202777482443389.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05414946377277374, "optim/lr": 0.0029478039841201236, "optim/total_tokens": 5050466304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 2.950427293777466, "created_at": "2025-01-16T08:45:59.901742+00:00"} {"global_step": 9634, "acc_step": 0, "speed/wps": 12907.744092222716, "speed/FLOPS": 202733858067455.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056142475455999374, "optim/lr": 0.002947781576242791, "optim/total_tokens": 5050990592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.9895989894866943, "created_at": "2025-01-16T08:46:10.059822+00:00"} {"global_step": 9635, "acc_step": 0, "speed/wps": 12903.685210542764, "speed/FLOPS": 202670107753184.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06900139898061752, "optim/lr": 0.0029477591636418166, "optim/total_tokens": 5051514880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.064480781555176, "created_at": "2025-01-16T08:46:20.225079+00:00"} {"global_step": 9636, "acc_step": 0, "speed/wps": 12900.009834432833, "speed/FLOPS": 202612380920881.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056119706481695175, "optim/lr": 0.0029477367463172735, "optim/total_tokens": 5052039168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8374338150024414, "created_at": "2025-01-16T08:46:30.390225+00:00"} {"global_step": 9637, "acc_step": 0, "speed/wps": 12905.51022616287, "speed/FLOPS": 202698772131332.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04966943711042404, "optim/lr": 0.002947714324269235, "optim/total_tokens": 5052563456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 2.813570261001587, "created_at": "2025-01-16T08:46:40.549978+00:00"} {"global_step": 9638, "acc_step": 0, "speed/wps": 12908.417324145968, "speed/FLOPS": 202744432099927.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06463980674743652, "optim/lr": 0.0029476918974977737, "optim/total_tokens": 5053087744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 2.935304641723633, "created_at": "2025-01-16T08:46:50.708431+00:00"} {"global_step": 9639, "acc_step": 0, "speed/wps": 12910.524199903479, "speed/FLOPS": 202777523478850.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05309507995843887, "optim/lr": 0.002947669466002963, "optim/total_tokens": 5053612032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.827037811279297, "created_at": "2025-01-16T08:47:00.867276+00:00"} {"global_step": 9640, "acc_step": 0, "speed/wps": 12887.917682763016, "speed/FLOPS": 202422457062549.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07100599259138107, "optim/lr": 0.002947647029784877, "optim/total_tokens": 5054136320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.886770248413086, "created_at": "2025-01-16T08:47:11.042043+00:00"} {"global_step": 9641, "acc_step": 0, "speed/wps": 12901.577868382103, "speed/FLOPS": 202637009048759.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055988237261772156, "optim/lr": 0.0029476245888435874, "optim/total_tokens": 5054660608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351231, "loss/out": 2.7849907875061035, "created_at": "2025-01-16T08:47:21.213480+00:00"} {"global_step": 9642, "acc_step": 0, "speed/wps": 12905.405479742183, "speed/FLOPS": 202697126944858.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.056730274111032486, "optim/lr": 0.0029476021431791687, "optim/total_tokens": 5055184896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9608922004699707, "created_at": "2025-01-16T08:47:31.375258+00:00"} {"global_step": 9643, "acc_step": 0, "speed/wps": 12902.475929226972, "speed/FLOPS": 202651114328392.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05756896734237671, "optim/lr": 0.0029475796927916935, "optim/total_tokens": 5055709184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460321, "loss/out": 2.8586926460266113, "created_at": "2025-01-16T08:47:41.539832+00:00"} {"global_step": 9644, "acc_step": 0, "speed/wps": 12897.17062856094, "speed/FLOPS": 202567787291185.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04864883050322533, "optim/lr": 0.0029475572376812345, "optim/total_tokens": 5056233472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9193201065063477, "created_at": "2025-01-16T08:47:51.715568+00:00"} {"global_step": 9645, "acc_step": 0, "speed/wps": 12908.437284520272, "speed/FLOPS": 202744745605034.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06498183310031891, "optim/lr": 0.002947534777847866, "optim/total_tokens": 5056757760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.8702211380004883, "created_at": "2025-01-16T08:48:01.877279+00:00"} {"global_step": 9646, "acc_step": 0, "speed/wps": 12908.750956666867, "speed/FLOPS": 202749672257127.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0456637442111969, "optim/lr": 0.0029475123132916606, "optim/total_tokens": 5057282048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.8785831928253174, "created_at": "2025-01-16T08:48:12.037047+00:00"} {"global_step": 9647, "acc_step": 0, "speed/wps": 12904.649422719347, "speed/FLOPS": 202685252030383.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.061990540474653244, "optim/lr": 0.0029474898440126924, "optim/total_tokens": 5057806336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8913817405700684, "created_at": "2025-01-16T08:48:22.199095+00:00"} {"global_step": 9648, "acc_step": 0, "speed/wps": 12906.334513865537, "speed/FLOPS": 202711718702392.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043767768889665604, "optim/lr": 0.002947467370011034, "optim/total_tokens": 5058330624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.836026668548584, "created_at": "2025-01-16T08:48:32.360713+00:00"} {"global_step": 9649, "acc_step": 0, "speed/wps": 12900.16569529455, "speed/FLOPS": 202614828929890.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06014690548181534, "optim/lr": 0.002947444891286759, "optim/total_tokens": 5058854912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282855, "loss/out": 2.9413962364196777, "created_at": "2025-01-16T08:48:42.527571+00:00"} {"global_step": 9650, "acc_step": 0, "speed/wps": 12909.93798677745, "speed/FLOPS": 202768316196164.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08763949573040009, "optim/lr": 0.00294742240783994, "optim/total_tokens": 5059379200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.843358039855957, "created_at": "2025-01-16T08:48:52.686171+00:00"} {"global_step": 9651, "acc_step": 0, "speed/wps": 12907.541985286713, "speed/FLOPS": 202730683700301.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07017168402671814, "optim/lr": 0.0029473999196706516, "optim/total_tokens": 5059903488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392822, "loss/out": 2.8775014877319336, "created_at": "2025-01-16T08:49:02.847262+00:00"} {"global_step": 9652, "acc_step": 0, "speed/wps": 12909.175514827617, "speed/FLOPS": 202756340526446.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05517130717635155, "optim/lr": 0.0029473774267789665, "optim/total_tokens": 5060427776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.0300588607788086, "created_at": "2025-01-16T08:49:13.006154+00:00"} {"global_step": 9653, "acc_step": 0, "speed/wps": 12902.928325895398, "speed/FLOPS": 202658219839728.7, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05186525359749794, "optim/lr": 0.002947354929164958, "optim/total_tokens": 5060952064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.8526511192321777, "created_at": "2025-01-16T08:49:23.170123+00:00"} {"global_step": 9654, "acc_step": 0, "speed/wps": 12906.482569047614, "speed/FLOPS": 202714044112476.12, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047656651586294174, "optim/lr": 0.0029473324268286995, "optim/total_tokens": 5061476352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.9498720169067383, "created_at": "2025-01-16T08:49:33.332004+00:00"} {"global_step": 9655, "acc_step": 0, "speed/wps": 12907.909779054362, "speed/FLOPS": 202736460406824.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045874763280153275, "optim/lr": 0.002947309919770265, "optim/total_tokens": 5062000640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8287320137023926, "created_at": "2025-01-16T08:49:43.490234+00:00"} {"global_step": 9656, "acc_step": 0, "speed/wps": 12903.705510030104, "speed/FLOPS": 202670426584527.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04604357108473778, "optim/lr": 0.002947287407989727, "optim/total_tokens": 5062524928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.960123062133789, "created_at": "2025-01-16T08:49:53.653292+00:00"} {"global_step": 9657, "acc_step": 0, "speed/wps": 12909.112767473905, "speed/FLOPS": 202755354993032.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05181140825152397, "optim/lr": 0.00294726489148716, "optim/total_tokens": 5063049216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.8575310707092285, "created_at": "2025-01-16T08:50:03.812431+00:00"} {"global_step": 9658, "acc_step": 0, "speed/wps": 12899.683663775619, "speed/FLOPS": 202607257962504.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05315205454826355, "optim/lr": 0.0029472423702626366, "optim/total_tokens": 5063573504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.015362501144409, "created_at": "2025-01-16T08:50:13.979737+00:00"} {"global_step": 9659, "acc_step": 0, "speed/wps": 12904.509035574625, "speed/FLOPS": 202683047057363.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.047132283449172974, "optim/lr": 0.002947219844316231, "optim/total_tokens": 5064097792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.8362836837768555, "created_at": "2025-01-16T08:50:24.143728+00:00"} {"global_step": 9660, "acc_step": 0, "speed/wps": 12910.29267896658, "speed/FLOPS": 202773887124393.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050314418971538544, "optim/lr": 0.002947197313648016, "optim/total_tokens": 5064622080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.890493154525757, "created_at": "2025-01-16T08:50:34.300028+00:00"} {"global_step": 9661, "acc_step": 0, "speed/wps": 12906.24206424666, "speed/FLOPS": 202710266654088.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05269373208284378, "optim/lr": 0.002947174778258065, "optim/total_tokens": 5065146368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0184803009033203, "created_at": "2025-01-16T08:50:44.460703+00:00"} {"global_step": 9662, "acc_step": 0, "speed/wps": 12907.16509421455, "speed/FLOPS": 202724764108110.72, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0478789322078228, "optim/lr": 0.002947152238146453, "optim/total_tokens": 5065670656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.866318702697754, "created_at": "2025-01-16T08:50:54.619603+00:00"} {"global_step": 9663, "acc_step": 0, "speed/wps": 12903.198615761663, "speed/FLOPS": 202662465113495.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0593169741332531, "optim/lr": 0.0029471296933132516, "optim/total_tokens": 5066194944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.918138027191162, "created_at": "2025-01-16T08:51:04.784267+00:00"} {"global_step": 9664, "acc_step": 0, "speed/wps": 12906.290573767457, "speed/FLOPS": 202711028562772.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04626470059156418, "optim/lr": 0.0029471071437585355, "optim/total_tokens": 5066719232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.0272884368896484, "created_at": "2025-01-16T08:51:14.943716+00:00"} {"global_step": 9665, "acc_step": 0, "speed/wps": 12908.249688563474, "speed/FLOPS": 202741799152749.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05733415111899376, "optim/lr": 0.002947084589482378, "optim/total_tokens": 5067243520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.865713119506836, "created_at": "2025-01-16T08:51:25.104581+00:00"} {"global_step": 9666, "acc_step": 0, "speed/wps": 12903.29224265827, "speed/FLOPS": 202663935652563.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.058072280138731, "optim/lr": 0.0029470620304848523, "optim/total_tokens": 5067767808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 2.91284441947937, "created_at": "2025-01-16T08:51:35.268561+00:00"} {"global_step": 9667, "acc_step": 0, "speed/wps": 12905.316898912757, "speed/FLOPS": 202695735661209.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05623655393719673, "optim/lr": 0.0029470394667660328, "optim/total_tokens": 5068292096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.826033115386963, "created_at": "2025-01-16T08:51:45.428708+00:00"} {"global_step": 9668, "acc_step": 0, "speed/wps": 12907.834181453309, "speed/FLOPS": 202735273042617.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04281844198703766, "optim/lr": 0.0029470168983259924, "optim/total_tokens": 5068816384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.915626049041748, "created_at": "2025-01-16T08:51:55.589315+00:00"} {"global_step": 9669, "acc_step": 0, "speed/wps": 12910.074330572928, "speed/FLOPS": 202770457662828.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05025247856974602, "optim/lr": 0.0029469943251648053, "optim/total_tokens": 5069340672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.8798747062683105, "created_at": "2025-01-16T08:52:05.753651+00:00"} {"global_step": 9670, "acc_step": 0, "speed/wps": 12905.112707177641, "speed/FLOPS": 202692528549420.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044999897480010986, "optim/lr": 0.002946971747282545, "optim/total_tokens": 5069864960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.8608758449554443, "created_at": "2025-01-16T08:52:15.917661+00:00"} {"global_step": 9671, "acc_step": 0, "speed/wps": 12903.283638575711, "speed/FLOPS": 202663800513623.78, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0508589893579483, "optim/lr": 0.002946949164679285, "optim/total_tokens": 5070389248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.8022146224975586, "created_at": "2025-01-16T08:52:26.079456+00:00"} {"global_step": 9672, "acc_step": 0, "speed/wps": 12897.658447213129, "speed/FLOPS": 202575449153451.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0484338253736496, "optim/lr": 0.002946926577355099, "optim/total_tokens": 5070913536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9365787506103516, "created_at": "2025-01-16T08:52:36.246677+00:00"} {"global_step": 9673, "acc_step": 0, "speed/wps": 12902.141285595959, "speed/FLOPS": 202645858290319.28, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05001293867826462, "optim/lr": 0.002946903985310061, "optim/total_tokens": 5071437824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 2.8248519897460938, "created_at": "2025-01-16T08:52:46.413556+00:00"} {"global_step": 9674, "acc_step": 0, "speed/wps": 12909.852657166008, "speed/FLOPS": 202766975977361.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04800194501876831, "optim/lr": 0.002946881388544244, "optim/total_tokens": 5071962112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.974123001098633, "created_at": "2025-01-16T08:52:56.570253+00:00"} {"global_step": 9675, "acc_step": 0, "speed/wps": 12897.913167636783, "speed/FLOPS": 202579449887727.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048352062702178955, "optim/lr": 0.0029468587870577225, "optim/total_tokens": 5072486400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.7669782638549805, "created_at": "2025-01-16T08:53:06.742113+00:00"} {"global_step": 9676, "acc_step": 0, "speed/wps": 12901.495181383807, "speed/FLOPS": 202635710335827.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05461006984114647, "optim/lr": 0.0029468361808505693, "optim/total_tokens": 5073010688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.9720497131347656, "created_at": "2025-01-16T08:53:16.907369+00:00"} {"global_step": 9677, "acc_step": 0, "speed/wps": 12908.369284055738, "speed/FLOPS": 202743677564296.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05367210507392883, "optim/lr": 0.0029468135699228593, "optim/total_tokens": 5073534976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9200363159179688, "created_at": "2025-01-16T08:53:27.065264+00:00"} {"global_step": 9678, "acc_step": 0, "speed/wps": 12901.657236550049, "speed/FLOPS": 202638255634900.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04547593370079994, "optim/lr": 0.0029467909542746653, "optim/total_tokens": 5074059264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368524, "loss/out": 2.9080889225006104, "created_at": "2025-01-16T08:53:37.228528+00:00"} {"global_step": 9679, "acc_step": 0, "speed/wps": 12898.278223618001, "speed/FLOPS": 202585183593552.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04361364617943764, "optim/lr": 0.0029467683339060627, "optim/total_tokens": 5074583552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.903832197189331, "created_at": "2025-01-16T08:53:47.397488+00:00"} {"global_step": 9680, "acc_step": 0, "speed/wps": 12907.865051619685, "speed/FLOPS": 202735757901000.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055758822709321976, "optim/lr": 0.002946745708817123, "optim/total_tokens": 5075107840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.032505512237549, "created_at": "2025-01-16T08:53:57.557822+00:00"} {"global_step": 9681, "acc_step": 0, "speed/wps": 12909.035003139665, "speed/FLOPS": 202754133597303.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05051789432764053, "optim/lr": 0.0029467230790079213, "optim/total_tokens": 5075632128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 2.9139909744262695, "created_at": "2025-01-16T08:54:07.715335+00:00"} {"global_step": 9682, "acc_step": 0, "speed/wps": 12895.652401280526, "speed/FLOPS": 202543941445483.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052996452897787094, "optim/lr": 0.0029467004444785313, "optim/total_tokens": 5076156416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9383983612060547, "created_at": "2025-01-16T08:54:17.884126+00:00"} {"global_step": 9683, "acc_step": 0, "speed/wps": 12886.311408535148, "speed/FLOPS": 202397228318548.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08463635295629501, "optim/lr": 0.0029466778052290267, "optim/total_tokens": 5076680704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 2.9640746116638184, "created_at": "2025-01-16T08:54:28.059660+00:00"} {"global_step": 9684, "acc_step": 0, "speed/wps": 12897.064845049908, "speed/FLOPS": 202566125815783.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06021829694509506, "optim/lr": 0.0029466551612594826, "optim/total_tokens": 5077204992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.9477250576019287, "created_at": "2025-01-16T08:54:38.228435+00:00"} {"global_step": 9685, "acc_step": 0, "speed/wps": 12895.695874309125, "speed/FLOPS": 202544624249136.3, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06433403491973877, "optim/lr": 0.0029466325125699707, "optim/total_tokens": 5077729280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.0664029121398926, "created_at": "2025-01-16T08:54:48.396956+00:00"} {"global_step": 9686, "acc_step": 0, "speed/wps": 12887.63644245449, "speed/FLOPS": 202418039797039.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.048012468963861465, "optim/lr": 0.002946609859160566, "optim/total_tokens": 5078253568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.873185157775879, "created_at": "2025-01-16T08:54:58.572364+00:00"} {"global_step": 9687, "acc_step": 0, "speed/wps": 12891.218020681983, "speed/FLOPS": 202474293404706.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04820626974105835, "optim/lr": 0.0029465872010313427, "optim/total_tokens": 5078777856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8026180267333984, "created_at": "2025-01-16T08:55:08.744804+00:00"} {"global_step": 9688, "acc_step": 0, "speed/wps": 12898.820953809402, "speed/FLOPS": 202593707916997.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05415849760174751, "optim/lr": 0.002946564538182374, "optim/total_tokens": 5079302144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9147682189941406, "created_at": "2025-01-16T08:55:18.911413+00:00"} {"global_step": 9689, "acc_step": 0, "speed/wps": 12898.609882897608, "speed/FLOPS": 202590392758285.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050065670162439346, "optim/lr": 0.002946541870613735, "optim/total_tokens": 5079826432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9246463775634766, "created_at": "2025-01-16T08:55:29.080490+00:00"} {"global_step": 9690, "acc_step": 0, "speed/wps": 12894.416828024627, "speed/FLOPS": 202524535069644.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.05465902015566826, "optim/lr": 0.0029465191983254983, "optim/total_tokens": 5080350720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312903, "loss/out": 2.8294730186462402, "created_at": "2025-01-16T08:55:39.249991+00:00"} {"global_step": 9691, "acc_step": 0, "speed/wps": 12890.36265639141, "speed/FLOPS": 202460858733129.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05837375670671463, "optim/lr": 0.002946496521317739, "optim/total_tokens": 5080875008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8727645874023438, "created_at": "2025-01-16T08:55:49.424229+00:00"} {"global_step": 9692, "acc_step": 0, "speed/wps": 12894.924000175177, "speed/FLOPS": 202532500905196.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06969572603702545, "optim/lr": 0.0029464738395905297, "optim/total_tokens": 5081399296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.860931396484375, "created_at": "2025-01-16T08:55:59.598279+00:00"} {"global_step": 9693, "acc_step": 0, "speed/wps": 12895.73779850985, "speed/FLOPS": 202545282726318.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05911291018128395, "optim/lr": 0.0029464511531439455, "optim/total_tokens": 5081923584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485528, "loss/out": 2.939499855041504, "created_at": "2025-01-16T08:56:09.766207+00:00"} {"global_step": 9694, "acc_step": 0, "speed/wps": 12903.956806361917, "speed/FLOPS": 202674373538735.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07356053590774536, "optim/lr": 0.002946428461978061, "optim/total_tokens": 5082447872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333025, "loss/out": 2.7892565727233887, "created_at": "2025-01-16T08:56:19.933789+00:00"} {"global_step": 9695, "acc_step": 0, "speed/wps": 12909.246977834526, "speed/FLOPS": 202757462951170.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05181115120649338, "optim/lr": 0.0029464057660929487, "optim/total_tokens": 5082972160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.0223259925842285, "created_at": "2025-01-16T08:56:30.091020+00:00"} {"global_step": 9696, "acc_step": 0, "speed/wps": 12901.194507636625, "speed/FLOPS": 202630987841458.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060154713690280914, "optim/lr": 0.002946383065488684, "optim/total_tokens": 5083496448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.9096083641052246, "created_at": "2025-01-16T08:56:40.256422+00:00"} {"global_step": 9697, "acc_step": 0, "speed/wps": 12903.934504218254, "speed/FLOPS": 202674023252923.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06732603907585144, "optim/lr": 0.0029463603601653393, "optim/total_tokens": 5084020736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 2.693848133087158, "created_at": "2025-01-16T08:56:50.419840+00:00"} {"global_step": 9698, "acc_step": 0, "speed/wps": 12890.49667669353, "speed/FLOPS": 202462963706137.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044106438755989075, "optim/lr": 0.0029463376501229904, "optim/total_tokens": 5084545024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.024700164794922, "created_at": "2025-01-16T08:57:00.592232+00:00"} {"global_step": 9699, "acc_step": 0, "speed/wps": 12894.894603743336, "speed/FLOPS": 202532039193839.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06278488785028458, "optim/lr": 0.0029463149353617106, "optim/total_tokens": 5085069312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 2.9512221813201904, "created_at": "2025-01-16T08:57:10.765137+00:00"} {"global_step": 9700, "acc_step": 0, "speed/wps": 12908.395978607277, "speed/FLOPS": 202744096838910.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04786672815680504, "optim/lr": 0.0029462922158815744, "optim/total_tokens": 5085593600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.0810818672180176, "created_at": "2025-01-16T08:57:20.924124+00:00"} {"global_step": 9701, "acc_step": 0, "speed/wps": 12904.514214622155, "speed/FLOPS": 202683128401421.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0591772235929966, "optim/lr": 0.0029462694916826556, "optim/total_tokens": 5086117888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.7712161540985107, "created_at": "2025-01-16T08:57:31.088454+00:00"} {"global_step": 9702, "acc_step": 0, "speed/wps": 12904.301365938496, "speed/FLOPS": 202679785320360.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05427107214927673, "optim/lr": 0.0029462467627650282, "optim/total_tokens": 5086642176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 2.9441123008728027, "created_at": "2025-01-16T08:57:41.250147+00:00"} {"global_step": 9703, "acc_step": 0, "speed/wps": 12909.024302698514, "speed/FLOPS": 202753965532171.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04435129091143608, "optim/lr": 0.0029462240291287666, "optim/total_tokens": 5087166464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 2.8094825744628906, "created_at": "2025-01-16T08:57:51.411095+00:00"} {"global_step": 9704, "acc_step": 0, "speed/wps": 12910.827921472252, "speed/FLOPS": 202782293843445.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052466411143541336, "optim/lr": 0.0029462012907739447, "optim/total_tokens": 5087690752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.851578950881958, "created_at": "2025-01-16T08:58:01.567887+00:00"} {"global_step": 9705, "acc_step": 0, "speed/wps": 12902.182498262833, "speed/FLOPS": 202646505591884.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05932420492172241, "optim/lr": 0.0029461785477006373, "optim/total_tokens": 5088215040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.8189823627471924, "created_at": "2025-01-16T08:58:11.734435+00:00"} {"global_step": 9706, "acc_step": 0, "speed/wps": 12906.015635679105, "speed/FLOPS": 202706710282290.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04831249266862869, "optim/lr": 0.002946155799908918, "optim/total_tokens": 5088739328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9817826747894287, "created_at": "2025-01-16T08:58:21.894719+00:00"} {"global_step": 9707, "acc_step": 0, "speed/wps": 12906.410119940256, "speed/FLOPS": 202712906199689.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06088307872414589, "optim/lr": 0.002946133047398861, "optim/total_tokens": 5089263616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.844287157058716, "created_at": "2025-01-16T08:58:32.056912+00:00"} {"global_step": 9708, "acc_step": 0, "speed/wps": 12905.993674371122, "speed/FLOPS": 202706365349770.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06281030178070068, "optim/lr": 0.002946110290170541, "optim/total_tokens": 5089787904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.9579126834869385, "created_at": "2025-01-16T08:58:42.216535+00:00"} {"global_step": 9709, "acc_step": 0, "speed/wps": 12903.628512394595, "speed/FLOPS": 202669217230855.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058380935341119766, "optim/lr": 0.0029460875282240322, "optim/total_tokens": 5090312192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.952761650085449, "created_at": "2025-01-16T08:58:52.379077+00:00"} {"global_step": 9710, "acc_step": 0, "speed/wps": 12904.486377734269, "speed/FLOPS": 202682691184845.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05183280259370804, "optim/lr": 0.002946064761559408, "optim/total_tokens": 5090836480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.856717109680176, "created_at": "2025-01-16T08:59:02.539784+00:00"} {"global_step": 9711, "acc_step": 0, "speed/wps": 12906.217593032577, "speed/FLOPS": 202709882300044.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.046597737818956375, "optim/lr": 0.002946041990176744, "optim/total_tokens": 5091360768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 2.9517154693603516, "created_at": "2025-01-16T08:59:12.700497+00:00"} {"global_step": 9712, "acc_step": 0, "speed/wps": 12901.435370749434, "speed/FLOPS": 202634770927625.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06509177386760712, "optim/lr": 0.0029460192140761134, "optim/total_tokens": 5091885056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8664774894714355, "created_at": "2025-01-16T08:59:22.864030+00:00"} {"global_step": 9713, "acc_step": 0, "speed/wps": 12905.241878399313, "speed/FLOPS": 202694557360956.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0534827820956707, "optim/lr": 0.002945996433257592, "optim/total_tokens": 5092409344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.946310043334961, "created_at": "2025-01-16T08:59:33.024212+00:00"} {"global_step": 9714, "acc_step": 0, "speed/wps": 12908.403951357215, "speed/FLOPS": 202744222061904.4, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06404575705528259, "optim/lr": 0.0029459736477212522, "optim/total_tokens": 5092933632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 2.857276439666748, "created_at": "2025-01-16T08:59:43.182192+00:00"} {"global_step": 9715, "acc_step": 0, "speed/wps": 12903.048177694116, "speed/FLOPS": 202660102276921.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06930922716856003, "optim/lr": 0.002945950857467169, "optim/total_tokens": 5093457920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.8527820110321045, "created_at": "2025-01-16T08:59:53.350779+00:00"} {"global_step": 9716, "acc_step": 0, "speed/wps": 12891.258632229474, "speed/FLOPS": 202474931264865.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06731387972831726, "optim/lr": 0.0029459280624954177, "optim/total_tokens": 5093982208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 2.913158893585205, "created_at": "2025-01-16T09:00:03.523383+00:00"} {"global_step": 9717, "acc_step": 0, "speed/wps": 12900.627410297715, "speed/FLOPS": 202622080798477.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053800661116838455, "optim/lr": 0.0029459052628060716, "optim/total_tokens": 5094506496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498399, "loss/out": 2.9315030574798584, "created_at": "2025-01-16T09:00:13.689474+00:00"} {"global_step": 9718, "acc_step": 0, "speed/wps": 12905.676022517395, "speed/FLOPS": 202701376190907.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04841892421245575, "optim/lr": 0.0029458824583992052, "optim/total_tokens": 5095030784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8458447456359863, "created_at": "2025-01-16T09:00:23.852479+00:00"} {"global_step": 9719, "acc_step": 0, "speed/wps": 12906.348990196098, "speed/FLOPS": 202711946073056.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06156359985470772, "optim/lr": 0.0029458596492748937, "optim/total_tokens": 5095555072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308039, "loss/out": 2.8801965713500977, "created_at": "2025-01-16T09:00:34.013034+00:00"} {"global_step": 9720, "acc_step": 0, "speed/wps": 12907.625007883684, "speed/FLOPS": 202731987684272.1, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04239026457071304, "optim/lr": 0.002945836835433211, "optim/total_tokens": 5096079360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8653628826141357, "created_at": "2025-01-16T09:00:44.173916+00:00"} {"global_step": 9721, "acc_step": 0, "speed/wps": 12909.76764841801, "speed/FLOPS": 202765640798160.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05183924362063408, "optim/lr": 0.0029458140168742313, "optim/total_tokens": 5096603648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.870837450027466, "created_at": "2025-01-16T09:00:54.332067+00:00"} {"global_step": 9722, "acc_step": 0, "speed/wps": 12911.334041343078, "speed/FLOPS": 202790243151497.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05023466795682907, "optim/lr": 0.0029457911935980292, "optim/total_tokens": 5097127936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.0283596515655518, "created_at": "2025-01-16T09:01:04.487604+00:00"} {"global_step": 9723, "acc_step": 0, "speed/wps": 12903.806375459857, "speed/FLOPS": 202672010814705.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04953667148947716, "optim/lr": 0.002945768365604679, "optim/total_tokens": 5097652224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.0463719367980957, "created_at": "2025-01-16T09:01:14.650619+00:00"} {"global_step": 9724, "acc_step": 0, "speed/wps": 12903.49281579303, "speed/FLOPS": 202667085929261.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04823431000113487, "optim/lr": 0.0029457455328942557, "optim/total_tokens": 5098176512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.973616600036621, "created_at": "2025-01-16T09:01:24.812054+00:00"} {"global_step": 9725, "acc_step": 0, "speed/wps": 12907.526996064382, "speed/FLOPS": 202730448273967.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05570568889379501, "optim/lr": 0.002945722695466834, "optim/total_tokens": 5098700800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 2.847970724105835, "created_at": "2025-01-16T09:01:34.971551+00:00"} {"global_step": 9726, "acc_step": 0, "speed/wps": 12905.172434760081, "speed/FLOPS": 202693466653177.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06112760305404663, "optim/lr": 0.0029456998533224873, "optim/total_tokens": 5099225088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 2.9615285396575928, "created_at": "2025-01-16T09:01:45.131633+00:00"} {"global_step": 9727, "acc_step": 0, "speed/wps": 12904.76173345111, "speed/FLOPS": 202687016024757.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04957267642021179, "optim/lr": 0.0029456770064612905, "optim/total_tokens": 5099749376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.7209460735321045, "created_at": "2025-01-16T09:01:55.295617+00:00"} {"global_step": 9728, "acc_step": 0, "speed/wps": 12907.673639080149, "speed/FLOPS": 202732751504038.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05798700824379921, "optim/lr": 0.0029456541548833187, "optim/total_tokens": 5100273664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 2.8867366313934326, "created_at": "2025-01-16T09:02:05.456335+00:00"} {"global_step": 9729, "acc_step": 0, "speed/wps": 12895.12913029568, "speed/FLOPS": 202535722755617.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050629787147045135, "optim/lr": 0.0029456312985886466, "optim/total_tokens": 5100797952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.0230610370635986, "created_at": "2025-01-16T09:02:15.624739+00:00"} {"global_step": 9730, "acc_step": 0, "speed/wps": 12891.197900980364, "speed/FLOPS": 202473977397145.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05704134702682495, "optim/lr": 0.0029456084375773474, "optim/total_tokens": 5101322240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.888526439666748, "created_at": "2025-01-16T09:02:25.798382+00:00"} {"global_step": 9731, "acc_step": 0, "speed/wps": 12889.50008538109, "speed/FLOPS": 202447310870115.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062353380024433136, "optim/lr": 0.0029455855718494976, "optim/total_tokens": 5101846528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.9525911808013916, "created_at": "2025-01-16T09:02:35.970981+00:00"} {"global_step": 9732, "acc_step": 0, "speed/wps": 12878.482775529894, "speed/FLOPS": 202274268879532.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056917864829301834, "optim/lr": 0.0029455627014051697, "optim/total_tokens": 5102370816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372699, "loss/out": 2.9234986305236816, "created_at": "2025-01-16T09:02:46.154633+00:00"} {"global_step": 9733, "acc_step": 0, "speed/wps": 12889.090380981515, "speed/FLOPS": 202440875899527.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.058620523661375046, "optim/lr": 0.0029455398262444405, "optim/total_tokens": 5102895104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.832749128341675, "created_at": "2025-01-16T09:02:56.330402+00:00"} {"global_step": 9734, "acc_step": 0, "speed/wps": 12885.71473758818, "speed/FLOPS": 202387856781414.72, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0688989907503128, "optim/lr": 0.002945516946367383, "optim/total_tokens": 5103419392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.907075881958008, "created_at": "2025-01-16T09:03:06.509938+00:00"} {"global_step": 9735, "acc_step": 0, "speed/wps": 12883.78542393072, "speed/FLOPS": 202357554259269.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04916929453611374, "optim/lr": 0.0029454940617740723, "optim/total_tokens": 5103943680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9900448322296143, "created_at": "2025-01-16T09:03:16.689711+00:00"} {"global_step": 9736, "acc_step": 0, "speed/wps": 12896.5708421036, "speed/FLOPS": 202558366820680.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05138714239001274, "optim/lr": 0.0029454711724645833, "optim/total_tokens": 5104467968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.9201605319976807, "created_at": "2025-01-16T09:03:26.864840+00:00"} {"global_step": 9737, "acc_step": 0, "speed/wps": 12904.985762015302, "speed/FLOPS": 202690534701205.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047249943017959595, "optim/lr": 0.0029454482784389904, "optim/total_tokens": 5104992256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.9824628829956055, "created_at": "2025-01-16T09:03:37.027810+00:00"} {"global_step": 9738, "acc_step": 0, "speed/wps": 12900.204046814844, "speed/FLOPS": 202615431293216.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050335727632045746, "optim/lr": 0.0029454253796973684, "optim/total_tokens": 5105516544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.887216567993164, "created_at": "2025-01-16T09:03:47.192051+00:00"} {"global_step": 9739, "acc_step": 0, "speed/wps": 12900.689031005415, "speed/FLOPS": 202623048636367.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05633864924311638, "optim/lr": 0.0029454024762397926, "optim/total_tokens": 5106040832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.8867812156677246, "created_at": "2025-01-16T09:03:57.357526+00:00"} {"global_step": 9740, "acc_step": 0, "speed/wps": 12901.508211073738, "speed/FLOPS": 202635914985012.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05045728757977486, "optim/lr": 0.0029453795680663367, "optim/total_tokens": 5106565120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9619698524475098, "created_at": "2025-01-16T09:04:07.520447+00:00"} {"global_step": 9741, "acc_step": 0, "speed/wps": 12904.425827060817, "speed/FLOPS": 202681740153313.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050308242440223694, "optim/lr": 0.0029453566551770763, "optim/total_tokens": 5107089408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349594, "loss/out": 2.807250499725342, "created_at": "2025-01-16T09:04:17.682351+00:00"} {"global_step": 9742, "acc_step": 0, "speed/wps": 12899.858570913993, "speed/FLOPS": 202610005119460.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05533702298998833, "optim/lr": 0.0029453337375720855, "optim/total_tokens": 5107613696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8384177684783936, "created_at": "2025-01-16T09:04:27.853477+00:00"} {"global_step": 9743, "acc_step": 0, "speed/wps": 12890.980434145815, "speed/FLOPS": 202470561781677.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04785647615790367, "optim/lr": 0.0029453108152514393, "optim/total_tokens": 5108137984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.915125846862793, "created_at": "2025-01-16T09:04:38.026926+00:00"} {"global_step": 9744, "acc_step": 0, "speed/wps": 12892.749360419519, "speed/FLOPS": 202498345199565.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05103779956698418, "optim/lr": 0.0029452878882152127, "optim/total_tokens": 5108662272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315344, "loss/out": 2.8023624420166016, "created_at": "2025-01-16T09:04:48.200087+00:00"} {"global_step": 9745, "acc_step": 0, "speed/wps": 12893.303848327836, "speed/FLOPS": 202507054194112.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06556672602891922, "optim/lr": 0.0029452649564634804, "optim/total_tokens": 5109186560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8846755027770996, "created_at": "2025-01-16T09:04:58.370613+00:00"} {"global_step": 9746, "acc_step": 0, "speed/wps": 12891.04077132951, "speed/FLOPS": 202471509460059.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052976883947849274, "optim/lr": 0.0029452420199963174, "optim/total_tokens": 5109710848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.995051622390747, "created_at": "2025-01-16T09:05:08.541898+00:00"} {"global_step": 9747, "acc_step": 0, "speed/wps": 12908.600252944014, "speed/FLOPS": 202747305248071.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05046362429857254, "optim/lr": 0.0029452190788137977, "optim/total_tokens": 5110235136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 2.9472904205322266, "created_at": "2025-01-16T09:05:18.699655+00:00"} {"global_step": 9748, "acc_step": 0, "speed/wps": 12899.544237112872, "speed/FLOPS": 202605068075173.8, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06785020232200623, "optim/lr": 0.0029451961329159965, "optim/total_tokens": 5110759424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.957953453063965, "created_at": "2025-01-16T09:05:28.871018+00:00"} {"global_step": 9749, "acc_step": 0, "speed/wps": 12891.30108493099, "speed/FLOPS": 202475598042879.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.07362808287143707, "optim/lr": 0.0029451731823029897, "optim/total_tokens": 5111283712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.0347251892089844, "created_at": "2025-01-16T09:05:39.043391+00:00"} {"global_step": 9750, "acc_step": 0, "speed/wps": 12885.011217131432, "speed/FLOPS": 202376807025903.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05499439314007759, "optim/lr": 0.002945150226974852, "optim/total_tokens": 5111808000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.930257797241211, "created_at": "2025-01-16T09:05:49.220230+00:00"} {"global_step": 9751, "acc_step": 0, "speed/wps": 12899.009357550874, "speed/FLOPS": 202596667056651.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055656153708696365, "optim/lr": 0.0029451272669316565, "optim/total_tokens": 5112332288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8659915924072266, "created_at": "2025-01-16T09:05:59.386907+00:00"} {"global_step": 9752, "acc_step": 0, "speed/wps": 12881.93517809261, "speed/FLOPS": 202328493605877.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.0741354376077652, "optim/lr": 0.0029451043021734794, "optim/total_tokens": 5112856576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9333086013793945, "created_at": "2025-01-16T09:06:09.570142+00:00"} {"global_step": 9753, "acc_step": 0, "speed/wps": 12890.825266270289, "speed/FLOPS": 202468124656969.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04720260947942734, "optim/lr": 0.002945081332700396, "optim/total_tokens": 5113380864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 2.8970224857330322, "created_at": "2025-01-16T09:06:19.745943+00:00"} {"global_step": 9754, "acc_step": 0, "speed/wps": 12897.542828845537, "speed/FLOPS": 202573633208112.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08211065083742142, "optim/lr": 0.0029450583585124805, "optim/total_tokens": 5113905152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 2.883310556411743, "created_at": "2025-01-16T09:06:29.919332+00:00"} {"global_step": 9755, "acc_step": 0, "speed/wps": 12899.336664578193, "speed/FLOPS": 202601807863282.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05947430059313774, "optim/lr": 0.002945035379609809, "optim/total_tokens": 5114429440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 2.8331642150878906, "created_at": "2025-01-16T09:06:40.092532+00:00"} {"global_step": 9756, "acc_step": 0, "speed/wps": 12898.857135081711, "speed/FLOPS": 202594276193597.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062054507434368134, "optim/lr": 0.002945012395992454, "optim/total_tokens": 5114953728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367469, "loss/out": 3.0723154544830322, "created_at": "2025-01-16T09:06:50.260479+00:00"} {"global_step": 9757, "acc_step": 0, "speed/wps": 12895.325712461154, "speed/FLOPS": 202538810348658.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05887765437364578, "optim/lr": 0.0029449894076604933, "optim/total_tokens": 5115478016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 2.905214786529541, "created_at": "2025-01-16T09:07:00.429279+00:00"} {"global_step": 9758, "acc_step": 0, "speed/wps": 12902.802355851061, "speed/FLOPS": 202656241307081.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04889523610472679, "optim/lr": 0.0029449664146139996, "optim/total_tokens": 5116002304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.918689727783203, "created_at": "2025-01-16T09:07:10.591419+00:00"} {"global_step": 9759, "acc_step": 0, "speed/wps": 12903.659896019206, "speed/FLOPS": 202669710153805.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05103540048003197, "optim/lr": 0.00294494341685305, "optim/total_tokens": 5116526592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 2.896657705307007, "created_at": "2025-01-16T09:07:20.754971+00:00"} {"global_step": 9760, "acc_step": 0, "speed/wps": 12902.017619605436, "speed/FLOPS": 202643915946003.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041830819100141525, "optim/lr": 0.0029449204143777185, "optim/total_tokens": 5117050880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 2.822434663772583, "created_at": "2025-01-16T09:07:30.918666+00:00"} {"global_step": 9761, "acc_step": 0, "speed/wps": 12902.455925775586, "speed/FLOPS": 202650800146701.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04269435629248619, "optim/lr": 0.00294489740718808, "optim/total_tokens": 5117575168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.0324902534484863, "created_at": "2025-01-16T09:07:41.081127+00:00"} {"global_step": 9762, "acc_step": 0, "speed/wps": 12900.464987228852, "speed/FLOPS": 202619529720988.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04382650554180145, "optim/lr": 0.0029448743952842095, "optim/total_tokens": 5118099456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484762, "loss/out": 3.097266674041748, "created_at": "2025-01-16T09:07:51.248326+00:00"} {"global_step": 9763, "acc_step": 0, "speed/wps": 12905.244412006334, "speed/FLOPS": 202694597154736.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05218319222331047, "optim/lr": 0.0029448513786661827, "optim/total_tokens": 5118623744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.88077712059021, "created_at": "2025-01-16T09:08:01.409485+00:00"} {"global_step": 9764, "acc_step": 0, "speed/wps": 12906.45905090184, "speed/FLOPS": 202713674727679.44, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04591767489910126, "optim/lr": 0.0029448283573340743, "optim/total_tokens": 5119148032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 3.0009384155273438, "created_at": "2025-01-16T09:08:11.569469+00:00"} {"global_step": 9765, "acc_step": 0, "speed/wps": 12897.967592149205, "speed/FLOPS": 202580304699482.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04292149096727371, "optim/lr": 0.0029448053312879593, "optim/total_tokens": 5119672320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.965322971343994, "created_at": "2025-01-16T09:08:21.736860+00:00"} {"global_step": 9766, "acc_step": 0, "speed/wps": 12904.284730128229, "speed/FLOPS": 202679524032100.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045582495629787445, "optim/lr": 0.0029447823005279133, "optim/total_tokens": 5120196608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.786147117614746, "created_at": "2025-01-16T09:08:31.902433+00:00"} {"global_step": 9767, "acc_step": 0, "speed/wps": 12906.217987156155, "speed/FLOPS": 202709888490297.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06524918973445892, "optim/lr": 0.0029447592650540108, "optim/total_tokens": 5120720896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299229, "loss/out": 2.93424916267395, "created_at": "2025-01-16T09:08:42.061930+00:00"} {"global_step": 9768, "acc_step": 0, "speed/wps": 12907.265708330782, "speed/FLOPS": 202726344391064.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05034998431801796, "optim/lr": 0.0029447362248663273, "optim/total_tokens": 5121245184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.0086917877197266, "created_at": "2025-01-16T09:08:52.220368+00:00"} {"global_step": 9769, "acc_step": 0, "speed/wps": 12911.149636209679, "speed/FLOPS": 202787346815478.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06199375540018082, "optim/lr": 0.0029447131799649378, "optim/total_tokens": 5121769472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8461365699768066, "created_at": "2025-01-16T09:09:02.380236+00:00"} {"global_step": 9770, "acc_step": 0, "speed/wps": 12899.693622769382, "speed/FLOPS": 202607414382186.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04992899298667908, "optim/lr": 0.002944690130349918, "optim/total_tokens": 5122293760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.8675003051757812, "created_at": "2025-01-16T09:09:12.547453+00:00"} {"global_step": 9771, "acc_step": 0, "speed/wps": 12902.753147418927, "speed/FLOPS": 202655468421035.25, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04340765252709389, "optim/lr": 0.0029446670760213426, "optim/total_tokens": 5122818048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.008120536804199, "created_at": "2025-01-16T09:09:22.710433+00:00"} {"global_step": 9772, "acc_step": 0, "speed/wps": 12905.058769053167, "speed/FLOPS": 202691681377057.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04955922067165375, "optim/lr": 0.002944644016979287, "optim/total_tokens": 5123342336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.9794514179229736, "created_at": "2025-01-16T09:09:32.872230+00:00"} {"global_step": 9773, "acc_step": 0, "speed/wps": 12903.042185938528, "speed/FLOPS": 202660008168166.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05720444768667221, "optim/lr": 0.0029446209532238265, "optim/total_tokens": 5123866624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9424068927764893, "created_at": "2025-01-16T09:09:43.035419+00:00"} {"global_step": 9774, "acc_step": 0, "speed/wps": 12900.756848876093, "speed/FLOPS": 202624113809218.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05065842717885971, "optim/lr": 0.0029445978847550355, "optim/total_tokens": 5124390912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.9449901580810547, "created_at": "2025-01-16T09:09:53.204085+00:00"} {"global_step": 9775, "acc_step": 0, "speed/wps": 12910.15164556196, "speed/FLOPS": 202771672000973.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054012179374694824, "optim/lr": 0.002944574811572991, "optim/total_tokens": 5124915200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 2.962667942047119, "created_at": "2025-01-16T09:10:03.364073+00:00"} {"global_step": 9776, "acc_step": 0, "speed/wps": 12901.157637970788, "speed/FLOPS": 202630408752692.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052430905401706696, "optim/lr": 0.0029445517336777667, "optim/total_tokens": 5125439488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465898, "loss/out": 2.906489849090576, "created_at": "2025-01-16T09:10:13.530824+00:00"} {"global_step": 9777, "acc_step": 0, "speed/wps": 12905.873092073649, "speed/FLOPS": 202704471439090.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04961339011788368, "optim/lr": 0.002944528651069439, "optim/total_tokens": 5125963776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 2.7796337604522705, "created_at": "2025-01-16T09:10:23.690307+00:00"} {"global_step": 9778, "acc_step": 0, "speed/wps": 12906.052278659487, "speed/FLOPS": 202707285810649.66, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07862301915884018, "optim/lr": 0.0029445055637480817, "optim/total_tokens": 5126488064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.777052402496338, "created_at": "2025-01-16T09:10:33.852816+00:00"} {"global_step": 9779, "acc_step": 0, "speed/wps": 12903.016275356465, "speed/FLOPS": 202659601206869.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05307278409600258, "optim/lr": 0.002944482471713772, "optim/total_tokens": 5127012352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395284, "loss/out": 2.8645992279052734, "created_at": "2025-01-16T09:10:44.016186+00:00"} {"global_step": 9780, "acc_step": 0, "speed/wps": 12901.84684085755, "speed/FLOPS": 202641233631093.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06573418527841568, "optim/lr": 0.0029444593749665842, "optim/total_tokens": 5127536640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.872042179107666, "created_at": "2025-01-16T09:10:54.178921+00:00"} {"global_step": 9781, "acc_step": 0, "speed/wps": 12906.312868669529, "speed/FLOPS": 202711378734845.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05151110514998436, "optim/lr": 0.0029444362735065936, "optim/total_tokens": 5128060928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.85626220703125, "created_at": "2025-01-16T09:11:04.338053+00:00"} {"global_step": 9782, "acc_step": 0, "speed/wps": 12895.34134684804, "speed/FLOPS": 202539055908188.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.05410339683294296, "optim/lr": 0.0029444131673338762, "optim/total_tokens": 5128585216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.9186244010925293, "created_at": "2025-01-16T09:11:14.506089+00:00"} {"global_step": 9783, "acc_step": 0, "speed/wps": 12904.858923481255, "speed/FLOPS": 202688542527733.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054166775196790695, "optim/lr": 0.0029443900564485067, "optim/total_tokens": 5129109504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8457841873168945, "created_at": "2025-01-16T09:11:24.667959+00:00"} {"global_step": 9784, "acc_step": 0, "speed/wps": 12910.567009352695, "speed/FLOPS": 202778195860076.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049242209643125534, "optim/lr": 0.002944366940850561, "optim/total_tokens": 5129633792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.8923208713531494, "created_at": "2025-01-16T09:11:34.825649+00:00"} {"global_step": 9785, "acc_step": 0, "speed/wps": 12909.55247240064, "speed/FLOPS": 202762261163124.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05421656370162964, "optim/lr": 0.002944343820540114, "optim/total_tokens": 5130158080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8561782836914062, "created_at": "2025-01-16T09:11:44.988223+00:00"} {"global_step": 9786, "acc_step": 0, "speed/wps": 12902.833112051083, "speed/FLOPS": 202656724375466.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04873575270175934, "optim/lr": 0.0029443206955172417, "optim/total_tokens": 5130682368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.954299211502075, "created_at": "2025-01-16T09:11:55.153161+00:00"} {"global_step": 9787, "acc_step": 0, "speed/wps": 12907.408003330242, "speed/FLOPS": 202728579329564.78, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06092151999473572, "optim/lr": 0.002944297565782019, "optim/total_tokens": 5131206656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.9570183753967285, "created_at": "2025-01-16T09:12:05.313558+00:00"} {"global_step": 9788, "acc_step": 0, "speed/wps": 12904.031681788478, "speed/FLOPS": 202675549560199.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04725988954305649, "optim/lr": 0.002944274431334522, "optim/total_tokens": 5131730944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.8062639236450195, "created_at": "2025-01-16T09:12:15.478062+00:00"} {"global_step": 9789, "acc_step": 0, "speed/wps": 12904.695121822564, "speed/FLOPS": 202685969797596.84, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05864877253770828, "optim/lr": 0.0029442512921748257, "optim/total_tokens": 5132255232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 2.967482328414917, "created_at": "2025-01-16T09:12:25.640974+00:00"} {"global_step": 9790, "acc_step": 0, "speed/wps": 12907.392416615347, "speed/FLOPS": 202728334518788.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047323163598775864, "optim/lr": 0.002944228148303006, "optim/total_tokens": 5132779520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 2.993114948272705, "created_at": "2025-01-16T09:12:35.800917+00:00"} {"global_step": 9791, "acc_step": 0, "speed/wps": 12896.919815112195, "speed/FLOPS": 202563847921319.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06335383653640747, "optim/lr": 0.0029442049997191375, "optim/total_tokens": 5133303808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8724918365478516, "created_at": "2025-01-16T09:12:45.968575+00:00"} {"global_step": 9792, "acc_step": 0, "speed/wps": 12908.317381127657, "speed/FLOPS": 202742862357489.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04887782782316208, "optim/lr": 0.002944181846423296, "optim/total_tokens": 5133828096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.826420783996582, "created_at": "2025-01-16T09:12:56.128547+00:00"} {"global_step": 9793, "acc_step": 0, "speed/wps": 12905.988058603514, "speed/FLOPS": 202706277146423.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05505793169140816, "optim/lr": 0.002944158688415559, "optim/total_tokens": 5134352384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.860722064971924, "created_at": "2025-01-16T09:13:06.290726+00:00"} {"global_step": 9794, "acc_step": 0, "speed/wps": 12901.273031408078, "speed/FLOPS": 202632221165192.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060738157480955124, "optim/lr": 0.0029441355256959994, "optim/total_tokens": 5134876672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375761, "loss/out": 2.9056625366210938, "created_at": "2025-01-16T09:13:16.453935+00:00"} {"global_step": 9795, "acc_step": 0, "speed/wps": 12900.69970580719, "speed/FLOPS": 202623216298798.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045924827456474304, "optim/lr": 0.002944112358264694, "optim/total_tokens": 5135400960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.849224328994751, "created_at": "2025-01-16T09:13:26.620141+00:00"} {"global_step": 9796, "acc_step": 0, "speed/wps": 12901.84943818312, "speed/FLOPS": 202641274425660.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053720008581876755, "optim/lr": 0.002944089186121718, "optim/total_tokens": 5135925248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413950, "loss/out": 2.8917973041534424, "created_at": "2025-01-16T09:13:36.786346+00:00"} {"global_step": 9797, "acc_step": 0, "speed/wps": 12908.877310484424, "speed/FLOPS": 202751656817460.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04657605662941933, "optim/lr": 0.0029440660092671478, "optim/total_tokens": 5136449536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.910456895828247, "created_at": "2025-01-16T09:13:46.947924+00:00"} {"global_step": 9798, "acc_step": 0, "speed/wps": 12906.399448705588, "speed/FLOPS": 202712738593285.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045299090445041656, "optim/lr": 0.002944042827701058, "optim/total_tokens": 5136973824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.9000024795532227, "created_at": "2025-01-16T09:13:57.107708+00:00"} {"global_step": 9799, "acc_step": 0, "speed/wps": 12906.496766379414, "speed/FLOPS": 202714267101081.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051998384296894073, "optim/lr": 0.002944019641423525, "optim/total_tokens": 5137498112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.9700427055358887, "created_at": "2025-01-16T09:14:07.267510+00:00"} {"global_step": 9800, "acc_step": 0, "speed/wps": 12904.811163098686, "speed/FLOPS": 202687792385295.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053406503051519394, "optim/lr": 0.0029439964504346236, "optim/total_tokens": 5138022400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.831873655319214, "created_at": "2025-01-16T09:14:17.430400+00:00"} {"global_step": 9801, "acc_step": 0, "speed/wps": 12900.01297611283, "speed/FLOPS": 202612430265283.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049805764108896255, "optim/lr": 0.0029439732547344305, "optim/total_tokens": 5138546688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.953334331512451, "created_at": "2025-01-16T09:14:27.594605+00:00"} {"global_step": 9802, "acc_step": 0, "speed/wps": 12895.255244761029, "speed/FLOPS": 202537703556595.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.053800106048583984, "optim/lr": 0.0029439500543230203, "optim/total_tokens": 5139070976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.907249927520752, "created_at": "2025-01-16T09:14:37.765423+00:00"} {"global_step": 9803, "acc_step": 0, "speed/wps": 12892.979389163587, "speed/FLOPS": 202501958117083.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06797010451555252, "optim/lr": 0.00294392684920047, "optim/total_tokens": 5139595264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8709917068481445, "created_at": "2025-01-16T09:14:47.936332+00:00"} {"global_step": 9804, "acc_step": 0, "speed/wps": 12904.090372378338, "speed/FLOPS": 202676471376562.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08391939103603363, "optim/lr": 0.0029439036393668545, "optim/total_tokens": 5140119552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367755, "loss/out": 2.8475189208984375, "created_at": "2025-01-16T09:14:58.100659+00:00"} {"global_step": 9805, "acc_step": 0, "speed/wps": 12900.769800208634, "speed/FLOPS": 202624317227692.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04795605689287186, "optim/lr": 0.0029438804248222493, "optim/total_tokens": 5140643840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.857602596282959, "created_at": "2025-01-16T09:15:08.269301+00:00"} {"global_step": 9806, "acc_step": 0, "speed/wps": 12905.587570519143, "speed/FLOPS": 202699986930729.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09515047818422318, "optim/lr": 0.0029438572055667305, "optim/total_tokens": 5141168128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.0181632041931152, "created_at": "2025-01-16T09:15:18.431069+00:00"} {"global_step": 9807, "acc_step": 0, "speed/wps": 12908.657417480668, "speed/FLOPS": 202748203095672.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043608393520116806, "optim/lr": 0.0029438339816003738, "optim/total_tokens": 5141692416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.922443389892578, "created_at": "2025-01-16T09:15:28.588351+00:00"} {"global_step": 9808, "acc_step": 0, "speed/wps": 12899.8262125762, "speed/FLOPS": 202609496887300.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08631570637226105, "optim/lr": 0.002943810752923255, "optim/total_tokens": 5142216704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.841597318649292, "created_at": "2025-01-16T09:15:38.762101+00:00"} {"global_step": 9809, "acc_step": 0, "speed/wps": 12897.930399317043, "speed/FLOPS": 202579720534944.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06412092596292496, "optim/lr": 0.0029437875195354496, "optim/total_tokens": 5142740992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 2.9051425457000732, "created_at": "2025-01-16T09:15:48.928982+00:00"} {"global_step": 9810, "acc_step": 0, "speed/wps": 12902.474340635312, "speed/FLOPS": 202651089377377.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06306514143943787, "optim/lr": 0.002943764281437034, "optim/total_tokens": 5143265280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9095139503479004, "created_at": "2025-01-16T09:15:59.091148+00:00"} {"global_step": 9811, "acc_step": 0, "speed/wps": 12903.632828269832, "speed/FLOPS": 202669285017606.88, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06982940435409546, "optim/lr": 0.002943741038628084, "optim/total_tokens": 5143789568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.0114340782165527, "created_at": "2025-01-16T09:16:09.258547+00:00"} {"global_step": 9812, "acc_step": 0, "speed/wps": 12900.980828287025, "speed/FLOPS": 202627631713647.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05864288657903671, "optim/lr": 0.002943717791108675, "optim/total_tokens": 5144313856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341862, "loss/out": 2.976433038711548, "created_at": "2025-01-16T09:16:19.422916+00:00"} {"global_step": 9813, "acc_step": 0, "speed/wps": 12896.70268109119, "speed/FLOPS": 202560437533146.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06926354765892029, "optim/lr": 0.0029436945388788822, "optim/total_tokens": 5144838144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.7888383865356445, "created_at": "2025-01-16T09:16:29.589628+00:00"} {"global_step": 9814, "acc_step": 0, "speed/wps": 12906.201207928334, "speed/FLOPS": 202709624949467.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057906247675418854, "optim/lr": 0.002943671281938783, "optim/total_tokens": 5145362432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8584089279174805, "created_at": "2025-01-16T09:16:39.749079+00:00"} {"global_step": 9815, "acc_step": 0, "speed/wps": 12903.130391294239, "speed/FLOPS": 202661393554484.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09497351944446564, "optim/lr": 0.0029436480202884524, "optim/total_tokens": 5145886720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.898049831390381, "created_at": "2025-01-16T09:16:49.910833+00:00"} {"global_step": 9816, "acc_step": 0, "speed/wps": 12898.736663640417, "speed/FLOPS": 202592384024066.66, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045508887618780136, "optim/lr": 0.0029436247539279662, "optim/total_tokens": 5146411008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.90586519241333, "created_at": "2025-01-16T09:17:00.076626+00:00"} {"global_step": 9817, "acc_step": 0, "speed/wps": 12890.63208710185, "speed/FLOPS": 202465090512673.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.09951473027467728, "optim/lr": 0.0029436014828574004, "optim/total_tokens": 5146935296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425156, "loss/out": 2.8160464763641357, "created_at": "2025-01-16T09:17:10.248341+00:00"} {"global_step": 9818, "acc_step": 0, "speed/wps": 12896.112636374184, "speed/FLOPS": 202551170070059.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045922622084617615, "optim/lr": 0.0029435782070768315, "optim/total_tokens": 5147459584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.057673215866089, "created_at": "2025-01-16T09:17:20.417835+00:00"} {"global_step": 9819, "acc_step": 0, "speed/wps": 12895.208992185231, "speed/FLOPS": 202536977096334.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09214702993631363, "optim/lr": 0.0029435549265863347, "optim/total_tokens": 5147983872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.999706268310547, "created_at": "2025-01-16T09:17:30.586159+00:00"} {"global_step": 9820, "acc_step": 0, "speed/wps": 12899.736529501637, "speed/FLOPS": 202608088291376.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053903982043266296, "optim/lr": 0.002943531641385986, "optim/total_tokens": 5148508160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.916628122329712, "created_at": "2025-01-16T09:17:40.754763+00:00"} {"global_step": 9821, "acc_step": 0, "speed/wps": 12896.502372172052, "speed/FLOPS": 202557291406317.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05870572850108147, "optim/lr": 0.0029435083514758624, "optim/total_tokens": 5149032448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 2.976027011871338, "created_at": "2025-01-16T09:17:50.924634+00:00"} {"global_step": 9822, "acc_step": 0, "speed/wps": 12896.595762465056, "speed/FLOPS": 202558758229201.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05052895471453667, "optim/lr": 0.0029434850568560386, "optim/total_tokens": 5149556736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.907017707824707, "created_at": "2025-01-16T09:18:01.092582+00:00"} {"global_step": 9823, "acc_step": 0, "speed/wps": 12908.667797632595, "speed/FLOPS": 202748366130222.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05270783230662346, "optim/lr": 0.0029434617575265914, "optim/total_tokens": 5150081024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.79534649848938, "created_at": "2025-01-16T09:18:11.250745+00:00"} {"global_step": 9824, "acc_step": 0, "speed/wps": 12900.943006677273, "speed/FLOPS": 202627037673293.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0644676685333252, "optim/lr": 0.002943438453487596, "optim/total_tokens": 5150605312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8195290565490723, "created_at": "2025-01-16T09:18:21.415575+00:00"} {"global_step": 9825, "acc_step": 0, "speed/wps": 12903.415579355582, "speed/FLOPS": 202665872824877.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06311403214931488, "optim/lr": 0.0029434151447391295, "optim/total_tokens": 5151129600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8718223571777344, "created_at": "2025-01-16T09:18:31.579349+00:00"} {"global_step": 9826, "acc_step": 0, "speed/wps": 12899.253931821859, "speed/FLOPS": 202600508431657.2, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05316246300935745, "optim/lr": 0.002943391831281267, "optim/total_tokens": 5151653888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.8695638179779053, "created_at": "2025-01-16T09:18:41.746990+00:00"} {"global_step": 9827, "acc_step": 0, "speed/wps": 12905.103427340957, "speed/FLOPS": 202692382796833.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06536664068698883, "optim/lr": 0.002943368513114085, "optim/total_tokens": 5152178176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8688502311706543, "created_at": "2025-01-16T09:18:51.910160+00:00"} {"global_step": 9828, "acc_step": 0, "speed/wps": 12900.96803026768, "speed/FLOPS": 202627430703166.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051031820476055145, "optim/lr": 0.00294334519023766, "optim/total_tokens": 5152702464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9730896949768066, "created_at": "2025-01-16T09:19:02.077951+00:00"} {"global_step": 9829, "acc_step": 0, "speed/wps": 12905.864412356, "speed/FLOPS": 202704335112198.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07492393255233765, "optim/lr": 0.002943321862652067, "optim/total_tokens": 5153226752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.945486545562744, "created_at": "2025-01-16T09:19:12.239357+00:00"} {"global_step": 9830, "acc_step": 0, "speed/wps": 12901.746008425083, "speed/FLOPS": 202639649919182.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04854913800954819, "optim/lr": 0.0029432985303573835, "optim/total_tokens": 5153751040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8518471717834473, "created_at": "2025-01-16T09:19:22.402385+00:00"} {"global_step": 9831, "acc_step": 0, "speed/wps": 12902.311577712691, "speed/FLOPS": 202648532962018.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06521093100309372, "optim/lr": 0.0029432751933536846, "optim/total_tokens": 5154275328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354275, "loss/out": 2.803626298904419, "created_at": "2025-01-16T09:19:32.566540+00:00"} {"global_step": 9832, "acc_step": 0, "speed/wps": 12904.209915657078, "speed/FLOPS": 202678348968024.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046020686626434326, "optim/lr": 0.0029432518516410463, "optim/total_tokens": 5154799616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.0165727138519287, "created_at": "2025-01-16T09:19:42.727562+00:00"} {"global_step": 9833, "acc_step": 0, "speed/wps": 12900.268173847311, "speed/FLOPS": 202616438496381.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07079992443323135, "optim/lr": 0.002943228505219545, "optim/total_tokens": 5155323904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9762043952941895, "created_at": "2025-01-16T09:19:52.893897+00:00"} {"global_step": 9834, "acc_step": 0, "speed/wps": 12899.036142267576, "speed/FLOPS": 202597087747432.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05622218921780586, "optim/lr": 0.0029432051540892582, "optim/total_tokens": 5155848192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.8519234657287598, "created_at": "2025-01-16T09:20:03.062308+00:00"} {"global_step": 9835, "acc_step": 0, "speed/wps": 12900.629471446417, "speed/FLOPS": 202622113171650.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06601760536432266, "optim/lr": 0.0029431817982502597, "optim/total_tokens": 5156372480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9140689373016357, "created_at": "2025-01-16T09:20:13.229844+00:00"} {"global_step": 9836, "acc_step": 0, "speed/wps": 12904.815415867039, "speed/FLOPS": 202687859180866.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05618010088801384, "optim/lr": 0.002943158437702628, "optim/total_tokens": 5156896768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348945, "loss/out": 2.87983775138855, "created_at": "2025-01-16T09:20:23.394591+00:00"} {"global_step": 9837, "acc_step": 0, "speed/wps": 12902.858756919346, "speed/FLOPS": 202657127163362.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06269911676645279, "optim/lr": 0.0029431350724464377, "optim/total_tokens": 5157421056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.966437816619873, "created_at": "2025-01-16T09:20:33.556737+00:00"} {"global_step": 9838, "acc_step": 0, "speed/wps": 12900.119383837877, "speed/FLOPS": 202614101544824.44, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05133179947733879, "optim/lr": 0.0029431117024817655, "optim/total_tokens": 5157945344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.9115114212036133, "created_at": "2025-01-16T09:20:43.726268+00:00"} {"global_step": 9839, "acc_step": 0, "speed/wps": 12901.729366192687, "speed/FLOPS": 202639388530054.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04802463948726654, "optim/lr": 0.0029430883278086877, "optim/total_tokens": 5158469632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369627, "loss/out": 2.8455772399902344, "created_at": "2025-01-16T09:20:53.896937+00:00"} {"global_step": 9840, "acc_step": 0, "speed/wps": 12905.912385280197, "speed/FLOPS": 202705088592894.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05263739079236984, "optim/lr": 0.002943064948427281, "optim/total_tokens": 5158993920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.897540330886841, "created_at": "2025-01-16T09:21:04.059760+00:00"} {"global_step": 9841, "acc_step": 0, "speed/wps": 12902.439079442373, "speed/FLOPS": 202650535551888.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045151881873607635, "optim/lr": 0.0029430415643376213, "optim/total_tokens": 5159518208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.915367603302002, "created_at": "2025-01-16T09:21:14.222033+00:00"} {"global_step": 9842, "acc_step": 0, "speed/wps": 12899.674089062408, "speed/FLOPS": 202607107578476.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050909992307424545, "optim/lr": 0.0029430181755397853, "optim/total_tokens": 5160042496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.8762872219085693, "created_at": "2025-01-16T09:21:24.387381+00:00"} {"global_step": 9843, "acc_step": 0, "speed/wps": 12901.535643642113, "speed/FLOPS": 202636345851194.88, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04997827857732773, "optim/lr": 0.002942994782033848, "optim/total_tokens": 5160566784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.8861207962036133, "created_at": "2025-01-16T09:21:34.550720+00:00"} {"global_step": 9844, "acc_step": 0, "speed/wps": 12899.842062430514, "speed/FLOPS": 202609745831042.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05698160454630852, "optim/lr": 0.002942971383819887, "optim/total_tokens": 5161091072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427550, "loss/out": 2.8856749534606934, "created_at": "2025-01-16T09:21:44.715236+00:00"} {"global_step": 9845, "acc_step": 0, "speed/wps": 12908.500154360543, "speed/FLOPS": 202745733062268.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04845645651221275, "optim/lr": 0.0029429479808979783, "optim/total_tokens": 5161615360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.96492862701416, "created_at": "2025-01-16T09:21:54.875199+00:00"} {"global_step": 9846, "acc_step": 0, "speed/wps": 12907.76790746006, "speed/FLOPS": 202734232118483.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06196599081158638, "optim/lr": 0.002942924573268198, "optim/total_tokens": 5162139648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8853089809417725, "created_at": "2025-01-16T09:22:05.034992+00:00"} {"global_step": 9847, "acc_step": 0, "speed/wps": 12907.609291609446, "speed/FLOPS": 202731740838588.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07026390731334686, "optim/lr": 0.0029429011609306225, "optim/total_tokens": 5162663936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.9565110206604004, "created_at": "2025-01-16T09:22:15.193179+00:00"} {"global_step": 9848, "acc_step": 0, "speed/wps": 12904.997297110205, "speed/FLOPS": 202690715875721.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05469367280602455, "optim/lr": 0.0029428777438853288, "optim/total_tokens": 5163188224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.833217144012451, "created_at": "2025-01-16T09:22:25.359368+00:00"} {"global_step": 9849, "acc_step": 0, "speed/wps": 12902.828661856453, "speed/FLOPS": 202656654479044.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055554475635290146, "optim/lr": 0.002942854322132393, "optim/total_tokens": 5163712512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.970895767211914, "created_at": "2025-01-16T09:22:35.522817+00:00"} {"global_step": 9850, "acc_step": 0, "speed/wps": 12904.59163578461, "speed/FLOPS": 202684344407166.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06734362244606018, "optim/lr": 0.0029428308956718913, "optim/total_tokens": 5164236800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.959138870239258, "created_at": "2025-01-16T09:22:45.684469+00:00"} {"global_step": 9851, "acc_step": 0, "speed/wps": 12907.29009710583, "speed/FLOPS": 202726727450290.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07758383452892303, "optim/lr": 0.0029428074645039, "optim/total_tokens": 5164761088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 2.9204249382019043, "created_at": "2025-01-16T09:22:55.844383+00:00"} {"global_step": 9852, "acc_step": 0, "speed/wps": 12906.94976702345, "speed/FLOPS": 202721382098683.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04882609099149704, "optim/lr": 0.0029427840286284958, "optim/total_tokens": 5165285376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8615052700042725, "created_at": "2025-01-16T09:23:06.005041+00:00"} {"global_step": 9853, "acc_step": 0, "speed/wps": 12903.480613779875, "speed/FLOPS": 202666894279877.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07387329638004303, "optim/lr": 0.0029427605880457553, "optim/total_tokens": 5165809664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.8766238689422607, "created_at": "2025-01-16T09:23:16.169798+00:00"} {"global_step": 9854, "acc_step": 0, "speed/wps": 12902.365122643814, "speed/FLOPS": 202649373958740.2, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049897562712430954, "optim/lr": 0.0029427371427557542, "optim/total_tokens": 5166333952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351996, "loss/out": 2.89717173576355, "created_at": "2025-01-16T09:23:26.334534+00:00"} {"global_step": 9855, "acc_step": 0, "speed/wps": 12906.683065016123, "speed/FLOPS": 202717193177171.9, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06871189177036285, "optim/lr": 0.0029427136927585706, "optim/total_tokens": 5166858240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.095529794692993, "created_at": "2025-01-16T09:23:36.493387+00:00"} {"global_step": 9856, "acc_step": 0, "speed/wps": 12907.740056533055, "speed/FLOPS": 202733794681403.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.054389405995607376, "optim/lr": 0.0029426902380542794, "optim/total_tokens": 5167382528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 2.9978604316711426, "created_at": "2025-01-16T09:23:46.655529+00:00"} {"global_step": 9857, "acc_step": 0, "speed/wps": 12909.985598005338, "speed/FLOPS": 202769063995923.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049160100519657135, "optim/lr": 0.002942666778642958, "optim/total_tokens": 5167906816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.020015239715576, "created_at": "2025-01-16T09:23:56.815135+00:00"} {"global_step": 9858, "acc_step": 0, "speed/wps": 12903.447689517849, "speed/FLOPS": 202666377159099.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057118333876132965, "optim/lr": 0.0029426433145246828, "optim/total_tokens": 5168431104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9784202575683594, "created_at": "2025-01-16T09:24:06.976717+00:00"} {"global_step": 9859, "acc_step": 0, "speed/wps": 12904.66443484083, "speed/FLOPS": 202685487816380.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04997643455862999, "optim/lr": 0.00294261984569953, "optim/total_tokens": 5168955392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 2.8804025650024414, "created_at": "2025-01-16T09:24:17.137160+00:00"} {"global_step": 9860, "acc_step": 0, "speed/wps": 12906.995519115126, "speed/FLOPS": 202722100698153.66, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05450640618801117, "optim/lr": 0.0029425963721675763, "optim/total_tokens": 5169479680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9562337398529053, "created_at": "2025-01-16T09:24:27.296388+00:00"} {"global_step": 9861, "acc_step": 0, "speed/wps": 12904.177484608836, "speed/FLOPS": 202677839593847.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061486244201660156, "optim/lr": 0.0029425728939288984, "optim/total_tokens": 5170003968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 2.9286577701568604, "created_at": "2025-01-16T09:24:37.463438+00:00"} {"global_step": 9862, "acc_step": 0, "speed/wps": 12901.585417582572, "speed/FLOPS": 202637127619326.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.040891967713832855, "optim/lr": 0.0029425494109835736, "optim/total_tokens": 5170528256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.785696029663086, "created_at": "2025-01-16T09:24:47.633847+00:00"} {"global_step": 9863, "acc_step": 0, "speed/wps": 12902.28811379869, "speed/FLOPS": 202648164429006.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051858317106962204, "optim/lr": 0.002942525923331677, "optim/total_tokens": 5171052544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.906376838684082, "created_at": "2025-01-16T09:24:57.798806+00:00"} {"global_step": 9864, "acc_step": 0, "speed/wps": 12904.26766486046, "speed/FLOPS": 202679255998619.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05091219022870064, "optim/lr": 0.002942502430973286, "optim/total_tokens": 5171576832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 2.9541380405426025, "created_at": "2025-01-16T09:25:07.961001+00:00"} {"global_step": 9865, "acc_step": 0, "speed/wps": 12904.091441958803, "speed/FLOPS": 202676488175793.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0694599375128746, "optim/lr": 0.002942478933908478, "optim/total_tokens": 5172101120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.929583787918091, "created_at": "2025-01-16T09:25:18.128329+00:00"} {"global_step": 9866, "acc_step": 0, "speed/wps": 12908.72977578687, "speed/FLOPS": 202749339582301.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07438630610704422, "optim/lr": 0.002942455432137328, "optim/total_tokens": 5172625408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.807973623275757, "created_at": "2025-01-16T09:25:28.285591+00:00"} {"global_step": 9867, "acc_step": 0, "speed/wps": 12904.477241969771, "speed/FLOPS": 202682547695110.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06954403221607208, "optim/lr": 0.0029424319256599147, "optim/total_tokens": 5173149696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 2.8988053798675537, "created_at": "2025-01-16T09:25:38.448756+00:00"} {"global_step": 9868, "acc_step": 0, "speed/wps": 12904.362326019194, "speed/FLOPS": 202680742782195.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05460737273097038, "optim/lr": 0.0029424084144763136, "optim/total_tokens": 5173673984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9484457969665527, "created_at": "2025-01-16T09:25:48.610124+00:00"} {"global_step": 9869, "acc_step": 0, "speed/wps": 12904.942445243732, "speed/FLOPS": 202689854351784.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07014866918325424, "optim/lr": 0.002942384898586601, "optim/total_tokens": 5174198272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 2.9351816177368164, "created_at": "2025-01-16T09:25:58.771776+00:00"} {"global_step": 9870, "acc_step": 0, "speed/wps": 12908.074390427479, "speed/FLOPS": 202739045854638.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050536684691905975, "optim/lr": 0.0029423613779908544, "optim/total_tokens": 5174722560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481564, "loss/out": 2.93988037109375, "created_at": "2025-01-16T09:26:08.934998+00:00"} {"global_step": 9871, "acc_step": 0, "speed/wps": 12906.010279082435, "speed/FLOPS": 202706626149578.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0713750571012497, "optim/lr": 0.0029423378526891504, "optim/total_tokens": 5175246848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446221, "loss/out": 2.8784866333007812, "created_at": "2025-01-16T09:26:19.094396+00:00"} {"global_step": 9872, "acc_step": 0, "speed/wps": 12906.625129667527, "speed/FLOPS": 202716283222910.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05286336690187454, "optim/lr": 0.0029423143226815656, "optim/total_tokens": 5175771136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.825178623199463, "created_at": "2025-01-16T09:26:29.253589+00:00"} {"global_step": 9873, "acc_step": 0, "speed/wps": 12909.55848915076, "speed/FLOPS": 202762355664453.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08335486799478531, "optim/lr": 0.0029422907879681766, "optim/total_tokens": 5176295424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.820647954940796, "created_at": "2025-01-16T09:26:39.412277+00:00"} {"global_step": 9874, "acc_step": 0, "speed/wps": 12901.877991092035, "speed/FLOPS": 202641722888330.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05221018195152283, "optim/lr": 0.0029422672485490613, "optim/total_tokens": 5176819712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 2.930673122406006, "created_at": "2025-01-16T09:26:49.575083+00:00"} {"global_step": 9875, "acc_step": 0, "speed/wps": 12903.046411515632, "speed/FLOPS": 202660074536661.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055531103163957596, "optim/lr": 0.0029422437044242947, "optim/total_tokens": 5177344000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.904567241668701, "created_at": "2025-01-16T09:26:59.743570+00:00"} {"global_step": 9876, "acc_step": 0, "speed/wps": 12907.718037267525, "speed/FLOPS": 202733448838581.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044194623827934265, "optim/lr": 0.002942220155593955, "optim/total_tokens": 5177868288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 2.903071165084839, "created_at": "2025-01-16T09:27:09.904205+00:00"} {"global_step": 9877, "acc_step": 0, "speed/wps": 12909.861896465218, "speed/FLOPS": 202767121093252.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06422111392021179, "optim/lr": 0.0029421966020581184, "optim/total_tokens": 5178392576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.907243251800537, "created_at": "2025-01-16T09:27:20.066272+00:00"} {"global_step": 9878, "acc_step": 0, "speed/wps": 12909.942587538671, "speed/FLOPS": 202768388457441.84, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057698026299476624, "optim/lr": 0.0029421730438168618, "optim/total_tokens": 5178916864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.8508310317993164, "created_at": "2025-01-16T09:27:30.223104+00:00"} {"global_step": 9879, "acc_step": 0, "speed/wps": 12893.54712545677, "speed/FLOPS": 202510875195719.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05190468207001686, "optim/lr": 0.0029421494808702623, "optim/total_tokens": 5179441152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292143, "loss/out": 2.9587063789367676, "created_at": "2025-01-16T09:27:40.394910+00:00"} {"global_step": 9880, "acc_step": 0, "speed/wps": 12904.499096358657, "speed/FLOPS": 202682890948318.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04628210887312889, "optim/lr": 0.002942125913218397, "optim/total_tokens": 5179965440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8607468605041504, "created_at": "2025-01-16T09:27:50.555975+00:00"} {"global_step": 9881, "acc_step": 0, "speed/wps": 12908.331037883092, "speed/FLOPS": 202743076855599.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04963240399956703, "optim/lr": 0.002942102340861342, "optim/total_tokens": 5180489728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 2.992220878601074, "created_at": "2025-01-16T09:28:00.714984+00:00"} {"global_step": 9882, "acc_step": 0, "speed/wps": 12906.489918041892, "speed/FLOPS": 202714159538529.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05357860028743744, "optim/lr": 0.002942078763799175, "optim/total_tokens": 5181014016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.799170970916748, "created_at": "2025-01-16T09:28:10.877103+00:00"} {"global_step": 9883, "acc_step": 0, "speed/wps": 12911.15440365892, "speed/FLOPS": 202787421694820.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05008929967880249, "optim/lr": 0.002942055182031972, "optim/total_tokens": 5181538304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 2.7716028690338135, "created_at": "2025-01-16T09:28:21.035759+00:00"} {"global_step": 9884, "acc_step": 0, "speed/wps": 12900.908479958314, "speed/FLOPS": 202626495383727.62, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05091331899166107, "optim/lr": 0.0029420315955598114, "optim/total_tokens": 5182062592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.864212989807129, "created_at": "2025-01-16T09:28:31.203341+00:00"} {"global_step": 9885, "acc_step": 0, "speed/wps": 12901.646652415431, "speed/FLOPS": 202638089396521.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051376551389694214, "optim/lr": 0.0029420080043827688, "optim/total_tokens": 5182586880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.923266887664795, "created_at": "2025-01-16T09:28:41.367341+00:00"} {"global_step": 9886, "acc_step": 0, "speed/wps": 12907.460954879278, "speed/FLOPS": 202729411006405.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051134780049324036, "optim/lr": 0.002941984408500922, "optim/total_tokens": 5183111168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381609, "loss/out": 2.894533157348633, "created_at": "2025-01-16T09:28:51.527465+00:00"} {"global_step": 9887, "acc_step": 0, "speed/wps": 12908.209482528355, "speed/FLOPS": 202741167661718.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06822788715362549, "optim/lr": 0.002941960807914347, "optim/total_tokens": 5183635456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.874084949493408, "created_at": "2025-01-16T09:29:01.685186+00:00"} {"global_step": 9888, "acc_step": 0, "speed/wps": 12901.772131529722, "speed/FLOPS": 202640060218438.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04632675275206566, "optim/lr": 0.0029419372026231218, "optim/total_tokens": 5184159744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370920, "loss/out": 2.7780215740203857, "created_at": "2025-01-16T09:29:11.857042+00:00"} {"global_step": 9889, "acc_step": 0, "speed/wps": 12905.965145848922, "speed/FLOPS": 202705917270126.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05284005030989647, "optim/lr": 0.0029419135926273226, "optim/total_tokens": 5184684032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.806918144226074, "created_at": "2025-01-16T09:29:22.017350+00:00"} {"global_step": 9890, "acc_step": 0, "speed/wps": 12904.952246693962, "speed/FLOPS": 202690008297028.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05816888064146042, "optim/lr": 0.0029418899779270276, "optim/total_tokens": 5185208320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9284844398498535, "created_at": "2025-01-16T09:29:32.180614+00:00"} {"global_step": 9891, "acc_step": 0, "speed/wps": 12902.044494880956, "speed/FLOPS": 202644338059135.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04262165352702141, "optim/lr": 0.002941866358522313, "optim/total_tokens": 5185732608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390204, "loss/out": 2.8932743072509766, "created_at": "2025-01-16T09:29:42.343560+00:00"} {"global_step": 9892, "acc_step": 0, "speed/wps": 12905.240149488225, "speed/FLOPS": 202694530206032.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051375824958086014, "optim/lr": 0.002941842734413256, "optim/total_tokens": 5186256896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.7954368591308594, "created_at": "2025-01-16T09:29:52.504000+00:00"} {"global_step": 9893, "acc_step": 0, "speed/wps": 12910.592417308393, "speed/FLOPS": 202778594926935.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044231709092855453, "optim/lr": 0.0029418191055999338, "optim/total_tokens": 5186781184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9244799613952637, "created_at": "2025-01-16T09:30:02.660424+00:00"} {"global_step": 9894, "acc_step": 0, "speed/wps": 12903.906262834485, "speed/FLOPS": 202673579683183.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04553793743252754, "optim/lr": 0.0029417954720824227, "optim/total_tokens": 5187305472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 2.952631950378418, "created_at": "2025-01-16T09:30:12.824471+00:00"} {"global_step": 9895, "acc_step": 0, "speed/wps": 12903.270584005151, "speed/FLOPS": 202663595473654.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07110120356082916, "optim/lr": 0.002941771833860801, "optim/total_tokens": 5187829760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9320952892303467, "created_at": "2025-01-16T09:30:22.993864+00:00"} {"global_step": 9896, "acc_step": 0, "speed/wps": 12907.267292183782, "speed/FLOPS": 202726369267652.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07489313930273056, "optim/lr": 0.0029417481909351454, "optim/total_tokens": 5188354048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359782, "loss/out": 2.836970329284668, "created_at": "2025-01-16T09:30:33.153246+00:00"} {"global_step": 9897, "acc_step": 0, "speed/wps": 12900.151781484632, "speed/FLOPS": 202614610394385.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047907859086990356, "optim/lr": 0.0029417245433055327, "optim/total_tokens": 5188878336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.8871943950653076, "created_at": "2025-01-16T09:30:43.319984+00:00"} {"global_step": 9898, "acc_step": 0, "speed/wps": 12908.526760655031, "speed/FLOPS": 202746150950684.25, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07201246172189713, "optim/lr": 0.0029417008909720407, "optim/total_tokens": 5189402624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.947983503341675, "created_at": "2025-01-16T09:30:53.478473+00:00"} {"global_step": 9899, "acc_step": 0, "speed/wps": 12907.267301653756, "speed/FLOPS": 202726369416391.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0674554780125618, "optim/lr": 0.002941677233934745, "optim/total_tokens": 5189926912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.7852344512939453, "created_at": "2025-01-16T09:31:03.638132+00:00"} {"global_step": 9900, "acc_step": 0, "speed/wps": 12895.426969183427, "speed/FLOPS": 202540400724622.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05059346556663513, "optim/lr": 0.002941653572193726, "optim/total_tokens": 5190451200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.849480628967285, "created_at": "2025-01-16T09:31:13.808944+00:00"} {"global_step": 9901, "acc_step": 0, "speed/wps": 12897.46165681317, "speed/FLOPS": 202572358289801.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06764701753854752, "optim/lr": 0.0029416299057490568, "optim/total_tokens": 5190975488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 2.896660327911377, "created_at": "2025-01-16T09:31:23.975457+00:00"} {"global_step": 9902, "acc_step": 0, "speed/wps": 12896.316138061833, "speed/FLOPS": 202554366343704.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04969409853219986, "optim/lr": 0.0029416062346008178, "optim/total_tokens": 5191499776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.8866920471191406, "created_at": "2025-01-16T09:31:34.143279+00:00"} {"global_step": 9903, "acc_step": 0, "speed/wps": 12894.044736076592, "speed/FLOPS": 202518690854292.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08049359917640686, "optim/lr": 0.0029415825587490846, "optim/total_tokens": 5192024064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9568686485290527, "created_at": "2025-01-16T09:31:44.312195+00:00"} {"global_step": 9904, "acc_step": 0, "speed/wps": 12896.397385489114, "speed/FLOPS": 202555642446195.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05671076104044914, "optim/lr": 0.002941558878193935, "optim/total_tokens": 5192548352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423841, "loss/out": 2.922168493270874, "created_at": "2025-01-16T09:31:54.479239+00:00"} {"global_step": 9905, "acc_step": 0, "speed/wps": 12897.50796455473, "speed/FLOPS": 202573085616516.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08326680213212967, "optim/lr": 0.002941535192935447, "optim/total_tokens": 5193072640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.0585780143737793, "created_at": "2025-01-16T09:32:04.649410+00:00"} {"global_step": 9906, "acc_step": 0, "speed/wps": 12898.11393056144, "speed/FLOPS": 202582603145336.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05087030306458473, "optim/lr": 0.0029415115029736965, "optim/total_tokens": 5193596928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9638452529907227, "created_at": "2025-01-16T09:32:14.820400+00:00"} {"global_step": 9907, "acc_step": 0, "speed/wps": 12912.696255866396, "speed/FLOPS": 202811638602465.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05754125863313675, "optim/lr": 0.002941487808308761, "optim/total_tokens": 5194121216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 2.7660651206970215, "created_at": "2025-01-16T09:32:24.977972+00:00"} {"global_step": 9908, "acc_step": 0, "speed/wps": 12906.113841133922, "speed/FLOPS": 202708252733906.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05743033066391945, "optim/lr": 0.002941464108940719, "optim/total_tokens": 5194645504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.748929262161255, "created_at": "2025-01-16T09:32:35.144968+00:00"} {"global_step": 9909, "acc_step": 0, "speed/wps": 12903.207857369043, "speed/FLOPS": 202662610265638.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04966999217867851, "optim/lr": 0.0029414404048696465, "optim/total_tokens": 5195169792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.834561347961426, "created_at": "2025-01-16T09:32:45.310188+00:00"} {"global_step": 9910, "acc_step": 0, "speed/wps": 12904.757152947797, "speed/FLOPS": 202686944081658.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0811588242650032, "optim/lr": 0.0029414166960956213, "optim/total_tokens": 5195694080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.888758897781372, "created_at": "2025-01-16T09:32:55.474915+00:00"} {"global_step": 9911, "acc_step": 0, "speed/wps": 12906.916378807138, "speed/FLOPS": 202720857690866.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06400786340236664, "optim/lr": 0.002941392982618721, "optim/total_tokens": 5196218368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.00738525390625, "created_at": "2025-01-16T09:33:05.633996+00:00"} {"global_step": 9912, "acc_step": 0, "speed/wps": 12903.829859093761, "speed/FLOPS": 202672379657445.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05986858159303665, "optim/lr": 0.002941369264439023, "optim/total_tokens": 5196742656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.933023452758789, "created_at": "2025-01-16T09:33:15.799475+00:00"} {"global_step": 9913, "acc_step": 0, "speed/wps": 12908.399954325425, "speed/FLOPS": 202744159283027.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05679568648338318, "optim/lr": 0.002941345541556604, "optim/total_tokens": 5197266944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0160486698150635, "created_at": "2025-01-16T09:33:25.958041+00:00"} {"global_step": 9914, "acc_step": 0, "speed/wps": 12906.177691991486, "speed/FLOPS": 202709255599364.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055020663887262344, "optim/lr": 0.002941321813971542, "optim/total_tokens": 5197791232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.995213508605957, "created_at": "2025-01-16T09:33:36.120562+00:00"} {"global_step": 9915, "acc_step": 0, "speed/wps": 12900.38573660528, "speed/FLOPS": 202618284981043.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05098818987607956, "optim/lr": 0.002941298081683914, "optim/total_tokens": 5198315520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 2.921053409576416, "created_at": "2025-01-16T09:33:46.286308+00:00"} {"global_step": 9916, "acc_step": 0, "speed/wps": 12907.55072526712, "speed/FLOPS": 202730820973703.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04927723482251167, "optim/lr": 0.002941274344693798, "optim/total_tokens": 5198839808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.8732552528381348, "created_at": "2025-01-16T09:33:56.448751+00:00"} {"global_step": 9917, "acc_step": 0, "speed/wps": 12908.28894414984, "speed/FLOPS": 202742415715676.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0461171492934227, "optim/lr": 0.0029412506030012716, "optim/total_tokens": 5199364096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374658, "loss/out": 2.8363704681396484, "created_at": "2025-01-16T09:34:06.610184+00:00"} {"global_step": 9918, "acc_step": 0, "speed/wps": 12907.245641491456, "speed/FLOPS": 202726029213778.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053943783044815063, "optim/lr": 0.002941226856606411, "optim/total_tokens": 5199888384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.999086856842041, "created_at": "2025-01-16T09:34:16.768713+00:00"} {"global_step": 9919, "acc_step": 0, "speed/wps": 12905.330204953083, "speed/FLOPS": 202695944650857.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049930550158023834, "optim/lr": 0.002941203105509295, "optim/total_tokens": 5200412672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.035374402999878, "created_at": "2025-01-16T09:34:26.928877+00:00"} {"global_step": 9920, "acc_step": 0, "speed/wps": 12900.543693687245, "speed/FLOPS": 202620765914071.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049056246876716614, "optim/lr": 0.0029411793497100005, "optim/total_tokens": 5200936960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.847506046295166, "created_at": "2025-01-16T09:34:37.094700+00:00"} {"global_step": 9921, "acc_step": 0, "speed/wps": 12905.07058473108, "speed/FLOPS": 202691866958515.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0684518814086914, "optim/lr": 0.002941155589208605, "optim/total_tokens": 5201461248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.041124105453491, "created_at": "2025-01-16T09:34:47.260938+00:00"} {"global_step": 9922, "acc_step": 0, "speed/wps": 12902.330918325217, "speed/FLOPS": 202648836732914.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05264484882354736, "optim/lr": 0.0029411318240051866, "optim/total_tokens": 5201985536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 2.9730749130249023, "created_at": "2025-01-16T09:34:57.424913+00:00"} {"global_step": 9923, "acc_step": 0, "speed/wps": 12901.285081415223, "speed/FLOPS": 202632410427113.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06236462667584419, "optim/lr": 0.002941108054099822, "optim/total_tokens": 5202509824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9250826835632324, "created_at": "2025-01-16T09:35:07.588181+00:00"} {"global_step": 9924, "acc_step": 0, "speed/wps": 12895.819623565842, "speed/FLOPS": 202546567901262.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060677915811538696, "optim/lr": 0.002941084279492589, "optim/total_tokens": 5203034112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.975052833557129, "created_at": "2025-01-16T09:35:17.760249+00:00"} {"global_step": 9925, "acc_step": 0, "speed/wps": 12895.581334481341, "speed/FLOPS": 202542825243746.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0706624686717987, "optim/lr": 0.0029410605001835653, "optim/total_tokens": 5203558400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427836, "loss/out": 2.844383955001831, "created_at": "2025-01-16T09:35:27.929271+00:00"} {"global_step": 9926, "acc_step": 0, "speed/wps": 12893.320875160156, "speed/FLOPS": 202507321623910.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052046336233615875, "optim/lr": 0.0029410367161728287, "optim/total_tokens": 5204082688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.8771510124206543, "created_at": "2025-01-16T09:35:38.099883+00:00"} {"global_step": 9927, "acc_step": 0, "speed/wps": 12887.876686276539, "speed/FLOPS": 202421813156394.22, "speed/curr_iter_time": 1.2882, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.04964959993958473, "optim/lr": 0.0029410129274604563, "optim/total_tokens": 5204606976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8998587131500244, "created_at": "2025-01-16T09:35:48.274157+00:00"} {"global_step": 9928, "acc_step": 0, "speed/wps": 12900.28752252378, "speed/FLOPS": 202616742393933.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06023356691002846, "optim/lr": 0.0029409891340465254, "optim/total_tokens": 5205131264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375623, "loss/out": 2.9118399620056152, "created_at": "2025-01-16T09:35:58.440638+00:00"} {"global_step": 9929, "acc_step": 0, "speed/wps": 12911.013415386069, "speed/FLOPS": 202785207280256.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05562923476099968, "optim/lr": 0.002940965335931115, "optim/total_tokens": 5205655552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.871377944946289, "created_at": "2025-01-16T09:36:08.598564+00:00"} {"global_step": 9930, "acc_step": 0, "speed/wps": 12908.185193279382, "speed/FLOPS": 202740786165686.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05888854339718819, "optim/lr": 0.002940941533114302, "optim/total_tokens": 5206179840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8782825469970703, "created_at": "2025-01-16T09:36:18.758301+00:00"} {"global_step": 9931, "acc_step": 0, "speed/wps": 12906.11622122369, "speed/FLOPS": 202708290116486.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06423396617174149, "optim/lr": 0.002940917725596164, "optim/total_tokens": 5206704128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 3.0052971839904785, "created_at": "2025-01-16T09:36:28.920624+00:00"} {"global_step": 9932, "acc_step": 0, "speed/wps": 12904.77816933193, "speed/FLOPS": 202687274172850.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044244520366191864, "optim/lr": 0.0029408939133767783, "optim/total_tokens": 5207228416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.9104251861572266, "created_at": "2025-01-16T09:36:39.084094+00:00"} {"global_step": 9933, "acc_step": 0, "speed/wps": 12903.578215151376, "speed/FLOPS": 202668427243535.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05776834115386009, "optim/lr": 0.002940870096456223, "optim/total_tokens": 5207752704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.906541585922241, "created_at": "2025-01-16T09:36:49.247981+00:00"} {"global_step": 9934, "acc_step": 0, "speed/wps": 12894.477974743799, "speed/FLOPS": 202525495462893.5, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045218631625175476, "optim/lr": 0.002940846274834576, "optim/total_tokens": 5208276992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.880009651184082, "created_at": "2025-01-16T09:36:59.418175+00:00"} {"global_step": 9935, "acc_step": 0, "speed/wps": 12897.300953953702, "speed/FLOPS": 202569834230564.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.002, "optim/grad_norm": 0.04815491661429405, "optim/lr": 0.002940822448511915, "optim/total_tokens": 5208801280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8775815963745117, "created_at": "2025-01-16T09:37:09.591648+00:00"} {"global_step": 9936, "acc_step": 0, "speed/wps": 12904.932108948411, "speed/FLOPS": 202689692006062.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05659738928079605, "optim/lr": 0.0029407986174883175, "optim/total_tokens": 5209325568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.9732935428619385, "created_at": "2025-01-16T09:37:19.752301+00:00"} {"global_step": 9937, "acc_step": 0, "speed/wps": 12904.781454146536, "speed/FLOPS": 202687325765377.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05012197047472, "optim/lr": 0.0029407747817638604, "optim/total_tokens": 5209849856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.935091972351074, "created_at": "2025-01-16T09:37:29.915029+00:00"} {"global_step": 9938, "acc_step": 0, "speed/wps": 12905.607723328269, "speed/FLOPS": 202700303458289.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06554152071475983, "optim/lr": 0.0029407509413386236, "optim/total_tokens": 5210374144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.016108989715576, "created_at": "2025-01-16T09:37:40.075606+00:00"} {"global_step": 9939, "acc_step": 0, "speed/wps": 12908.408998560551, "speed/FLOPS": 202744301335168.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052564170211553574, "optim/lr": 0.0029407270962126834, "optim/total_tokens": 5210898432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 2.95896577835083, "created_at": "2025-01-16T09:37:50.236378+00:00"} {"global_step": 9940, "acc_step": 0, "speed/wps": 12905.638850305297, "speed/FLOPS": 202700792350237.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06100783124566078, "optim/lr": 0.002940703246386117, "optim/total_tokens": 5211422720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9108877182006836, "created_at": "2025-01-16T09:38:00.396179+00:00"} {"global_step": 9941, "acc_step": 0, "speed/wps": 12905.340618806702, "speed/FLOPS": 202696108214738.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06347601860761642, "optim/lr": 0.002940679391859004, "optim/total_tokens": 5211947008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9073238372802734, "created_at": "2025-01-16T09:38:10.557810+00:00"} {"global_step": 9942, "acc_step": 0, "speed/wps": 12895.296645032307, "speed/FLOPS": 202538353804745.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0486944280564785, "optim/lr": 0.0029406555326314207, "optim/total_tokens": 5212471296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.97346568107605, "created_at": "2025-01-16T09:38:20.728121+00:00"} {"global_step": 9943, "acc_step": 0, "speed/wps": 12906.571383665825, "speed/FLOPS": 202715439068099.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058379098773002625, "optim/lr": 0.002940631668703446, "optim/total_tokens": 5212995584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406246, "loss/out": 2.8168766498565674, "created_at": "2025-01-16T09:38:30.889366+00:00"} {"global_step": 9944, "acc_step": 0, "speed/wps": 12902.29939922452, "speed/FLOPS": 202648341682126.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040034763514995575, "optim/lr": 0.002940607800075157, "optim/total_tokens": 5213519872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.9476943016052246, "created_at": "2025-01-16T09:38:41.053658+00:00"} {"global_step": 9945, "acc_step": 0, "speed/wps": 12901.32417756945, "speed/FLOPS": 202633024485938.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05114240199327469, "optim/lr": 0.002940583926746632, "optim/total_tokens": 5214044160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9577481746673584, "created_at": "2025-01-16T09:38:51.219740+00:00"} {"global_step": 9946, "acc_step": 0, "speed/wps": 12903.539372706966, "speed/FLOPS": 202667817169570.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04992484673857689, "optim/lr": 0.0029405600487179486, "optim/total_tokens": 5214568448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418245, "loss/out": 3.025210380554199, "created_at": "2025-01-16T09:39:01.381781+00:00"} {"global_step": 9947, "acc_step": 0, "speed/wps": 12898.130525641587, "speed/FLOPS": 202582863793874.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04921454191207886, "optim/lr": 0.002940536165989185, "optim/total_tokens": 5215092736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.900583267211914, "created_at": "2025-01-16T09:39:11.547876+00:00"} {"global_step": 9948, "acc_step": 0, "speed/wps": 12904.836330677063, "speed/FLOPS": 202688187676698.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05170712247490883, "optim/lr": 0.0029405122785604192, "optim/total_tokens": 5215617024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417160, "loss/out": 2.914407253265381, "created_at": "2025-01-16T09:39:21.711882+00:00"} {"global_step": 9949, "acc_step": 0, "speed/wps": 12905.12230534972, "speed/FLOPS": 202692679301902.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05359230190515518, "optim/lr": 0.0029404883864317284, "optim/total_tokens": 5216141312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.9815046787261963, "created_at": "2025-01-16T09:39:31.874299+00:00"} {"global_step": 9950, "acc_step": 0, "speed/wps": 12904.937430333872, "speed/FLOPS": 202689775585733.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05716972425580025, "optim/lr": 0.0029404644896031916, "optim/total_tokens": 5216665600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.8841304779052734, "created_at": "2025-01-16T09:39:42.034662+00:00"} {"global_step": 9951, "acc_step": 0, "speed/wps": 12899.567513697742, "speed/FLOPS": 202605433665924.6, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044407300651073456, "optim/lr": 0.002940440588074886, "optim/total_tokens": 5217189888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373603, "loss/out": 2.8911492824554443, "created_at": "2025-01-16T09:39:52.199167+00:00"} {"global_step": 9952, "acc_step": 0, "speed/wps": 12907.339136543374, "speed/FLOPS": 202727497682044.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05149480700492859, "optim/lr": 0.00294041668184689, "optim/total_tokens": 5217714176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.906733751296997, "created_at": "2025-01-16T09:40:02.363625+00:00"} {"global_step": 9953, "acc_step": 0, "speed/wps": 12903.020477257238, "speed/FLOPS": 202659667203495.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05170963332056999, "optim/lr": 0.0029403927709192815, "optim/total_tokens": 5218238464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.857419490814209, "created_at": "2025-01-16T09:40:12.530943+00:00"} {"global_step": 9954, "acc_step": 0, "speed/wps": 12891.20143275747, "speed/FLOPS": 202474032868557.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08673466742038727, "optim/lr": 0.0029403688552921384, "optim/total_tokens": 5218762752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 2.795593500137329, "created_at": "2025-01-16T09:40:22.708689+00:00"} {"global_step": 9955, "acc_step": 0, "speed/wps": 12891.701772052751, "speed/FLOPS": 202481891384728.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05955890566110611, "optim/lr": 0.0029403449349655383, "optim/total_tokens": 5219287040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.957899570465088, "created_at": "2025-01-16T09:40:32.880614+00:00"} {"global_step": 9956, "acc_step": 0, "speed/wps": 12902.444486303828, "speed/FLOPS": 202650620474077.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04903919994831085, "optim/lr": 0.00294032100993956, "optim/total_tokens": 5219811328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9893293380737305, "created_at": "2025-01-16T09:40:43.047500+00:00"} {"global_step": 9957, "acc_step": 0, "speed/wps": 12901.70470984216, "speed/FLOPS": 202639001268187.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06994668394327164, "optim/lr": 0.002940297080214282, "optim/total_tokens": 5220335616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 2.7804622650146484, "created_at": "2025-01-16T09:40:53.210289+00:00"} {"global_step": 9958, "acc_step": 0, "speed/wps": 12902.654870080074, "speed/FLOPS": 202653924840381.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05204389616847038, "optim/lr": 0.0029402731457897806, "optim/total_tokens": 5220859904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.9500536918640137, "created_at": "2025-01-16T09:41:03.372804+00:00"} {"global_step": 9959, "acc_step": 0, "speed/wps": 12906.115764378943, "speed/FLOPS": 202708282941112.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06446853280067444, "optim/lr": 0.002940249206666135, "optim/total_tokens": 5221384192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.896580219268799, "created_at": "2025-01-16T09:41:13.534402+00:00"} {"global_step": 9960, "acc_step": 0, "speed/wps": 12899.607172878703, "speed/FLOPS": 202606056567859.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058816250413656235, "optim/lr": 0.002940225262843424, "optim/total_tokens": 5221908480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 2.8891425132751465, "created_at": "2025-01-16T09:41:23.698689+00:00"} {"global_step": 9961, "acc_step": 0, "speed/wps": 12901.13771784256, "speed/FLOPS": 202630095879705.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048601970076560974, "optim/lr": 0.0029402013143217244, "optim/total_tokens": 5222432768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.920125961303711, "created_at": "2025-01-16T09:41:33.866857+00:00"} {"global_step": 9962, "acc_step": 0, "speed/wps": 12901.970315512955, "speed/FLOPS": 202643172970228.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053024739027023315, "optim/lr": 0.002940177361101115, "optim/total_tokens": 5222957056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346666, "loss/out": 2.9563260078430176, "created_at": "2025-01-16T09:41:44.030524+00:00"} {"global_step": 9963, "acc_step": 0, "speed/wps": 12890.1672786199, "speed/FLOPS": 202457790056750.44, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.002, "optim/grad_norm": 0.05971447005867958, "optim/lr": 0.0029401534031816743, "optim/total_tokens": 5223481344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475383, "loss/out": 2.87814998626709, "created_at": "2025-01-16T09:41:54.211653+00:00"} {"global_step": 9964, "acc_step": 0, "speed/wps": 12897.99710779587, "speed/FLOPS": 202580768283271.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053221385926008224, "optim/lr": 0.0029401294405634795, "optim/total_tokens": 5224005632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477322, "loss/out": 2.894575595855713, "created_at": "2025-01-16T09:42:04.380785+00:00"} {"global_step": 9965, "acc_step": 0, "speed/wps": 12889.66342714492, "speed/FLOPS": 202449876376971.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05134350433945656, "optim/lr": 0.002940105473246609, "optim/total_tokens": 5224529920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.831589937210083, "created_at": "2025-01-16T09:42:14.553951+00:00"} {"global_step": 9966, "acc_step": 0, "speed/wps": 12891.928522930099, "speed/FLOPS": 202485452818845.4, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.002, "optim/grad_norm": 0.06800593435764313, "optim/lr": 0.0029400815012311427, "optim/total_tokens": 5225054208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377259, "loss/out": 2.9607784748077393, "created_at": "2025-01-16T09:42:24.725437+00:00"} {"global_step": 9967, "acc_step": 0, "speed/wps": 12905.217679603114, "speed/FLOPS": 202694177285610.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05045969784259796, "optim/lr": 0.002940057524517156, "optim/total_tokens": 5225578496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.812570333480835, "created_at": "2025-01-16T09:42:34.885578+00:00"} {"global_step": 9968, "acc_step": 0, "speed/wps": 12902.12327029166, "speed/FLOPS": 202645575335209.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06249115988612175, "optim/lr": 0.0029400335431047286, "optim/total_tokens": 5226102784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386324, "loss/out": 2.9644739627838135, "created_at": "2025-01-16T09:42:45.048553+00:00"} {"global_step": 9969, "acc_step": 0, "speed/wps": 12901.724047452604, "speed/FLOPS": 202639304991932.62, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058475006371736526, "optim/lr": 0.0029400095569939396, "optim/total_tokens": 5226627072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9195938110351562, "created_at": "2025-01-16T09:42:55.211627+00:00"} {"global_step": 9970, "acc_step": 0, "speed/wps": 12903.358754626606, "speed/FLOPS": 202664980314423.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05655563622713089, "optim/lr": 0.002939985566184866, "optim/total_tokens": 5227151360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8475162982940674, "created_at": "2025-01-16T09:43:05.373307+00:00"} {"global_step": 9971, "acc_step": 0, "speed/wps": 12899.247484866411, "speed/FLOPS": 202600407173362.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.003, "optim/grad_norm": 0.051170844584703445, "optim/lr": 0.0029399615706775864, "optim/total_tokens": 5227675648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.8160555362701416, "created_at": "2025-01-16T09:43:15.541193+00:00"} {"global_step": 9972, "acc_step": 0, "speed/wps": 12902.451715962454, "speed/FLOPS": 202650734025801.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05656563490629196, "optim/lr": 0.002939937570472179, "optim/total_tokens": 5228199936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.909881591796875, "created_at": "2025-01-16T09:43:25.703521+00:00"} {"global_step": 9973, "acc_step": 0, "speed/wps": 12902.70104951388, "speed/FLOPS": 202654650151846.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04981778562068939, "optim/lr": 0.0029399135655687222, "optim/total_tokens": 5228724224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.845393180847168, "created_at": "2025-01-16T09:43:35.865758+00:00"} {"global_step": 9974, "acc_step": 0, "speed/wps": 12901.585657671654, "speed/FLOPS": 202637131390255.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0453181192278862, "optim/lr": 0.0029398895559672946, "optim/total_tokens": 5229248512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466629, "loss/out": 2.880796432495117, "created_at": "2025-01-16T09:43:46.031386+00:00"} {"global_step": 9975, "acc_step": 0, "speed/wps": 12903.900115292005, "speed/FLOPS": 202673483127580.97, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04491744935512543, "optim/lr": 0.0029398655416679738, "optim/total_tokens": 5229772800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9519710540771484, "created_at": "2025-01-16T09:43:56.192447+00:00"} {"global_step": 9976, "acc_step": 0, "speed/wps": 12903.585495769594, "speed/FLOPS": 202668541595648.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046241678297519684, "optim/lr": 0.0029398415226708392, "optim/total_tokens": 5230297088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 2.839393138885498, "created_at": "2025-01-16T09:44:06.354367+00:00"} {"global_step": 9977, "acc_step": 0, "speed/wps": 12897.349766228945, "speed/FLOPS": 202570600894422.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04056365042924881, "optim/lr": 0.002939817498975968, "optim/total_tokens": 5230821376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.96212100982666, "created_at": "2025-01-16T09:44:16.520538+00:00"} {"global_step": 9978, "acc_step": 0, "speed/wps": 12904.745392274897, "speed/FLOPS": 202686759364129.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04956365376710892, "optim/lr": 0.00293979347058344, "optim/total_tokens": 5231345664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.887455940246582, "created_at": "2025-01-16T09:44:26.681239+00:00"} {"global_step": 9979, "acc_step": 0, "speed/wps": 12902.769029307996, "speed/FLOPS": 202655717867927.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04236166179180145, "optim/lr": 0.002939769437493332, "optim/total_tokens": 5231869952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 2.9841084480285645, "created_at": "2025-01-16T09:44:36.846260+00:00"} {"global_step": 9980, "acc_step": 0, "speed/wps": 12895.323363525407, "speed/FLOPS": 202538773455394.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06058725714683533, "optim/lr": 0.0029397453997057235, "optim/total_tokens": 5232394240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.967264175415039, "created_at": "2025-01-16T09:44:47.014848+00:00"} {"global_step": 9981, "acc_step": 0, "speed/wps": 12904.103189608117, "speed/FLOPS": 202676672688768.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05734128877520561, "optim/lr": 0.0029397213572206924, "optim/total_tokens": 5232918528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382133, "loss/out": 2.9576797485351562, "created_at": "2025-01-16T09:44:57.182998+00:00"} {"global_step": 9982, "acc_step": 0, "speed/wps": 12898.55049503208, "speed/FLOPS": 202589459990250.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05316748842597008, "optim/lr": 0.0029396973100383173, "optim/total_tokens": 5233442816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.11118221282959, "created_at": "2025-01-16T09:45:07.349892+00:00"} {"global_step": 9983, "acc_step": 0, "speed/wps": 12897.09042917379, "speed/FLOPS": 202566527649605.06, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04920141398906708, "optim/lr": 0.002939673258158677, "optim/total_tokens": 5233967104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 2.8803226947784424, "created_at": "2025-01-16T09:45:17.516512+00:00"} {"global_step": 9984, "acc_step": 0, "speed/wps": 12902.97083301545, "speed/FLOPS": 202658887472460.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05148414522409439, "optim/lr": 0.0029396492015818496, "optim/total_tokens": 5234491392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489248, "loss/out": 2.8722920417785645, "created_at": "2025-01-16T09:45:27.678481+00:00"} {"global_step": 9985, "acc_step": 0, "speed/wps": 12907.390928616964, "speed/FLOPS": 202728311147729.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05831846222281456, "optim/lr": 0.0029396251403079137, "optim/total_tokens": 5235015680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9746265411376953, "created_at": "2025-01-16T09:45:37.839687+00:00"} {"global_step": 9986, "acc_step": 0, "speed/wps": 12901.44523428577, "speed/FLOPS": 202634925848018.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05459881201386452, "optim/lr": 0.0029396010743369476, "optim/total_tokens": 5235539968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.816830635070801, "created_at": "2025-01-16T09:45:48.005122+00:00"} {"global_step": 9987, "acc_step": 0, "speed/wps": 12904.161051441253, "speed/FLOPS": 202677581488368.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04827594384551048, "optim/lr": 0.0029395770036690304, "optim/total_tokens": 5236064256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 2.9110448360443115, "created_at": "2025-01-16T09:45:58.166638+00:00"} {"global_step": 9988, "acc_step": 0, "speed/wps": 12902.158044806421, "speed/FLOPS": 202646121516749.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05519188940525055, "optim/lr": 0.0029395529283042397, "optim/total_tokens": 5236588544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 2.988555431365967, "created_at": "2025-01-16T09:46:08.333440+00:00"} {"global_step": 9989, "acc_step": 0, "speed/wps": 12891.008787828736, "speed/FLOPS": 202471007115230.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051531463861465454, "optim/lr": 0.002939528848242655, "optim/total_tokens": 5237112832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.941260814666748, "created_at": "2025-01-16T09:46:18.506008+00:00"} {"global_step": 9990, "acc_step": 0, "speed/wps": 12901.539230759161, "speed/FLOPS": 202636402191797.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060562971979379654, "optim/lr": 0.002939504763484354, "optim/total_tokens": 5237637120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.809631109237671, "created_at": "2025-01-16T09:46:28.668929+00:00"} {"global_step": 9991, "acc_step": 0, "speed/wps": 12899.634863391404, "speed/FLOPS": 202606491485411.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05560465157032013, "optim/lr": 0.0029394806740294164, "optim/total_tokens": 5238161408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.872964859008789, "created_at": "2025-01-16T09:46:38.834946+00:00"} {"global_step": 9992, "acc_step": 0, "speed/wps": 12903.848836308096, "speed/FLOPS": 202672677720673.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06761856377124786, "optim/lr": 0.0029394565798779196, "optim/total_tokens": 5238685696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 3.074608325958252, "created_at": "2025-01-16T09:46:48.997455+00:00"} {"global_step": 9993, "acc_step": 0, "speed/wps": 12899.446021656275, "speed/FLOPS": 202603525466465.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06630914658308029, "optim/lr": 0.002939432481029943, "optim/total_tokens": 5239209984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.968597412109375, "created_at": "2025-01-16T09:46:59.162320+00:00"} {"global_step": 9994, "acc_step": 0, "speed/wps": 12900.215485861503, "speed/FLOPS": 202615610959163.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04764851555228233, "optim/lr": 0.0029394083774855646, "optim/total_tokens": 5239734272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.8316516876220703, "created_at": "2025-01-16T09:47:09.328558+00:00"} {"global_step": 9995, "acc_step": 0, "speed/wps": 12898.470733604534, "speed/FLOPS": 202588207227426.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06693017482757568, "optim/lr": 0.002939384269244864, "optim/total_tokens": 5240258560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.958874225616455, "created_at": "2025-01-16T09:47:19.495732+00:00"} {"global_step": 9996, "acc_step": 0, "speed/wps": 12900.528555162848, "speed/FLOPS": 202620528142743.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06691914051771164, "optim/lr": 0.002939360156307919, "optim/total_tokens": 5240782848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 2.892993927001953, "created_at": "2025-01-16T09:47:29.659481+00:00"} {"global_step": 9997, "acc_step": 0, "speed/wps": 12901.519192365937, "speed/FLOPS": 202636087461296.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047955822199583054, "optim/lr": 0.002939336038674809, "optim/total_tokens": 5241307136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.860900640487671, "created_at": "2025-01-16T09:47:39.824988+00:00"} {"global_step": 9998, "acc_step": 0, "speed/wps": 12889.711559487112, "speed/FLOPS": 202450632361546.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07049544155597687, "optim/lr": 0.0029393119163456114, "optim/total_tokens": 5241831424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 2.87227463722229, "created_at": "2025-01-16T09:47:50.002150+00:00"} {"global_step": 9999, "acc_step": 0, "speed/wps": 12897.624422684057, "speed/FLOPS": 202574914751467.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04938818886876106, "optim/lr": 0.002939287789320406, "optim/total_tokens": 5242355712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.854382038116455, "created_at": "2025-01-16T09:48:00.168458+00:00"} {"global_step": 10000, "acc_step": 0, "speed/wps": 12899.555810957672, "speed/FLOPS": 202605249858310.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05784224718809128, "optim/lr": 0.002939263657599271, "optim/total_tokens": 5242880000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8648669719696045, "created_at": "2025-01-16T09:48:10.335452+00:00"} {"global_step": 10001, "acc_step": 0, "speed/wps": 5181.7708986574635, "speed/FLOPS": 81386832462805.23, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.2481, "optim/grad_norm": 0.05103256180882454, "optim/lr": 0.002939239521182286, "optim/total_tokens": 5243404288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9433860778808594, "created_at": "2025-01-16T09:48:35.636464+00:00"} {"global_step": 10002, "acc_step": 0, "speed/wps": 12947.706548717204, "speed/FLOPS": 203361523360871.38, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060307059437036514, "optim/lr": 0.002939215380069528, "optim/total_tokens": 5243928576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 2.864008665084839, "created_at": "2025-01-16T09:48:45.765060+00:00"} {"global_step": 10003, "acc_step": 0, "speed/wps": 12915.213343891373, "speed/FLOPS": 202851172928739.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06708858162164688, "optim/lr": 0.0029391912342610783, "optim/total_tokens": 5244452864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398145, "loss/out": 2.8679771423339844, "created_at": "2025-01-16T09:48:55.918687+00:00"} {"global_step": 10004, "acc_step": 0, "speed/wps": 12909.057510001303, "speed/FLOPS": 202754487098493.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05775841325521469, "optim/lr": 0.0029391670837570134, "optim/total_tokens": 5244977152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.8772900104522705, "created_at": "2025-01-16T09:49:06.078813+00:00"} {"global_step": 10005, "acc_step": 0, "speed/wps": 12908.78984825404, "speed/FLOPS": 202750283102946.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05567999556660652, "optim/lr": 0.0029391429285574133, "optim/total_tokens": 5245501440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.884047508239746, "created_at": "2025-01-16T09:49:16.236887+00:00"} {"global_step": 10006, "acc_step": 0, "speed/wps": 12914.850850905234, "speed/FLOPS": 202845479478274.34, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06064530462026596, "optim/lr": 0.0029391187686623565, "optim/total_tokens": 5246025728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.019425392150879, "created_at": "2025-01-16T09:49:26.390322+00:00"} {"global_step": 10007, "acc_step": 0, "speed/wps": 12910.649653971572, "speed/FLOPS": 202779493907381.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04761384427547455, "optim/lr": 0.0029390946040719216, "optim/total_tokens": 5246550016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8138954639434814, "created_at": "2025-01-16T09:49:36.549166+00:00"} {"global_step": 10008, "acc_step": 0, "speed/wps": 12910.21731517886, "speed/FLOPS": 202772703432546.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053951505571603775, "optim/lr": 0.0029390704347861877, "optim/total_tokens": 5247074304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.8419511318206787, "created_at": "2025-01-16T09:49:46.705802+00:00"} {"global_step": 10009, "acc_step": 0, "speed/wps": 12900.850064283244, "speed/FLOPS": 202625577885279.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048115964978933334, "optim/lr": 0.002939046260805234, "optim/total_tokens": 5247598592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.897667646408081, "created_at": "2025-01-16T09:49:56.874523+00:00"} {"global_step": 10010, "acc_step": 0, "speed/wps": 12906.051432439932, "speed/FLOPS": 202707272519608.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04595139995217323, "optim/lr": 0.002939022082129139, "optim/total_tokens": 5248122880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 2.7702720165252686, "created_at": "2025-01-16T09:50:07.034222+00:00"} {"global_step": 10011, "acc_step": 0, "speed/wps": 12905.899299353056, "speed/FLOPS": 202704883060426.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05364993214607239, "optim/lr": 0.002938997898757981, "optim/total_tokens": 5248647168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.9597320556640625, "created_at": "2025-01-16T09:50:17.196732+00:00"} {"global_step": 10012, "acc_step": 0, "speed/wps": 12900.186436587612, "speed/FLOPS": 202615154700399.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05187895521521568, "optim/lr": 0.0029389737106918393, "optim/total_tokens": 5249171456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.986907482147217, "created_at": "2025-01-16T09:50:27.360737+00:00"} {"global_step": 10013, "acc_step": 0, "speed/wps": 12893.446220724978, "speed/FLOPS": 202509290348249.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05163579061627388, "optim/lr": 0.002938949517930794, "optim/total_tokens": 5249695744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 3.0008201599121094, "created_at": "2025-01-16T09:50:37.530834+00:00"} {"global_step": 10014, "acc_step": 0, "speed/wps": 12889.298570607218, "speed/FLOPS": 202444145803681.38, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0601254366338253, "optim/lr": 0.002938925320474922, "optim/total_tokens": 5250220032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8930530548095703, "created_at": "2025-01-16T09:50:47.704159+00:00"} {"global_step": 10015, "acc_step": 0, "speed/wps": 12888.418366773254, "speed/FLOPS": 202430320992942.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.044400863349437714, "optim/lr": 0.0029389011183243038, "optim/total_tokens": 5250744320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.754265546798706, "created_at": "2025-01-16T09:50:57.879554+00:00"} {"global_step": 10016, "acc_step": 0, "speed/wps": 12902.856253822405, "speed/FLOPS": 202657087848784.94, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058768268674612045, "optim/lr": 0.0029388769114790173, "optim/total_tokens": 5251268608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348249, "loss/out": 2.9318108558654785, "created_at": "2025-01-16T09:51:08.046253+00:00"} {"global_step": 10017, "acc_step": 0, "speed/wps": 12902.744602018254, "speed/FLOPS": 202655334203775.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05487446486949921, "optim/lr": 0.0029388526999391427, "optim/total_tokens": 5251792896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.945587396621704, "created_at": "2025-01-16T09:51:18.208346+00:00"} {"global_step": 10018, "acc_step": 0, "speed/wps": 12897.907577811582, "speed/FLOPS": 202579362091841.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057266172021627426, "optim/lr": 0.0029388284837047578, "optim/total_tokens": 5252317184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.917877435684204, "created_at": "2025-01-16T09:51:28.378579+00:00"} {"global_step": 10019, "acc_step": 0, "speed/wps": 12898.864721293512, "speed/FLOPS": 202594395345478.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05719758942723274, "optim/lr": 0.0029388042627759422, "optim/total_tokens": 5252841472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 274078, "loss/out": 2.88303804397583, "created_at": "2025-01-16T09:51:38.547568+00:00"} {"global_step": 10020, "acc_step": 0, "speed/wps": 12901.666389611071, "speed/FLOPS": 202638399396301.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05037137120962143, "optim/lr": 0.002938780037152775, "optim/total_tokens": 5253365760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 2.82570743560791, "created_at": "2025-01-16T09:51:48.711927+00:00"} {"global_step": 10021, "acc_step": 0, "speed/wps": 12906.67092206865, "speed/FLOPS": 202717002455495.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04813559725880623, "optim/lr": 0.002938755806835335, "optim/total_tokens": 5253890048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477309, "loss/out": 3.008270740509033, "created_at": "2025-01-16T09:51:58.874464+00:00"} {"global_step": 10022, "acc_step": 0, "speed/wps": 12896.632620242652, "speed/FLOPS": 202559337131246.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05413905158638954, "optim/lr": 0.0029387315718237017, "optim/total_tokens": 5254414336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.9264121055603027, "created_at": "2025-01-16T09:52:09.044649+00:00"} {"global_step": 10023, "acc_step": 0, "speed/wps": 12901.336077474427, "speed/FLOPS": 202633211390298.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05023648589849472, "optim/lr": 0.0029387073321179534, "optim/total_tokens": 5254938624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.8849663734436035, "created_at": "2025-01-16T09:52:19.211262+00:00"} {"global_step": 10024, "acc_step": 0, "speed/wps": 12902.086310155753, "speed/FLOPS": 202644994825486.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05116257444024086, "optim/lr": 0.0029386830877181696, "optim/total_tokens": 5255462912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.9577436447143555, "created_at": "2025-01-16T09:52:29.376026+00:00"} {"global_step": 10025, "acc_step": 0, "speed/wps": 12895.480602322119, "speed/FLOPS": 202541243106765.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04670696705579758, "optim/lr": 0.0029386588386244297, "optim/total_tokens": 5255987200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 3.0067524909973145, "created_at": "2025-01-16T09:52:39.543709+00:00"} {"global_step": 10026, "acc_step": 0, "speed/wps": 12902.29722872132, "speed/FLOPS": 202648307591391.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07201632112264633, "optim/lr": 0.002938634584836812, "optim/total_tokens": 5256511488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.95725154876709, "created_at": "2025-01-16T09:52:49.708252+00:00"} {"global_step": 10027, "acc_step": 0, "speed/wps": 12902.451401320728, "speed/FLOPS": 202650729083920.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06270086765289307, "optim/lr": 0.0029386103263553966, "optim/total_tokens": 5257035776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 2.9349658489227295, "created_at": "2025-01-16T09:52:59.870817+00:00"} {"global_step": 10028, "acc_step": 0, "speed/wps": 12898.40860331355, "speed/FLOPS": 202587231385831.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048363927751779556, "optim/lr": 0.0029385860631802618, "optim/total_tokens": 5257560064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.9285237789154053, "created_at": "2025-01-16T09:53:10.039398+00:00"} {"global_step": 10029, "acc_step": 0, "speed/wps": 12899.913919927376, "speed/FLOPS": 202610874451773.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05368826538324356, "optim/lr": 0.002938561795311488, "optim/total_tokens": 5258084352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9067296981811523, "created_at": "2025-01-16T09:53:20.203976+00:00"} {"global_step": 10030, "acc_step": 0, "speed/wps": 12904.654214951068, "speed/FLOPS": 202685327298968.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04865780845284462, "optim/lr": 0.0029385375227491523, "optim/total_tokens": 5258608640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.946305990219116, "created_at": "2025-01-16T09:53:30.366486+00:00"} {"global_step": 10031, "acc_step": 0, "speed/wps": 12899.607519304964, "speed/FLOPS": 202606062008959.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.058714259415864944, "optim/lr": 0.002938513245493336, "optim/total_tokens": 5259132928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8769097328186035, "created_at": "2025-01-16T09:53:40.532195+00:00"} {"global_step": 10032, "acc_step": 0, "speed/wps": 12903.096098926346, "speed/FLOPS": 202660854945724.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043284155428409576, "optim/lr": 0.002938488963544117, "optim/total_tokens": 5259657216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.932595729827881, "created_at": "2025-01-16T09:53:50.693823+00:00"} {"global_step": 10033, "acc_step": 0, "speed/wps": 12900.224213557944, "speed/FLOPS": 202615748039629.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052181001752614975, "optim/lr": 0.002938464676901575, "optim/total_tokens": 5260181504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.7310705184936523, "created_at": "2025-01-16T09:54:00.859535+00:00"} {"global_step": 10034, "acc_step": 0, "speed/wps": 12904.666588391508, "speed/FLOPS": 202685521640853.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04450954496860504, "optim/lr": 0.00293844038556579, "optim/total_tokens": 5260705792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.8528811931610107, "created_at": "2025-01-16T09:54:11.021644+00:00"} {"global_step": 10035, "acc_step": 0, "speed/wps": 12903.497427341339, "speed/FLOPS": 202667158359964.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05058574303984642, "optim/lr": 0.002938416089536839, "optim/total_tokens": 5261230080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9628982543945312, "created_at": "2025-01-16T09:54:21.183057+00:00"} {"global_step": 10036, "acc_step": 0, "speed/wps": 12904.346125473397, "speed/FLOPS": 202680488330361.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05188042297959328, "optim/lr": 0.0029383917888148037, "optim/total_tokens": 5261754368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 2.928445339202881, "created_at": "2025-01-16T09:54:31.345006+00:00"} {"global_step": 10037, "acc_step": 0, "speed/wps": 12901.687522480732, "speed/FLOPS": 202638731317058.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05638136342167854, "optim/lr": 0.0029383674833997624, "optim/total_tokens": 5262278656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 2.8388264179229736, "created_at": "2025-01-16T09:54:41.511364+00:00"} {"global_step": 10038, "acc_step": 0, "speed/wps": 12890.574795192735, "speed/FLOPS": 202464190664513.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04492677375674248, "optim/lr": 0.0029383431732917936, "optim/total_tokens": 5262802944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8842687606811523, "created_at": "2025-01-16T09:54:51.683305+00:00"} {"global_step": 10039, "acc_step": 0, "speed/wps": 12904.321331333847, "speed/FLOPS": 202680098904329.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04614337533712387, "optim/lr": 0.0029383188584909785, "optim/total_tokens": 5263327232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9138572216033936, "created_at": "2025-01-16T09:55:01.845159+00:00"} {"global_step": 10040, "acc_step": 0, "speed/wps": 12901.995410507267, "speed/FLOPS": 202643567121598.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.058606743812561035, "optim/lr": 0.0029382945389973946, "optim/total_tokens": 5263851520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8108341693878174, "created_at": "2025-01-16T09:55:12.008058+00:00"} {"global_step": 10041, "acc_step": 0, "speed/wps": 12909.025163516793, "speed/FLOPS": 202753979052505.53, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05565987527370453, "optim/lr": 0.0029382702148111227, "optim/total_tokens": 5264375808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9280624389648438, "created_at": "2025-01-16T09:55:22.165143+00:00"} {"global_step": 10042, "acc_step": 0, "speed/wps": 12904.721045064065, "speed/FLOPS": 202686376957726.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05468643829226494, "optim/lr": 0.002938245885932241, "optim/total_tokens": 5264900096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 2.9864871501922607, "created_at": "2025-01-16T09:55:32.326443+00:00"} {"global_step": 10043, "acc_step": 0, "speed/wps": 12901.931701560285, "speed/FLOPS": 202642566485040.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07750760763883591, "optim/lr": 0.0029382215523608294, "optim/total_tokens": 5265424384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.0620737075805664, "created_at": "2025-01-16T09:55:42.490636+00:00"} {"global_step": 10044, "acc_step": 0, "speed/wps": 12907.669356016046, "speed/FLOPS": 202732684232631.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057914748787879944, "optim/lr": 0.0029381972140969676, "optim/total_tokens": 5265948672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8186378479003906, "created_at": "2025-01-16T09:55:52.651733+00:00"} {"global_step": 10045, "acc_step": 0, "speed/wps": 12903.83788181079, "speed/FLOPS": 202672505665240.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08560305088758469, "optim/lr": 0.002938172871140734, "optim/total_tokens": 5266472960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.780404567718506, "created_at": "2025-01-16T09:56:02.813573+00:00"} {"global_step": 10046, "acc_step": 0, "speed/wps": 12899.185446402442, "speed/FLOPS": 202599432774036.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10027757287025452, "optim/lr": 0.002938148523492209, "optim/total_tokens": 5266997248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.869009256362915, "created_at": "2025-01-16T09:56:12.985537+00:00"} {"global_step": 10047, "acc_step": 0, "speed/wps": 12908.087704522439, "speed/FLOPS": 202739254970795.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04984806105494499, "optim/lr": 0.0029381241711514718, "optim/total_tokens": 5267521536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291378, "loss/out": 2.9525299072265625, "created_at": "2025-01-16T09:56:23.143337+00:00"} {"global_step": 10048, "acc_step": 0, "speed/wps": 12894.019843292726, "speed/FLOPS": 202518299878915.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07553271949291229, "optim/lr": 0.0029380998141186015, "optim/total_tokens": 5268045824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.977229595184326, "created_at": "2025-01-16T09:56:33.315945+00:00"} {"global_step": 10049, "acc_step": 0, "speed/wps": 12898.360774816589, "speed/FLOPS": 202586480173563.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06564570963382721, "optim/lr": 0.0029380754523936777, "optim/total_tokens": 5268570112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.867514133453369, "created_at": "2025-01-16T09:56:43.483175+00:00"} {"global_step": 10050, "acc_step": 0, "speed/wps": 12905.533022509611, "speed/FLOPS": 202699130179283.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05761830881237984, "optim/lr": 0.0029380510859767805, "optim/total_tokens": 5269094400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.8403759002685547, "created_at": "2025-01-16T09:56:53.645886+00:00"} {"global_step": 10051, "acc_step": 0, "speed/wps": 12903.456142389068, "speed/FLOPS": 202666509923057.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0717562660574913, "optim/lr": 0.002938026714867989, "optim/total_tokens": 5269618688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9148497581481934, "created_at": "2025-01-16T09:57:03.815371+00:00"} {"global_step": 10052, "acc_step": 0, "speed/wps": 12902.029892184575, "speed/FLOPS": 202644108703722.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04824903607368469, "optim/lr": 0.002938002339067382, "optim/total_tokens": 5270142976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.8561043739318848, "created_at": "2025-01-16T09:57:13.979348+00:00"} {"global_step": 10053, "acc_step": 0, "speed/wps": 12899.207245038535, "speed/FLOPS": 202599775151570.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05182197690010071, "optim/lr": 0.00293797795857504, "optim/total_tokens": 5270667264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8434383869171143, "created_at": "2025-01-16T09:57:24.145260+00:00"} {"global_step": 10054, "acc_step": 0, "speed/wps": 12887.773408435329, "speed/FLOPS": 202420191035979.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04304343834519386, "optim/lr": 0.002937953573391042, "optim/total_tokens": 5271191552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 2.7568421363830566, "created_at": "2025-01-16T09:57:34.320350+00:00"} {"global_step": 10055, "acc_step": 0, "speed/wps": 12902.783666799445, "speed/FLOPS": 202655947769844.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.039621807634830475, "optim/lr": 0.002937929183515468, "optim/total_tokens": 5271715840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.954188823699951, "created_at": "2025-01-16T09:57:44.482692+00:00"} {"global_step": 10056, "acc_step": 0, "speed/wps": 12904.16006230528, "speed/FLOPS": 202677565952629.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04442726820707321, "optim/lr": 0.002937904788948398, "optim/total_tokens": 5272240128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.897915840148926, "created_at": "2025-01-16T09:57:54.644921+00:00"} {"global_step": 10057, "acc_step": 0, "speed/wps": 12906.09432585116, "speed/FLOPS": 202707946219573.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04507703334093094, "optim/lr": 0.0029378803896899096, "optim/total_tokens": 5272764416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8679072856903076, "created_at": "2025-01-16T09:58:04.805987+00:00"} {"global_step": 10058, "acc_step": 0, "speed/wps": 12906.722223484596, "speed/FLOPS": 202717808214728.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04893176630139351, "optim/lr": 0.0029378559857400847, "optim/total_tokens": 5273288704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 2.932237386703491, "created_at": "2025-01-16T09:58:14.973852+00:00"} {"global_step": 10059, "acc_step": 0, "speed/wps": 12900.896766506969, "speed/FLOPS": 202626311407878.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045114915817976, "optim/lr": 0.002937831577099001, "optim/total_tokens": 5273812992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.8899295330047607, "created_at": "2025-01-16T09:58:25.142748+00:00"} {"global_step": 10060, "acc_step": 0, "speed/wps": 12898.92514294588, "speed/FLOPS": 202595344350556.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0598624013364315, "optim/lr": 0.0029378071637667404, "optim/total_tokens": 5274337280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.870467185974121, "created_at": "2025-01-16T09:58:35.311863+00:00"} {"global_step": 10061, "acc_step": 0, "speed/wps": 12903.779858423874, "speed/FLOPS": 202671594328216.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043197017163038254, "optim/lr": 0.0029377827457433806, "optim/total_tokens": 5274861568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.92276930809021, "created_at": "2025-01-16T09:58:45.473005+00:00"} {"global_step": 10062, "acc_step": 0, "speed/wps": 12896.618470603426, "speed/FLOPS": 202559114891718.47, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07743480801582336, "optim/lr": 0.0029377583230290015, "optim/total_tokens": 5275385856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.1534080505371094, "created_at": "2025-01-16T09:58:55.640327+00:00"} {"global_step": 10063, "acc_step": 0, "speed/wps": 12905.640942641889, "speed/FLOPS": 202700825213258.34, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05561868101358414, "optim/lr": 0.002937733895623684, "optim/total_tokens": 5275910144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458205, "loss/out": 2.9325881004333496, "created_at": "2025-01-16T09:59:05.802957+00:00"} {"global_step": 10064, "acc_step": 0, "speed/wps": 12906.993844184182, "speed/FLOPS": 202722074391061.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0641193613409996, "optim/lr": 0.0029377094635275065, "optim/total_tokens": 5276434432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.9642653465270996, "created_at": "2025-01-16T09:59:15.964139+00:00"} {"global_step": 10065, "acc_step": 0, "speed/wps": 12892.551004910632, "speed/FLOPS": 202495229753729.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048082005232572556, "optim/lr": 0.002937685026740549, "optim/total_tokens": 5276958720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.8322653770446777, "created_at": "2025-01-16T09:59:26.136856+00:00"} {"global_step": 10066, "acc_step": 0, "speed/wps": 12906.562231929127, "speed/FLOPS": 202715295327498.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08977598696947098, "optim/lr": 0.002937660585262892, "optim/total_tokens": 5277483008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.8346149921417236, "created_at": "2025-01-16T09:59:36.299091+00:00"} {"global_step": 10067, "acc_step": 0, "speed/wps": 12905.782501202322, "speed/FLOPS": 202703048584971.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05491799861192703, "optim/lr": 0.0029376361390946143, "optim/total_tokens": 5278007296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9158759117126465, "created_at": "2025-01-16T09:59:46.459248+00:00"} {"global_step": 10068, "acc_step": 0, "speed/wps": 12907.088213903962, "speed/FLOPS": 202723556597188.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07388279587030411, "optim/lr": 0.0029376116882357966, "optim/total_tokens": 5278531584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 2.9677014350891113, "created_at": "2025-01-16T09:59:56.622739+00:00"} {"global_step": 10069, "acc_step": 0, "speed/wps": 12901.442757756164, "speed/FLOPS": 202634886950717.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05517880246043205, "optim/lr": 0.0029375872326865174, "optim/total_tokens": 5279055872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.901370048522949, "created_at": "2025-01-16T10:00:06.787809+00:00"} {"global_step": 10070, "acc_step": 0, "speed/wps": 12903.034364132895, "speed/FLOPS": 202659885315960.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0705428421497345, "optim/lr": 0.0029375627724468577, "optim/total_tokens": 5279580160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8839974403381348, "created_at": "2025-01-16T10:00:16.951475+00:00"} {"global_step": 10071, "acc_step": 0, "speed/wps": 12900.337199330996, "speed/FLOPS": 202617522636453.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04540947079658508, "optim/lr": 0.002937538307516897, "optim/total_tokens": 5280104448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 2.8193464279174805, "created_at": "2025-01-16T10:00:27.116002+00:00"} {"global_step": 10072, "acc_step": 0, "speed/wps": 12900.991174810044, "speed/FLOPS": 202627794220008.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06012120842933655, "optim/lr": 0.0029375138378967144, "optim/total_tokens": 5280628736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9211249351501465, "created_at": "2025-01-16T10:00:37.279524+00:00"} {"global_step": 10073, "acc_step": 0, "speed/wps": 12898.70065209511, "speed/FLOPS": 202591818413262.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06240583956241608, "optim/lr": 0.0029374893635863908, "optim/total_tokens": 5281153024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 2.885272979736328, "created_at": "2025-01-16T10:00:47.448486+00:00"} {"global_step": 10074, "acc_step": 0, "speed/wps": 12904.972313502809, "speed/FLOPS": 202690323473836.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04750989004969597, "optim/lr": 0.002937464884586005, "optim/total_tokens": 5281677312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8955109119415283, "created_at": "2025-01-16T10:00:57.612164+00:00"} {"global_step": 10075, "acc_step": 0, "speed/wps": 12903.011433473635, "speed/FLOPS": 202659525158446.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046723995357751846, "optim/lr": 0.0029374404008956374, "optim/total_tokens": 5282201600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.920748233795166, "created_at": "2025-01-16T10:01:07.777012+00:00"} {"global_step": 10076, "acc_step": 0, "speed/wps": 12901.49601399641, "speed/FLOPS": 202635723413152.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06175246462225914, "optim/lr": 0.0029374159125153685, "optim/total_tokens": 5282725888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 2.9422531127929688, "created_at": "2025-01-16T10:01:17.941824+00:00"} {"global_step": 10077, "acc_step": 0, "speed/wps": 12902.410706209781, "speed/FLOPS": 202650089911282.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04991840198636055, "optim/lr": 0.0029373914194452773, "optim/total_tokens": 5283250176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8491411209106445, "created_at": "2025-01-16T10:01:28.104347+00:00"} {"global_step": 10078, "acc_step": 0, "speed/wps": 12888.322975101628, "speed/FLOPS": 202428822735657.38, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04605327546596527, "optim/lr": 0.002937366921685444, "optim/total_tokens": 5283774464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.838918685913086, "created_at": "2025-01-16T10:01:38.279020+00:00"} {"global_step": 10079, "acc_step": 0, "speed/wps": 12896.065796254614, "speed/FLOPS": 202550434381616.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05549811199307442, "optim/lr": 0.002937342419235948, "optim/total_tokens": 5284298752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.962132215499878, "created_at": "2025-01-16T10:01:48.449260+00:00"} {"global_step": 10080, "acc_step": 0, "speed/wps": 12901.374879453835, "speed/FLOPS": 202633820828704.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05036362260580063, "optim/lr": 0.0029373179120968703, "optim/total_tokens": 5284823040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9427242279052734, "created_at": "2025-01-16T10:01:58.612354+00:00"} {"global_step": 10081, "acc_step": 0, "speed/wps": 12904.174039189284, "speed/FLOPS": 202677785478798.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04274436831474304, "optim/lr": 0.0029372934002682904, "optim/total_tokens": 5285347328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 2.9505615234375, "created_at": "2025-01-16T10:02:08.773266+00:00"} {"global_step": 10082, "acc_step": 0, "speed/wps": 12897.055340408875, "speed/FLOPS": 202565976532335.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051362089812755585, "optim/lr": 0.0029372688837502876, "optim/total_tokens": 5285871616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8859703540802, "created_at": "2025-01-16T10:02:18.940614+00:00"} {"global_step": 10083, "acc_step": 0, "speed/wps": 12905.84909920385, "speed/FLOPS": 202704094598101.03, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060255542397499084, "optim/lr": 0.002937244362542943, "optim/total_tokens": 5286395904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.8954145908355713, "created_at": "2025-01-16T10:02:29.103874+00:00"} {"global_step": 10084, "acc_step": 0, "speed/wps": 12901.436458812654, "speed/FLOPS": 202634788017153.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05022577941417694, "optim/lr": 0.002937219836646336, "optim/total_tokens": 5286920192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 2.8677968978881836, "created_at": "2025-01-16T10:02:39.269432+00:00"} {"global_step": 10085, "acc_step": 0, "speed/wps": 12905.895889702344, "speed/FLOPS": 202704829507176.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05045366287231445, "optim/lr": 0.002937195306060547, "optim/total_tokens": 5287444480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9643619060516357, "created_at": "2025-01-16T10:02:49.429178+00:00"} {"global_step": 10086, "acc_step": 0, "speed/wps": 12896.922034626077, "speed/FLOPS": 202563882781834.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07513109594583511, "optim/lr": 0.0029371707707856556, "optim/total_tokens": 5287968768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422470, "loss/out": 2.8655142784118652, "created_at": "2025-01-16T10:02:59.597056+00:00"} {"global_step": 10087, "acc_step": 0, "speed/wps": 12900.515760405833, "speed/FLOPS": 202620327183502.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05730365216732025, "optim/lr": 0.0029371462308217418, "optim/total_tokens": 5288493056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9889092445373535, "created_at": "2025-01-16T10:03:09.763926+00:00"} {"global_step": 10088, "acc_step": 0, "speed/wps": 12907.62735420028, "speed/FLOPS": 202732024536398.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053532809019088745, "optim/lr": 0.002937121686168886, "optim/total_tokens": 5289017344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9669222831726074, "created_at": "2025-01-16T10:03:19.923018+00:00"} {"global_step": 10089, "acc_step": 0, "speed/wps": 12900.799879755139, "speed/FLOPS": 202624789668304.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054095152765512466, "optim/lr": 0.002937097136827168, "optim/total_tokens": 5289541632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415004, "loss/out": 2.891077756881714, "created_at": "2025-01-16T10:03:30.094990+00:00"} {"global_step": 10090, "acc_step": 0, "speed/wps": 12897.863199806874, "speed/FLOPS": 202578665074295.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05159163475036621, "optim/lr": 0.0029370725827966683, "optim/total_tokens": 5290065920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.15621280670166, "created_at": "2025-01-16T10:03:40.263735+00:00"} {"global_step": 10091, "acc_step": 0, "speed/wps": 12904.200626470052, "speed/FLOPS": 202678203068577.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044871505349874496, "optim/lr": 0.0029370480240774664, "optim/total_tokens": 5290590208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 2.8562731742858887, "created_at": "2025-01-16T10:03:50.427997+00:00"} {"global_step": 10092, "acc_step": 0, "speed/wps": 12903.91116809487, "speed/FLOPS": 202673656727038.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047393277287483215, "optim/lr": 0.002937023460669643, "optim/total_tokens": 5291114496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 2.833397626876831, "created_at": "2025-01-16T10:04:00.591279+00:00"} {"global_step": 10093, "acc_step": 0, "speed/wps": 12902.447266030907, "speed/FLOPS": 202650664133511.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04510969668626785, "optim/lr": 0.0029369988925732776, "optim/total_tokens": 5291638784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.189770221710205, "created_at": "2025-01-16T10:04:10.753845+00:00"} {"global_step": 10094, "acc_step": 0, "speed/wps": 12899.648385910858, "speed/FLOPS": 202606703875161.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046724747866392136, "optim/lr": 0.0029369743197884515, "optim/total_tokens": 5292163072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.9003543853759766, "created_at": "2025-01-16T10:04:20.918389+00:00"} {"global_step": 10095, "acc_step": 0, "speed/wps": 12898.068779251389, "speed/FLOPS": 202581893981967.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049128949642181396, "optim/lr": 0.0029369497423152435, "optim/total_tokens": 5292687360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.9102745056152344, "created_at": "2025-01-16T10:04:31.089576+00:00"} {"global_step": 10096, "acc_step": 0, "speed/wps": 12901.36102689924, "speed/FLOPS": 202633603255299.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04789843410253525, "optim/lr": 0.0029369251601537353, "optim/total_tokens": 5293211648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8969712257385254, "created_at": "2025-01-16T10:04:41.252726+00:00"} {"global_step": 10097, "acc_step": 0, "speed/wps": 12902.159221702294, "speed/FLOPS": 202646140001516.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056035082787275314, "optim/lr": 0.0029369005733040055, "optim/total_tokens": 5293735936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 2.8988871574401855, "created_at": "2025-01-16T10:04:51.415583+00:00"} {"global_step": 10098, "acc_step": 0, "speed/wps": 12895.969830493796, "speed/FLOPS": 202548927107471.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050609201192855835, "optim/lr": 0.0029368759817661352, "optim/total_tokens": 5294260224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.912475109100342, "created_at": "2025-01-16T10:05:01.584129+00:00"} {"global_step": 10099, "acc_step": 0, "speed/wps": 12902.104811487638, "speed/FLOPS": 202645285414327.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.03841174393892288, "optim/lr": 0.0029368513855402047, "optim/total_tokens": 5294784512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459779, "loss/out": 3.0200555324554443, "created_at": "2025-01-16T10:05:11.749595+00:00"} {"global_step": 10100, "acc_step": 0, "speed/wps": 12899.976184144538, "speed/FLOPS": 202611852396863.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05754278972744942, "optim/lr": 0.002936826784626294, "optim/total_tokens": 5295308800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 2.8324570655822754, "created_at": "2025-01-16T10:05:21.916117+00:00"} {"global_step": 10101, "acc_step": 0, "speed/wps": 12899.939511808823, "speed/FLOPS": 202611276407438.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04680418595671654, "optim/lr": 0.0029368021790244832, "optim/total_tokens": 5295833088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.985185384750366, "created_at": "2025-01-16T10:05:32.083710+00:00"} {"global_step": 10102, "acc_step": 0, "speed/wps": 12897.515520812627, "speed/FLOPS": 202573204297930.12, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053106971085071564, "optim/lr": 0.002936777568734853, "optim/total_tokens": 5296357376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.0158534049987793, "created_at": "2025-01-16T10:05:42.249905+00:00"} {"global_step": 10103, "acc_step": 0, "speed/wps": 12900.43643347881, "speed/FLOPS": 202619081245107.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07151395082473755, "optim/lr": 0.0029367529537574835, "optim/total_tokens": 5296881664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.8495583534240723, "created_at": "2025-01-16T10:05:52.414533+00:00"} {"global_step": 10104, "acc_step": 0, "speed/wps": 12901.295860094137, "speed/FLOPS": 202632579721076.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05237200856208801, "optim/lr": 0.002936728334092455, "optim/total_tokens": 5297405952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.957643747329712, "created_at": "2025-01-16T10:06:02.581225+00:00"} {"global_step": 10105, "acc_step": 0, "speed/wps": 12901.923354769702, "speed/FLOPS": 202642435387224.6, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05268336459994316, "optim/lr": 0.0029367037097398478, "optim/total_tokens": 5297930240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384005, "loss/out": 2.8385326862335205, "created_at": "2025-01-16T10:06:12.744618+00:00"} {"global_step": 10106, "acc_step": 0, "speed/wps": 12896.878717606056, "speed/FLOPS": 202563202428511.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050913941115140915, "optim/lr": 0.002936679080699742, "optim/total_tokens": 5298454528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8784618377685547, "created_at": "2025-01-16T10:06:22.911405+00:00"} {"global_step": 10107, "acc_step": 0, "speed/wps": 12901.783509489476, "speed/FLOPS": 202640238924931.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0819845199584961, "optim/lr": 0.0029366544469722187, "optim/total_tokens": 5298978816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8759331703186035, "created_at": "2025-01-16T10:06:33.078131+00:00"} {"global_step": 10108, "acc_step": 0, "speed/wps": 12900.16065217055, "speed/FLOPS": 202614749720697.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08008699864149094, "optim/lr": 0.002936629808557357, "optim/total_tokens": 5299503104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.8566694259643555, "created_at": "2025-01-16T10:06:43.243597+00:00"} {"global_step": 10109, "acc_step": 0, "speed/wps": 12899.648169540864, "speed/FLOPS": 202606700476773.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044676605612039566, "optim/lr": 0.0029366051654552395, "optim/total_tokens": 5300027392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.895972728729248, "created_at": "2025-01-16T10:06:53.408520+00:00"} {"global_step": 10110, "acc_step": 0, "speed/wps": 12901.70322552033, "speed/FLOPS": 202638977954873.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0653652474284172, "optim/lr": 0.002936580517665944, "optim/total_tokens": 5300551680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507980, "loss/out": 2.888571262359619, "created_at": "2025-01-16T10:07:03.571339+00:00"} {"global_step": 10111, "acc_step": 0, "speed/wps": 12900.410636237477, "speed/FLOPS": 202618676063982.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04376253858208656, "optim/lr": 0.0029365558651895524, "optim/total_tokens": 5301075968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 2.8829185962677, "created_at": "2025-01-16T10:07:13.735240+00:00"} {"global_step": 10112, "acc_step": 0, "speed/wps": 12895.646801598577, "speed/FLOPS": 202543853494784.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07417485862970352, "optim/lr": 0.002936531208026145, "optim/total_tokens": 5301600256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.982743740081787, "created_at": "2025-01-16T10:07:23.904178+00:00"} {"global_step": 10113, "acc_step": 0, "speed/wps": 12903.39218950636, "speed/FLOPS": 202665505455154.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05653733015060425, "optim/lr": 0.0029365065461758023, "optim/total_tokens": 5302124544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.7533974647521973, "created_at": "2025-01-16T10:07:34.065743+00:00"} {"global_step": 10114, "acc_step": 0, "speed/wps": 12894.912504347885, "speed/FLOPS": 202532320347431.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05853212997317314, "optim/lr": 0.002936481879638604, "optim/total_tokens": 5302648832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.952767848968506, "created_at": "2025-01-16T10:07:44.236913+00:00"} {"global_step": 10115, "acc_step": 0, "speed/wps": 12900.992282902474, "speed/FLOPS": 202627811624122.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06215333938598633, "optim/lr": 0.0029364572084146314, "optim/total_tokens": 5303173120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.9567158222198486, "created_at": "2025-01-16T10:07:54.400292+00:00"} {"global_step": 10116, "acc_step": 0, "speed/wps": 12897.753933392407, "speed/FLOPS": 202576948895108.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05093442648649216, "optim/lr": 0.002936432532503964, "optim/total_tokens": 5303697408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416384, "loss/out": 2.853579521179199, "created_at": "2025-01-16T10:08:04.566517+00:00"} {"global_step": 10117, "acc_step": 0, "speed/wps": 12900.812856007906, "speed/FLOPS": 202624993478185.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06358294188976288, "optim/lr": 0.0029364078519066843, "optim/total_tokens": 5304221696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9005887508392334, "created_at": "2025-01-16T10:08:14.730663+00:00"} {"global_step": 10118, "acc_step": 0, "speed/wps": 12899.80004059691, "speed/FLOPS": 202609085820401.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04685653746128082, "optim/lr": 0.0029363831666228708, "optim/total_tokens": 5304745984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9662935733795166, "created_at": "2025-01-16T10:08:24.896205+00:00"} {"global_step": 10119, "acc_step": 0, "speed/wps": 12903.408889265624, "speed/FLOPS": 202665767747821.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06957273930311203, "optim/lr": 0.002936358476652605, "optim/total_tokens": 5305270272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.93660306930542, "created_at": "2025-01-16T10:08:35.060248+00:00"} {"global_step": 10120, "acc_step": 0, "speed/wps": 12904.15106306796, "speed/FLOPS": 202677424607240.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058949828147888184, "optim/lr": 0.002936333781995967, "optim/total_tokens": 5305794560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8499398231506348, "created_at": "2025-01-16T10:08:45.221139+00:00"} {"global_step": 10121, "acc_step": 0, "speed/wps": 12902.698891887541, "speed/FLOPS": 202654616263359.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057082001119852066, "optim/lr": 0.0029363090826530383, "optim/total_tokens": 5306318848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 3.0229601860046387, "created_at": "2025-01-16T10:08:55.386269+00:00"} {"global_step": 10122, "acc_step": 0, "speed/wps": 12903.312764500295, "speed/FLOPS": 202664257976292.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05126943439245224, "optim/lr": 0.0029362843786238982, "optim/total_tokens": 5306843136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.834428548812866, "created_at": "2025-01-16T10:09:05.547920+00:00"} {"global_step": 10123, "acc_step": 0, "speed/wps": 12904.073256748696, "speed/FLOPS": 202676202552079.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051045823842287064, "optim/lr": 0.0029362596699086282, "optim/total_tokens": 5307367424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9370062351226807, "created_at": "2025-01-16T10:09:15.709756+00:00"} {"global_step": 10124, "acc_step": 0, "speed/wps": 12897.903967891096, "speed/FLOPS": 202579305393079.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0510515496134758, "optim/lr": 0.0029362349565073085, "optim/total_tokens": 5307891712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 2.956422805786133, "created_at": "2025-01-16T10:09:25.879549+00:00"} {"global_step": 10125, "acc_step": 0, "speed/wps": 12903.15818920849, "speed/FLOPS": 202661830158926.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04846276342868805, "optim/lr": 0.0029362102384200197, "optim/total_tokens": 5308416000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.0117454528808594, "created_at": "2025-01-16T10:09:36.048617+00:00"} {"global_step": 10126, "acc_step": 0, "speed/wps": 12902.214227334352, "speed/FLOPS": 202647003940553.22, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04545390233397484, "optim/lr": 0.002936185515646843, "optim/total_tokens": 5308940288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 2.846872091293335, "created_at": "2025-01-16T10:09:46.211016+00:00"} {"global_step": 10127, "acc_step": 0, "speed/wps": 12900.003468308396, "speed/FLOPS": 202612280932149.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052154116332530975, "optim/lr": 0.002936160788187858, "optim/total_tokens": 5309464576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.860470771789551, "created_at": "2025-01-16T10:09:56.375420+00:00"} {"global_step": 10128, "acc_step": 0, "speed/wps": 12900.765136212254, "speed/FLOPS": 202624243973220.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06406483799219131, "optim/lr": 0.0029361360560431465, "optim/total_tokens": 5309988864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8959548473358154, "created_at": "2025-01-16T10:10:06.540913+00:00"} {"global_step": 10129, "acc_step": 0, "speed/wps": 12901.909512864899, "speed/FLOPS": 202642217981088.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05058220773935318, "optim/lr": 0.0029361113192127894, "optim/total_tokens": 5310513152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.879638910293579, "created_at": "2025-01-16T10:10:16.706484+00:00"} {"global_step": 10130, "acc_step": 0, "speed/wps": 12897.1184929955, "speed/FLOPS": 202566968430488.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057717837393283844, "optim/lr": 0.0029360865776968657, "optim/total_tokens": 5311037440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9971883296966553, "created_at": "2025-01-16T10:10:26.872836+00:00"} {"global_step": 10131, "acc_step": 0, "speed/wps": 12906.909035245599, "speed/FLOPS": 202720742350141.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07241250574588776, "optim/lr": 0.0029360618314954578, "optim/total_tokens": 5311561728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9691808223724365, "created_at": "2025-01-16T10:10:37.031691+00:00"} {"global_step": 10132, "acc_step": 0, "speed/wps": 12895.353555922142, "speed/FLOPS": 202539247668474.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08058938384056091, "optim/lr": 0.0029360370806086455, "optim/total_tokens": 5312086016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9842841625213623, "created_at": "2025-01-16T10:10:47.199657+00:00"} {"global_step": 10133, "acc_step": 0, "speed/wps": 12903.01743939627, "speed/FLOPS": 202659619489714.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04566505551338196, "optim/lr": 0.00293601232503651, "optim/total_tokens": 5312610304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 2.9105310440063477, "created_at": "2025-01-16T10:10:57.365098+00:00"} {"global_step": 10134, "acc_step": 0, "speed/wps": 12900.021073306414, "speed/FLOPS": 202612557442835.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05259228125214577, "optim/lr": 0.0029359875647791316, "optim/total_tokens": 5313134592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 2.868401527404785, "created_at": "2025-01-16T10:11:07.530771+00:00"} {"global_step": 10135, "acc_step": 0, "speed/wps": 12897.601005364259, "speed/FLOPS": 202574546950281.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059108294546604156, "optim/lr": 0.002935962799836592, "optim/total_tokens": 5313658880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 2.941413402557373, "created_at": "2025-01-16T10:11:17.699789+00:00"} {"global_step": 10136, "acc_step": 0, "speed/wps": 12906.674895540611, "speed/FLOPS": 202717064864332.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048573434352874756, "optim/lr": 0.002935938030208971, "optim/total_tokens": 5314183168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0439841747283936, "created_at": "2025-01-16T10:11:27.863312+00:00"} {"global_step": 10137, "acc_step": 0, "speed/wps": 12903.015593967408, "speed/FLOPS": 202659590504718.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.062035851180553436, "optim/lr": 0.00293591325589635, "optim/total_tokens": 5314707456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 2.9178223609924316, "created_at": "2025-01-16T10:11:38.029883+00:00"} {"global_step": 10138, "acc_step": 0, "speed/wps": 12903.715341545065, "speed/FLOPS": 202670581001979.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04497743397951126, "optim/lr": 0.0029358884768988097, "optim/total_tokens": 5315231744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.973801612854004, "created_at": "2025-01-16T10:11:48.195936+00:00"} {"global_step": 10139, "acc_step": 0, "speed/wps": 12898.895549754527, "speed/FLOPS": 202594879548821.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07085592299699783, "optim/lr": 0.002935863693216431, "optim/total_tokens": 5315756032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.941904067993164, "created_at": "2025-01-16T10:11:58.361026+00:00"} {"global_step": 10140, "acc_step": 0, "speed/wps": 12909.593102447263, "speed/FLOPS": 202762899313838.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0627366453409195, "optim/lr": 0.002935838904849294, "optim/total_tokens": 5316280320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.8445539474487305, "created_at": "2025-01-16T10:12:08.517841+00:00"} {"global_step": 10141, "acc_step": 0, "speed/wps": 12904.895375296972, "speed/FLOPS": 202689115053589.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06167467683553696, "optim/lr": 0.002935814111797481, "optim/total_tokens": 5316804608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.0130884647369385, "created_at": "2025-01-16T10:12:18.679433+00:00"} {"global_step": 10142, "acc_step": 0, "speed/wps": 12904.680184175839, "speed/FLOPS": 202685735181328.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0745813325047493, "optim/lr": 0.0029357893140610717, "optim/total_tokens": 5317328896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.9071786403656006, "created_at": "2025-01-16T10:12:28.839813+00:00"} {"global_step": 10143, "acc_step": 0, "speed/wps": 12899.132715241287, "speed/FLOPS": 202598604558690.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05834006518125534, "optim/lr": 0.002935764511640148, "optim/total_tokens": 5317853184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.808417320251465, "created_at": "2025-01-16T10:12:39.007859+00:00"} {"global_step": 10144, "acc_step": 0, "speed/wps": 12901.027498344638, "speed/FLOPS": 202628364731033.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08183609694242477, "optim/lr": 0.0029357397045347896, "optim/total_tokens": 5318377472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.800812005996704, "created_at": "2025-01-16T10:12:49.171197+00:00"} {"global_step": 10145, "acc_step": 0, "speed/wps": 12901.489972826728, "speed/FLOPS": 202635628528281.0, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04888610914349556, "optim/lr": 0.0029357148927450787, "optim/total_tokens": 5318901760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 281836, "loss/out": 2.953425884246826, "created_at": "2025-01-16T10:12:59.338140+00:00"} {"global_step": 10146, "acc_step": 0, "speed/wps": 12906.281261464017, "speed/FLOPS": 202710882300250.12, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06401991099119186, "optim/lr": 0.0029356900762710953, "optim/total_tokens": 5319426048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.810551643371582, "created_at": "2025-01-16T10:13:09.497555+00:00"} {"global_step": 10147, "acc_step": 0, "speed/wps": 12903.775135461692, "speed/FLOPS": 202671520147605.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056914735585451126, "optim/lr": 0.002935665255112921, "optim/total_tokens": 5319950336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408877, "loss/out": 2.85678768157959, "created_at": "2025-01-16T10:13:19.658848+00:00"} {"global_step": 10148, "acc_step": 0, "speed/wps": 12905.91517478623, "speed/FLOPS": 202705132405919.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06131673604249954, "optim/lr": 0.002935640429270636, "optim/total_tokens": 5320474624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.036501407623291, "created_at": "2025-01-16T10:13:29.819868+00:00"} {"global_step": 10149, "acc_step": 0, "speed/wps": 12904.378533704972, "speed/FLOPS": 202680997346172.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06728664040565491, "optim/lr": 0.0029356155987443227, "optim/total_tokens": 5320998912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9384655952453613, "created_at": "2025-01-16T10:13:39.981756+00:00"} {"global_step": 10150, "acc_step": 0, "speed/wps": 12897.789703249855, "speed/FLOPS": 202577510709872.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05828185752034187, "optim/lr": 0.0029355907635340603, "optim/total_tokens": 5321523200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379658, "loss/out": 2.9587717056274414, "created_at": "2025-01-16T10:13:50.171248+00:00"} {"global_step": 10151, "acc_step": 0, "speed/wps": 12905.99161742554, "speed/FLOPS": 202706333042613.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08201722055673599, "optim/lr": 0.0029355659236399313, "optim/total_tokens": 5322047488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.892033815383911, "created_at": "2025-01-16T10:14:00.336323+00:00"} {"global_step": 10152, "acc_step": 0, "speed/wps": 12901.03971109258, "speed/FLOPS": 202628556549021.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05824819952249527, "optim/lr": 0.002935541079062016, "optim/total_tokens": 5322571776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8546433448791504, "created_at": "2025-01-16T10:14:10.503471+00:00"} {"global_step": 10153, "acc_step": 0, "speed/wps": 12899.43113162841, "speed/FLOPS": 202603291598117.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052940838038921356, "optim/lr": 0.0029355162298003957, "optim/total_tokens": 5323096064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 2.903611898422241, "created_at": "2025-01-16T10:14:20.668061+00:00"} {"global_step": 10154, "acc_step": 0, "speed/wps": 12906.937316999383, "speed/FLOPS": 202721186553947.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057510532438755035, "optim/lr": 0.0029354913758551514, "optim/total_tokens": 5323620352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 3.0094943046569824, "created_at": "2025-01-16T10:14:30.827379+00:00"} {"global_step": 10155, "acc_step": 0, "speed/wps": 12898.030235930877, "speed/FLOPS": 202581288606154.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054918140172958374, "optim/lr": 0.002935466517226364, "optim/total_tokens": 5324144640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9922709465026855, "created_at": "2025-01-16T10:14:40.994357+00:00"} {"global_step": 10156, "acc_step": 0, "speed/wps": 12899.411709853188, "speed/FLOPS": 202602986552448.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05657878890633583, "optim/lr": 0.0029354416539141156, "optim/total_tokens": 5324668928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.9341917037963867, "created_at": "2025-01-16T10:14:51.161335+00:00"} {"global_step": 10157, "acc_step": 0, "speed/wps": 12908.28404384811, "speed/FLOPS": 202742338749703.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04474053531885147, "optim/lr": 0.002935416785918486, "optim/total_tokens": 5325193216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9103031158447266, "created_at": "2025-01-16T10:15:01.322238+00:00"} {"global_step": 10158, "acc_step": 0, "speed/wps": 12899.847791089418, "speed/FLOPS": 202609835807502.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05437217280268669, "optim/lr": 0.0029353919132395568, "optim/total_tokens": 5325717504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.7946155071258545, "created_at": "2025-01-16T10:15:11.486498+00:00"} {"global_step": 10159, "acc_step": 0, "speed/wps": 12899.905384199074, "speed/FLOPS": 202610740386430.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04725487902760506, "optim/lr": 0.0029353670358774095, "optim/total_tokens": 5326241792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.798257827758789, "created_at": "2025-01-16T10:15:21.654370+00:00"} {"global_step": 10160, "acc_step": 0, "speed/wps": 12903.606638475925, "speed/FLOPS": 202668873670905.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057779956609010696, "optim/lr": 0.0029353421538321246, "optim/total_tokens": 5326766080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.0749778747558594, "created_at": "2025-01-16T10:15:31.816799+00:00"} {"global_step": 10161, "acc_step": 0, "speed/wps": 12903.77738809653, "speed/FLOPS": 202671555528331.2, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04827326536178589, "optim/lr": 0.002935317267103784, "optim/total_tokens": 5327290368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.0571205615997314, "created_at": "2025-01-16T10:15:41.981015+00:00"} {"global_step": 10162, "acc_step": 0, "speed/wps": 12901.368036550699, "speed/FLOPS": 202633713351507.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052322328090667725, "optim/lr": 0.0029352923756924684, "optim/total_tokens": 5327814656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 2.872286558151245, "created_at": "2025-01-16T10:15:52.145802+00:00"} {"global_step": 10163, "acc_step": 0, "speed/wps": 12908.650236513884, "speed/FLOPS": 202748090308721.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05138463154435158, "optim/lr": 0.0029352674795982596, "optim/total_tokens": 5328338944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.7572219371795654, "created_at": "2025-01-16T10:16:02.303210+00:00"} {"global_step": 10164, "acc_step": 0, "speed/wps": 12903.852544206742, "speed/FLOPS": 202672735958317.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04867539554834366, "optim/lr": 0.0029352425788212386, "optim/total_tokens": 5328863232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 2.9306020736694336, "created_at": "2025-01-16T10:16:12.472663+00:00"} {"global_step": 10165, "acc_step": 0, "speed/wps": 12901.23810082461, "speed/FLOPS": 202631672532379.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04778226464986801, "optim/lr": 0.002935217673361486, "optim/total_tokens": 5329387520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.9326634407043457, "created_at": "2025-01-16T10:16:22.644874+00:00"} {"global_step": 10166, "acc_step": 0, "speed/wps": 12897.297297096024, "speed/FLOPS": 202569776794590.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04255915805697441, "optim/lr": 0.0029351927632190835, "optim/total_tokens": 5329911808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.940483570098877, "created_at": "2025-01-16T10:16:32.815045+00:00"} {"global_step": 10167, "acc_step": 0, "speed/wps": 12901.39350765641, "speed/FLOPS": 202634113410224.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054120175540447235, "optim/lr": 0.0029351678483941126, "optim/total_tokens": 5330436096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.872227668762207, "created_at": "2025-01-16T10:16:42.981721+00:00"} {"global_step": 10168, "acc_step": 0, "speed/wps": 12898.512448878324, "speed/FLOPS": 202588862423124.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051292382180690765, "optim/lr": 0.002935142928886654, "optim/total_tokens": 5330960384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8534109592437744, "created_at": "2025-01-16T10:16:53.147129+00:00"} {"global_step": 10169, "acc_step": 0, "speed/wps": 12906.088610579389, "speed/FLOPS": 202707856453377.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04972716420888901, "optim/lr": 0.00293511800469679, "optim/total_tokens": 5331484672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.873861789703369, "created_at": "2025-01-16T10:17:03.308699+00:00"} {"global_step": 10170, "acc_step": 0, "speed/wps": 12898.332480987654, "speed/FLOPS": 202586035780099.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06379278749227524, "optim/lr": 0.0029350930758246007, "optim/total_tokens": 5332008960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305392, "loss/out": 2.858130931854248, "created_at": "2025-01-16T10:17:13.474442+00:00"} {"global_step": 10171, "acc_step": 0, "speed/wps": 12902.49634437977, "speed/FLOPS": 202651434976420.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06342273205518723, "optim/lr": 0.002935068142270168, "optim/total_tokens": 5332533248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8612794876098633, "created_at": "2025-01-16T10:17:23.637075+00:00"} {"global_step": 10172, "acc_step": 0, "speed/wps": 12903.457072264731, "speed/FLOPS": 202666524528032.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05261748284101486, "optim/lr": 0.0029350432040335744, "optim/total_tokens": 5333057536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 2.8990352153778076, "created_at": "2025-01-16T10:17:33.798578+00:00"} {"global_step": 10173, "acc_step": 0, "speed/wps": 12901.659761653453, "speed/FLOPS": 202638295295119.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05349653214216232, "optim/lr": 0.0029350182611148994, "optim/total_tokens": 5333581824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9232494831085205, "created_at": "2025-01-16T10:17:43.964186+00:00"} {"global_step": 10174, "acc_step": 0, "speed/wps": 12903.328608727245, "speed/FLOPS": 202664506831649.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04217953979969025, "optim/lr": 0.0029349933135142253, "optim/total_tokens": 5334106112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.9171183109283447, "created_at": "2025-01-16T10:17:54.125716+00:00"} {"global_step": 10175, "acc_step": 0, "speed/wps": 12904.207131620893, "speed/FLOPS": 202678305240910.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052422333508729935, "optim/lr": 0.0029349683612316328, "optim/total_tokens": 5334630400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.898294687271118, "created_at": "2025-01-16T10:18:04.290315+00:00"} {"global_step": 10176, "acc_step": 0, "speed/wps": 12905.42217524585, "speed/FLOPS": 202697389170685.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042919427156448364, "optim/lr": 0.002934943404267204, "optim/total_tokens": 5335154688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8037710189819336, "created_at": "2025-01-16T10:18:14.452022+00:00"} {"global_step": 10177, "acc_step": 0, "speed/wps": 12907.962676710627, "speed/FLOPS": 202737291237204.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046173762530088425, "optim/lr": 0.0029349184426210205, "optim/total_tokens": 5335678976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382133, "loss/out": 2.7654521465301514, "created_at": "2025-01-16T10:18:24.610802+00:00"} {"global_step": 10178, "acc_step": 0, "speed/wps": 12899.064623613564, "speed/FLOPS": 202597535086109.16, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04816265404224396, "optim/lr": 0.002934893476293164, "optim/total_tokens": 5336203264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9420366287231445, "created_at": "2025-01-16T10:18:34.776340+00:00"} {"global_step": 10179, "acc_step": 0, "speed/wps": 12902.399624040137, "speed/FLOPS": 202649915850579.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0410740040242672, "optim/lr": 0.002934868505283715, "optim/total_tokens": 5336727552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400642, "loss/out": 2.9156105518341064, "created_at": "2025-01-16T10:18:44.938585+00:00"} {"global_step": 10180, "acc_step": 0, "speed/wps": 12904.195269013537, "speed/FLOPS": 202678118922361.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04521297290921211, "optim/lr": 0.002934843529592754, "optim/total_tokens": 5337251840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9210996627807617, "created_at": "2025-01-16T10:18:55.102657+00:00"} {"global_step": 10181, "acc_step": 0, "speed/wps": 12893.089237319076, "speed/FLOPS": 202503683433313.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051734887063503265, "optim/lr": 0.0029348185492203654, "optim/total_tokens": 5337776128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.946375608444214, "created_at": "2025-01-16T10:19:05.277160+00:00"} {"global_step": 10182, "acc_step": 0, "speed/wps": 12886.925137606271, "speed/FLOPS": 202406867776962.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048411622643470764, "optim/lr": 0.002934793564166629, "optim/total_tokens": 5338300416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.8152480125427246, "created_at": "2025-01-16T10:19:15.453452+00:00"} {"global_step": 10183, "acc_step": 0, "speed/wps": 12885.680233342135, "speed/FLOPS": 202387314844816.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04780816659331322, "optim/lr": 0.002934768574431626, "optim/total_tokens": 5338824704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381609, "loss/out": 2.8980419635772705, "created_at": "2025-01-16T10:19:25.630275+00:00"} {"global_step": 10184, "acc_step": 0, "speed/wps": 12896.154246586673, "speed/FLOPS": 202551823615624.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048534464091062546, "optim/lr": 0.0029347435800154385, "optim/total_tokens": 5339348992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8819286823272705, "created_at": "2025-01-16T10:19:35.804821+00:00"} {"global_step": 10185, "acc_step": 0, "speed/wps": 12902.779493443077, "speed/FLOPS": 202655882221547.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04948876425623894, "optim/lr": 0.002934718580918148, "optim/total_tokens": 5339873280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9842913150787354, "created_at": "2025-01-16T10:19:45.972387+00:00"} {"global_step": 10186, "acc_step": 0, "speed/wps": 12898.457512633242, "speed/FLOPS": 202587999573905.3, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06750672310590744, "optim/lr": 0.0029346935771398363, "optim/total_tokens": 5340397568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 2.9313011169433594, "created_at": "2025-01-16T10:19:56.138496+00:00"} {"global_step": 10187, "acc_step": 0, "speed/wps": 12903.534221652313, "speed/FLOPS": 202667736265179.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06870745867490768, "optim/lr": 0.002934668568680584, "optim/total_tokens": 5340921856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.7285780906677246, "created_at": "2025-01-16T10:20:06.301020+00:00"} {"global_step": 10188, "acc_step": 0, "speed/wps": 12904.235171996026, "speed/FLOPS": 202678745653533.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05182410031557083, "optim/lr": 0.002934643555540474, "optim/total_tokens": 5341446144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.8827149868011475, "created_at": "2025-01-16T10:20:16.462300+00:00"} {"global_step": 10189, "acc_step": 0, "speed/wps": 12903.393629261458, "speed/FLOPS": 202665528068486.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046811703592538834, "optim/lr": 0.0029346185377195867, "optim/total_tokens": 5341970432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 2.9954943656921387, "created_at": "2025-01-16T10:20:26.624167+00:00"} {"global_step": 10190, "acc_step": 0, "speed/wps": 12902.251114867877, "speed/FLOPS": 202647583309956.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05181199684739113, "optim/lr": 0.002934593515218005, "optim/total_tokens": 5342494720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9033167362213135, "created_at": "2025-01-16T10:20:36.790030+00:00"} {"global_step": 10191, "acc_step": 0, "speed/wps": 12901.475570076573, "speed/FLOPS": 202635402313298.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05321582779288292, "optim/lr": 0.00293456848803581, "optim/total_tokens": 5343019008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.891986846923828, "created_at": "2025-01-16T10:20:46.955525+00:00"} {"global_step": 10192, "acc_step": 0, "speed/wps": 12900.208264620984, "speed/FLOPS": 202615497539658.12, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050202444195747375, "optim/lr": 0.0029345434561730823, "optim/total_tokens": 5343543296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.8045763969421387, "created_at": "2025-01-16T10:20:57.119852+00:00"} {"global_step": 10193, "acc_step": 0, "speed/wps": 12902.660893414308, "speed/FLOPS": 202654019445122.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05175131931900978, "optim/lr": 0.002934518419629905, "optim/total_tokens": 5344067584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.990328311920166, "created_at": "2025-01-16T10:21:07.285133+00:00"} {"global_step": 10194, "acc_step": 0, "speed/wps": 12906.073879190899, "speed/FLOPS": 202707625076677.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048800159245729446, "optim/lr": 0.002934493378406359, "optim/total_tokens": 5344591872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 2.8869714736938477, "created_at": "2025-01-16T10:21:17.445675+00:00"} {"global_step": 10195, "acc_step": 0, "speed/wps": 12897.690941243309, "speed/FLOPS": 202575959516846.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05907473713159561, "optim/lr": 0.002934468332502527, "optim/total_tokens": 5345116160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.9820902347564697, "created_at": "2025-01-16T10:21:27.614593+00:00"} {"global_step": 10196, "acc_step": 0, "speed/wps": 12896.215712305499, "speed/FLOPS": 202552789019201.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05967143550515175, "optim/lr": 0.002934443281918489, "optim/total_tokens": 5345640448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8759565353393555, "created_at": "2025-01-16T10:21:37.786574+00:00"} {"global_step": 10197, "acc_step": 0, "speed/wps": 12899.032828456775, "speed/FLOPS": 202597035699480.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04704369604587555, "optim/lr": 0.0029344182266543284, "optim/total_tokens": 5346164736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 2.8655974864959717, "created_at": "2025-01-16T10:21:47.952814+00:00"} {"global_step": 10198, "acc_step": 0, "speed/wps": 12903.194369973871, "speed/FLOPS": 202662398427563.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048823196440935135, "optim/lr": 0.002934393166710126, "optim/total_tokens": 5346689024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.8707947731018066, "created_at": "2025-01-16T10:21:58.121044+00:00"} {"global_step": 10199, "acc_step": 0, "speed/wps": 12900.563299796486, "speed/FLOPS": 202621073854958.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0497836172580719, "optim/lr": 0.002934368102085964, "optim/total_tokens": 5347213312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471956, "loss/out": 2.9873616695404053, "created_at": "2025-01-16T10:22:08.285666+00:00"} {"global_step": 10200, "acc_step": 0, "speed/wps": 12895.326252432922, "speed/FLOPS": 202538818829656.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04919794574379921, "optim/lr": 0.0029343430327819237, "optim/total_tokens": 5347737600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458694, "loss/out": 2.847865581512451, "created_at": "2025-01-16T10:22:18.456909+00:00"} {"global_step": 10201, "acc_step": 0, "speed/wps": 12904.16832325177, "speed/FLOPS": 202677695702145.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04732396453619003, "optim/lr": 0.0029343179587980875, "optim/total_tokens": 5348261888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9298529624938965, "created_at": "2025-01-16T10:22:28.619534+00:00"} {"global_step": 10202, "acc_step": 0, "speed/wps": 12903.418522760752, "speed/FLOPS": 202665919055100.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05045969784259796, "optim/lr": 0.0029342928801345367, "optim/total_tokens": 5348786176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432352, "loss/out": 2.891239643096924, "created_at": "2025-01-16T10:22:38.783958+00:00"} {"global_step": 10203, "acc_step": 0, "speed/wps": 12904.215684874858, "speed/FLOPS": 202678439581517.72, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04718494042754173, "optim/lr": 0.0029342677967913538, "optim/total_tokens": 5349310464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.916151285171509, "created_at": "2025-01-16T10:22:48.947692+00:00"} {"global_step": 10204, "acc_step": 0, "speed/wps": 12902.04439670947, "speed/FLOPS": 202644336517217.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05338464677333832, "optim/lr": 0.00293424270876862, "optim/total_tokens": 5349834752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9989819526672363, "created_at": "2025-01-16T10:22:59.117115+00:00"} {"global_step": 10205, "acc_step": 0, "speed/wps": 12897.357829349363, "speed/FLOPS": 202570727536808.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05112306401133537, "optim/lr": 0.002934217616066417, "optim/total_tokens": 5350359040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 2.8833673000335693, "created_at": "2025-01-16T10:23:09.292905+00:00"} {"global_step": 10206, "acc_step": 0, "speed/wps": 12899.272491239208, "speed/FLOPS": 202600799932809.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054527733474969864, "optim/lr": 0.002934192518684827, "optim/total_tokens": 5350883328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.913097858428955, "created_at": "2025-01-16T10:23:19.460551+00:00"} {"global_step": 10207, "acc_step": 0, "speed/wps": 12907.649155334118, "speed/FLOPS": 202732366953163.25, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05249250680208206, "optim/lr": 0.002934167416623932, "optim/total_tokens": 5351407616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.988966941833496, "created_at": "2025-01-16T10:23:29.618919+00:00"} {"global_step": 10208, "acc_step": 0, "speed/wps": 12904.64308987933, "speed/FLOPS": 202685152564428.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04764694347977638, "optim/lr": 0.0029341423098838137, "optim/total_tokens": 5351931904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 2.9111223220825195, "created_at": "2025-01-16T10:23:39.779337+00:00"} {"global_step": 10209, "acc_step": 0, "speed/wps": 12900.856013795687, "speed/FLOPS": 202625671330547.47, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04390426725149155, "optim/lr": 0.0029341171984645542, "optim/total_tokens": 5352456192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.906757354736328, "created_at": "2025-01-16T10:23:49.945574+00:00"} {"global_step": 10210, "acc_step": 0, "speed/wps": 12905.742595750651, "speed/FLOPS": 202702421815017.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0544871985912323, "optim/lr": 0.002934092082366235, "optim/total_tokens": 5352980480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.8515634536743164, "created_at": "2025-01-16T10:24:00.110729+00:00"} {"global_step": 10211, "acc_step": 0, "speed/wps": 12902.456739586365, "speed/FLOPS": 202650812928717.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05604392662644386, "optim/lr": 0.002934066961588939, "optim/total_tokens": 5353504768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.003056287765503, "created_at": "2025-01-16T10:24:10.273512+00:00"} {"global_step": 10212, "acc_step": 0, "speed/wps": 12898.858636481811, "speed/FLOPS": 202594299775149.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0464446097612381, "optim/lr": 0.0029340418361327473, "optim/total_tokens": 5354029056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.787316083908081, "created_at": "2025-01-16T10:24:20.441100+00:00"} {"global_step": 10213, "acc_step": 0, "speed/wps": 12905.13838240869, "speed/FLOPS": 202692931814205.84, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04326435178518295, "optim/lr": 0.0029340167059977423, "optim/total_tokens": 5354553344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.7817978858947754, "created_at": "2025-01-16T10:24:30.606732+00:00"} {"global_step": 10214, "acc_step": 0, "speed/wps": 12904.037828273162, "speed/FLOPS": 202675646099187.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05467979609966278, "optim/lr": 0.002933991571184005, "optim/total_tokens": 5355077632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.948305368423462, "created_at": "2025-01-16T10:24:40.770723+00:00"} {"global_step": 10215, "acc_step": 0, "speed/wps": 12895.761185431586, "speed/FLOPS": 202545650050061.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05520397424697876, "optim/lr": 0.0029339664316916187, "optim/total_tokens": 5355601920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8953211307525635, "created_at": "2025-01-16T10:24:50.941167+00:00"} {"global_step": 10216, "acc_step": 0, "speed/wps": 12908.684648394985, "speed/FLOPS": 202748630794600.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05309758335351944, "optim/lr": 0.002933941287520665, "optim/total_tokens": 5356126208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297240, "loss/out": 2.9471845626831055, "created_at": "2025-01-16T10:25:01.101585+00:00"} {"global_step": 10217, "acc_step": 0, "speed/wps": 12903.028452832055, "speed/FLOPS": 202659792470857.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048908378928899765, "optim/lr": 0.002933916138671226, "optim/total_tokens": 5356650496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.895103693008423, "created_at": "2025-01-16T10:25:11.268822+00:00"} {"global_step": 10218, "acc_step": 0, "speed/wps": 12901.015813059574, "speed/FLOPS": 202628181197574.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054373305290937424, "optim/lr": 0.0029338909851433833, "optim/total_tokens": 5357174784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.8840744495391846, "created_at": "2025-01-16T10:25:21.435551+00:00"} {"global_step": 10219, "acc_step": 0, "speed/wps": 12893.361953537964, "speed/FLOPS": 202507966816282.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06517834961414337, "optim/lr": 0.0029338658269372196, "optim/total_tokens": 5357699072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.9408297538757324, "created_at": "2025-01-16T10:25:31.605582+00:00"} {"global_step": 10220, "acc_step": 0, "speed/wps": 12898.435022622292, "speed/FLOPS": 202587646337378.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06897947192192078, "optim/lr": 0.0029338406640528163, "optim/total_tokens": 5358223360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.946653127670288, "created_at": "2025-01-16T10:25:41.771480+00:00"} {"global_step": 10221, "acc_step": 0, "speed/wps": 12900.775166646776, "speed/FLOPS": 202624401514978.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05058232694864273, "optim/lr": 0.0029338154964902567, "optim/total_tokens": 5358747648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.851792812347412, "created_at": "2025-01-16T10:25:51.938342+00:00"} {"global_step": 10222, "acc_step": 0, "speed/wps": 12900.158572252563, "speed/FLOPS": 202614717052727.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07204809039831161, "optim/lr": 0.0029337903242496214, "optim/total_tokens": 5359271936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.9033241271972656, "created_at": "2025-01-16T10:26:02.103509+00:00"} {"global_step": 10223, "acc_step": 0, "speed/wps": 12900.571385855761, "speed/FLOPS": 202621200857631.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05216808244585991, "optim/lr": 0.0029337651473309934, "optim/total_tokens": 5359796224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.9834022521972656, "created_at": "2025-01-16T10:26:12.270231+00:00"} {"global_step": 10224, "acc_step": 0, "speed/wps": 12908.636960330383, "speed/FLOPS": 202747881788015.88, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05745327100157738, "optim/lr": 0.002933739965734454, "optim/total_tokens": 5360320512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.8460097312927246, "created_at": "2025-01-16T10:26:22.430956+00:00"} {"global_step": 10225, "acc_step": 0, "speed/wps": 12898.60370257152, "speed/FLOPS": 202590295687771.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06666730344295502, "optim/lr": 0.0029337147794600868, "optim/total_tokens": 5360844800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.83400559425354, "created_at": "2025-01-16T10:26:32.598141+00:00"} {"global_step": 10226, "acc_step": 0, "speed/wps": 12898.902819178413, "speed/FLOPS": 202594993725112.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06863725930452347, "optim/lr": 0.002933689588507973, "optim/total_tokens": 5361369088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.956679582595825, "created_at": "2025-01-16T10:26:42.763213+00:00"} {"global_step": 10227, "acc_step": 0, "speed/wps": 12896.940298961552, "speed/FLOPS": 202564169648321.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06539477407932281, "optim/lr": 0.002933664392878195, "optim/total_tokens": 5361893376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 2.8623557090759277, "created_at": "2025-01-16T10:26:52.930853+00:00"} {"global_step": 10228, "acc_step": 0, "speed/wps": 12903.746569422696, "speed/FLOPS": 202671071478709.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05889136344194412, "optim/lr": 0.002933639192570835, "optim/total_tokens": 5362417664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9369564056396484, "created_at": "2025-01-16T10:27:03.096603+00:00"} {"global_step": 10229, "acc_step": 0, "speed/wps": 12899.845773938769, "speed/FLOPS": 202609804125379.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05965404585003853, "optim/lr": 0.002933613987585975, "optim/total_tokens": 5362941952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9379091262817383, "created_at": "2025-01-16T10:27:13.266026+00:00"} {"global_step": 10230, "acc_step": 0, "speed/wps": 12902.535176996304, "speed/FLOPS": 202652044896025.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06820268929004669, "optim/lr": 0.0029335887779236977, "optim/total_tokens": 5363466240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 2.927813768386841, "created_at": "2025-01-16T10:27:23.431406+00:00"} {"global_step": 10231, "acc_step": 0, "speed/wps": 12898.43965183443, "speed/FLOPS": 202587719045516.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05237048864364624, "optim/lr": 0.002933563563584085, "optim/total_tokens": 5363990528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 2.791172504425049, "created_at": "2025-01-16T10:27:33.603428+00:00"} {"global_step": 10232, "acc_step": 0, "speed/wps": 12899.996224839384, "speed/FLOPS": 202612167163514.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0734449252486229, "optim/lr": 0.0029335383445672192, "optim/total_tokens": 5364514816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 2.842885732650757, "created_at": "2025-01-16T10:27:43.775118+00:00"} {"global_step": 10233, "acc_step": 0, "speed/wps": 12903.113400058539, "speed/FLOPS": 202661126683779.38, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057576388120651245, "optim/lr": 0.002933513120873182, "optim/total_tokens": 5365039104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.8875319957733154, "created_at": "2025-01-16T10:27:53.936721+00:00"} {"global_step": 10234, "acc_step": 0, "speed/wps": 12904.704299310248, "speed/FLOPS": 202686113942652.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08157367259263992, "optim/lr": 0.0029334878925020573, "optim/total_tokens": 5365563392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8745245933532715, "created_at": "2025-01-16T10:28:04.099678+00:00"} {"global_step": 10235, "acc_step": 0, "speed/wps": 12901.669132338568, "speed/FLOPS": 202638442474605.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06270550936460495, "optim/lr": 0.0029334626594539268, "optim/total_tokens": 5366087680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.931359052658081, "created_at": "2025-01-16T10:28:14.262624+00:00"} {"global_step": 10236, "acc_step": 0, "speed/wps": 12902.160398598378, "speed/FLOPS": 202646158486286.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057123247534036636, "optim/lr": 0.002933437421728871, "optim/total_tokens": 5366611968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8819212913513184, "created_at": "2025-01-16T10:28:24.425154+00:00"} {"global_step": 10237, "acc_step": 0, "speed/wps": 12900.996042376506, "speed/FLOPS": 202627870671828.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07279369235038757, "optim/lr": 0.002933412179326975, "optim/total_tokens": 5367136256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.905423164367676, "created_at": "2025-01-16T10:28:34.595465+00:00"} {"global_step": 10238, "acc_step": 0, "speed/wps": 12896.909021289064, "speed/FLOPS": 202563678389494.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05821957811713219, "optim/lr": 0.0029333869322483194, "optim/total_tokens": 5367660544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 2.923844337463379, "created_at": "2025-01-16T10:28:44.763781+00:00"} {"global_step": 10239, "acc_step": 0, "speed/wps": 12903.601683757935, "speed/FLOPS": 202668795850250.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07287023216485977, "optim/lr": 0.0029333616804929873, "optim/total_tokens": 5368184832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.876880645751953, "created_at": "2025-01-16T10:28:54.928650+00:00"} {"global_step": 10240, "acc_step": 0, "speed/wps": 12903.965928374988, "speed/FLOPS": 202674516812485.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049116965383291245, "optim/lr": 0.0029333364240610605, "optim/total_tokens": 5368709120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8915085792541504, "created_at": "2025-01-16T10:29:05.091343+00:00"} {"global_step": 10241, "acc_step": 0, "speed/wps": 12896.1104336827, "speed/FLOPS": 202551135473762.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05493847280740738, "optim/lr": 0.0029333111629526215, "optim/total_tokens": 5369233408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.937840700149536, "created_at": "2025-01-16T10:29:15.259606+00:00"} {"global_step": 10242, "acc_step": 0, "speed/wps": 12901.600465153588, "speed/FLOPS": 202637363962106.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05490753427147865, "optim/lr": 0.0029332858971677536, "optim/total_tokens": 5369757696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8272385597229004, "created_at": "2025-01-16T10:29:25.425634+00:00"} {"global_step": 10243, "acc_step": 0, "speed/wps": 12903.306293375714, "speed/FLOPS": 202664156338388.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04742812737822533, "optim/lr": 0.0029332606267065383, "optim/total_tokens": 5370281984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.6641592979431152, "created_at": "2025-01-16T10:29:35.588602+00:00"} {"global_step": 10244, "acc_step": 0, "speed/wps": 12901.137830191721, "speed/FLOPS": 202630097644303.5, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05293818190693855, "optim/lr": 0.0029332353515690583, "optim/total_tokens": 5370806272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9316039085388184, "created_at": "2025-01-16T10:29:45.751938+00:00"} {"global_step": 10245, "acc_step": 0, "speed/wps": 12897.864186795721, "speed/FLOPS": 202578680576311.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051402498036623, "optim/lr": 0.0029332100717553967, "optim/total_tokens": 5371330560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9435105323791504, "created_at": "2025-01-16T10:29:55.923374+00:00"} {"global_step": 10246, "acc_step": 0, "speed/wps": 12899.922413100578, "speed/FLOPS": 202611007848729.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04906514659523964, "optim/lr": 0.0029331847872656343, "optim/total_tokens": 5371854848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 2.954425811767578, "created_at": "2025-01-16T10:30:06.087658+00:00"} {"global_step": 10247, "acc_step": 0, "speed/wps": 12904.982517329829, "speed/FLOPS": 202690483738961.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04733363166451454, "optim/lr": 0.0029331594980998555, "optim/total_tokens": 5372379136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8592617511749268, "created_at": "2025-01-16T10:30:16.250585+00:00"} {"global_step": 10248, "acc_step": 0, "speed/wps": 12899.311531451774, "speed/FLOPS": 202601413112995.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05422971397638321, "optim/lr": 0.0029331342042581417, "optim/total_tokens": 5372903424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.00502872467041, "created_at": "2025-01-16T10:30:26.415323+00:00"} {"global_step": 10249, "acc_step": 0, "speed/wps": 12899.564358046708, "speed/FLOPS": 202605384102088.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05304577946662903, "optim/lr": 0.0029331089057405757, "optim/total_tokens": 5373427712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.9861414432525635, "created_at": "2025-01-16T10:30:36.583758+00:00"} {"global_step": 10250, "acc_step": 0, "speed/wps": 12902.771877784415, "speed/FLOPS": 202655762607163.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0512009933590889, "optim/lr": 0.00293308360254724, "optim/total_tokens": 5373952000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.934904098510742, "created_at": "2025-01-16T10:30:46.745652+00:00"} {"global_step": 10251, "acc_step": 0, "speed/wps": 12896.647781533406, "speed/FLOPS": 202559575260151.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06554685533046722, "optim/lr": 0.0029330582946782173, "optim/total_tokens": 5374476288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371444, "loss/out": 2.934678077697754, "created_at": "2025-01-16T10:30:56.913479+00:00"} {"global_step": 10252, "acc_step": 0, "speed/wps": 12904.485424045517, "speed/FLOPS": 202682676205853.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04829556867480278, "optim/lr": 0.00293303298213359, "optim/total_tokens": 5375000576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9097416400909424, "created_at": "2025-01-16T10:31:07.074306+00:00"} {"global_step": 10253, "acc_step": 0, "speed/wps": 12901.711984798163, "speed/FLOPS": 202639115531368.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04662491753697395, "optim/lr": 0.002933007664913441, "optim/total_tokens": 5375524864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9008913040161133, "created_at": "2025-01-16T10:31:17.240223+00:00"} {"global_step": 10254, "acc_step": 0, "speed/wps": 12903.289761870466, "speed/FLOPS": 202663896688382.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054185934364795685, "optim/lr": 0.0029329823430178526, "optim/total_tokens": 5376049152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324702, "loss/out": 2.9135336875915527, "created_at": "2025-01-16T10:31:27.402461+00:00"} {"global_step": 10255, "acc_step": 0, "speed/wps": 12901.097131540351, "speed/FLOPS": 202629458416057.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06515273451805115, "optim/lr": 0.002932957016446907, "optim/total_tokens": 5376573440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.921604871749878, "created_at": "2025-01-16T10:31:37.569274+00:00"} {"global_step": 10256, "acc_step": 0, "speed/wps": 12901.96817468985, "speed/FLOPS": 202643139345659.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055799130350351334, "optim/lr": 0.0029329316852006885, "optim/total_tokens": 5377097728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8766489028930664, "created_at": "2025-01-16T10:31:47.733939+00:00"} {"global_step": 10257, "acc_step": 0, "speed/wps": 12896.208245017777, "speed/FLOPS": 202552671735186.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05288078635931015, "optim/lr": 0.0029329063492792777, "optim/total_tokens": 5377622016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 2.94806170463562, "created_at": "2025-01-16T10:31:57.901054+00:00"} {"global_step": 10258, "acc_step": 0, "speed/wps": 12899.06592644515, "speed/FLOPS": 202597555548869.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07130651921033859, "optim/lr": 0.002932881008682758, "optim/total_tokens": 5378146304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8991458415985107, "created_at": "2025-01-16T10:32:08.069006+00:00"} {"global_step": 10259, "acc_step": 0, "speed/wps": 12900.677988451836, "speed/FLOPS": 202622875197889.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05112074688076973, "optim/lr": 0.002932855663411213, "optim/total_tokens": 5378670592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 2.8608217239379883, "created_at": "2025-01-16T10:32:18.232634+00:00"} {"global_step": 10260, "acc_step": 0, "speed/wps": 12898.922620095014, "speed/FLOPS": 202595304725717.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0525985024869442, "optim/lr": 0.0029328303134647238, "optim/total_tokens": 5379194880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9576377868652344, "created_at": "2025-01-16T10:32:28.401137+00:00"} {"global_step": 10261, "acc_step": 0, "speed/wps": 12909.342516083087, "speed/FLOPS": 202758963510647.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07150387763977051, "optim/lr": 0.0029328049588433743, "optim/total_tokens": 5379719168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.907106876373291, "created_at": "2025-01-16T10:32:38.560085+00:00"} {"global_step": 10262, "acc_step": 0, "speed/wps": 12899.756763360469, "speed/FLOPS": 202608406091934.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057040024548769, "optim/lr": 0.002932779599547247, "optim/total_tokens": 5380243456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 2.9188785552978516, "created_at": "2025-01-16T10:32:48.724441+00:00"} {"global_step": 10263, "acc_step": 0, "speed/wps": 12905.952462206687, "speed/FLOPS": 202705718056096.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07033366709947586, "optim/lr": 0.002932754235576424, "optim/total_tokens": 5380767744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.9803590774536133, "created_at": "2025-01-16T10:32:58.886351+00:00"} {"global_step": 10264, "acc_step": 0, "speed/wps": 12903.222806986483, "speed/FLOPS": 202662845069923.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08369710296392441, "optim/lr": 0.002932728866930989, "optim/total_tokens": 5381292032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386402, "loss/out": 2.943686008453369, "created_at": "2025-01-16T10:33:09.057504+00:00"} {"global_step": 10265, "acc_step": 0, "speed/wps": 12903.23703618686, "speed/FLOPS": 202663068559068.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05419459566473961, "optim/lr": 0.0029327034936110245, "optim/total_tokens": 5381816320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9470431804656982, "created_at": "2025-01-16T10:33:19.220732+00:00"} {"global_step": 10266, "acc_step": 0, "speed/wps": 12903.333566784602, "speed/FLOPS": 202664584704752.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07042169570922852, "optim/lr": 0.0029326781156166124, "optim/total_tokens": 5382340608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9438111782073975, "created_at": "2025-01-16T10:33:29.385698+00:00"} {"global_step": 10267, "acc_step": 0, "speed/wps": 12902.737531725064, "speed/FLOPS": 202655223155104.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07078159600496292, "optim/lr": 0.0029326527329478365, "optim/total_tokens": 5382864896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290558, "loss/out": 2.8801329135894775, "created_at": "2025-01-16T10:33:39.550162+00:00"} {"global_step": 10268, "acc_step": 0, "speed/wps": 12898.148950616054, "speed/FLOPS": 202583153183416.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04744228348135948, "optim/lr": 0.0029326273456047793, "optim/total_tokens": 5383389184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.925342559814453, "created_at": "2025-01-16T10:33:49.718852+00:00"} {"global_step": 10269, "acc_step": 0, "speed/wps": 12902.4403368241, "speed/FLOPS": 202650555300796.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05389193445444107, "optim/lr": 0.002932601953587524, "optim/total_tokens": 5383913472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.7677459716796875, "created_at": "2025-01-16T10:33:59.881520+00:00"} {"global_step": 10270, "acc_step": 0, "speed/wps": 12901.532423162533, "speed/FLOPS": 202636295269137.66, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059580057859420776, "optim/lr": 0.002932576556896152, "optim/total_tokens": 5384437760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287430, "loss/out": 2.825780153274536, "created_at": "2025-01-16T10:34:10.046637+00:00"} {"global_step": 10271, "acc_step": 0, "speed/wps": 12900.4340862283, "speed/FLOPS": 202619044378312.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050475239753723145, "optim/lr": 0.0029325511555307483, "optim/total_tokens": 5384962048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.8031036853790283, "created_at": "2025-01-16T10:34:20.212557+00:00"} {"global_step": 10272, "acc_step": 0, "speed/wps": 12886.96335379865, "speed/FLOPS": 202407468014778.66, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05407876893877983, "optim/lr": 0.002932525749491395, "optim/total_tokens": 5385486336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 2.9036271572113037, "created_at": "2025-01-16T10:34:30.389720+00:00"} {"global_step": 10273, "acc_step": 0, "speed/wps": 12893.43941812271, "speed/FLOPS": 202509183504032.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0518152117729187, "optim/lr": 0.002932500338778174, "optim/total_tokens": 5386010624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.844686269760132, "created_at": "2025-01-16T10:34:40.559420+00:00"} {"global_step": 10274, "acc_step": 0, "speed/wps": 12893.708825086362, "speed/FLOPS": 202513414910601.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04806877300143242, "optim/lr": 0.0029324749233911687, "optim/total_tokens": 5386534912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.956169605255127, "created_at": "2025-01-16T10:34:50.734039+00:00"} {"global_step": 10275, "acc_step": 0, "speed/wps": 12886.291671096054, "speed/FLOPS": 202396918314945.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.03938842937350273, "optim/lr": 0.0029324495033304625, "optim/total_tokens": 5387059200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8909473419189453, "created_at": "2025-01-16T10:35:00.909482+00:00"} {"global_step": 10276, "acc_step": 0, "speed/wps": 12900.458803942403, "speed/FLOPS": 202619432603978.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05009599030017853, "optim/lr": 0.0029324240785961386, "optim/total_tokens": 5387583488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306453, "loss/out": 2.857600212097168, "created_at": "2025-01-16T10:35:11.074061+00:00"} {"global_step": 10277, "acc_step": 0, "speed/wps": 12903.16145427054, "speed/FLOPS": 202661881441212.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05218353867530823, "optim/lr": 0.002932398649188279, "optim/total_tokens": 5388107776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9289510250091553, "created_at": "2025-01-16T10:35:21.236653+00:00"} {"global_step": 10278, "acc_step": 0, "speed/wps": 12898.071075994692, "speed/FLOPS": 202581930055477.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07284665107727051, "optim/lr": 0.002932373215106967, "optim/total_tokens": 5388632064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.798443078994751, "created_at": "2025-01-16T10:35:31.408221+00:00"} {"global_step": 10279, "acc_step": 0, "speed/wps": 12902.685193816105, "speed/FLOPS": 202654401116324.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056708924472332, "optim/lr": 0.002932347776352286, "optim/total_tokens": 5389156352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8927698135375977, "created_at": "2025-01-16T10:35:41.571293+00:00"} {"global_step": 10280, "acc_step": 0, "speed/wps": 12903.469080213126, "speed/FLOPS": 202666713129362.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05130363628268242, "optim/lr": 0.002932322332924318, "optim/total_tokens": 5389680640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9263572692871094, "created_at": "2025-01-16T10:35:51.734935+00:00"} {"global_step": 10281, "acc_step": 0, "speed/wps": 12896.142394053608, "speed/FLOPS": 202551637455305.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07808349281549454, "optim/lr": 0.0029322968848231475, "optim/total_tokens": 5390204928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 2.8816745281219482, "created_at": "2025-01-16T10:36:01.903250+00:00"} {"global_step": 10282, "acc_step": 0, "speed/wps": 12904.071862985196, "speed/FLOPS": 202676180661108.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06080474704504013, "optim/lr": 0.0029322714320488565, "optim/total_tokens": 5390729216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 2.926218032836914, "created_at": "2025-01-16T10:36:12.064728+00:00"} {"global_step": 10283, "acc_step": 0, "speed/wps": 12899.895852905336, "speed/FLOPS": 202610590684365.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06396058946847916, "optim/lr": 0.002932245974601528, "optim/total_tokens": 5391253504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9203944206237793, "created_at": "2025-01-16T10:36:22.233303+00:00"} {"global_step": 10284, "acc_step": 0, "speed/wps": 12903.248604829087, "speed/FLOPS": 202663250260491.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06598549336194992, "optim/lr": 0.0029322205124812455, "optim/total_tokens": 5391777792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.971102237701416, "created_at": "2025-01-16T10:36:32.398094+00:00"} {"global_step": 10285, "acc_step": 0, "speed/wps": 12903.115155608493, "speed/FLOPS": 202661154257103.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0679958388209343, "optim/lr": 0.002932195045688092, "optim/total_tokens": 5392302080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.9297075271606445, "created_at": "2025-01-16T10:36:42.559831+00:00"} {"global_step": 10286, "acc_step": 0, "speed/wps": 12899.976573157373, "speed/FLOPS": 202611858506844.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058791905641555786, "optim/lr": 0.00293216957422215, "optim/total_tokens": 5392826368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 3.003347396850586, "created_at": "2025-01-16T10:36:52.726535+00:00"} {"global_step": 10287, "acc_step": 0, "speed/wps": 12903.921345398086, "speed/FLOPS": 202673816575570.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06645455956459045, "optim/lr": 0.0029321440980835042, "optim/total_tokens": 5393350656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9432990550994873, "created_at": "2025-01-16T10:37:02.887651+00:00"} {"global_step": 10288, "acc_step": 0, "speed/wps": 12903.255170536684, "speed/FLOPS": 202663353383951.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05456690490245819, "optim/lr": 0.0029321186172722355, "optim/total_tokens": 5393874944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506128, "loss/out": 2.8478615283966064, "created_at": "2025-01-16T10:37:13.050510+00:00"} {"global_step": 10289, "acc_step": 0, "speed/wps": 12898.03455515434, "speed/FLOPS": 202581356445494.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052483074367046356, "optim/lr": 0.0029320931317884285, "optim/total_tokens": 5394399232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 2.9010872840881348, "created_at": "2025-01-16T10:37:23.217329+00:00"} {"global_step": 10290, "acc_step": 0, "speed/wps": 12894.110803068395, "speed/FLOPS": 202519728527184.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04750756919384003, "optim/lr": 0.0029320676416321662, "optim/total_tokens": 5394923520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.969022750854492, "created_at": "2025-01-16T10:37:33.386133+00:00"} {"global_step": 10291, "acc_step": 0, "speed/wps": 12901.587501508806, "speed/FLOPS": 202637160350251.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05477327108383179, "optim/lr": 0.0029320421468035313, "optim/total_tokens": 5395447808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420369, "loss/out": 2.9757864475250244, "created_at": "2025-01-16T10:37:43.553697+00:00"} {"global_step": 10292, "acc_step": 0, "speed/wps": 12899.991595692292, "speed/FLOPS": 202612094456398.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06066306307911873, "optim/lr": 0.0029320166473026073, "optim/total_tokens": 5395972096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.033735513687134, "created_at": "2025-01-16T10:37:53.720923+00:00"} {"global_step": 10293, "acc_step": 0, "speed/wps": 12902.105722243743, "speed/FLOPS": 202645299719003.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04991995915770531, "optim/lr": 0.0029319911431294776, "optim/total_tokens": 5396496384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8533806800842285, "created_at": "2025-01-16T10:38:03.888450+00:00"} {"global_step": 10294, "acc_step": 0, "speed/wps": 12900.894395437375, "speed/FLOPS": 202626274166972.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05480393394827843, "optim/lr": 0.002931965634284225, "optim/total_tokens": 5397020672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.928652286529541, "created_at": "2025-01-16T10:38:14.051989+00:00"} {"global_step": 10295, "acc_step": 0, "speed/wps": 12905.194815735136, "speed/FLOPS": 202693818177145.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05757933855056763, "optim/lr": 0.002931940120766933, "optim/total_tokens": 5397544960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 2.927466869354248, "created_at": "2025-01-16T10:38:24.249674+00:00"} {"global_step": 10296, "acc_step": 0, "speed/wps": 12902.914952647658, "speed/FLOPS": 202658009794496.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.044636353850364685, "optim/lr": 0.0029319146025776847, "optim/total_tokens": 5398069248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8215856552124023, "created_at": "2025-01-16T10:38:34.413402+00:00"} {"global_step": 10297, "acc_step": 0, "speed/wps": 12896.374525683595, "speed/FLOPS": 202555283401537.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06623031944036484, "optim/lr": 0.0029318890797165632, "optim/total_tokens": 5398593536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.9519083499908447, "created_at": "2025-01-16T10:38:44.581800+00:00"} {"global_step": 10298, "acc_step": 0, "speed/wps": 12899.909602992313, "speed/FLOPS": 202610806648375.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04355079308152199, "optim/lr": 0.0029318635521836524, "optim/total_tokens": 5399117824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 2.849736213684082, "created_at": "2025-01-16T10:38:54.748792+00:00"} {"global_step": 10299, "acc_step": 0, "speed/wps": 12901.144637371935, "speed/FLOPS": 202630204560422.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05584922432899475, "optim/lr": 0.0029318380199790345, "optim/total_tokens": 5399642112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.9383327960968018, "created_at": "2025-01-16T10:39:04.915834+00:00"} {"global_step": 10300, "acc_step": 0, "speed/wps": 12908.462834457761, "speed/FLOPS": 202745146901912.34, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04854588583111763, "optim/lr": 0.002931812483102794, "optim/total_tokens": 5400166400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.920620918273926, "created_at": "2025-01-16T10:39:15.073338+00:00"} {"global_step": 10301, "acc_step": 0, "speed/wps": 12900.861077567108, "speed/FLOPS": 202625750864036.0, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051565103232860565, "optim/lr": 0.0029317869415550132, "optim/total_tokens": 5400690688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.7845146656036377, "created_at": "2025-01-16T10:39:25.236832+00:00"} {"global_step": 10302, "acc_step": 0, "speed/wps": 12899.332238076342, "speed/FLOPS": 202601738338987.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056045420467853546, "optim/lr": 0.0029317613953357764, "optim/total_tokens": 5401214976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.916062831878662, "created_at": "2025-01-16T10:39:35.401678+00:00"} {"global_step": 10303, "acc_step": 0, "speed/wps": 12896.357884306668, "speed/FLOPS": 202555022025844.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05077376589179039, "optim/lr": 0.002931735844445166, "optim/total_tokens": 5401739264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.9650535583496094, "created_at": "2025-01-16T10:39:45.573360+00:00"} {"global_step": 10304, "acc_step": 0, "speed/wps": 12901.367624983302, "speed/FLOPS": 202633706887276.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05670478194952011, "optim/lr": 0.0029317102888832664, "optim/total_tokens": 5402263552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9997434616088867, "created_at": "2025-01-16T10:39:55.740650+00:00"} {"global_step": 10305, "acc_step": 0, "speed/wps": 12903.43250395365, "speed/FLOPS": 202666138648946.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0491194985806942, "optim/lr": 0.0029316847286501597, "optim/total_tokens": 5402787840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.893887996673584, "created_at": "2025-01-16T10:40:05.902260+00:00"} {"global_step": 10306, "acc_step": 0, "speed/wps": 12897.98864907291, "speed/FLOPS": 202580635427404.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05302336812019348, "optim/lr": 0.0029316591637459305, "optim/total_tokens": 5403312128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286083, "loss/out": 2.8857035636901855, "created_at": "2025-01-16T10:40:16.070579+00:00"} {"global_step": 10307, "acc_step": 0, "speed/wps": 12896.423343945937, "speed/FLOPS": 202556050159430.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05328601226210594, "optim/lr": 0.0029316335941706614, "optim/total_tokens": 5403836416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9152939319610596, "created_at": "2025-01-16T10:40:26.238855+00:00"} {"global_step": 10308, "acc_step": 0, "speed/wps": 12902.329060081473, "speed/FLOPS": 202648807546643.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047444019466638565, "optim/lr": 0.0029316080199244363, "optim/total_tokens": 5404360704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 2.8115029335021973, "created_at": "2025-01-16T10:40:36.401500+00:00"} {"global_step": 10309, "acc_step": 0, "speed/wps": 12900.12286611299, "speed/FLOPS": 202614156238740.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05212243273854256, "optim/lr": 0.0029315824410073386, "optim/total_tokens": 5404884992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341330, "loss/out": 2.9836807250976562, "created_at": "2025-01-16T10:40:46.569553+00:00"} {"global_step": 10310, "acc_step": 0, "speed/wps": 12899.160411992887, "speed/FLOPS": 202599039574233.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05870717018842697, "optim/lr": 0.0029315568574194515, "optim/total_tokens": 5405409280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.968153476715088, "created_at": "2025-01-16T10:40:56.737869+00:00"} {"global_step": 10311, "acc_step": 0, "speed/wps": 12899.907745445958, "speed/FLOPS": 202610777473057.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052878182381391525, "optim/lr": 0.0029315312691608583, "optim/total_tokens": 5405933568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398956, "loss/out": 2.845827102661133, "created_at": "2025-01-16T10:41:06.904730+00:00"} {"global_step": 10312, "acc_step": 0, "speed/wps": 12890.103267318462, "speed/FLOPS": 202456784671300.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0031, "optim/grad_norm": 0.046013668179512024, "optim/lr": 0.0029315056762316428, "optim/total_tokens": 5406457856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.0082180500030518, "created_at": "2025-01-16T10:41:17.081449+00:00"} {"global_step": 10313, "acc_step": 0, "speed/wps": 12897.259794922851, "speed/FLOPS": 202569187771427.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04783758148550987, "optim/lr": 0.0029314800786318894, "optim/total_tokens": 5406982144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9791641235351562, "created_at": "2025-01-16T10:41:27.249672+00:00"} {"global_step": 10314, "acc_step": 0, "speed/wps": 12893.96722641625, "speed/FLOPS": 202517473458567.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05549751967191696, "optim/lr": 0.00293145447636168, "optim/total_tokens": 5407506432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.823432445526123, "created_at": "2025-01-16T10:41:37.419738+00:00"} {"global_step": 10315, "acc_step": 0, "speed/wps": 12886.688909313596, "speed/FLOPS": 202403157487014.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052938640117645264, "optim/lr": 0.0029314288694210987, "optim/total_tokens": 5408030720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 2.8981661796569824, "created_at": "2025-01-16T10:41:47.598125+00:00"} {"global_step": 10316, "acc_step": 0, "speed/wps": 12904.947778470932, "speed/FLOPS": 202689938117446.1, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05445428565144539, "optim/lr": 0.0029314032578102297, "optim/total_tokens": 5408555008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.904500722885132, "created_at": "2025-01-16T10:41:57.758484+00:00"} {"global_step": 10317, "acc_step": 0, "speed/wps": 12897.186495528125, "speed/FLOPS": 202568036503708.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04430652782320976, "optim/lr": 0.0029313776415291555, "optim/total_tokens": 5409079296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.867551326751709, "created_at": "2025-01-16T10:42:07.925940+00:00"} {"global_step": 10318, "acc_step": 0, "speed/wps": 12904.032037917652, "speed/FLOPS": 202675555153697.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04787867143750191, "optim/lr": 0.0029313520205779603, "optim/total_tokens": 5409603584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8544507026672363, "created_at": "2025-01-16T10:42:18.088276+00:00"} {"global_step": 10319, "acc_step": 0, "speed/wps": 12896.56909897005, "speed/FLOPS": 202558339442372.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04582718759775162, "optim/lr": 0.002931326394956728, "optim/total_tokens": 5410127872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8696866035461426, "created_at": "2025-01-16T10:42:28.255170+00:00"} {"global_step": 10320, "acc_step": 0, "speed/wps": 12892.64787882155, "speed/FLOPS": 202496751291619.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05165794491767883, "optim/lr": 0.0029313007646655414, "optim/total_tokens": 5410652160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.81843900680542, "created_at": "2025-01-16T10:42:38.426861+00:00"} {"global_step": 10321, "acc_step": 0, "speed/wps": 12895.224382892084, "speed/FLOPS": 202537218828534.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0628587082028389, "optim/lr": 0.0029312751297044847, "optim/total_tokens": 5411176448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 2.879237651824951, "created_at": "2025-01-16T10:42:48.597482+00:00"} {"global_step": 10322, "acc_step": 0, "speed/wps": 12901.438515489177, "speed/FLOPS": 202634820320084.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045763127505779266, "optim/lr": 0.0029312494900736413, "optim/total_tokens": 5411700736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.977750301361084, "created_at": "2025-01-16T10:42:58.760980+00:00"} {"global_step": 10323, "acc_step": 0, "speed/wps": 12905.444474248274, "speed/FLOPS": 202697739407160.75, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06084812432527542, "optim/lr": 0.0029312238457730954, "optim/total_tokens": 5412225024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.850309133529663, "created_at": "2025-01-16T10:43:08.923680+00:00"} {"global_step": 10324, "acc_step": 0, "speed/wps": 12906.258909860178, "speed/FLOPS": 202710531237597.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06255483627319336, "optim/lr": 0.0029311981968029297, "optim/total_tokens": 5412749312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.801384210586548, "created_at": "2025-01-16T10:43:19.085041+00:00"} {"global_step": 10325, "acc_step": 0, "speed/wps": 12901.775399438562, "speed/FLOPS": 202640111545437.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04324573278427124, "optim/lr": 0.0029311725431632285, "optim/total_tokens": 5413273600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8975396156311035, "created_at": "2025-01-16T10:43:29.249112+00:00"} {"global_step": 10326, "acc_step": 0, "speed/wps": 12900.04326378375, "speed/FLOPS": 202612905974774.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04933139681816101, "optim/lr": 0.002931146884854076, "optim/total_tokens": 5413797888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.8910725116729736, "created_at": "2025-01-16T10:43:39.417619+00:00"} {"global_step": 10327, "acc_step": 0, "speed/wps": 12905.662299188687, "speed/FLOPS": 202701160647172.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04535326361656189, "optim/lr": 0.002931121221875554, "optim/total_tokens": 5414322176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8713204860687256, "created_at": "2025-01-16T10:43:49.577397+00:00"} {"global_step": 10328, "acc_step": 0, "speed/wps": 12896.755413659153, "speed/FLOPS": 202561265770588.16, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04970874637365341, "optim/lr": 0.0029310955542277486, "optim/total_tokens": 5414846464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.9595823287963867, "created_at": "2025-01-16T10:43:59.744135+00:00"} {"global_step": 10329, "acc_step": 0, "speed/wps": 12892.164032826062, "speed/FLOPS": 202489151825386.03, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04090987145900726, "optim/lr": 0.0029310698819107416, "optim/total_tokens": 5415370752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 275631, "loss/out": 2.950561046600342, "created_at": "2025-01-16T10:44:09.914346+00:00"} {"global_step": 10330, "acc_step": 0, "speed/wps": 12901.959204530487, "speed/FLOPS": 202642998456980.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04226921126246452, "optim/lr": 0.0029310442049246188, "optim/total_tokens": 5415895040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.751431703567505, "created_at": "2025-01-16T10:44:20.079748+00:00"} {"global_step": 10331, "acc_step": 0, "speed/wps": 12900.625449664623, "speed/FLOPS": 202622050004040.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050586868077516556, "optim/lr": 0.002931018523269462, "optim/total_tokens": 5416419328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 2.875403881072998, "created_at": "2025-01-16T10:44:30.252287+00:00"} {"global_step": 10332, "acc_step": 0, "speed/wps": 12897.842073553678, "speed/FLOPS": 202578333257458.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05098310485482216, "optim/lr": 0.002930992836945356, "optim/total_tokens": 5416943616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.0031707286834717, "created_at": "2025-01-16T10:44:40.419822+00:00"} {"global_step": 10333, "acc_step": 0, "speed/wps": 12903.772520797242, "speed/FLOPS": 202671479080707.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09862487763166428, "optim/lr": 0.0029309671459523843, "optim/total_tokens": 5417467904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8792033195495605, "created_at": "2025-01-16T10:44:50.583778+00:00"} {"global_step": 10334, "acc_step": 0, "speed/wps": 12896.56471218452, "speed/FLOPS": 202558270541877.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08937795460224152, "optim/lr": 0.002930941450290631, "optim/total_tokens": 5417992192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347841, "loss/out": 2.8891658782958984, "created_at": "2025-01-16T10:45:00.750577+00:00"} {"global_step": 10335, "acc_step": 0, "speed/wps": 12902.131462393465, "speed/FLOPS": 202645704003425.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060915347188711166, "optim/lr": 0.00293091574996018, "optim/total_tokens": 5418516480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498635, "loss/out": 2.929896831512451, "created_at": "2025-01-16T10:45:10.913086+00:00"} {"global_step": 10336, "acc_step": 0, "speed/wps": 12902.241427523028, "speed/FLOPS": 202647431156893.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0604153573513031, "optim/lr": 0.0029308900449611145, "optim/total_tokens": 5419040768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.9487197399139404, "created_at": "2025-01-16T10:45:21.075509+00:00"} {"global_step": 10337, "acc_step": 0, "speed/wps": 12903.182719802055, "speed/FLOPS": 202662215445606.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06056586652994156, "optim/lr": 0.002930864335293519, "optim/total_tokens": 5419565056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.917553424835205, "created_at": "2025-01-16T10:45:31.237706+00:00"} {"global_step": 10338, "acc_step": 0, "speed/wps": 12900.827530451148, "speed/FLOPS": 202625223960481.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07478151470422745, "optim/lr": 0.002930838620957477, "optim/total_tokens": 5420089344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.9743716716766357, "created_at": "2025-01-16T10:45:41.403615+00:00"} {"global_step": 10339, "acc_step": 0, "speed/wps": 12896.198404849805, "speed/FLOPS": 202552517181827.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06945932656526566, "optim/lr": 0.002930812901953073, "optim/total_tokens": 5420613632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470674, "loss/out": 2.8886821269989014, "created_at": "2025-01-16T10:45:51.573121+00:00"} {"global_step": 10340, "acc_step": 0, "speed/wps": 12903.102468103563, "speed/FLOPS": 202660954982404.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06237310171127319, "optim/lr": 0.0029307871782803903, "optim/total_tokens": 5421137920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9141578674316406, "created_at": "2025-01-16T10:46:01.738767+00:00"} {"global_step": 10341, "acc_step": 0, "speed/wps": 12905.425849729882, "speed/FLOPS": 202697446883506.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07466049492359161, "optim/lr": 0.0029307614499395123, "optim/total_tokens": 5421662208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.995967388153076, "created_at": "2025-01-16T10:46:11.898958+00:00"} {"global_step": 10342, "acc_step": 0, "speed/wps": 12901.641803281449, "speed/FLOPS": 202638013234209.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06270492076873779, "optim/lr": 0.0029307357169305243, "optim/total_tokens": 5422186496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408354, "loss/out": 2.951533079147339, "created_at": "2025-01-16T10:46:22.064914+00:00"} {"global_step": 10343, "acc_step": 0, "speed/wps": 12901.839003942565, "speed/FLOPS": 202641110541574.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08281895518302917, "optim/lr": 0.00293070997925351, "optim/total_tokens": 5422710784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8469834327697754, "created_at": "2025-01-16T10:46:32.227818+00:00"} {"global_step": 10344, "acc_step": 0, "speed/wps": 12903.180926377798, "speed/FLOPS": 202662187277414.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050380051136016846, "optim/lr": 0.0029306842369085525, "optim/total_tokens": 5423235072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 2.8627893924713135, "created_at": "2025-01-16T10:46:42.389896+00:00"} {"global_step": 10345, "acc_step": 0, "speed/wps": 12899.576097454716, "speed/FLOPS": 202605568485623.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07031995058059692, "optim/lr": 0.0029306584898957364, "optim/total_tokens": 5423759360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9098289012908936, "created_at": "2025-01-16T10:46:52.555565+00:00"} {"global_step": 10346, "acc_step": 0, "speed/wps": 12904.777188384376, "speed/FLOPS": 202687258765721.38, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050257712602615356, "optim/lr": 0.0029306327382151456, "optim/total_tokens": 5424283648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.900074005126953, "created_at": "2025-01-16T10:47:02.719132+00:00"} {"global_step": 10347, "acc_step": 0, "speed/wps": 12908.267419078331, "speed/FLOPS": 202742077634849.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07191559672355652, "optim/lr": 0.002930606981866864, "optim/total_tokens": 5424807936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 2.926483631134033, "created_at": "2025-01-16T10:47:12.879013+00:00"} {"global_step": 10348, "acc_step": 0, "speed/wps": 12896.37275896813, "speed/FLOPS": 202555255652842.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04977783188223839, "optim/lr": 0.0029305812208509758, "optim/total_tokens": 5425332224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.7421560287475586, "created_at": "2025-01-16T10:47:23.046009+00:00"} {"global_step": 10349, "acc_step": 0, "speed/wps": 12901.95792003875, "speed/FLOPS": 202642978282272.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0518515408039093, "optim/lr": 0.002930555455167565, "optim/total_tokens": 5425856512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9241857528686523, "created_at": "2025-01-16T10:47:33.212236+00:00"} {"global_step": 10350, "acc_step": 0, "speed/wps": 12894.90142796636, "speed/FLOPS": 202532146377639.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051447223871946335, "optim/lr": 0.0029305296848167158, "optim/total_tokens": 5426380800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 2.7413268089294434, "created_at": "2025-01-16T10:47:43.383507+00:00"} {"global_step": 10351, "acc_step": 0, "speed/wps": 12903.808360722189, "speed/FLOPS": 202672041995978.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05654403194785118, "optim/lr": 0.002930503909798512, "optim/total_tokens": 5426905088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.8966684341430664, "created_at": "2025-01-16T10:47:53.545548+00:00"} {"global_step": 10352, "acc_step": 0, "speed/wps": 12897.016175573506, "speed/FLOPS": 202565361394777.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38670703768730164, "optim/lr": 0.0029304781301130378, "optim/total_tokens": 5427429376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 2.8482630252838135, "created_at": "2025-01-16T10:48:03.715768+00:00"} {"global_step": 10353, "acc_step": 0, "speed/wps": 12901.156837332768, "speed/FLOPS": 202630396177572.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08833363652229309, "optim/lr": 0.002930452345760377, "optim/total_tokens": 5427953664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 2.899674654006958, "created_at": "2025-01-16T10:48:13.881985+00:00"} {"global_step": 10354, "acc_step": 0, "speed/wps": 12897.745587278483, "speed/FLOPS": 202576817807920.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06938038021326065, "optim/lr": 0.0029304265567406144, "optim/total_tokens": 5428477952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.971693754196167, "created_at": "2025-01-16T10:48:24.051739+00:00"} {"global_step": 10355, "acc_step": 0, "speed/wps": 12902.067435114266, "speed/FLOPS": 202644698367023.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055993158370256424, "optim/lr": 0.0029304007630538338, "optim/total_tokens": 5429002240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8644661903381348, "created_at": "2025-01-16T10:48:34.221716+00:00"} {"global_step": 10356, "acc_step": 0, "speed/wps": 12904.6729981672, "speed/FLOPS": 202685622315188.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050172850489616394, "optim/lr": 0.002930374964700119, "optim/total_tokens": 5429526528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 3.0126967430114746, "created_at": "2025-01-16T10:48:44.382106+00:00"} {"global_step": 10357, "acc_step": 0, "speed/wps": 12897.327476280232, "speed/FLOPS": 202570250800148.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06132901459932327, "optim/lr": 0.0029303491616795552, "optim/total_tokens": 5430050816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8632678985595703, "created_at": "2025-01-16T10:48:54.553111+00:00"} {"global_step": 10358, "acc_step": 0, "speed/wps": 12900.777903082575, "speed/FLOPS": 202624444494462.7, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07081499695777893, "optim/lr": 0.002930323353992226, "optim/total_tokens": 5430575104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 2.830376148223877, "created_at": "2025-01-16T10:49:04.718578+00:00"} {"global_step": 10359, "acc_step": 0, "speed/wps": 12902.586776748696, "speed/FLOPS": 202652855341042.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05375867709517479, "optim/lr": 0.0029302975416382144, "optim/total_tokens": 5431099392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8448896408081055, "created_at": "2025-01-16T10:49:14.880914+00:00"} {"global_step": 10360, "acc_step": 0, "speed/wps": 12904.378017822391, "speed/FLOPS": 202680989243527.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08002234995365143, "optim/lr": 0.002930271724617607, "optim/total_tokens": 5431623680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8997693061828613, "created_at": "2025-01-16T10:49:25.041916+00:00"} {"global_step": 10361, "acc_step": 0, "speed/wps": 12901.611102434352, "speed/FLOPS": 202637531035218.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05697737634181976, "optim/lr": 0.002930245902930486, "optim/total_tokens": 5432147968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 2.931823492050171, "created_at": "2025-01-16T10:49:35.205698+00:00"} {"global_step": 10362, "acc_step": 0, "speed/wps": 12901.87578761339, "speed/FLOPS": 202641688279670.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05889732018113136, "optim/lr": 0.0029302200765769365, "optim/total_tokens": 5432672256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.939476251602173, "created_at": "2025-01-16T10:49:45.372305+00:00"} {"global_step": 10363, "acc_step": 0, "speed/wps": 12906.832425494094, "speed/FLOPS": 202719539088721.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0498788021504879, "optim/lr": 0.002930194245557043, "optim/total_tokens": 5433196544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.855560779571533, "created_at": "2025-01-16T10:49:55.531080+00:00"} {"global_step": 10364, "acc_step": 0, "speed/wps": 12904.747453556394, "speed/FLOPS": 202686791739388.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05908632650971413, "optim/lr": 0.0029301684098708893, "optim/total_tokens": 5433720832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.960242748260498, "created_at": "2025-01-16T10:50:05.691486+00:00"} {"global_step": 10365, "acc_step": 0, "speed/wps": 12905.085273595494, "speed/FLOPS": 202692097667315.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05972081422805786, "optim/lr": 0.00293014256951856, "optim/total_tokens": 5434245120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.8990488052368164, "created_at": "2025-01-16T10:50:15.853224+00:00"} {"global_step": 10366, "acc_step": 0, "speed/wps": 12908.693971271043, "speed/FLOPS": 202748777223180.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04638992249965668, "optim/lr": 0.0029301167245001387, "optim/total_tokens": 5434769408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9945340156555176, "created_at": "2025-01-16T10:50:26.012363+00:00"} {"global_step": 10367, "acc_step": 0, "speed/wps": 12900.600771517744, "speed/FLOPS": 202621662399832.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04487038776278496, "optim/lr": 0.002930090874815711, "optim/total_tokens": 5435293696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.871854782104492, "created_at": "2025-01-16T10:50:36.176019+00:00"} {"global_step": 10368, "acc_step": 0, "speed/wps": 12904.199222030382, "speed/FLOPS": 202678181009922.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053417086601257324, "optim/lr": 0.0029300650204653602, "optim/total_tokens": 5435817984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446221, "loss/out": 2.928062677383423, "created_at": "2025-01-16T10:50:46.340779+00:00"} {"global_step": 10369, "acc_step": 0, "speed/wps": 12906.64897746863, "speed/FLOPS": 202716657785397.34, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04966738447546959, "optim/lr": 0.0029300391614491715, "optim/total_tokens": 5436342272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.924393653869629, "created_at": "2025-01-16T10:50:56.499951+00:00"} {"global_step": 10370, "acc_step": 0, "speed/wps": 12902.5597099526, "speed/FLOPS": 202652430219815.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05006634071469307, "optim/lr": 0.002930013297767228, "optim/total_tokens": 5436866560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.859086036682129, "created_at": "2025-01-16T10:51:06.665578+00:00"} {"global_step": 10371, "acc_step": 0, "speed/wps": 12903.452827489897, "speed/FLOPS": 202666457858011.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0557774119079113, "optim/lr": 0.002929987429419615, "optim/total_tokens": 5437390848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8784875869750977, "created_at": "2025-01-16T10:51:16.827698+00:00"} {"global_step": 10372, "acc_step": 0, "speed/wps": 12903.837227546663, "speed/FLOPS": 202672495389123.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06425256282091141, "optim/lr": 0.002929961556406417, "optim/total_tokens": 5437915136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.8282713890075684, "created_at": "2025-01-16T10:51:26.988968+00:00"} {"global_step": 10373, "acc_step": 0, "speed/wps": 12904.119428588532, "speed/FLOPS": 202676927744270.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0656263679265976, "optim/lr": 0.0029299356787277175, "optim/total_tokens": 5438439424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 2.8631269931793213, "created_at": "2025-01-16T10:51:37.150031+00:00"} {"global_step": 10374, "acc_step": 0, "speed/wps": 12905.533575170837, "speed/FLOPS": 202699138859587.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05611550435423851, "optim/lr": 0.0029299097963836024, "optim/total_tokens": 5438963712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.7810893058776855, "created_at": "2025-01-16T10:51:47.313190+00:00"} {"global_step": 10375, "acc_step": 0, "speed/wps": 12909.881193403622, "speed/FLOPS": 202767424178186.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054683148860931396, "optim/lr": 0.0029298839093741553, "optim/total_tokens": 5439488000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.933016538619995, "created_at": "2025-01-16T10:51:57.469594+00:00"} {"global_step": 10376, "acc_step": 0, "speed/wps": 12908.918722490258, "speed/FLOPS": 202752307249918.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06948404014110565, "optim/lr": 0.00292985801769946, "optim/total_tokens": 5440012288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8713791370391846, "created_at": "2025-01-16T10:52:07.635778+00:00"} {"global_step": 10377, "acc_step": 0, "speed/wps": 12902.975718637319, "speed/FLOPS": 202658964207865.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06077083200216293, "optim/lr": 0.0029298321213596023, "optim/total_tokens": 5440536576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.7630112171173096, "created_at": "2025-01-16T10:52:17.798046+00:00"} {"global_step": 10378, "acc_step": 0, "speed/wps": 12911.08382563409, "speed/FLOPS": 202786313169955.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06339684873819351, "optim/lr": 0.002929806220354666, "optim/total_tokens": 5441060864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 2.942519187927246, "created_at": "2025-01-16T10:52:27.953360+00:00"} {"global_step": 10379, "acc_step": 0, "speed/wps": 12908.011588664127, "speed/FLOPS": 202738059466646.6, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05279259756207466, "optim/lr": 0.0029297803146847356, "optim/total_tokens": 5441585152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8389334678649902, "created_at": "2025-01-16T10:52:38.111653+00:00"} {"global_step": 10380, "acc_step": 0, "speed/wps": 12905.312303814668, "speed/FLOPS": 202695663488879.78, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05330080911517143, "optim/lr": 0.0029297544043498957, "optim/total_tokens": 5442109440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 2.969681739807129, "created_at": "2025-01-16T10:52:48.274608+00:00"} {"global_step": 10381, "acc_step": 0, "speed/wps": 12903.515736332718, "speed/FLOPS": 202667445927833.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05113430693745613, "optim/lr": 0.0029297284893502304, "optim/total_tokens": 5442633728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.900587797164917, "created_at": "2025-01-16T10:52:58.439380+00:00"} {"global_step": 10382, "acc_step": 0, "speed/wps": 12908.538247617958, "speed/FLOPS": 202746331369221.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05228270962834358, "optim/lr": 0.0029297025696858247, "optim/total_tokens": 5443158016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.9005703926086426, "created_at": "2025-01-16T10:53:08.596729+00:00"} {"global_step": 10383, "acc_step": 0, "speed/wps": 12902.710389773067, "speed/FLOPS": 202654796853451.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048352599143981934, "optim/lr": 0.0029296766453567637, "optim/total_tokens": 5443682304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 2.779850482940674, "created_at": "2025-01-16T10:53:18.759170+00:00"} {"global_step": 10384, "acc_step": 0, "speed/wps": 12901.876868654348, "speed/FLOPS": 202641705258904.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05246658995747566, "optim/lr": 0.0029296507163631314, "optim/total_tokens": 5444206592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8602170944213867, "created_at": "2025-01-16T10:53:28.921987+00:00"} {"global_step": 10385, "acc_step": 0, "speed/wps": 12902.994373939364, "speed/FLOPS": 202659257215018.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06652417778968811, "optim/lr": 0.002929624782705012, "optim/total_tokens": 5444730880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 2.7789225578308105, "created_at": "2025-01-16T10:53:39.084978+00:00"} {"global_step": 10386, "acc_step": 0, "speed/wps": 12896.683709526713, "speed/FLOPS": 202560139558656.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06113440543413162, "optim/lr": 0.002929598844382491, "optim/total_tokens": 5445255168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.842438220977783, "created_at": "2025-01-16T10:53:49.252851+00:00"} {"global_step": 10387, "acc_step": 0, "speed/wps": 12902.022416979391, "speed/FLOPS": 202643991295353.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06098632141947746, "optim/lr": 0.002929572901395652, "optim/total_tokens": 5445779456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8912065029144287, "created_at": "2025-01-16T10:53:59.418178+00:00"} {"global_step": 10388, "acc_step": 0, "speed/wps": 12912.204111144052, "speed/FLOPS": 202803908793323.44, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07063981890678406, "optim/lr": 0.0029295469537445806, "optim/total_tokens": 5446303744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 2.8492465019226074, "created_at": "2025-01-16T10:54:09.573703+00:00"} {"global_step": 10389, "acc_step": 0, "speed/wps": 12902.233587763023, "speed/FLOPS": 202647308022689.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04949319362640381, "optim/lr": 0.0029295210014293606, "optim/total_tokens": 5446828032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.921468734741211, "created_at": "2025-01-16T10:54:19.736083+00:00"} {"global_step": 10390, "acc_step": 0, "speed/wps": 12909.280517835445, "speed/FLOPS": 202757989742973.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0701613575220108, "optim/lr": 0.0029294950444500776, "optim/total_tokens": 5447352320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344503, "loss/out": 2.883626937866211, "created_at": "2025-01-16T10:54:29.893155+00:00"} {"global_step": 10391, "acc_step": 0, "speed/wps": 12903.680068836386, "speed/FLOPS": 202670026995619.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05763046443462372, "optim/lr": 0.002929469082806816, "optim/total_tokens": 5447876608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.890427827835083, "created_at": "2025-01-16T10:54:40.055585+00:00"} {"global_step": 10392, "acc_step": 0, "speed/wps": 12903.725429807097, "speed/FLOPS": 202670739451997.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06667964160442352, "optim/lr": 0.00292944311649966, "optim/total_tokens": 5448400896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8625075817108154, "created_at": "2025-01-16T10:54:50.217938+00:00"} {"global_step": 10393, "acc_step": 0, "speed/wps": 12906.115683898524, "speed/FLOPS": 202708281677056.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05361354723572731, "optim/lr": 0.002929417145528695, "optim/total_tokens": 5448925184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.0506978034973145, "created_at": "2025-01-16T10:55:00.380878+00:00"} {"global_step": 10394, "acc_step": 0, "speed/wps": 12901.49150677237, "speed/FLOPS": 202635652621005.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06081240251660347, "optim/lr": 0.0029293911698940046, "optim/total_tokens": 5449449472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9610390663146973, "created_at": "2025-01-16T10:55:10.546150+00:00"} {"global_step": 10395, "acc_step": 0, "speed/wps": 12907.033998849745, "speed/FLOPS": 202722705075262.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055664803832769394, "optim/lr": 0.0029293651895956753, "optim/total_tokens": 5449973760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9845144748687744, "created_at": "2025-01-16T10:55:20.704717+00:00"} {"global_step": 10396, "acc_step": 0, "speed/wps": 12906.336641934986, "speed/FLOPS": 202711752126647.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04786672815680504, "optim/lr": 0.0029293392046337904, "optim/total_tokens": 5450498048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.02260160446167, "created_at": "2025-01-16T10:55:30.864410+00:00"} {"global_step": 10397, "acc_step": 0, "speed/wps": 12900.114992271874, "speed/FLOPS": 202614032569245.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05487145483493805, "optim/lr": 0.0029293132150084353, "optim/total_tokens": 5451022336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.861889362335205, "created_at": "2025-01-16T10:55:41.031530+00:00"} {"global_step": 10398, "acc_step": 0, "speed/wps": 12902.346771969444, "speed/FLOPS": 202649085736182.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059755269438028336, "optim/lr": 0.0029292872207196947, "optim/total_tokens": 5451546624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8939433097839355, "created_at": "2025-01-16T10:55:51.196536+00:00"} {"global_step": 10399, "acc_step": 0, "speed/wps": 12902.015101458652, "speed/FLOPS": 202643876395047.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0411054901778698, "optim/lr": 0.002929261221767653, "optim/total_tokens": 5452070912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.780154228210449, "created_at": "2025-01-16T10:56:01.359037+00:00"} {"global_step": 10400, "acc_step": 0, "speed/wps": 12902.964799928173, "speed/FLOPS": 202658792714534.0, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04821860045194626, "optim/lr": 0.0029292352181523964, "optim/total_tokens": 5452595200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 2.913205623626709, "created_at": "2025-01-16T10:56:11.520797+00:00"} {"global_step": 10401, "acc_step": 0, "speed/wps": 12900.732995645325, "speed/FLOPS": 202623739161451.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042803436517715454, "optim/lr": 0.002929209209874008, "optim/total_tokens": 5453119488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.8219187259674072, "created_at": "2025-01-16T10:56:21.688472+00:00"} {"global_step": 10402, "acc_step": 0, "speed/wps": 12906.421656398026, "speed/FLOPS": 202713087395611.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05981418862938881, "optim/lr": 0.0029291831969325732, "optim/total_tokens": 5453643776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.835775375366211, "created_at": "2025-01-16T10:56:31.847575+00:00"} {"global_step": 10403, "acc_step": 0, "speed/wps": 12900.919443673441, "speed/FLOPS": 202626667583939.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06504110246896744, "optim/lr": 0.0029291571793281777, "optim/total_tokens": 5454168064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 264162, "loss/out": 2.895890712738037, "created_at": "2025-01-16T10:56:42.016361+00:00"} {"global_step": 10404, "acc_step": 0, "speed/wps": 12898.328163928869, "speed/FLOPS": 202585967974759.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055321503430604935, "optim/lr": 0.0029291311570609053, "optim/total_tokens": 5454692352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8223483562469482, "created_at": "2025-01-16T10:56:52.182141+00:00"} {"global_step": 10405, "acc_step": 0, "speed/wps": 12904.924532164545, "speed/FLOPS": 202689573002260.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07615973055362701, "optim/lr": 0.002929105130130842, "optim/total_tokens": 5455216640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 2.912733554840088, "created_at": "2025-01-16T10:57:02.344744+00:00"} {"global_step": 10406, "acc_step": 0, "speed/wps": 12902.391397295072, "speed/FLOPS": 202649786638243.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04659644514322281, "optim/lr": 0.002929079098538071, "optim/total_tokens": 5455740928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8895654678344727, "created_at": "2025-01-16T10:57:12.508126+00:00"} {"global_step": 10407, "acc_step": 0, "speed/wps": 12898.573282166017, "speed/FLOPS": 202589817893501.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04450651630759239, "optim/lr": 0.0029290530622826795, "optim/total_tokens": 5456265216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8719358444213867, "created_at": "2025-01-16T10:57:22.673830+00:00"} {"global_step": 10408, "acc_step": 0, "speed/wps": 12906.302250856319, "speed/FLOPS": 202711211967498.7, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06863203644752502, "optim/lr": 0.0029290270213647505, "optim/total_tokens": 5456789504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.859463691711426, "created_at": "2025-01-16T10:57:32.833825+00:00"} {"global_step": 10409, "acc_step": 0, "speed/wps": 12901.5629959035, "speed/FLOPS": 202636775456045.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060692138969898224, "optim/lr": 0.00292900097578437, "optim/total_tokens": 5457313792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.0963521003723145, "created_at": "2025-01-16T10:57:42.999136+00:00"} {"global_step": 10410, "acc_step": 0, "speed/wps": 12901.608787872212, "speed/FLOPS": 202637494681839.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.072867751121521, "optim/lr": 0.0029289749255416225, "optim/total_tokens": 5457838080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 2.926356792449951, "created_at": "2025-01-16T10:57:53.163825+00:00"} {"global_step": 10411, "acc_step": 0, "speed/wps": 12905.976851936675, "speed/FLOPS": 202706101130320.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07294125109910965, "optim/lr": 0.002928948870636594, "optim/total_tokens": 5458362368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.0245258808135986, "created_at": "2025-01-16T10:58:03.323895+00:00"} {"global_step": 10412, "acc_step": 0, "speed/wps": 12899.189526399721, "speed/FLOPS": 202599496856000.2, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05639192461967468, "optim/lr": 0.002928922811069368, "optim/total_tokens": 5458886656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.942181348800659, "created_at": "2025-01-16T10:58:13.489516+00:00"} {"global_step": 10413, "acc_step": 0, "speed/wps": 12899.97977039446, "speed/FLOPS": 202611908723846.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08505203574895859, "optim/lr": 0.0029288967468400306, "optim/total_tokens": 5459410944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 2.957503318786621, "created_at": "2025-01-16T10:58:23.655266+00:00"} {"global_step": 10414, "acc_step": 0, "speed/wps": 12907.576654063683, "speed/FLOPS": 202731228221083.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05759970471262932, "optim/lr": 0.002928870677948666, "optim/total_tokens": 5459935232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.928574562072754, "created_at": "2025-01-16T10:58:33.819519+00:00"} {"global_step": 10415, "acc_step": 0, "speed/wps": 12900.845789294091, "speed/FLOPS": 202625510740700.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.06263432651758194, "optim/lr": 0.0029288446043953605, "optim/total_tokens": 5460459520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.7417521476745605, "created_at": "2025-01-16T10:58:43.983628+00:00"} {"global_step": 10416, "acc_step": 0, "speed/wps": 12904.430769396242, "speed/FLOPS": 202681817779483.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0625075250864029, "optim/lr": 0.0029288185261801976, "optim/total_tokens": 5460983808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.8821396827697754, "created_at": "2025-01-16T10:58:54.145605+00:00"} {"global_step": 10417, "acc_step": 0, "speed/wps": 12901.241202893083, "speed/FLOPS": 202631721254627.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.047536950558423996, "optim/lr": 0.0029287924433032634, "optim/total_tokens": 5461508096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.936893939971924, "created_at": "2025-01-16T10:59:04.309809+00:00"} {"global_step": 10418, "acc_step": 0, "speed/wps": 12902.351032576436, "speed/FLOPS": 202649152654870.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05117090418934822, "optim/lr": 0.0029287663557646436, "optim/total_tokens": 5462032384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 2.921875476837158, "created_at": "2025-01-16T10:59:14.477261+00:00"} {"global_step": 10419, "acc_step": 0, "speed/wps": 12902.550146346923, "speed/FLOPS": 202652280010246.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.046187564730644226, "optim/lr": 0.0029287402635644216, "optim/total_tokens": 5462556672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.7824108600616455, "created_at": "2025-01-16T10:59:24.641089+00:00"} {"global_step": 10420, "acc_step": 0, "speed/wps": 12897.811636646336, "speed/FLOPS": 202577855204004.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04874676465988159, "optim/lr": 0.0029287141667026836, "optim/total_tokens": 5463080960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.877227544784546, "created_at": "2025-01-16T10:59:34.807107+00:00"} {"global_step": 10421, "acc_step": 0, "speed/wps": 12895.25369811944, "speed/FLOPS": 202537679264463.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045574236661195755, "optim/lr": 0.0029286880651795146, "optim/total_tokens": 5463605248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396273, "loss/out": 2.9376096725463867, "created_at": "2025-01-16T10:59:44.977237+00:00"} {"global_step": 10422, "acc_step": 0, "speed/wps": 12896.534420755534, "speed/FLOPS": 202557794773360.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.045612633228302, "optim/lr": 0.002928661958995, "optim/total_tokens": 5464129536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.045356273651123, "created_at": "2025-01-16T10:59:55.148149+00:00"} {"global_step": 10423, "acc_step": 0, "speed/wps": 12904.459739580248, "speed/FLOPS": 202682272796032.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05110128968954086, "optim/lr": 0.002928635848149224, "optim/total_tokens": 5464653824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.0560123920440674, "created_at": "2025-01-16T11:00:05.309994+00:00"} {"global_step": 10424, "acc_step": 0, "speed/wps": 12898.577305032688, "speed/FLOPS": 202589881078150.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04231536388397217, "optim/lr": 0.0029286097326422732, "optim/total_tokens": 5465178112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.960146427154541, "created_at": "2025-01-16T11:00:15.478278+00:00"} {"global_step": 10425, "acc_step": 0, "speed/wps": 12882.642362638895, "speed/FLOPS": 202339600910950.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04291941970586777, "optim/lr": 0.0029285836124742326, "optim/total_tokens": 5465702400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9071404933929443, "created_at": "2025-01-16T11:00:25.656875+00:00"} {"global_step": 10426, "acc_step": 0, "speed/wps": 12890.095104651129, "speed/FLOPS": 202456656465393.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041356831789016724, "optim/lr": 0.002928557487645186, "optim/total_tokens": 5466226688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.929417610168457, "created_at": "2025-01-16T11:00:35.832044+00:00"} {"global_step": 10427, "acc_step": 0, "speed/wps": 12898.265425049683, "speed/FLOPS": 202584982574449.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05584132671356201, "optim/lr": 0.00292853135815522, "optim/total_tokens": 5466750976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 2.8802309036254883, "created_at": "2025-01-16T11:00:45.999706+00:00"} {"global_step": 10428, "acc_step": 0, "speed/wps": 12896.530327074579, "speed/FLOPS": 202557730476476.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055347442626953125, "optim/lr": 0.002928505224004419, "optim/total_tokens": 5467275264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9372329711914062, "created_at": "2025-01-16T11:00:56.167690+00:00"} {"global_step": 10429, "acc_step": 0, "speed/wps": 12908.633345585426, "speed/FLOPS": 202747825013479.16, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06183192506432533, "optim/lr": 0.0029284790851928696, "optim/total_tokens": 5467799552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8790791034698486, "created_at": "2025-01-16T11:01:06.328511+00:00"} {"global_step": 10430, "acc_step": 0, "speed/wps": 12901.995903726543, "speed/FLOPS": 202643574868284.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07820073515176773, "optim/lr": 0.0029284529417206552, "optim/total_tokens": 5468323840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8547749519348145, "created_at": "2025-01-16T11:01:16.491185+00:00"} {"global_step": 10431, "acc_step": 0, "speed/wps": 12907.12558504481, "speed/FLOPS": 202724143562308.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06869633495807648, "optim/lr": 0.0029284267935878626, "optim/total_tokens": 5468848128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.924468517303467, "created_at": "2025-01-16T11:01:26.651093+00:00"} {"global_step": 10432, "acc_step": 0, "speed/wps": 12903.286545247169, "speed/FLOPS": 202663846166893.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07509087026119232, "optim/lr": 0.0029284006407945765, "optim/total_tokens": 5469372416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342633, "loss/out": 2.9227089881896973, "created_at": "2025-01-16T11:01:36.812728+00:00"} {"global_step": 10433, "acc_step": 0, "speed/wps": 12908.400976078281, "speed/FLOPS": 202744175331060.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055590856820344925, "optim/lr": 0.002928374483340882, "optim/total_tokens": 5469896704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.8448002338409424, "created_at": "2025-01-16T11:01:46.972955+00:00"} {"global_step": 10434, "acc_step": 0, "speed/wps": 12901.094057923914, "speed/FLOPS": 202629410140688.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04975706338882446, "optim/lr": 0.0029283483212268657, "optim/total_tokens": 5470420992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9605698585510254, "created_at": "2025-01-16T11:01:57.138662+00:00"} {"global_step": 10435, "acc_step": 0, "speed/wps": 12900.683934257948, "speed/FLOPS": 202622968584945.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061890341341495514, "optim/lr": 0.0029283221544526114, "optim/total_tokens": 5470945280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305392, "loss/out": 2.9170305728912354, "created_at": "2025-01-16T11:02:07.303959+00:00"} {"global_step": 10436, "acc_step": 0, "speed/wps": 12908.681963068182, "speed/FLOPS": 202748588617853.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04393059015274048, "optim/lr": 0.0029282959830182046, "optim/total_tokens": 5471469568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.852734088897705, "created_at": "2025-01-16T11:02:17.466564+00:00"} {"global_step": 10437, "acc_step": 0, "speed/wps": 12901.7223963614, "speed/FLOPS": 202639279059276.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052273377776145935, "optim/lr": 0.0029282698069237317, "optim/total_tokens": 5471993856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358724, "loss/out": 2.909576892852783, "created_at": "2025-01-16T11:02:27.632596+00:00"} {"global_step": 10438, "acc_step": 0, "speed/wps": 12900.842379947926, "speed/FLOPS": 202625457192233.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.058165475726127625, "optim/lr": 0.002928243626169278, "optim/total_tokens": 5472518144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.824315309524536, "created_at": "2025-01-16T11:02:37.798721+00:00"} {"global_step": 10439, "acc_step": 0, "speed/wps": 12904.286885918578, "speed/FLOPS": 202679557891749.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04570699483156204, "optim/lr": 0.002928217440754928, "optim/total_tokens": 5473042432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.893087387084961, "created_at": "2025-01-16T11:02:47.960734+00:00"} {"global_step": 10440, "acc_step": 0, "speed/wps": 12903.733680732323, "speed/FLOPS": 202670869044116.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06359470635652542, "optim/lr": 0.002928191250680768, "optim/total_tokens": 5473566720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290558, "loss/out": 2.7050013542175293, "created_at": "2025-01-16T11:02:58.125850+00:00"} {"global_step": 10441, "acc_step": 0, "speed/wps": 12906.09754505803, "speed/FLOPS": 202707996781641.3, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049659762531518936, "optim/lr": 0.002928165055946882, "optim/total_tokens": 5474091008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.836970806121826, "created_at": "2025-01-16T11:03:08.286575+00:00"} {"global_step": 10442, "acc_step": 0, "speed/wps": 12909.714947938002, "speed/FLOPS": 202764813064703.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05527530238032341, "optim/lr": 0.002928138856553358, "optim/total_tokens": 5474615296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.89454984664917, "created_at": "2025-01-16T11:03:18.443108+00:00"} {"global_step": 10443, "acc_step": 0, "speed/wps": 12907.716370435859, "speed/FLOPS": 202733422658699.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045370373874902725, "optim/lr": 0.002928112652500279, "optim/total_tokens": 5475139584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.849903106689453, "created_at": "2025-01-16T11:03:28.601108+00:00"} {"global_step": 10444, "acc_step": 0, "speed/wps": 12902.220521123247, "speed/FLOPS": 202647102793156.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04745911806821823, "optim/lr": 0.002928086443787732, "optim/total_tokens": 5475663872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460071, "loss/out": 2.8921828269958496, "created_at": "2025-01-16T11:03:38.768545+00:00"} {"global_step": 10445, "acc_step": 0, "speed/wps": 12899.735081103612, "speed/FLOPS": 202608065542295.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05070101097226143, "optim/lr": 0.002928060230415802, "optim/total_tokens": 5476188160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.891663074493408, "created_at": "2025-01-16T11:03:48.936544+00:00"} {"global_step": 10446, "acc_step": 0, "speed/wps": 12905.915403201501, "speed/FLOPS": 202705135993495.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04585595801472664, "optim/lr": 0.0029280340123845746, "optim/total_tokens": 5476712448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9443538188934326, "created_at": "2025-01-16T11:03:59.096056+00:00"} {"global_step": 10447, "acc_step": 0, "speed/wps": 12905.268094949035, "speed/FLOPS": 202694969127895.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04749080911278725, "optim/lr": 0.002928007789694135, "optim/total_tokens": 5477236736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366026, "loss/out": 2.9465837478637695, "created_at": "2025-01-16T11:04:09.256643+00:00"} {"global_step": 10448, "acc_step": 0, "speed/wps": 12906.175006524056, "speed/FLOPS": 202709213420408.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057695869356393814, "optim/lr": 0.0029279815623445694, "optim/total_tokens": 5477761024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.747053623199463, "created_at": "2025-01-16T11:04:19.424232+00:00"} {"global_step": 10449, "acc_step": 0, "speed/wps": 12899.738858762817, "speed/FLOPS": 202608124875624.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05246307700872421, "optim/lr": 0.0029279553303359633, "optim/total_tokens": 5478285312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9783360958099365, "created_at": "2025-01-16T11:04:29.589098+00:00"} {"global_step": 10450, "acc_step": 0, "speed/wps": 12899.237845775597, "speed/FLOPS": 202600255778196.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06393495947122574, "optim/lr": 0.0029279290936684017, "optim/total_tokens": 5478809600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9451887607574463, "created_at": "2025-01-16T11:04:39.754018+00:00"} {"global_step": 10451, "acc_step": 0, "speed/wps": 12906.473609180259, "speed/FLOPS": 202713903385447.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0536813847720623, "optim/lr": 0.0029279028523419705, "optim/total_tokens": 5479333888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8159000873565674, "created_at": "2025-01-16T11:04:49.913363+00:00"} {"global_step": 10452, "acc_step": 0, "speed/wps": 12905.14841727811, "speed/FLOPS": 202693089425619.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04164754971861839, "optim/lr": 0.0029278766063567554, "optim/total_tokens": 5479858176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306165, "loss/out": 2.9796011447906494, "created_at": "2025-01-16T11:05:00.076207+00:00"} {"global_step": 10453, "acc_step": 0, "speed/wps": 12904.775077395214, "speed/FLOPS": 202687225609735.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05188615247607231, "optim/lr": 0.002927850355712842, "optim/total_tokens": 5480382464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9043118953704834, "created_at": "2025-01-16T11:05:10.237548+00:00"} {"global_step": 10454, "acc_step": 0, "speed/wps": 12904.258641809158, "speed/FLOPS": 202679114279199.9, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.066765196621418, "optim/lr": 0.002927824100410316, "optim/total_tokens": 5480906752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9996252059936523, "created_at": "2025-01-16T11:05:20.402136+00:00"} {"global_step": 10455, "acc_step": 0, "speed/wps": 12904.43218335213, "speed/FLOPS": 202681839987603.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060068320482969284, "optim/lr": 0.002927797840449263, "optim/total_tokens": 5481431040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 2.97244930267334, "created_at": "2025-01-16T11:05:30.562714+00:00"} {"global_step": 10456, "acc_step": 0, "speed/wps": 12905.12566962424, "speed/FLOPS": 202692732142456.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048758529126644135, "optim/lr": 0.002927771575829768, "optim/total_tokens": 5481955328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.875609874725342, "created_at": "2025-01-16T11:05:40.730905+00:00"} {"global_step": 10457, "acc_step": 0, "speed/wps": 12904.897625956326, "speed/FLOPS": 202689150403287.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05085546895861626, "optim/lr": 0.002927745306551918, "optim/total_tokens": 5482479616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309769, "loss/out": 2.7928268909454346, "created_at": "2025-01-16T11:05:50.891151+00:00"} {"global_step": 10458, "acc_step": 0, "speed/wps": 12902.833359284208, "speed/FLOPS": 202656728258602.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06640947610139847, "optim/lr": 0.002927719032615798, "optim/total_tokens": 5483003904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8968048095703125, "created_at": "2025-01-16T11:06:01.053698+00:00"} {"global_step": 10459, "acc_step": 0, "speed/wps": 12902.29395699962, "speed/FLOPS": 202648256204506.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04325837269425392, "optim/lr": 0.0029276927540214936, "optim/total_tokens": 5483528192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.7928357124328613, "created_at": "2025-01-16T11:06:11.216084+00:00"} {"global_step": 10460, "acc_step": 0, "speed/wps": 12903.089900104631, "speed/FLOPS": 202660757584710.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07138057798147202, "optim/lr": 0.0029276664707690906, "optim/total_tokens": 5484052480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 2.91591477394104, "created_at": "2025-01-16T11:06:21.377673+00:00"} {"global_step": 10461, "acc_step": 0, "speed/wps": 12906.86889674533, "speed/FLOPS": 202720111919838.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05912341922521591, "optim/lr": 0.0029276401828586753, "optim/total_tokens": 5484576768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.837254047393799, "created_at": "2025-01-16T11:06:31.539586+00:00"} {"global_step": 10462, "acc_step": 0, "speed/wps": 12902.837239307699, "speed/FLOPS": 202656789199702.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06771913170814514, "optim/lr": 0.002927613890290333, "optim/total_tokens": 5485101056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 2.8425045013427734, "created_at": "2025-01-16T11:06:41.701660+00:00"} {"global_step": 10463, "acc_step": 0, "speed/wps": 12905.85026967227, "speed/FLOPS": 202704112981915.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048960164189338684, "optim/lr": 0.002927587593064149, "optim/total_tokens": 5485625344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.7880711555480957, "created_at": "2025-01-16T11:06:51.866491+00:00"} {"global_step": 10464, "acc_step": 0, "speed/wps": 12907.328338294054, "speed/FLOPS": 202727328080700.22, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05571358650922775, "optim/lr": 0.00292756129118021, "optim/total_tokens": 5486149632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443300, "loss/out": 2.944802761077881, "created_at": "2025-01-16T11:07:02.026645+00:00"} {"global_step": 10465, "acc_step": 0, "speed/wps": 12905.338494615879, "speed/FLOPS": 202696074851402.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.096341073513031, "optim/lr": 0.0029275349846386023, "optim/total_tokens": 5486673920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.85396671295166, "created_at": "2025-01-16T11:07:12.189684+00:00"} {"global_step": 10466, "acc_step": 0, "speed/wps": 12907.261319000283, "speed/FLOPS": 202726275450597.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06754649430513382, "optim/lr": 0.00292750867343941, "optim/total_tokens": 5487198208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8960580825805664, "created_at": "2025-01-16T11:07:22.349805+00:00"} {"global_step": 10467, "acc_step": 0, "speed/wps": 12899.715861779385, "speed/FLOPS": 202607763676398.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05442225933074951, "optim/lr": 0.0029274823575827194, "optim/total_tokens": 5487722496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 2.9331600666046143, "created_at": "2025-01-16T11:07:32.514524+00:00"} {"global_step": 10468, "acc_step": 0, "speed/wps": 12893.112934727427, "speed/FLOPS": 202504055633674.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05073424056172371, "optim/lr": 0.0029274560370686174, "optim/total_tokens": 5488246784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8595705032348633, "created_at": "2025-01-16T11:07:42.684555+00:00"} {"global_step": 10469, "acc_step": 0, "speed/wps": 12904.85752954803, "speed/FLOPS": 202688520634096.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04223059117794037, "optim/lr": 0.002927429711897189, "optim/total_tokens": 5488771072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.840817451477051, "created_at": "2025-01-16T11:07:52.844973+00:00"} {"global_step": 10470, "acc_step": 0, "speed/wps": 12903.744077812287, "speed/FLOPS": 202671032344544.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052680183202028275, "optim/lr": 0.0029274033820685208, "optim/total_tokens": 5489295360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483427, "loss/out": 2.925729513168335, "created_at": "2025-01-16T11:08:03.009411+00:00"} {"global_step": 10471, "acc_step": 0, "speed/wps": 12906.471546163686, "speed/FLOPS": 202713870982937.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05319512262940407, "optim/lr": 0.0029273770475826974, "optim/total_tokens": 5489819648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503672, "loss/out": 2.836731433868408, "created_at": "2025-01-16T11:08:13.168968+00:00"} {"global_step": 10472, "acc_step": 0, "speed/wps": 12904.80364443829, "speed/FLOPS": 202687674294402.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05294063687324524, "optim/lr": 0.0029273507084398064, "optim/total_tokens": 5490343936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.8964405059814453, "created_at": "2025-01-16T11:08:23.329415+00:00"} {"global_step": 10473, "acc_step": 0, "speed/wps": 12902.148542119388, "speed/FLOPS": 202645972263991.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.058406539261341095, "optim/lr": 0.0029273243646399326, "optim/total_tokens": 5490868224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.916175127029419, "created_at": "2025-01-16T11:08:33.493255+00:00"} {"global_step": 10474, "acc_step": 0, "speed/wps": 12902.479914310823, "speed/FLOPS": 202651176919610.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05846533551812172, "optim/lr": 0.002927298016183162, "optim/total_tokens": 5491392512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.912871837615967, "created_at": "2025-01-16T11:08:43.656040+00:00"} {"global_step": 10475, "acc_step": 0, "speed/wps": 12896.978319295218, "speed/FLOPS": 202564766809907.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05252108722925186, "optim/lr": 0.0029272716630695814, "optim/total_tokens": 5491916800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 2.908679962158203, "created_at": "2025-01-16T11:08:53.822950+00:00"} {"global_step": 10476, "acc_step": 0, "speed/wps": 12902.751889976229, "speed/FLOPS": 202655448671169.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05323050916194916, "optim/lr": 0.002927245305299276, "optim/total_tokens": 5492441088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.8667054176330566, "created_at": "2025-01-16T11:09:03.996248+00:00"} {"global_step": 10477, "acc_step": 0, "speed/wps": 12900.255099363538, "speed/FLOPS": 202616233143647.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07695656269788742, "optim/lr": 0.002927218942872332, "optim/total_tokens": 5492965376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488719, "loss/out": 2.886321544647217, "created_at": "2025-01-16T11:09:14.161550+00:00"} {"global_step": 10478, "acc_step": 0, "speed/wps": 12906.865904418892, "speed/FLOPS": 202720064921240.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.041892848908901215, "optim/lr": 0.0029271925757888353, "optim/total_tokens": 5493489664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.964933395385742, "created_at": "2025-01-16T11:09:24.323046+00:00"} {"global_step": 10479, "acc_step": 0, "speed/wps": 12906.530092060624, "speed/FLOPS": 202714790526698.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04848801717162132, "optim/lr": 0.0029271662040488726, "optim/total_tokens": 5494013952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.013077735900879, "created_at": "2025-01-16T11:09:34.482267+00:00"} {"global_step": 10480, "acc_step": 0, "speed/wps": 12905.797739637364, "speed/FLOPS": 202703287925533.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04374169185757637, "optim/lr": 0.0029271398276525286, "optim/total_tokens": 5494538240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 2.8138556480407715, "created_at": "2025-01-16T11:09:44.643495+00:00"} {"global_step": 10481, "acc_step": 0, "speed/wps": 12909.499214755653, "speed/FLOPS": 202761424678626.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.045508094131946564, "optim/lr": 0.0029271134465998906, "optim/total_tokens": 5495062528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.8340485095977783, "created_at": "2025-01-16T11:09:54.804669+00:00"} {"global_step": 10482, "acc_step": 0, "speed/wps": 12902.941221257663, "speed/FLOPS": 202658422379113.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04849444329738617, "optim/lr": 0.0029270870608910436, "optim/total_tokens": 5495586816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8845231533050537, "created_at": "2025-01-16T11:10:04.966886+00:00"} {"global_step": 10483, "acc_step": 0, "speed/wps": 12896.649566051226, "speed/FLOPS": 202559603288455.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.046890269964933395, "optim/lr": 0.002927060670526075, "optim/total_tokens": 5496111104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.7703628540039062, "created_at": "2025-01-16T11:10:15.134806+00:00"} {"global_step": 10484, "acc_step": 0, "speed/wps": 12904.9301635731, "speed/FLOPS": 202689661451270.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07414794713258743, "optim/lr": 0.00292703427550507, "optim/total_tokens": 5496635392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.937525510787964, "created_at": "2025-01-16T11:10:25.296127+00:00"} {"global_step": 10485, "acc_step": 0, "speed/wps": 12904.41219868219, "speed/FLOPS": 202681526100900.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06823724508285522, "optim/lr": 0.0029270078758281146, "optim/total_tokens": 5497159680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 2.9679789543151855, "created_at": "2025-01-16T11:10:35.458466+00:00"} {"global_step": 10486, "acc_step": 0, "speed/wps": 12905.345457704178, "speed/FLOPS": 202696184216272.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05935584008693695, "optim/lr": 0.002926981471495296, "optim/total_tokens": 5497683968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8984737396240234, "created_at": "2025-01-16T11:10:45.621738+00:00"} {"global_step": 10487, "acc_step": 0, "speed/wps": 12907.685397988, "speed/FLOPS": 202732936193845.06, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07004199177026749, "optim/lr": 0.0029269550625066992, "optim/total_tokens": 5498208256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.9360411167144775, "created_at": "2025-01-16T11:10:55.780275+00:00"} {"global_step": 10488, "acc_step": 0, "speed/wps": 12903.654721206643, "speed/FLOPS": 202669628876263.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05379173159599304, "optim/lr": 0.0029269286488624105, "optim/total_tokens": 5498732544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 2.9355454444885254, "created_at": "2025-01-16T11:11:05.941490+00:00"} {"global_step": 10489, "acc_step": 0, "speed/wps": 12903.425914404737, "speed/FLOPS": 202666035151026.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05026380717754364, "optim/lr": 0.002926902230562517, "optim/total_tokens": 5499256832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9568605422973633, "created_at": "2025-01-16T11:11:16.103125+00:00"} {"global_step": 10490, "acc_step": 0, "speed/wps": 12908.417570409181, "speed/FLOPS": 202744435967830.06, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056131914258003235, "optim/lr": 0.002926875807607104, "optim/total_tokens": 5499781120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.9168009757995605, "created_at": "2025-01-16T11:11:26.276991+00:00"} {"global_step": 10491, "acc_step": 0, "speed/wps": 12901.926125992868, "speed/FLOPS": 202642478913092.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054524198174476624, "optim/lr": 0.0029268493799962576, "optim/total_tokens": 5500305408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.9587066173553467, "created_at": "2025-01-16T11:11:36.439716+00:00"} {"global_step": 10492, "acc_step": 0, "speed/wps": 12899.51834412432, "speed/FLOPS": 202604661390207.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048534125089645386, "optim/lr": 0.0029268229477300645, "optim/total_tokens": 5500829696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.976501226425171, "created_at": "2025-01-16T11:11:46.606313+00:00"} {"global_step": 10493, "acc_step": 0, "speed/wps": 12904.464767134412, "speed/FLOPS": 202682351760678.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04811158776283264, "optim/lr": 0.0029267965108086113, "optim/total_tokens": 5501353984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.9113147258758545, "created_at": "2025-01-16T11:11:56.768794+00:00"} {"global_step": 10494, "acc_step": 0, "speed/wps": 12903.995512891332, "speed/FLOPS": 202674981477968.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06013546884059906, "optim/lr": 0.0029267700692319834, "optim/total_tokens": 5501878272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.087167263031006, "created_at": "2025-01-16T11:12:06.933634+00:00"} {"global_step": 10495, "acc_step": 0, "speed/wps": 12905.017323825796, "speed/FLOPS": 202691030422809.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05844854563474655, "optim/lr": 0.0029267436230002676, "optim/total_tokens": 5502402560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.8262953758239746, "created_at": "2025-01-16T11:12:17.096619+00:00"} {"global_step": 10496, "acc_step": 0, "speed/wps": 12903.259870641019, "speed/FLOPS": 202663427205548.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050763536244630814, "optim/lr": 0.0029267171721135497, "optim/total_tokens": 5502926848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.824528217315674, "created_at": "2025-01-16T11:12:27.261496+00:00"} {"global_step": 10497, "acc_step": 0, "speed/wps": 12903.24401356978, "speed/FLOPS": 202663178148455.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07345211505889893, "optim/lr": 0.0029266907165719164, "optim/total_tokens": 5503451136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8901631832122803, "created_at": "2025-01-16T11:12:37.425223+00:00"} {"global_step": 10498, "acc_step": 0, "speed/wps": 12901.271317752005, "speed/FLOPS": 202632194249868.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05204279348254204, "optim/lr": 0.002926664256375454, "optim/total_tokens": 5503975424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316975, "loss/out": 3.0095720291137695, "created_at": "2025-01-16T11:12:47.590589+00:00"} {"global_step": 10499, "acc_step": 0, "speed/wps": 12905.371706626513, "speed/FLOPS": 202696596491667.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044787850230932236, "optim/lr": 0.002926637791524249, "optim/total_tokens": 5504499712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8375582695007324, "created_at": "2025-01-16T11:12:57.753412+00:00"} {"global_step": 10500, "acc_step": 0, "speed/wps": 12903.249612753494, "speed/FLOPS": 202663266091329.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059481486678123474, "optim/lr": 0.0029266113220183876, "optim/total_tokens": 5505024000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.880749225616455, "created_at": "2025-01-16T11:13:07.916511+00:00"} {"global_step": 10501, "acc_step": 0, "speed/wps": 12898.724203448028, "speed/FLOPS": 202592188319623.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0625036209821701, "optim/lr": 0.002926584847857956, "optim/total_tokens": 5505548288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 2.8136799335479736, "created_at": "2025-01-16T11:13:18.083901+00:00"} {"global_step": 10502, "acc_step": 0, "speed/wps": 12905.629278584109, "speed/FLOPS": 202700642013202.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0594579353928566, "optim/lr": 0.0029265583690430406, "optim/total_tokens": 5506072576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472483, "loss/out": 2.8529036045074463, "created_at": "2025-01-16T11:13:28.244635+00:00"} {"global_step": 10503, "acc_step": 0, "speed/wps": 12903.266208038567, "speed/FLOPS": 202663526743085.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047115325927734375, "optim/lr": 0.0029265318855737276, "optim/total_tokens": 5506596864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284007, "loss/out": 2.93221378326416, "created_at": "2025-01-16T11:13:38.410866+00:00"} {"global_step": 10504, "acc_step": 0, "speed/wps": 12884.55518187391, "speed/FLOPS": 202369644365525.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05403011292219162, "optim/lr": 0.002926505397450104, "optim/total_tokens": 5507121152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.805072784423828, "created_at": "2025-01-16T11:13:48.596442+00:00"} {"global_step": 10505, "acc_step": 0, "speed/wps": 12900.52551729264, "speed/FLOPS": 202620480428817.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05364236235618591, "optim/lr": 0.0029264789046722557, "optim/total_tokens": 5507645440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.975618839263916, "created_at": "2025-01-16T11:13:58.761424+00:00"} {"global_step": 10506, "acc_step": 0, "speed/wps": 12892.806246465441, "speed/FLOPS": 202499238673086.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.05341970920562744, "optim/lr": 0.00292645240724027, "optim/total_tokens": 5508169728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 2.7586545944213867, "created_at": "2025-01-16T11:14:08.931369+00:00"} {"global_step": 10507, "acc_step": 0, "speed/wps": 12896.910183045356, "speed/FLOPS": 202563696636473.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04409594461321831, "optim/lr": 0.002926425905154232, "optim/total_tokens": 5508694016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9391655921936035, "created_at": "2025-01-16T11:14:19.100197+00:00"} {"global_step": 10508, "acc_step": 0, "speed/wps": 12897.709441791534, "speed/FLOPS": 202576250093379.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050992079079151154, "optim/lr": 0.002926399398414229, "optim/total_tokens": 5509218304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.9180831909179688, "created_at": "2025-01-16T11:14:29.267172+00:00"} {"global_step": 10509, "acc_step": 0, "speed/wps": 12898.116832540107, "speed/FLOPS": 202582648724899.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051525913178920746, "optim/lr": 0.002926372887020347, "optim/total_tokens": 5509742592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382419, "loss/out": 2.931461811065674, "created_at": "2025-01-16T11:14:39.433209+00:00"} {"global_step": 10510, "acc_step": 0, "speed/wps": 12901.44012156595, "speed/FLOPS": 202634845545727.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04499334469437599, "optim/lr": 0.0029263463709726734, "optim/total_tokens": 5510266880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.036785840988159, "created_at": "2025-01-16T11:14:49.597729+00:00"} {"global_step": 10511, "acc_step": 0, "speed/wps": 12902.862756434952, "speed/FLOPS": 202657189981250.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048441555351018906, "optim/lr": 0.0029263198502712936, "optim/total_tokens": 5510791168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.842071294784546, "created_at": "2025-01-16T11:14:59.761691+00:00"} {"global_step": 10512, "acc_step": 0, "speed/wps": 12898.326591725812, "speed/FLOPS": 202585943281149.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04081291705369949, "optim/lr": 0.002926293324916295, "optim/total_tokens": 5511315456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 2.9238080978393555, "created_at": "2025-01-16T11:15:09.928097+00:00"} {"global_step": 10513, "acc_step": 0, "speed/wps": 12900.64032589442, "speed/FLOPS": 202622283655671.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04844038188457489, "optim/lr": 0.002926266794907764, "optim/total_tokens": 5511839744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.986611843109131, "created_at": "2025-01-16T11:15:20.098392+00:00"} {"global_step": 10514, "acc_step": 0, "speed/wps": 12905.910717731345, "speed/FLOPS": 202705062401748.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05109703540802002, "optim/lr": 0.002926240260245787, "optim/total_tokens": 5512364032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.9897985458374023, "created_at": "2025-01-16T11:15:30.262892+00:00"} {"global_step": 10515, "acc_step": 0, "speed/wps": 12905.247636706106, "speed/FLOPS": 202694647803077.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05429042875766754, "optim/lr": 0.00292621372093045, "optim/total_tokens": 5512888320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352289, "loss/out": 2.7455856800079346, "created_at": "2025-01-16T11:15:40.434738+00:00"} {"global_step": 10516, "acc_step": 0, "speed/wps": 12899.382556609746, "speed/FLOPS": 202602528660699.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06272994726896286, "optim/lr": 0.0029261871769618404, "optim/total_tokens": 5513412608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9192044734954834, "created_at": "2025-01-16T11:15:50.602445+00:00"} {"global_step": 10517, "acc_step": 0, "speed/wps": 12901.634326160533, "speed/FLOPS": 202637895795750.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05039740353822708, "optim/lr": 0.002926160628340045, "optim/total_tokens": 5513936896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412896, "loss/out": 2.846282720565796, "created_at": "2025-01-16T11:16:00.767039+00:00"} {"global_step": 10518, "acc_step": 0, "speed/wps": 12904.613683432355, "speed/FLOPS": 202684690695770.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04893915727734566, "optim/lr": 0.002926134075065149, "optim/total_tokens": 5514461184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.773597240447998, "created_at": "2025-01-16T11:16:10.932981+00:00"} {"global_step": 10519, "acc_step": 0, "speed/wps": 12901.874136483348, "speed/FLOPS": 202641662346404.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04621445760130882, "optim/lr": 0.002926107517137241, "optim/total_tokens": 5514985472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 2.9672834873199463, "created_at": "2025-01-16T11:16:21.097159+00:00"} {"global_step": 10520, "acc_step": 0, "speed/wps": 12900.28472480776, "speed/FLOPS": 202616698451958.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.06551328301429749, "optim/lr": 0.002926080954556406, "optim/total_tokens": 5515509760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8089888095855713, "created_at": "2025-01-16T11:16:31.265208+00:00"} {"global_step": 10521, "acc_step": 0, "speed/wps": 12902.728745030523, "speed/FLOPS": 202655085147992.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0754375234246254, "optim/lr": 0.002926054387322731, "optim/total_tokens": 5516034048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9550933837890625, "created_at": "2025-01-16T11:16:41.427365+00:00"} {"global_step": 10522, "acc_step": 0, "speed/wps": 12901.347869779182, "speed/FLOPS": 202633396604649.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04738997295498848, "optim/lr": 0.0029260278154363043, "optim/total_tokens": 5516558336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455507, "loss/out": 2.942347526550293, "created_at": "2025-01-16T11:16:51.593211+00:00"} {"global_step": 10523, "acc_step": 0, "speed/wps": 12901.360299561116, "speed/FLOPS": 202633591831454.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059030964970588684, "optim/lr": 0.00292600123889721, "optim/total_tokens": 5517082624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 3.1013333797454834, "created_at": "2025-01-16T11:17:01.758489+00:00"} {"global_step": 10524, "acc_step": 0, "speed/wps": 12904.542007850003, "speed/FLOPS": 202683564932257.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06477343291044235, "optim/lr": 0.0029259746577055368, "optim/total_tokens": 5517606912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.9587340354919434, "created_at": "2025-01-16T11:17:11.919641+00:00"} {"global_step": 10525, "acc_step": 0, "speed/wps": 12902.570769922426, "speed/FLOPS": 202652603931839.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05467216670513153, "optim/lr": 0.00292594807186137, "optim/total_tokens": 5518131200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9510750770568848, "created_at": "2025-01-16T11:17:22.081737+00:00"} {"global_step": 10526, "acc_step": 0, "speed/wps": 12902.735143417973, "speed/FLOPS": 202655185643460.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060630299150943756, "optim/lr": 0.002925921481364798, "optim/total_tokens": 5518655488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.898808240890503, "created_at": "2025-01-16T11:17:32.253690+00:00"} {"global_step": 10527, "acc_step": 0, "speed/wps": 12903.893945275948, "speed/FLOPS": 202673386219000.3, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057009417563676834, "optim/lr": 0.002925894886215906, "optim/total_tokens": 5519179776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.913520097732544, "created_at": "2025-01-16T11:17:42.414782+00:00"} {"global_step": 10528, "acc_step": 0, "speed/wps": 12905.316126159747, "speed/FLOPS": 202695723524061.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06432700157165527, "optim/lr": 0.0029258682864147815, "optim/total_tokens": 5519704064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.897508144378662, "created_at": "2025-01-16T11:17:52.574828+00:00"} {"global_step": 10529, "acc_step": 0, "speed/wps": 12910.267062570047, "speed/FLOPS": 202773484783685.0, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.055932264775037766, "optim/lr": 0.0029258416819615113, "optim/total_tokens": 5520228352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.810275077819824, "created_at": "2025-01-16T11:18:02.730949+00:00"} {"global_step": 10530, "acc_step": 0, "speed/wps": 12909.126006745328, "speed/FLOPS": 202755562933982.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04731617495417595, "optim/lr": 0.002925815072856182, "optim/total_tokens": 5520752640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 3.007593870162964, "created_at": "2025-01-16T11:18:12.892771+00:00"} {"global_step": 10531, "acc_step": 0, "speed/wps": 12910.95554301541, "speed/FLOPS": 202784298315150.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06625102460384369, "optim/lr": 0.002925788459098881, "optim/total_tokens": 5521276928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.8551502227783203, "created_at": "2025-01-16T11:18:23.051443+00:00"} {"global_step": 10532, "acc_step": 0, "speed/wps": 12905.471606402749, "speed/FLOPS": 202698165554930.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04588713496923447, "optim/lr": 0.002925761840689694, "optim/total_tokens": 5521801216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.0110700130462646, "created_at": "2025-01-16T11:18:33.211293+00:00"} {"global_step": 10533, "acc_step": 0, "speed/wps": 12901.91795780713, "speed/FLOPS": 202642350620511.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08133434504270554, "optim/lr": 0.0029257352176287085, "optim/total_tokens": 5522325504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.948164939880371, "created_at": "2025-01-16T11:18:43.376823+00:00"} {"global_step": 10534, "acc_step": 0, "speed/wps": 12904.135699555474, "speed/FLOPS": 202677183302165.47, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07373765856027603, "optim/lr": 0.0029257085899160114, "optim/total_tokens": 5522849792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8600049018859863, "created_at": "2025-01-16T11:18:53.538013+00:00"} {"global_step": 10535, "acc_step": 0, "speed/wps": 12903.358273133425, "speed/FLOPS": 202664972751911.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05916372314095497, "optim/lr": 0.0029256819575516897, "optim/total_tokens": 5523374080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.089473247528076, "created_at": "2025-01-16T11:19:03.700877+00:00"} {"global_step": 10536, "acc_step": 0, "speed/wps": 12905.428601156493, "speed/FLOPS": 202697490098442.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07756857573986053, "optim/lr": 0.0029256553205358297, "optim/total_tokens": 5523898368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9244675636291504, "created_at": "2025-01-16T11:19:13.862959+00:00"} {"global_step": 10537, "acc_step": 0, "speed/wps": 12907.740771568528, "speed/FLOPS": 202733805912018.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05017600581049919, "optim/lr": 0.002925628678868519, "optim/total_tokens": 5524422656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446744, "loss/out": 2.8370351791381836, "created_at": "2025-01-16T11:19:24.023642+00:00"} {"global_step": 10538, "acc_step": 0, "speed/wps": 12904.936896656813, "speed/FLOPS": 202689767203602.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06922783702611923, "optim/lr": 0.0029256020325498446, "optim/total_tokens": 5524946944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.87347149848938, "created_at": "2025-01-16T11:19:34.184381+00:00"} {"global_step": 10539, "acc_step": 0, "speed/wps": 12903.622101290299, "speed/FLOPS": 202669116535652.9, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.041375067085027695, "optim/lr": 0.002925575381579893, "optim/total_tokens": 5525471232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.881351947784424, "created_at": "2025-01-16T11:19:44.347057+00:00"} {"global_step": 10540, "acc_step": 0, "speed/wps": 12909.351843461805, "speed/FLOPS": 202759110009947.25, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05887915939092636, "optim/lr": 0.0029255487259587506, "optim/total_tokens": 5525995520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328312, "loss/out": 3.0266942977905273, "created_at": "2025-01-16T11:19:54.506446+00:00"} {"global_step": 10541, "acc_step": 0, "speed/wps": 12905.321715308035, "speed/FLOPS": 202695811309315.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04487768933176994, "optim/lr": 0.0029255220656865056, "optim/total_tokens": 5526519808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.872069835662842, "created_at": "2025-01-16T11:20:04.667498+00:00"} {"global_step": 10542, "acc_step": 0, "speed/wps": 12903.355765110555, "speed/FLOPS": 202664933359965.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.061306148767471313, "optim/lr": 0.002925495400763244, "optim/total_tokens": 5527044096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8855342864990234, "created_at": "2025-01-16T11:20:14.829695+00:00"} {"global_step": 10543, "acc_step": 0, "speed/wps": 12897.020769493085, "speed/FLOPS": 202565433548597.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045029062777757645, "optim/lr": 0.0029254687311890534, "optim/total_tokens": 5527568384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8511252403259277, "created_at": "2025-01-16T11:20:24.999030+00:00"} {"global_step": 10544, "acc_step": 0, "speed/wps": 12907.32692488718, "speed/FLOPS": 202727305881203.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0733230784535408, "optim/lr": 0.002925442056964021, "optim/total_tokens": 5528092672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8719115257263184, "created_at": "2025-01-16T11:20:35.162890+00:00"} {"global_step": 10545, "acc_step": 0, "speed/wps": 12904.595430487896, "speed/FLOPS": 202684404008195.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06782887876033783, "optim/lr": 0.0029254153780882324, "optim/total_tokens": 5528616960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 2.95297908782959, "created_at": "2025-01-16T11:20:45.325540+00:00"} {"global_step": 10546, "acc_step": 0, "speed/wps": 12901.66670657995, "speed/FLOPS": 202638404374732.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.046465273946523666, "optim/lr": 0.0029253886945617763, "optim/total_tokens": 5529141248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.903172492980957, "created_at": "2025-01-16T11:20:55.491613+00:00"} {"global_step": 10547, "acc_step": 0, "speed/wps": 12909.38152491178, "speed/FLOPS": 202759576197905.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05830258131027222, "optim/lr": 0.00292536200638474, "optim/total_tokens": 5529665536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.977921485900879, "created_at": "2025-01-16T11:21:05.648405+00:00"} {"global_step": 10548, "acc_step": 0, "speed/wps": 12902.297443997353, "speed/FLOPS": 202648310972597.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048782575875520706, "optim/lr": 0.0029253353135572082, "optim/total_tokens": 5530189824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.889742851257324, "created_at": "2025-01-16T11:21:15.811346+00:00"} {"global_step": 10549, "acc_step": 0, "speed/wps": 12907.262440007546, "speed/FLOPS": 202726293057556.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05381008982658386, "optim/lr": 0.0029253086160792702, "optim/total_tokens": 5530714112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9294862747192383, "created_at": "2025-01-16T11:21:25.970106+00:00"} {"global_step": 10550, "acc_step": 0, "speed/wps": 12901.046590308219, "speed/FLOPS": 202628664596556.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04912404343485832, "optim/lr": 0.002925281913951012, "optim/total_tokens": 5531238400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295655, "loss/out": 2.8171324729919434, "created_at": "2025-01-16T11:21:36.133808+00:00"} {"global_step": 10551, "acc_step": 0, "speed/wps": 12902.480438321734, "speed/FLOPS": 202651185149922.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04837241768836975, "optim/lr": 0.0029252552071725216, "optim/total_tokens": 5531762688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9534997940063477, "created_at": "2025-01-16T11:21:46.298919+00:00"} {"global_step": 10552, "acc_step": 0, "speed/wps": 12906.866184949437, "speed/FLOPS": 202720069327357.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044651445001363754, "optim/lr": 0.0029252284957438854, "optim/total_tokens": 5532286976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.9235167503356934, "created_at": "2025-01-16T11:21:56.457678+00:00"} {"global_step": 10553, "acc_step": 0, "speed/wps": 12909.174717930506, "speed/FLOPS": 202756328010082.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06380835175514221, "optim/lr": 0.002925201779665191, "optim/total_tokens": 5532811264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.7760281562805176, "created_at": "2025-01-16T11:22:06.614621+00:00"} {"global_step": 10554, "acc_step": 0, "speed/wps": 12904.600484179458, "speed/FLOPS": 202684483383366.12, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053395677357912064, "optim/lr": 0.0029251750589365253, "optim/total_tokens": 5533335552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9477317333221436, "created_at": "2025-01-16T11:22:16.779323+00:00"} {"global_step": 10555, "acc_step": 0, "speed/wps": 12903.834314711514, "speed/FLOPS": 202672449639044.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0461692251265049, "optim/lr": 0.0029251483335579753, "optim/total_tokens": 5533859840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387667, "loss/out": 2.8325233459472656, "created_at": "2025-01-16T11:22:26.941811+00:00"} {"global_step": 10556, "acc_step": 0, "speed/wps": 12904.292065970998, "speed/FLOPS": 202679639251591.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048001792281866074, "optim/lr": 0.0029251216035296288, "optim/total_tokens": 5534384128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485284, "loss/out": 2.889538526535034, "created_at": "2025-01-16T11:22:37.105131+00:00"} {"global_step": 10557, "acc_step": 0, "speed/wps": 12901.509475370185, "speed/FLOPS": 202635934842525.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04539844021201134, "optim/lr": 0.0029250948688515728, "optim/total_tokens": 5534908416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.86411714553833, "created_at": "2025-01-16T11:22:47.271373+00:00"} {"global_step": 10558, "acc_step": 0, "speed/wps": 12895.920495731609, "speed/FLOPS": 202548152237238.6, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05881905555725098, "optim/lr": 0.002925068129523894, "optim/total_tokens": 5535432704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.946103572845459, "created_at": "2025-01-16T11:22:57.438720+00:00"} {"global_step": 10559, "acc_step": 0, "speed/wps": 12901.99834616788, "speed/FLOPS": 202643613230181.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05663513392210007, "optim/lr": 0.00292504138554668, "optim/total_tokens": 5535956992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8909995555877686, "created_at": "2025-01-16T11:23:07.604011+00:00"} {"global_step": 10560, "acc_step": 0, "speed/wps": 12907.212445877187, "speed/FLOPS": 202725507831040.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.044220227748155594, "optim/lr": 0.0029250146369200185, "optim/total_tokens": 5536481280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9436278343200684, "created_at": "2025-01-16T11:23:17.762453+00:00"} {"global_step": 10561, "acc_step": 0, "speed/wps": 12904.653145277309, "speed/FLOPS": 202685310498271.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0749635323882103, "optim/lr": 0.002924987883643996, "optim/total_tokens": 5537005568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.944741725921631, "created_at": "2025-01-16T11:23:27.923783+00:00"} {"global_step": 10562, "acc_step": 0, "speed/wps": 12900.334755153603, "speed/FLOPS": 202617484247289.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05947526916861534, "optim/lr": 0.002924961125718701, "optim/total_tokens": 5537529856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.755514621734619, "created_at": "2025-01-16T11:23:38.089066+00:00"} {"global_step": 10563, "acc_step": 0, "speed/wps": 12904.759739612406, "speed/FLOPS": 202686984708780.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05510486662387848, "optim/lr": 0.002924934363144219, "optim/total_tokens": 5538054144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.863568067550659, "created_at": "2025-01-16T11:23:48.251517+00:00"} {"global_step": 10564, "acc_step": 0, "speed/wps": 12905.078872877002, "speed/FLOPS": 202691997135236.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04309101030230522, "optim/lr": 0.0029249075959206384, "optim/total_tokens": 5538578432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.797250747680664, "created_at": "2025-01-16T11:23:58.412239+00:00"} {"global_step": 10565, "acc_step": 0, "speed/wps": 12900.378895933873, "speed/FLOPS": 202618177538899.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04095378890633583, "optim/lr": 0.0029248808240480467, "optim/total_tokens": 5539102720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 2.8848679065704346, "created_at": "2025-01-16T11:24:08.576081+00:00"} {"global_step": 10566, "acc_step": 0, "speed/wps": 12904.37755045174, "speed/FLOPS": 202680981902829.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04262878745794296, "optim/lr": 0.002924854047526531, "optim/total_tokens": 5539627008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.924508571624756, "created_at": "2025-01-16T11:24:18.737157+00:00"} {"global_step": 10567, "acc_step": 0, "speed/wps": 12904.690759088993, "speed/FLOPS": 202685901274871.1, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05660899356007576, "optim/lr": 0.002924827266356178, "optim/total_tokens": 5540151296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.7707037925720215, "created_at": "2025-01-16T11:24:28.897626+00:00"} {"global_step": 10568, "acc_step": 0, "speed/wps": 12908.93151260792, "speed/FLOPS": 202752508136291.84, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04761552810668945, "optim/lr": 0.002924800480537076, "optim/total_tokens": 5540675584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.7896499633789062, "created_at": "2025-01-16T11:24:39.056419+00:00"} {"global_step": 10569, "acc_step": 0, "speed/wps": 12903.196722960965, "speed/FLOPS": 202662435384459.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0501297190785408, "optim/lr": 0.0029247736900693123, "optim/total_tokens": 5541199872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 2.8193068504333496, "created_at": "2025-01-16T11:24:49.218466+00:00"} {"global_step": 10570, "acc_step": 0, "speed/wps": 12904.543305872041, "speed/FLOPS": 202683585319476.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0449250265955925, "optim/lr": 0.002924746894952974, "optim/total_tokens": 5541724160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 2.9190006256103516, "created_at": "2025-01-16T11:24:59.379700+00:00"} {"global_step": 10571, "acc_step": 0, "speed/wps": 12902.829069968273, "speed/FLOPS": 202656660889001.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053557995706796646, "optim/lr": 0.0029247200951881486, "optim/total_tokens": 5542248448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.934627056121826, "created_at": "2025-01-16T11:25:09.541727+00:00"} {"global_step": 10572, "acc_step": 0, "speed/wps": 12905.44632747245, "speed/FLOPS": 202697768514593.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10544979572296143, "optim/lr": 0.0029246932907749236, "optim/total_tokens": 5542772736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.955169677734375, "created_at": "2025-01-16T11:25:19.702283+00:00"} {"global_step": 10573, "acc_step": 0, "speed/wps": 12900.206061712639, "speed/FLOPS": 202615462939955.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11756670475006104, "optim/lr": 0.0029246664817133864, "optim/total_tokens": 5543297024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.91595196723938, "created_at": "2025-01-16T11:25:29.875061+00:00"} {"global_step": 10574, "acc_step": 0, "speed/wps": 12905.048485821797, "speed/FLOPS": 202691519864777.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06411820650100708, "optim/lr": 0.0029246396680036243, "optim/total_tokens": 5543821312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8324718475341797, "created_at": "2025-01-16T11:25:40.038548+00:00"} {"global_step": 10575, "acc_step": 0, "speed/wps": 12904.416259505317, "speed/FLOPS": 202681589881707.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.13552325963974, "optim/lr": 0.0029246128496457253, "optim/total_tokens": 5544345600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423603, "loss/out": 2.9182605743408203, "created_at": "2025-01-16T11:25:50.201922+00:00"} {"global_step": 10576, "acc_step": 0, "speed/wps": 12899.546199781997, "speed/FLOPS": 202605098901589.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4562303125858307, "optim/lr": 0.002924586026639776, "optim/total_tokens": 5544869888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.838099718093872, "created_at": "2025-01-16T11:26:00.366483+00:00"} {"global_step": 10577, "acc_step": 0, "speed/wps": 12910.718363035508, "speed/FLOPS": 202780573077646.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.7843071222305298, "optim/lr": 0.0029245591989858653, "optim/total_tokens": 5545394176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435843, "loss/out": 2.975466251373291, "created_at": "2025-01-16T11:26:10.547354+00:00"} {"global_step": 10578, "acc_step": 0, "speed/wps": 12903.37587427607, "speed/FLOPS": 202665249202043.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13835695385932922, "optim/lr": 0.0029245323666840795, "optim/total_tokens": 5545918464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 2.871547222137451, "created_at": "2025-01-16T11:26:20.711471+00:00"} {"global_step": 10579, "acc_step": 0, "speed/wps": 12904.511877719911, "speed/FLOPS": 202683091697160.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17543131113052368, "optim/lr": 0.0029245055297345062, "optim/total_tokens": 5546442752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.902193546295166, "created_at": "2025-01-16T11:26:30.872400+00:00"} {"global_step": 10580, "acc_step": 0, "speed/wps": 12902.417779693369, "speed/FLOPS": 202650201010061.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17926344275474548, "optim/lr": 0.0029244786881372337, "optim/total_tokens": 5546967040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8962035179138184, "created_at": "2025-01-16T11:26:41.045153+00:00"} {"global_step": 10581, "acc_step": 0, "speed/wps": 12902.256395009217, "speed/FLOPS": 202647666241831.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16179531812667847, "optim/lr": 0.0029244518418923495, "optim/total_tokens": 5547491328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 2.978771686553955, "created_at": "2025-01-16T11:26:51.207856+00:00"} {"global_step": 10582, "acc_step": 0, "speed/wps": 12904.18213332515, "speed/FLOPS": 202677912608324.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09840893745422363, "optim/lr": 0.00292442499099994, "optim/total_tokens": 5548015616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 3.0653929710388184, "created_at": "2025-01-16T11:27:01.371332+00:00"} {"global_step": 10583, "acc_step": 0, "speed/wps": 12906.463333162767, "speed/FLOPS": 202713741986471.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.10140015929937363, "optim/lr": 0.0029243981354600945, "optim/total_tokens": 5548539904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369627, "loss/out": 2.938770055770874, "created_at": "2025-01-16T11:27:11.530518+00:00"} {"global_step": 10584, "acc_step": 0, "speed/wps": 12906.032689001118, "speed/FLOPS": 202706978128145.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07718250900506973, "optim/lr": 0.0029243712752728994, "optim/total_tokens": 5549064192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.882347583770752, "created_at": "2025-01-16T11:27:21.694211+00:00"} {"global_step": 10585, "acc_step": 0, "speed/wps": 12904.042963165723, "speed/FLOPS": 202675726749731.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09056289494037628, "optim/lr": 0.002924344410438443, "optim/total_tokens": 5549588480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9921655654907227, "created_at": "2025-01-16T11:27:31.860864+00:00"} {"global_step": 10586, "acc_step": 0, "speed/wps": 12904.050090498024, "speed/FLOPS": 202675838694279.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06719470024108887, "optim/lr": 0.0029243175409568124, "optim/total_tokens": 5550112768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 2.7948527336120605, "created_at": "2025-01-16T11:27:42.022686+00:00"} {"global_step": 10587, "acc_step": 0, "speed/wps": 12907.43006415397, "speed/FLOPS": 202728925825116.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09816927462816238, "optim/lr": 0.0029242906668280957, "optim/total_tokens": 5550637056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.837007999420166, "created_at": "2025-01-16T11:27:52.182049+00:00"} {"global_step": 10588, "acc_step": 0, "speed/wps": 12904.374508402076, "speed/FLOPS": 202680934123259.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05677520111203194, "optim/lr": 0.0029242637880523806, "optim/total_tokens": 5551161344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8338074684143066, "created_at": "2025-01-16T11:28:02.349675+00:00"} {"global_step": 10589, "acc_step": 0, "speed/wps": 12906.260224797503, "speed/FLOPS": 202710551890495.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0743304118514061, "optim/lr": 0.0029242369046297545, "optim/total_tokens": 5551685632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.879613161087036, "created_at": "2025-01-16T11:28:12.508872+00:00"} {"global_step": 10590, "acc_step": 0, "speed/wps": 12900.184092975418, "speed/FLOPS": 202615117890749.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056361038237810135, "optim/lr": 0.002924210016560305, "optim/total_tokens": 5552209920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.0126633644104004, "created_at": "2025-01-16T11:28:22.675464+00:00"} {"global_step": 10591, "acc_step": 0, "speed/wps": 12907.90235741125, "speed/FLOPS": 202736343839720.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06342979520559311, "optim/lr": 0.002924183123844121, "optim/total_tokens": 5552734208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305632, "loss/out": 2.994540214538574, "created_at": "2025-01-16T11:28:32.835686+00:00"} {"global_step": 10592, "acc_step": 0, "speed/wps": 12903.285808227785, "speed/FLOPS": 202663834590991.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04987873509526253, "optim/lr": 0.0029241562264812884, "optim/total_tokens": 5553258496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8693125247955322, "created_at": "2025-01-16T11:28:43.000711+00:00"} {"global_step": 10593, "acc_step": 0, "speed/wps": 12906.644357749235, "speed/FLOPS": 202716585226356.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06183459609746933, "optim/lr": 0.002924129324471896, "optim/total_tokens": 5553782784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.897423267364502, "created_at": "2025-01-16T11:28:53.159583+00:00"} {"global_step": 10594, "acc_step": 0, "speed/wps": 12906.923549559206, "speed/FLOPS": 202720970317381.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05153285339474678, "optim/lr": 0.002924102417816032, "optim/total_tokens": 5554307072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 2.951504945755005, "created_at": "2025-01-16T11:29:03.318861+00:00"} {"global_step": 10595, "acc_step": 0, "speed/wps": 12908.03312945685, "speed/FLOPS": 202738397794396.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05998995527625084, "optim/lr": 0.002924075506513783, "optim/total_tokens": 5554831360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8500967025756836, "created_at": "2025-01-16T11:29:13.484838+00:00"} {"global_step": 10596, "acc_step": 0, "speed/wps": 12903.982008468603, "speed/FLOPS": 202674769372452.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04676147550344467, "optim/lr": 0.002924048590565238, "optim/total_tokens": 5555355648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 3.0220141410827637, "created_at": "2025-01-16T11:29:23.648838+00:00"} {"global_step": 10597, "acc_step": 0, "speed/wps": 12900.006938696773, "speed/FLOPS": 202612335439367.6, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052707232534885406, "optim/lr": 0.0029240216699704838, "optim/total_tokens": 5555879936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.87986421585083, "created_at": "2025-01-16T11:29:33.815617+00:00"} {"global_step": 10598, "acc_step": 0, "speed/wps": 12911.446156932547, "speed/FLOPS": 202792004080893.53, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050605691969394684, "optim/lr": 0.0029239947447296088, "optim/total_tokens": 5556404224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.8631229400634766, "created_at": "2025-01-16T11:29:43.972439+00:00"} {"global_step": 10599, "acc_step": 0, "speed/wps": 12902.844087318264, "speed/FLOPS": 202656896757118.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05552881211042404, "optim/lr": 0.002923967814842701, "optim/total_tokens": 5556928512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.930992364883423, "created_at": "2025-01-16T11:29:54.134400+00:00"} {"global_step": 10600, "acc_step": 0, "speed/wps": 12901.913594585598, "speed/FLOPS": 202642282090121.34, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05029546841979027, "optim/lr": 0.002923940880309848, "optim/total_tokens": 5557452800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.844938278198242, "created_at": "2025-01-16T11:30:04.300441+00:00"} {"global_step": 10601, "acc_step": 0, "speed/wps": 12907.86315744968, "speed/FLOPS": 202735728150457.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05301734432578087, "optim/lr": 0.0029239139411311376, "optim/total_tokens": 5557977088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 2.931295871734619, "created_at": "2025-01-16T11:30:14.459624+00:00"} {"global_step": 10602, "acc_step": 0, "speed/wps": 12903.457616466836, "speed/FLOPS": 202666533075474.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047469474375247955, "optim/lr": 0.0029238869973066575, "optim/total_tokens": 5558501376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8964192867279053, "created_at": "2025-01-16T11:30:24.621419+00:00"} {"global_step": 10603, "acc_step": 0, "speed/wps": 12905.818579521863, "speed/FLOPS": 202703615244557.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04118512570858002, "optim/lr": 0.002923860048836496, "optim/total_tokens": 5559025664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 2.844209671020508, "created_at": "2025-01-16T11:30:34.780962+00:00"} {"global_step": 10604, "acc_step": 0, "speed/wps": 12898.351377019657, "speed/FLOPS": 202586332568248.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0494491383433342, "optim/lr": 0.0029238330957207406, "optim/total_tokens": 5559549952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.931210994720459, "created_at": "2025-01-16T11:30:44.950426+00:00"} {"global_step": 10605, "acc_step": 0, "speed/wps": 12905.072227212608, "speed/FLOPS": 202691892755944.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.038622502237558365, "optim/lr": 0.00292380613795948, "optim/total_tokens": 5560074240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.91070556640625, "created_at": "2025-01-16T11:30:55.110543+00:00"} {"global_step": 10606, "acc_step": 0, "speed/wps": 12903.82410915081, "speed/FLOPS": 202672289346690.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043536074459552765, "optim/lr": 0.002923779175552802, "optim/total_tokens": 5560598528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 2.971930742263794, "created_at": "2025-01-16T11:31:05.274499+00:00"} {"global_step": 10607, "acc_step": 0, "speed/wps": 12901.142605624203, "speed/FLOPS": 202630172649032.78, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040390074253082275, "optim/lr": 0.002923752208500794, "optim/total_tokens": 5561122816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.6987106800079346, "created_at": "2025-01-16T11:31:15.439898+00:00"} {"global_step": 10608, "acc_step": 0, "speed/wps": 12906.407564570896, "speed/FLOPS": 202712866064102.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044479142874479294, "optim/lr": 0.0029237252368035445, "optim/total_tokens": 5561647104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.7838187217712402, "created_at": "2025-01-16T11:31:25.601955+00:00"} {"global_step": 10609, "acc_step": 0, "speed/wps": 12903.755266404653, "speed/FLOPS": 202671208076761.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05069620907306671, "optim/lr": 0.0029236982604611407, "optim/total_tokens": 5562171392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.849325656890869, "created_at": "2025-01-16T11:31:35.766409+00:00"} {"global_step": 10610, "acc_step": 0, "speed/wps": 12905.053871203638, "speed/FLOPS": 202691604449599.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048867788165807724, "optim/lr": 0.0029236712794736717, "optim/total_tokens": 5562695680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.876801013946533, "created_at": "2025-01-16T11:31:45.929974+00:00"} {"global_step": 10611, "acc_step": 0, "speed/wps": 12908.350921034085, "speed/FLOPS": 202743389147808.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045854803174734116, "optim/lr": 0.0029236442938412243, "optim/total_tokens": 5563219968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 2.7775650024414062, "created_at": "2025-01-16T11:31:56.090801+00:00"} {"global_step": 10612, "acc_step": 0, "speed/wps": 12907.858966600515, "speed/FLOPS": 202735662327412.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048117879778146744, "optim/lr": 0.002923617303563888, "optim/total_tokens": 5563744256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.7960972785949707, "created_at": "2025-01-16T11:32:06.249965+00:00"} {"global_step": 10613, "acc_step": 0, "speed/wps": 12893.062401700898, "speed/FLOPS": 202503261943053.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06332317739725113, "optim/lr": 0.00292359030864175, "optim/total_tokens": 5564268544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8265178203582764, "created_at": "2025-01-16T11:32:16.423163+00:00"} {"global_step": 10614, "acc_step": 0, "speed/wps": 12908.246444604356, "speed/FLOPS": 202741748201913.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047837186604738235, "optim/lr": 0.002923563309074898, "optim/total_tokens": 5564792832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 2.876009464263916, "created_at": "2025-01-16T11:32:26.580799+00:00"} {"global_step": 10615, "acc_step": 0, "speed/wps": 12882.061729121173, "speed/FLOPS": 202330481263680.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06423377990722656, "optim/lr": 0.002923536304863421, "optim/total_tokens": 5565317120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8968441486358643, "created_at": "2025-01-16T11:32:36.762643+00:00"} {"global_step": 10616, "acc_step": 0, "speed/wps": 12892.57164495619, "speed/FLOPS": 202495553934007.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06856981664896011, "optim/lr": 0.0029235092960074064, "optim/total_tokens": 5565841408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.9117846488952637, "created_at": "2025-01-16T11:32:46.932815+00:00"} {"global_step": 10617, "acc_step": 0, "speed/wps": 12904.587378430728, "speed/FLOPS": 202684277539573.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05268211290240288, "optim/lr": 0.002923482282506943, "optim/total_tokens": 5566365696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.904613971710205, "created_at": "2025-01-16T11:32:57.097896+00:00"} {"global_step": 10618, "acc_step": 0, "speed/wps": 12902.74291044665, "speed/FLOPS": 202655307635318.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08084049075841904, "optim/lr": 0.002923455264362118, "optim/total_tokens": 5566889984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.869662046432495, "created_at": "2025-01-16T11:33:07.260430+00:00"} {"global_step": 10619, "acc_step": 0, "speed/wps": 12898.080343356318, "speed/FLOPS": 202582075612126.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04104316607117653, "optim/lr": 0.0029234282415730208, "optim/total_tokens": 5567414272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.856403350830078, "created_at": "2025-01-16T11:33:17.427361+00:00"} {"global_step": 10620, "acc_step": 0, "speed/wps": 12900.47276807966, "speed/FLOPS": 202619651929942.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05870684236288071, "optim/lr": 0.0029234012141397384, "optim/total_tokens": 5567938560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 2.841454267501831, "created_at": "2025-01-16T11:33:27.596262+00:00"} {"global_step": 10621, "acc_step": 0, "speed/wps": 12905.762288673806, "speed/FLOPS": 202702731119436.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05021629109978676, "optim/lr": 0.00292337418206236, "optim/total_tokens": 5568462848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.031651496887207, "created_at": "2025-01-16T11:33:37.756339+00:00"} {"global_step": 10622, "acc_step": 0, "speed/wps": 12900.506594787274, "speed/FLOPS": 202620183224868.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06218333542346954, "optim/lr": 0.0029233471453409725, "optim/total_tokens": 5568987136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285796, "loss/out": 2.8505845069885254, "created_at": "2025-01-16T11:33:47.924249+00:00"} {"global_step": 10623, "acc_step": 0, "speed/wps": 12903.730963155125, "speed/FLOPS": 202670826360831.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05385785922408104, "optim/lr": 0.002923320103975665, "optim/total_tokens": 5569511424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.0212302207946777, "created_at": "2025-01-16T11:33:58.085826+00:00"} {"global_step": 10624, "acc_step": 0, "speed/wps": 12906.528926203337, "speed/FLOPS": 202714772215308.16, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04860955849289894, "optim/lr": 0.0029232930579665257, "optim/total_tokens": 5570035712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.9125170707702637, "created_at": "2025-01-16T11:34:08.244832+00:00"} {"global_step": 10625, "acc_step": 0, "speed/wps": 12899.212488402985, "speed/FLOPS": 202599857505814.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049127113074064255, "optim/lr": 0.0029232660073136426, "optim/total_tokens": 5570560000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323796, "loss/out": 2.8057785034179688, "created_at": "2025-01-16T11:34:18.410101+00:00"} {"global_step": 10626, "acc_step": 0, "speed/wps": 12900.296304711246, "speed/FLOPS": 202616880330255.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06629671901464462, "optim/lr": 0.0029232389520171044, "optim/total_tokens": 5571084288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.889615058898926, "created_at": "2025-01-16T11:34:28.577273+00:00"} {"global_step": 10627, "acc_step": 0, "speed/wps": 12902.064225014323, "speed/FLOPS": 202644647947992.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04760422557592392, "optim/lr": 0.002923211892076999, "optim/total_tokens": 5571608576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.833376884460449, "created_at": "2025-01-16T11:34:38.742467+00:00"} {"global_step": 10628, "acc_step": 0, "speed/wps": 12904.400187795181, "speed/FLOPS": 202681337453414.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05549998953938484, "optim/lr": 0.0029231848274934146, "optim/total_tokens": 5572132864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.7820568084716797, "created_at": "2025-01-16T11:34:48.903133+00:00"} {"global_step": 10629, "acc_step": 0, "speed/wps": 12901.060260065866, "speed/FLOPS": 202628879298884.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04825567826628685, "optim/lr": 0.0029231577582664397, "optim/total_tokens": 5572657152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8309226036071777, "created_at": "2025-01-16T11:34:59.067760+00:00"} {"global_step": 10630, "acc_step": 0, "speed/wps": 12903.601080389959, "speed/FLOPS": 202668786373527.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.072186678647995, "optim/lr": 0.002923130684396162, "optim/total_tokens": 5573181440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458694, "loss/out": 2.834756851196289, "created_at": "2025-01-16T11:35:09.232100+00:00"} {"global_step": 10631, "acc_step": 0, "speed/wps": 12899.95396670265, "speed/FLOPS": 202611503441408.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04666709154844284, "optim/lr": 0.002923103605882671, "optim/total_tokens": 5573705728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.853231906890869, "created_at": "2025-01-16T11:35:19.396293+00:00"} {"global_step": 10632, "acc_step": 0, "speed/wps": 12904.829617814385, "speed/FLOPS": 202688082241965.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06017895042896271, "optim/lr": 0.0029230765227260544, "optim/total_tokens": 5574230016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 2.8182621002197266, "created_at": "2025-01-16T11:35:29.561062+00:00"} {"global_step": 10633, "acc_step": 0, "speed/wps": 12901.036838546222, "speed/FLOPS": 202628511431733.9, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04114259034395218, "optim/lr": 0.002923049434926401, "optim/total_tokens": 5574754304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.7927327156066895, "created_at": "2025-01-16T11:35:39.725587+00:00"} {"global_step": 10634, "acc_step": 0, "speed/wps": 12904.661056607561, "speed/FLOPS": 202685434756584.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05219130963087082, "optim/lr": 0.0029230223424837978, "optim/total_tokens": 5575278592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.854757308959961, "created_at": "2025-01-16T11:35:49.890607+00:00"} {"global_step": 10635, "acc_step": 0, "speed/wps": 12900.007276867598, "speed/FLOPS": 202612340750805.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052042990922927856, "optim/lr": 0.002922995245398335, "optim/total_tokens": 5575802880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382950, "loss/out": 2.8864734172821045, "created_at": "2025-01-16T11:36:00.062901+00:00"} {"global_step": 10636, "acc_step": 0, "speed/wps": 12900.928970566909, "speed/FLOPS": 202626817216892.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0486450232565403, "optim/lr": 0.0029229681436701, "optim/total_tokens": 5576327168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.9166524410247803, "created_at": "2025-01-16T11:36:10.231766+00:00"} {"global_step": 10637, "acc_step": 0, "speed/wps": 12903.608144531121, "speed/FLOPS": 202668897325571.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0513065941631794, "optim/lr": 0.002922941037299181, "optim/total_tokens": 5576851456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.0395264625549316, "created_at": "2025-01-16T11:36:20.393161+00:00"} {"global_step": 10638, "acc_step": 0, "speed/wps": 12902.834140020455, "speed/FLOPS": 202656740521137.75, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05170143023133278, "optim/lr": 0.0029229139262856676, "optim/total_tokens": 5577375744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 2.8328428268432617, "created_at": "2025-01-16T11:36:30.558919+00:00"} {"global_step": 10639, "acc_step": 0, "speed/wps": 12903.444068214198, "speed/FLOPS": 202666320281549.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06636753678321838, "optim/lr": 0.002922886810629647, "optim/total_tokens": 5577900032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.963381767272949, "created_at": "2025-01-16T11:36:40.722265+00:00"} {"global_step": 10640, "acc_step": 0, "speed/wps": 12903.652663823468, "speed/FLOPS": 202669596562233.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05466049164533615, "optim/lr": 0.0029228596903312083, "optim/total_tokens": 5578424320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.7455830574035645, "created_at": "2025-01-16T11:36:50.887363+00:00"} {"global_step": 10641, "acc_step": 0, "speed/wps": 12905.286304827878, "speed/FLOPS": 202695255139065.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05852385237812996, "optim/lr": 0.00292283256539044, "optim/total_tokens": 5578948608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 2.862654209136963, "created_at": "2025-01-16T11:37:01.047965+00:00"} {"global_step": 10642, "acc_step": 0, "speed/wps": 12906.857978548283, "speed/FLOPS": 202719940434550.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05413639917969704, "optim/lr": 0.00292280543580743, "optim/total_tokens": 5579472896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8404598236083984, "created_at": "2025-01-16T11:37:11.213904+00:00"} {"global_step": 10643, "acc_step": 0, "speed/wps": 12904.999583305036, "speed/FLOPS": 202690751783553.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06160635128617287, "optim/lr": 0.002922778301582268, "optim/total_tokens": 5579997184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 2.8865113258361816, "created_at": "2025-01-16T11:37:21.377572+00:00"} {"global_step": 10644, "acc_step": 0, "speed/wps": 12901.331920428998, "speed/FLOPS": 202633146098187.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057645153254270554, "optim/lr": 0.002922751162715041, "optim/total_tokens": 5580521472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8784267902374268, "created_at": "2025-01-16T11:37:31.540879+00:00"} {"global_step": 10645, "acc_step": 0, "speed/wps": 12904.519177140413, "speed/FLOPS": 202683206344590.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04767501726746559, "optim/lr": 0.002922724019205839, "optim/total_tokens": 5581045760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.953744888305664, "created_at": "2025-01-16T11:37:41.704454+00:00"} {"global_step": 10646, "acc_step": 0, "speed/wps": 12903.022902341983, "speed/FLOPS": 202659705292783.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04318862780928612, "optim/lr": 0.0029226968710547494, "optim/total_tokens": 5581570048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 2.8913097381591797, "created_at": "2025-01-16T11:37:51.866097+00:00"} {"global_step": 10647, "acc_step": 0, "speed/wps": 12906.722279115995, "speed/FLOPS": 202717809088496.22, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04503355920314789, "optim/lr": 0.002922669718261862, "optim/total_tokens": 5582094336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.6911230087280273, "created_at": "2025-01-16T11:38:02.025515+00:00"} {"global_step": 10648, "acc_step": 0, "speed/wps": 12907.425089926452, "speed/FLOPS": 202728847698037.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05163950473070145, "optim/lr": 0.0029226425608272636, "optim/total_tokens": 5582618624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377259, "loss/out": 2.9855828285217285, "created_at": "2025-01-16T11:38:12.183832+00:00"} {"global_step": 10649, "acc_step": 0, "speed/wps": 12900.954184499707, "speed/FLOPS": 202627213236354.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05086183920502663, "optim/lr": 0.0029226153987510447, "optim/total_tokens": 5583142912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.84696888923645, "created_at": "2025-01-16T11:38:22.355347+00:00"} {"global_step": 10650, "acc_step": 0, "speed/wps": 12908.392015915064, "speed/FLOPS": 202744034599384.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0465799942612648, "optim/lr": 0.002922588232033293, "optim/total_tokens": 5583667200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456054, "loss/out": 2.9813857078552246, "created_at": "2025-01-16T11:38:32.514762+00:00"} {"global_step": 10651, "acc_step": 0, "speed/wps": 12909.211320830682, "speed/FLOPS": 202756902908927.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053714875131845474, "optim/lr": 0.0029225610606740974, "optim/total_tokens": 5584191488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.8386473655700684, "created_at": "2025-01-16T11:38:42.674987+00:00"} {"global_step": 10652, "acc_step": 0, "speed/wps": 12905.661200944947, "speed/FLOPS": 202701143397745.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04924677684903145, "optim/lr": 0.0029225338846735464, "optim/total_tokens": 5584715776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.010470151901245, "created_at": "2025-01-16T11:38:52.834720+00:00"} {"global_step": 10653, "acc_step": 0, "speed/wps": 12905.014961891322, "speed/FLOPS": 202690993325383.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05672863870859146, "optim/lr": 0.002922506704031728, "optim/total_tokens": 5585240064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286807, "loss/out": 2.869802713394165, "created_at": "2025-01-16T11:39:02.995909+00:00"} {"global_step": 10654, "acc_step": 0, "speed/wps": 12904.897594006898, "speed/FLOPS": 202689149901477.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04825728386640549, "optim/lr": 0.0029224795187487323, "optim/total_tokens": 5585764352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.865140199661255, "created_at": "2025-01-16T11:39:13.157058+00:00"} {"global_step": 10655, "acc_step": 0, "speed/wps": 12905.902079385462, "speed/FLOPS": 202704926724655.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049165405333042145, "optim/lr": 0.002922452328824647, "optim/total_tokens": 5586288640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386402, "loss/out": 2.972247838973999, "created_at": "2025-01-16T11:39:23.316536+00:00"} {"global_step": 10656, "acc_step": 0, "speed/wps": 12905.598538638355, "speed/FLOPS": 202700159200113.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04408423975110054, "optim/lr": 0.002922425134259561, "optim/total_tokens": 5586812928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.986645221710205, "created_at": "2025-01-16T11:39:33.481110+00:00"} {"global_step": 10657, "acc_step": 0, "speed/wps": 12905.207772429836, "speed/FLOPS": 202694021679839.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04638615623116493, "optim/lr": 0.002922397935053563, "optim/total_tokens": 5587337216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424126, "loss/out": 2.8553988933563232, "created_at": "2025-01-16T11:39:43.645007+00:00"} {"global_step": 10658, "acc_step": 0, "speed/wps": 12899.35426420936, "speed/FLOPS": 202602084289674.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0611422173678875, "optim/lr": 0.0029223707312067415, "optim/total_tokens": 5587861504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.0783047676086426, "created_at": "2025-01-16T11:39:53.810057+00:00"} {"global_step": 10659, "acc_step": 0, "speed/wps": 12901.24413466256, "speed/FLOPS": 202631767302095.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048281360417604446, "optim/lr": 0.002922343522719186, "optim/total_tokens": 5588385792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9485740661621094, "created_at": "2025-01-16T11:40:03.973554+00:00"} {"global_step": 10660, "acc_step": 0, "speed/wps": 12900.319241093244, "speed/FLOPS": 202617240577652.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059312932193279266, "optim/lr": 0.002922316309590985, "optim/total_tokens": 5588910080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9977526664733887, "created_at": "2025-01-16T11:40:14.145214+00:00"} {"global_step": 10661, "acc_step": 0, "speed/wps": 12904.85670596793, "speed/FLOPS": 202688507698639.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05501534417271614, "optim/lr": 0.002922289091822227, "optim/total_tokens": 5589434368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329658, "loss/out": 2.8385262489318848, "created_at": "2025-01-16T11:40:24.306161+00:00"} {"global_step": 10662, "acc_step": 0, "speed/wps": 12898.925516526504, "speed/FLOPS": 202595350218153.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05692199990153313, "optim/lr": 0.0029222618694130006, "optim/total_tokens": 5589958656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.967028856277466, "created_at": "2025-01-16T11:40:34.471132+00:00"} {"global_step": 10663, "acc_step": 0, "speed/wps": 12893.38585275316, "speed/FLOPS": 202508342186297.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07271980494260788, "optim/lr": 0.0029222346423633957, "optim/total_tokens": 5590482944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 2.933419942855835, "created_at": "2025-01-16T11:40:44.644388+00:00"} {"global_step": 10664, "acc_step": 0, "speed/wps": 12904.85978611071, "speed/FLOPS": 202688556076514.62, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04981228709220886, "optim/lr": 0.0029222074106734996, "optim/total_tokens": 5591007232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9572296142578125, "created_at": "2025-01-16T11:40:54.808739+00:00"} {"global_step": 10665, "acc_step": 0, "speed/wps": 12906.720863474558, "speed/FLOPS": 202717786853902.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07527641952037811, "optim/lr": 0.0029221801743434025, "optim/total_tokens": 5591531520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9523844718933105, "created_at": "2025-01-16T11:41:04.970256+00:00"} {"global_step": 10666, "acc_step": 0, "speed/wps": 12903.51046462504, "speed/FLOPS": 202667363128420.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06362265348434448, "optim/lr": 0.0029221529333731924, "optim/total_tokens": 5592055808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8535332679748535, "created_at": "2025-01-16T11:41:15.131672+00:00"} {"global_step": 10667, "acc_step": 0, "speed/wps": 12905.426214219877, "speed/FLOPS": 202697452608322.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05707408860325813, "optim/lr": 0.002922125687762959, "optim/total_tokens": 5592580096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.0257253646850586, "created_at": "2025-01-16T11:41:25.296931+00:00"} {"global_step": 10668, "acc_step": 0, "speed/wps": 12902.185856273492, "speed/FLOPS": 202646558334056.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05617256835103035, "optim/lr": 0.00292209843751279, "optim/total_tokens": 5593104384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.92291259765625, "created_at": "2025-01-16T11:41:35.463235+00:00"} {"global_step": 10669, "acc_step": 0, "speed/wps": 12903.76182671236, "speed/FLOPS": 202671311115409.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04876449704170227, "optim/lr": 0.0029220711826227755, "optim/total_tokens": 5593628672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.819289207458496, "created_at": "2025-01-16T11:41:45.626952+00:00"} {"global_step": 10670, "acc_step": 0, "speed/wps": 12899.839309838324, "speed/FLOPS": 202609702597799.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0524783581495285, "optim/lr": 0.0029220439230930035, "optim/total_tokens": 5594152960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.989750385284424, "created_at": "2025-01-16T11:41:55.791204+00:00"} {"global_step": 10671, "acc_step": 0, "speed/wps": 12902.632329942657, "speed/FLOPS": 202653570816549.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05906775966286659, "optim/lr": 0.0029220166589235632, "optim/total_tokens": 5594677248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 2.865929365158081, "created_at": "2025-01-16T11:42:05.953270+00:00"} {"global_step": 10672, "acc_step": 0, "speed/wps": 12900.64069839187, "speed/FLOPS": 202622289506256.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05898837000131607, "optim/lr": 0.0029219893901145447, "optim/total_tokens": 5595201536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.902787685394287, "created_at": "2025-01-16T11:42:16.118528+00:00"} {"global_step": 10673, "acc_step": 0, "speed/wps": 12896.061858854546, "speed/FLOPS": 202550372539337.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05156463757157326, "optim/lr": 0.0029219621166660346, "optim/total_tokens": 5595725824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.8304595947265625, "created_at": "2025-01-16T11:42:26.291595+00:00"} {"global_step": 10674, "acc_step": 0, "speed/wps": 12902.497685756576, "speed/FLOPS": 202651456044586.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.060682762414216995, "optim/lr": 0.002921934838578124, "optim/total_tokens": 5596250112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.816110610961914, "created_at": "2025-01-16T11:42:36.453976+00:00"} {"global_step": 10675, "acc_step": 0, "speed/wps": 12898.537811773744, "speed/FLOPS": 202589260782249.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06058579683303833, "optim/lr": 0.0029219075558509014, "optim/total_tokens": 5596774400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 2.8387610912323, "created_at": "2025-01-16T11:42:46.628072+00:00"} {"global_step": 10676, "acc_step": 0, "speed/wps": 12902.255917146786, "speed/FLOPS": 202647658736345.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05037305876612663, "optim/lr": 0.002921880268484455, "optim/total_tokens": 5597298688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377871, "loss/out": 2.9897420406341553, "created_at": "2025-01-16T11:42:56.790529+00:00"} {"global_step": 10677, "acc_step": 0, "speed/wps": 12902.122745126957, "speed/FLOPS": 202645567086776.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04843306541442871, "optim/lr": 0.002921852976478875, "optim/total_tokens": 5597822976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.8804686069488525, "created_at": "2025-01-16T11:43:06.954767+00:00"} {"global_step": 10678, "acc_step": 0, "speed/wps": 12903.061945148866, "speed/FLOPS": 202660318513716.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048261526972055435, "optim/lr": 0.0029218256798342495, "optim/total_tokens": 5598347264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.062595844268799, "created_at": "2025-01-16T11:43:17.119864+00:00"} {"global_step": 10679, "acc_step": 0, "speed/wps": 12902.033038396943, "speed/FLOPS": 202644158119311.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05407530069351196, "optim/lr": 0.002921798378550668, "optim/total_tokens": 5598871552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363200, "loss/out": 2.960756540298462, "created_at": "2025-01-16T11:43:27.282876+00:00"} {"global_step": 10680, "acc_step": 0, "speed/wps": 12902.765702932642, "speed/FLOPS": 202655665622631.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05069835111498833, "optim/lr": 0.002921771072628219, "optim/total_tokens": 5599395840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.835745334625244, "created_at": "2025-01-16T11:43:37.447235+00:00"} {"global_step": 10681, "acc_step": 0, "speed/wps": 12901.575294819932, "speed/FLOPS": 202636968627428.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06267409771680832, "optim/lr": 0.002921743762066993, "optim/total_tokens": 5599920128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9100348949432373, "created_at": "2025-01-16T11:43:47.612928+00:00"} {"global_step": 10682, "acc_step": 0, "speed/wps": 12891.58080482722, "speed/FLOPS": 202479991428224.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05135276913642883, "optim/lr": 0.0029217164468670776, "optim/total_tokens": 5600444416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.1023669242858887, "created_at": "2025-01-16T11:43:57.785958+00:00"} {"global_step": 10683, "acc_step": 0, "speed/wps": 12900.827820179795, "speed/FLOPS": 202625228511067.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058319274336099625, "optim/lr": 0.0029216891270285625, "optim/total_tokens": 5600968704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.8796324729919434, "created_at": "2025-01-16T11:44:07.953109+00:00"} {"global_step": 10684, "acc_step": 0, "speed/wps": 12902.710411065465, "speed/FLOPS": 202654797187878.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05515284463763237, "optim/lr": 0.0029216618025515373, "optim/total_tokens": 5601492992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8404197692871094, "created_at": "2025-01-16T11:44:18.118322+00:00"} {"global_step": 10685, "acc_step": 0, "speed/wps": 12898.739113123194, "speed/FLOPS": 202592422496559.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05991804972290993, "optim/lr": 0.00292163447343609, "optim/total_tokens": 5602017280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 2.8309359550476074, "created_at": "2025-01-16T11:44:28.283392+00:00"} {"global_step": 10686, "acc_step": 0, "speed/wps": 12908.716780050756, "speed/FLOPS": 202749135466408.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06487199664115906, "optim/lr": 0.0029216071396823105, "optim/total_tokens": 5602541568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.739664316177368, "created_at": "2025-01-16T11:44:38.440769+00:00"} {"global_step": 10687, "acc_step": 0, "speed/wps": 12901.915278843167, "speed/FLOPS": 202642308543700.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044398099184036255, "optim/lr": 0.002921579801290288, "optim/total_tokens": 5603065856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 2.8961100578308105, "created_at": "2025-01-16T11:44:48.607620+00:00"} {"global_step": 10688, "acc_step": 0, "speed/wps": 12903.35129089482, "speed/FLOPS": 202664863086259.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05637296661734581, "optim/lr": 0.002921552458260112, "optim/total_tokens": 5603590144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.9426817893981934, "created_at": "2025-01-16T11:44:58.773353+00:00"} {"global_step": 10689, "acc_step": 0, "speed/wps": 12903.976287968637, "speed/FLOPS": 202674679524139.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05119903013110161, "optim/lr": 0.002921525110591871, "optim/total_tokens": 5604114432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.8796987533569336, "created_at": "2025-01-16T11:45:08.934577+00:00"} {"global_step": 10690, "acc_step": 0, "speed/wps": 12893.957132143045, "speed/FLOPS": 202517314914135.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0026, "optim/grad_norm": 0.05178045853972435, "optim/lr": 0.002921497758285655, "optim/total_tokens": 5604638720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 2.907881259918213, "created_at": "2025-01-16T11:45:19.106778+00:00"} {"global_step": 10691, "acc_step": 0, "speed/wps": 12893.739700964223, "speed/FLOPS": 202513899858690.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0570230595767498, "optim/lr": 0.0029214704013415527, "optim/total_tokens": 5605163008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.994079828262329, "created_at": "2025-01-16T11:45:29.280170+00:00"} {"global_step": 10692, "acc_step": 0, "speed/wps": 12897.351667951982, "speed/FLOPS": 202570630763596.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058737948536872864, "optim/lr": 0.002921443039759653, "optim/total_tokens": 5605687296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 2.9034974575042725, "created_at": "2025-01-16T11:45:39.451997+00:00"} {"global_step": 10693, "acc_step": 0, "speed/wps": 12899.777532897897, "speed/FLOPS": 202608732306059.62, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045144468545913696, "optim/lr": 0.002921415673540046, "optim/total_tokens": 5606211584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484999, "loss/out": 2.907593250274658, "created_at": "2025-01-16T11:45:49.622627+00:00"} {"global_step": 10694, "acc_step": 0, "speed/wps": 12900.640215918986, "speed/FLOPS": 202622281928356.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04516847804188728, "optim/lr": 0.0029213883026828207, "optim/total_tokens": 5606735872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.841097116470337, "created_at": "2025-01-16T11:45:59.787375+00:00"} {"global_step": 10695, "acc_step": 0, "speed/wps": 12898.782034785363, "speed/FLOPS": 202593096640244.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04199712350964546, "optim/lr": 0.002921360927188066, "optim/total_tokens": 5607260160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8048553466796875, "created_at": "2025-01-16T11:46:09.952631+00:00"} {"global_step": 10696, "acc_step": 0, "speed/wps": 12891.69033392919, "speed/FLOPS": 202481711733280.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04621489346027374, "optim/lr": 0.002921333547055872, "optim/total_tokens": 5607784448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 2.9414100646972656, "created_at": "2025-01-16T11:46:20.124154+00:00"} {"global_step": 10697, "acc_step": 0, "speed/wps": 12888.677030589995, "speed/FLOPS": 202434383663625.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05340904742479324, "optim/lr": 0.0029213061622863274, "optim/total_tokens": 5608308736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.908889055252075, "created_at": "2025-01-16T11:46:30.298306+00:00"} {"global_step": 10698, "acc_step": 0, "speed/wps": 12888.41471495092, "speed/FLOPS": 202430263636054.28, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05691475421190262, "optim/lr": 0.0029212787728795215, "optim/total_tokens": 5608833024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.873112916946411, "created_at": "2025-01-16T11:46:40.471607+00:00"} {"global_step": 10699, "acc_step": 0, "speed/wps": 12900.834357407553, "speed/FLOPS": 202625331187212.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05249692499637604, "optim/lr": 0.002921251378835544, "optim/total_tokens": 5609357312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8331499099731445, "created_at": "2025-01-16T11:46:50.635773+00:00"} {"global_step": 10700, "acc_step": 0, "speed/wps": 12890.13446948707, "speed/FLOPS": 202457274744234.88, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07567182183265686, "optim/lr": 0.002921223980154484, "optim/total_tokens": 5609881600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.8776915073394775, "created_at": "2025-01-16T11:47:00.808421+00:00"} {"global_step": 10701, "acc_step": 0, "speed/wps": 12905.826734895065, "speed/FLOPS": 202703743335899.78, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04702024161815643, "optim/lr": 0.002921196576836432, "optim/total_tokens": 5610405888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.938140392303467, "created_at": "2025-01-16T11:47:10.968387+00:00"} {"global_step": 10702, "acc_step": 0, "speed/wps": 12893.547971218595, "speed/FLOPS": 202510888479571.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06032956391572952, "optim/lr": 0.002921169168881475, "optim/total_tokens": 5610930176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 275344, "loss/out": 2.8214688301086426, "created_at": "2025-01-16T11:47:21.137841+00:00"} {"global_step": 10703, "acc_step": 0, "speed/wps": 12895.155399114637, "speed/FLOPS": 202536135343516.16, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061371710151433945, "optim/lr": 0.002921141756289705, "optim/total_tokens": 5611454464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.940384864807129, "created_at": "2025-01-16T11:47:31.308622+00:00"} {"global_step": 10704, "acc_step": 0, "speed/wps": 12901.616774711762, "speed/FLOPS": 202637620126129.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048225630074739456, "optim/lr": 0.00292111433906121, "optim/total_tokens": 5611978752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.76171612739563, "created_at": "2025-01-16T11:47:41.474327+00:00"} {"global_step": 10705, "acc_step": 0, "speed/wps": 12900.23272960642, "speed/FLOPS": 202615881795872.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0686870738863945, "optim/lr": 0.0029210869171960798, "optim/total_tokens": 5612503040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 2.9869027137756348, "created_at": "2025-01-16T11:47:51.639399+00:00"} {"global_step": 10706, "acc_step": 0, "speed/wps": 12902.366938315412, "speed/FLOPS": 202649402476357.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05736652761697769, "optim/lr": 0.0029210594906944037, "optim/total_tokens": 5613027328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.92606782913208, "created_at": "2025-01-16T11:48:01.803644+00:00"} {"global_step": 10707, "acc_step": 0, "speed/wps": 12904.02424449068, "speed/FLOPS": 202675432747217.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06862387806177139, "optim/lr": 0.002921032059556271, "optim/total_tokens": 5613551616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8505401611328125, "created_at": "2025-01-16T11:48:11.968525+00:00"} {"global_step": 10708, "acc_step": 0, "speed/wps": 12902.360678686233, "speed/FLOPS": 202649304160279.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0743764340877533, "optim/lr": 0.0029210046237817724, "optim/total_tokens": 5614075904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 2.973905563354492, "created_at": "2025-01-16T11:48:22.134495+00:00"} {"global_step": 10709, "acc_step": 0, "speed/wps": 12900.734913739348, "speed/FLOPS": 202623769287753.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055028658360242844, "optim/lr": 0.002920977183370996, "optim/total_tokens": 5614600192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9843719005584717, "created_at": "2025-01-16T11:48:32.298123+00:00"} {"global_step": 10710, "acc_step": 0, "speed/wps": 12900.704063477553, "speed/FLOPS": 202623284741999.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06076561287045479, "optim/lr": 0.002920949738324032, "optim/total_tokens": 5615124480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 2.7637836933135986, "created_at": "2025-01-16T11:48:42.464571+00:00"} {"global_step": 10711, "acc_step": 0, "speed/wps": 12902.072179299534, "speed/FLOPS": 202644772880971.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0509786419570446, "optim/lr": 0.0029209222886409695, "optim/total_tokens": 5615648768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8298442363739014, "created_at": "2025-01-16T11:48:52.636853+00:00"} {"global_step": 10712, "acc_step": 0, "speed/wps": 12901.44364003451, "speed/FLOPS": 202634900808111.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06533806771039963, "optim/lr": 0.002920894834321899, "optim/total_tokens": 5616173056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.881654739379883, "created_at": "2025-01-16T11:49:02.800680+00:00"} {"global_step": 10713, "acc_step": 0, "speed/wps": 12901.68289685061, "speed/FLOPS": 202638658665181.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04598870500922203, "optim/lr": 0.002920867375366909, "optim/total_tokens": 5616697344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396331, "loss/out": 2.9025750160217285, "created_at": "2025-01-16T11:49:12.967614+00:00"} {"global_step": 10714, "acc_step": 0, "speed/wps": 12901.313167042852, "speed/FLOPS": 202632851550488.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06362730264663696, "optim/lr": 0.0029208399117760897, "optim/total_tokens": 5617221632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.93804931640625, "created_at": "2025-01-16T11:49:23.133416+00:00"} {"global_step": 10715, "acc_step": 0, "speed/wps": 12896.85473098458, "speed/FLOPS": 202562825685660.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05311833322048187, "optim/lr": 0.002920812443549531, "optim/total_tokens": 5617745920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8450026512145996, "created_at": "2025-01-16T11:49:33.301393+00:00"} {"global_step": 10716, "acc_step": 0, "speed/wps": 12901.651457792534, "speed/FLOPS": 202638164871572.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06545429676771164, "optim/lr": 0.0029207849706873214, "optim/total_tokens": 5618270208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.912461757659912, "created_at": "2025-01-16T11:49:43.465332+00:00"} {"global_step": 10717, "acc_step": 0, "speed/wps": 12897.45266691623, "speed/FLOPS": 202572217091116.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06297577917575836, "optim/lr": 0.0029207574931895517, "optim/total_tokens": 5618794496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9935083389282227, "created_at": "2025-01-16T11:49:53.632432+00:00"} {"global_step": 10718, "acc_step": 0, "speed/wps": 12901.999479271728, "speed/FLOPS": 202643631027134.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05753552168607712, "optim/lr": 0.0029207300110563107, "optim/total_tokens": 5619318784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.8091540336608887, "created_at": "2025-01-16T11:50:03.799779+00:00"} {"global_step": 10719, "acc_step": 0, "speed/wps": 12890.656375275148, "speed/FLOPS": 202465471991810.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0669778436422348, "optim/lr": 0.002920702524287688, "optim/total_tokens": 5619843072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432114, "loss/out": 2.950305461883545, "created_at": "2025-01-16T11:50:13.973721+00:00"} {"global_step": 10720, "acc_step": 0, "speed/wps": 12893.394399936293, "speed/FLOPS": 202508476431553.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04954897612333298, "optim/lr": 0.002920675032883775, "optim/total_tokens": 5620367360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 2.8779540061950684, "created_at": "2025-01-16T11:50:24.146230+00:00"} {"global_step": 10721, "acc_step": 0, "speed/wps": 12901.3995629278, "speed/FLOPS": 202634208516582.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0557706318795681, "optim/lr": 0.002920647536844659, "optim/total_tokens": 5620891648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8859198093414307, "created_at": "2025-01-16T11:50:34.312008+00:00"} {"global_step": 10722, "acc_step": 0, "speed/wps": 12893.56269410702, "speed/FLOPS": 202511119722765.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054871346801519394, "optim/lr": 0.002920620036170431, "optim/total_tokens": 5621415936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 2.896329402923584, "created_at": "2025-01-16T11:50:44.481208+00:00"} {"global_step": 10723, "acc_step": 0, "speed/wps": 12899.153662472605, "speed/FLOPS": 202598933563742.62, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04695546254515648, "optim/lr": 0.0029205925308611806, "optim/total_tokens": 5621940224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.868621826171875, "created_at": "2025-01-16T11:50:54.648283+00:00"} {"global_step": 10724, "acc_step": 0, "speed/wps": 12897.207708276415, "speed/FLOPS": 202568369679070.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052327532321214676, "optim/lr": 0.0029205650209169977, "optim/total_tokens": 5622464512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 2.933152437210083, "created_at": "2025-01-16T11:51:04.818721+00:00"} {"global_step": 10725, "acc_step": 0, "speed/wps": 12884.833113315832, "speed/FLOPS": 202374009660738.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.047337211668491364, "optim/lr": 0.002920537506337972, "optim/total_tokens": 5622988800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.8694677352905273, "created_at": "2025-01-16T11:51:14.995822+00:00"} {"global_step": 10726, "acc_step": 0, "speed/wps": 12889.874198275598, "speed/FLOPS": 202453186827208.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04859510436654091, "optim/lr": 0.0029205099871241927, "optim/total_tokens": 5623513088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330478, "loss/out": 2.94327974319458, "created_at": "2025-01-16T11:51:25.171554+00:00"} {"global_step": 10727, "acc_step": 0, "speed/wps": 12885.414905096639, "speed/FLOPS": 202383147500122.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0612262487411499, "optim/lr": 0.0029204824632757496, "optim/total_tokens": 5624037376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 2.856283187866211, "created_at": "2025-01-16T11:51:35.350533+00:00"} {"global_step": 10728, "acc_step": 0, "speed/wps": 12894.190618085899, "speed/FLOPS": 202520982131711.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05163736641407013, "optim/lr": 0.002920454934792734, "optim/total_tokens": 5624561664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.846019983291626, "created_at": "2025-01-16T11:51:45.525282+00:00"} {"global_step": 10729, "acc_step": 0, "speed/wps": 12900.728395533133, "speed/FLOPS": 202623666910368.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06216319277882576, "optim/lr": 0.0029204274016752336, "optim/total_tokens": 5625085952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 2.9209179878234863, "created_at": "2025-01-16T11:51:55.688912+00:00"} {"global_step": 10730, "acc_step": 0, "speed/wps": 12903.703006604619, "speed/FLOPS": 202670387264789.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050077978521585464, "optim/lr": 0.0029203998639233397, "optim/total_tokens": 5625610240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.6982667446136475, "created_at": "2025-01-16T11:52:05.852856+00:00"} {"global_step": 10731, "acc_step": 0, "speed/wps": 12903.167428391973, "speed/FLOPS": 202661975272998.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04759419709444046, "optim/lr": 0.0029203723215371416, "optim/total_tokens": 5626134528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.918308734893799, "created_at": "2025-01-16T11:52:16.015948+00:00"} {"global_step": 10732, "acc_step": 0, "speed/wps": 12903.44714295049, "speed/FLOPS": 202666368574508.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06487228721380234, "optim/lr": 0.002920344774516729, "optim/total_tokens": 5626658816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.8740549087524414, "created_at": "2025-01-16T11:52:26.180730+00:00"} {"global_step": 10733, "acc_step": 0, "speed/wps": 12903.967117432116, "speed/FLOPS": 202674535488262.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07145997881889343, "optim/lr": 0.0029203172228621926, "optim/total_tokens": 5627183104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.931288242340088, "created_at": "2025-01-16T11:52:36.342835+00:00"} {"global_step": 10734, "acc_step": 0, "speed/wps": 12909.452580484487, "speed/FLOPS": 202760692223315.62, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05328284204006195, "optim/lr": 0.0029202896665736218, "optim/total_tokens": 5627707392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.8468832969665527, "created_at": "2025-01-16T11:52:46.499655+00:00"} {"global_step": 10735, "acc_step": 0, "speed/wps": 12900.170979641374, "speed/FLOPS": 202614911927818.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07202606648206711, "optim/lr": 0.0029202621056511058, "optim/total_tokens": 5628231680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8292953968048096, "created_at": "2025-01-16T11:52:56.666767+00:00"} {"global_step": 10736, "acc_step": 0, "speed/wps": 12905.829848638445, "speed/FLOPS": 202703792241518.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05272983759641647, "optim/lr": 0.0029202345400947353, "optim/total_tokens": 5628755968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.7559685707092285, "created_at": "2025-01-16T11:53:06.826315+00:00"} {"global_step": 10737, "acc_step": 0, "speed/wps": 12901.146918654054, "speed/FLOPS": 202630240391093.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05899309739470482, "optim/lr": 0.0029202069699046, "optim/total_tokens": 5629280256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.9173054695129395, "created_at": "2025-01-16T11:53:16.994861+00:00"} {"global_step": 10738, "acc_step": 0, "speed/wps": 12902.90046154363, "speed/FLOPS": 202657782191794.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06668813526630402, "optim/lr": 0.00292017939508079, "optim/total_tokens": 5629804544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.868346691131592, "created_at": "2025-01-16T11:53:27.157586+00:00"} {"global_step": 10739, "acc_step": 0, "speed/wps": 12899.59837271205, "speed/FLOPS": 202605918349149.0, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06531713157892227, "optim/lr": 0.002920151815623396, "optim/total_tokens": 5630328832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 3.018348217010498, "created_at": "2025-01-16T11:53:37.331057+00:00"} {"global_step": 10740, "acc_step": 0, "speed/wps": 12903.424527879324, "speed/FLOPS": 202666013373739.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0724412053823471, "optim/lr": 0.002920124231532506, "optim/total_tokens": 5630853120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 2.7420525550842285, "created_at": "2025-01-16T11:53:47.493045+00:00"} {"global_step": 10741, "acc_step": 0, "speed/wps": 12905.508054579246, "speed/FLOPS": 202698738023627.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06851176917552948, "optim/lr": 0.0029200966428082117, "optim/total_tokens": 5631377408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.946802854537964, "created_at": "2025-01-16T11:53:57.661238+00:00"} {"global_step": 10742, "acc_step": 0, "speed/wps": 12899.088378459874, "speed/FLOPS": 202597908188613.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05403851717710495, "optim/lr": 0.002920069049450603, "optim/total_tokens": 5631901696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 2.9986159801483154, "created_at": "2025-01-16T11:54:07.827781+00:00"} {"global_step": 10743, "acc_step": 0, "speed/wps": 12903.896937407639, "speed/FLOPS": 202673433214540.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07591190189123154, "optim/lr": 0.002920041451459769, "optim/total_tokens": 5632425984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.891808032989502, "created_at": "2025-01-16T11:54:17.989636+00:00"} {"global_step": 10744, "acc_step": 0, "speed/wps": 12906.35118691789, "speed/FLOPS": 202711980575590.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07251815497875214, "optim/lr": 0.0029200138488358, "optim/total_tokens": 5632950272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8994717597961426, "created_at": "2025-01-16T11:54:28.148695+00:00"} {"global_step": 10745, "acc_step": 0, "speed/wps": 12902.897216723735, "speed/FLOPS": 202657731227439.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06206709146499634, "optim/lr": 0.0029199862415787868, "optim/total_tokens": 5633474560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 2.987852096557617, "created_at": "2025-01-16T11:54:38.314651+00:00"} {"global_step": 10746, "acc_step": 0, "speed/wps": 12900.806581330704, "speed/FLOPS": 202624894925757.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04268772155046463, "optim/lr": 0.002919958629688819, "optim/total_tokens": 5633998848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.018864393234253, "created_at": "2025-01-16T11:54:48.479282+00:00"} {"global_step": 10747, "acc_step": 0, "speed/wps": 12903.09308114045, "speed/FLOPS": 202660807547249.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05017443373799324, "optim/lr": 0.0029199310131659867, "optim/total_tokens": 5634523136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.868973731994629, "created_at": "2025-01-16T11:54:58.643320+00:00"} {"global_step": 10748, "acc_step": 0, "speed/wps": 12901.48879960101, "speed/FLOPS": 202635610101158.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047455690801143646, "optim/lr": 0.0029199033920103796, "optim/total_tokens": 5635047424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9343678951263428, "created_at": "2025-01-16T11:55:08.806318+00:00"} {"global_step": 10749, "acc_step": 0, "speed/wps": 12903.071474023951, "speed/FLOPS": 202660468177793.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0505070760846138, "optim/lr": 0.002919875766222089, "optim/total_tokens": 5635571712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.833266019821167, "created_at": "2025-01-16T11:55:18.969120+00:00"} {"global_step": 10750, "acc_step": 0, "speed/wps": 12904.282977809602, "speed/FLOPS": 202679496509528.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05536409839987755, "optim/lr": 0.002919848135801204, "optim/total_tokens": 5636096000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284448, "loss/out": 2.9200634956359863, "created_at": "2025-01-16T11:55:29.129860+00:00"} {"global_step": 10751, "acc_step": 0, "speed/wps": 12905.409582609685, "speed/FLOPS": 202697191386030.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06810108572244644, "optim/lr": 0.0029198205007478144, "optim/total_tokens": 5636620288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.907953977584839, "created_at": "2025-01-16T11:55:39.291436+00:00"} {"global_step": 10752, "acc_step": 0, "speed/wps": 12901.486752369949, "speed/FLOPS": 202635577946581.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05551350489258766, "optim/lr": 0.0029197928610620117, "optim/total_tokens": 5637144576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.8081650733947754, "created_at": "2025-01-16T11:55:49.456151+00:00"} {"global_step": 10753, "acc_step": 0, "speed/wps": 12902.247728435417, "speed/FLOPS": 202647530121380.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04270564019680023, "optim/lr": 0.0029197652167438845, "optim/total_tokens": 5637668864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.85880970954895, "created_at": "2025-01-16T11:55:59.618396+00:00"} {"global_step": 10754, "acc_step": 0, "speed/wps": 12900.512694161456, "speed/FLOPS": 202620279023921.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.057402223348617554, "optim/lr": 0.002919737567793525, "optim/total_tokens": 5638193152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467434, "loss/out": 2.691575050354004, "created_at": "2025-01-16T11:56:09.790305+00:00"} {"global_step": 10755, "acc_step": 0, "speed/wps": 12904.240338991822, "speed/FLOPS": 202678826808302.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0445144921541214, "optim/lr": 0.0029197099142110215, "optim/total_tokens": 5638717440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299229, "loss/out": 2.8124279975891113, "created_at": "2025-01-16T11:56:19.951515+00:00"} {"global_step": 10756, "acc_step": 0, "speed/wps": 12903.791475363849, "speed/FLOPS": 202671776788222.38, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04863201081752777, "optim/lr": 0.0029196822559964647, "optim/total_tokens": 5639241728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.895388603210449, "created_at": "2025-01-16T11:56:30.116168+00:00"} {"global_step": 10757, "acc_step": 0, "speed/wps": 12902.551823673399, "speed/FLOPS": 202652306354963.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05849221721291542, "optim/lr": 0.0029196545931499452, "optim/total_tokens": 5639766016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.93365478515625, "created_at": "2025-01-16T11:56:40.286383+00:00"} {"global_step": 10758, "acc_step": 0, "speed/wps": 12899.85533117103, "speed/FLOPS": 202609954234845.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04780939593911171, "optim/lr": 0.0029196269256715536, "optim/total_tokens": 5640290304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348781, "loss/out": 2.952319860458374, "created_at": "2025-01-16T11:56:50.450621+00:00"} {"global_step": 10759, "acc_step": 0, "speed/wps": 12904.86825976503, "speed/FLOPS": 202688689166899.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05144265294075012, "optim/lr": 0.0029195992535613795, "optim/total_tokens": 5640814592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.892082691192627, "created_at": "2025-01-16T11:57:00.613513+00:00"} {"global_step": 10760, "acc_step": 0, "speed/wps": 12907.17342766531, "speed/FLOPS": 202724894996406.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062070246785879135, "optim/lr": 0.0029195715768195133, "optim/total_tokens": 5641338880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.7999467849731445, "created_at": "2025-01-16T11:57:10.774554+00:00"} {"global_step": 10761, "acc_step": 0, "speed/wps": 12902.589491478195, "speed/FLOPS": 202652897979599.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04644836485385895, "optim/lr": 0.002919543895446046, "optim/total_tokens": 5641863168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 2.9199647903442383, "created_at": "2025-01-16T11:57:20.939470+00:00"} {"global_step": 10762, "acc_step": 0, "speed/wps": 12903.698797160745, "speed/FLOPS": 202670321149689.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05933481454849243, "optim/lr": 0.0029195162094410665, "optim/total_tokens": 5642387456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8125641345977783, "created_at": "2025-01-16T11:57:31.102627+00:00"} {"global_step": 10763, "acc_step": 0, "speed/wps": 12908.510996009754, "speed/FLOPS": 202745903345267.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050069428980350494, "optim/lr": 0.002919488518804667, "optim/total_tokens": 5642911744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.9712047576904297, "created_at": "2025-01-16T11:57:41.260791+00:00"} {"global_step": 10764, "acc_step": 0, "speed/wps": 12901.380160060682, "speed/FLOPS": 202633903767891.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.046028170734643936, "optim/lr": 0.002919460823536936, "optim/total_tokens": 5643436032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480466, "loss/out": 2.9262726306915283, "created_at": "2025-01-16T11:57:51.426901+00:00"} {"global_step": 10765, "acc_step": 0, "speed/wps": 12896.849868906336, "speed/FLOPS": 202562749320040.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06507943570613861, "optim/lr": 0.002919433123637965, "optim/total_tokens": 5643960320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.813253879547119, "created_at": "2025-01-16T11:58:01.594439+00:00"} {"global_step": 10766, "acc_step": 0, "speed/wps": 12902.674048480994, "speed/FLOPS": 202654226063521.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04850883409380913, "optim/lr": 0.0029194054191078443, "optim/total_tokens": 5644484608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 3.0255777835845947, "created_at": "2025-01-16T11:58:11.756868+00:00"} {"global_step": 10767, "acc_step": 0, "speed/wps": 12902.99450643124, "speed/FLOPS": 202659259295985.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06981881707906723, "optim/lr": 0.002919377709946663, "optim/total_tokens": 5645008896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0185165405273438, "created_at": "2025-01-16T11:58:21.921825+00:00"} {"global_step": 10768, "acc_step": 0, "speed/wps": 12903.36705358721, "speed/FLOPS": 202665110661003.62, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0486312136054039, "optim/lr": 0.002919349996154514, "optim/total_tokens": 5645533184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.9232840538024902, "created_at": "2025-01-16T11:58:32.085908+00:00"} {"global_step": 10769, "acc_step": 0, "speed/wps": 12899.624982523166, "speed/FLOPS": 202606336292798.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05525606498122215, "optim/lr": 0.002919322277731485, "optim/total_tokens": 5646057472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.768277645111084, "created_at": "2025-01-16T11:58:42.255249+00:00"} {"global_step": 10770, "acc_step": 0, "speed/wps": 12905.470907003082, "speed/FLOPS": 202698154569897.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05185804143548012, "optim/lr": 0.002919294554677668, "optim/total_tokens": 5646581760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498768, "loss/out": 2.897765636444092, "created_at": "2025-01-16T11:58:52.415223+00:00"} {"global_step": 10771, "acc_step": 0, "speed/wps": 12902.016531444187, "speed/FLOPS": 202643898854935.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04831398278474808, "optim/lr": 0.002919266826993154, "optim/total_tokens": 5647106048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 2.7549095153808594, "created_at": "2025-01-16T11:59:02.584060+00:00"} {"global_step": 10772, "acc_step": 0, "speed/wps": 12905.503765851665, "speed/FLOPS": 202698670663267.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04511946439743042, "optim/lr": 0.002919239094678032, "optim/total_tokens": 5647630336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.913743495941162, "created_at": "2025-01-16T11:59:12.746437+00:00"} {"global_step": 10773, "acc_step": 0, "speed/wps": 12908.545791942393, "speed/FLOPS": 202746449863204.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0589144267141819, "optim/lr": 0.002919211357732393, "optim/total_tokens": 5648154624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.8949804306030273, "created_at": "2025-01-16T11:59:22.904611+00:00"} {"global_step": 10774, "acc_step": 0, "speed/wps": 12900.501474533834, "speed/FLOPS": 202620102804252.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06402277946472168, "optim/lr": 0.0029191836161563275, "optim/total_tokens": 5648678912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.975428581237793, "created_at": "2025-01-16T11:59:33.069096+00:00"} {"global_step": 10775, "acc_step": 0, "speed/wps": 12896.227839133007, "speed/FLOPS": 202552979487691.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05931666120886803, "optim/lr": 0.0029191558699499264, "optim/total_tokens": 5649203200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8369970321655273, "created_at": "2025-01-16T11:59:43.240041+00:00"} {"global_step": 10776, "acc_step": 0, "speed/wps": 12903.798388275321, "speed/FLOPS": 202671885364996.38, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04769742488861084, "optim/lr": 0.00291912811911328, "optim/total_tokens": 5649727488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.815718173980713, "created_at": "2025-01-16T11:59:53.403258+00:00"} {"global_step": 10777, "acc_step": 0, "speed/wps": 12895.104007654943, "speed/FLOPS": 202535328170022.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06835831701755524, "optim/lr": 0.0029191003636464784, "optim/total_tokens": 5650251776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.833582639694214, "created_at": "2025-01-16T12:00:03.571900+00:00"} {"global_step": 10778, "acc_step": 0, "speed/wps": 12898.6721889068, "speed/FLOPS": 202591371359777.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05367710441350937, "optim/lr": 0.0029190726035496126, "optim/total_tokens": 5650776064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.887871742248535, "created_at": "2025-01-16T12:00:13.738435+00:00"} {"global_step": 10779, "acc_step": 0, "speed/wps": 12899.831571135046, "speed/FLOPS": 202609581050830.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07949250936508179, "optim/lr": 0.0029190448388227734, "optim/total_tokens": 5651300352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 2.8242030143737793, "created_at": "2025-01-16T12:00:23.905750+00:00"} {"global_step": 10780, "acc_step": 0, "speed/wps": 12900.703282999064, "speed/FLOPS": 202623272483512.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05763580650091171, "optim/lr": 0.0029190170694660513, "optim/total_tokens": 5651824640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 2.881352186203003, "created_at": "2025-01-16T12:00:34.069449+00:00"} {"global_step": 10781, "acc_step": 0, "speed/wps": 12895.672642416628, "speed/FLOPS": 202544259360340.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0750027447938919, "optim/lr": 0.002918989295479536, "optim/total_tokens": 5652348928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.047025203704834, "created_at": "2025-01-16T12:00:44.237896+00:00"} {"global_step": 10782, "acc_step": 0, "speed/wps": 12898.899147225755, "speed/FLOPS": 202594936052050.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08078739792108536, "optim/lr": 0.0029189615168633194, "optim/total_tokens": 5652873216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.921189546585083, "created_at": "2025-01-16T12:00:54.404993+00:00"} {"global_step": 10783, "acc_step": 0, "speed/wps": 12902.40477537183, "speed/FLOPS": 202649996759322.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04366707429289818, "optim/lr": 0.0029189337336174913, "optim/total_tokens": 5653397504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8197546005249023, "created_at": "2025-01-16T12:01:04.567875+00:00"} {"global_step": 10784, "acc_step": 0, "speed/wps": 12907.104422519229, "speed/FLOPS": 202723811175764.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08077830821275711, "optim/lr": 0.002918905945742143, "optim/total_tokens": 5653921792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376816, "loss/out": 2.9176673889160156, "created_at": "2025-01-16T12:01:14.731949+00:00"} {"global_step": 10785, "acc_step": 0, "speed/wps": 12902.886565492649, "speed/FLOPS": 202657563935218.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06988869607448578, "optim/lr": 0.002918878153237364, "optim/total_tokens": 5654446080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.814183235168457, "created_at": "2025-01-16T12:01:24.893929+00:00"} {"global_step": 10786, "acc_step": 0, "speed/wps": 12909.450910836518, "speed/FLOPS": 202760665999199.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.049831241369247437, "optim/lr": 0.002918850356103246, "optim/total_tokens": 5654970368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.931661367416382, "created_at": "2025-01-16T12:01:35.057398+00:00"} {"global_step": 10787, "acc_step": 0, "speed/wps": 12908.322926702192, "speed/FLOPS": 202742949458357.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053211480379104614, "optim/lr": 0.00291882255433988, "optim/total_tokens": 5655494656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 2.852578639984131, "created_at": "2025-01-16T12:01:45.216353+00:00"} {"global_step": 10788, "acc_step": 0, "speed/wps": 12887.836240017234, "speed/FLOPS": 202421177892312.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06087828800082207, "optim/lr": 0.0029187947479473557, "optim/total_tokens": 5656018944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.889639139175415, "created_at": "2025-01-16T12:01:55.392398+00:00"} {"global_step": 10789, "acc_step": 0, "speed/wps": 12908.214947503579, "speed/FLOPS": 202741253496664.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06614264100790024, "optim/lr": 0.0029187669369257646, "optim/total_tokens": 5656543232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7805731296539307, "created_at": "2025-01-16T12:02:05.550752+00:00"} {"global_step": 10790, "acc_step": 0, "speed/wps": 12899.19748896342, "speed/FLOPS": 202599621919004.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06402697414159775, "optim/lr": 0.0029187391212751965, "optim/total_tokens": 5657067520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 2.928236961364746, "created_at": "2025-01-16T12:02:15.731049+00:00"} {"global_step": 10791, "acc_step": 0, "speed/wps": 12896.461564495357, "speed/FLOPS": 202556650465679.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0519951730966568, "optim/lr": 0.0029187113009957436, "optim/total_tokens": 5657591808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.8968324661254883, "created_at": "2025-01-16T12:02:25.900593+00:00"} {"global_step": 10792, "acc_step": 0, "speed/wps": 12901.600635463481, "speed/FLOPS": 202637366637057.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05772639438509941, "optim/lr": 0.002918683476087495, "optim/total_tokens": 5658116096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9676241874694824, "created_at": "2025-01-16T12:02:36.067533+00:00"} {"global_step": 10793, "acc_step": 0, "speed/wps": 12904.284201237951, "speed/FLOPS": 202679515725151.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055541347712278366, "optim/lr": 0.0029186556465505427, "optim/total_tokens": 5658640384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356561, "loss/out": 2.934082508087158, "created_at": "2025-01-16T12:02:46.231011+00:00"} {"global_step": 10794, "acc_step": 0, "speed/wps": 12905.920293420508, "speed/FLOPS": 202705212801104.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04982617869973183, "optim/lr": 0.0029186278123849766, "optim/total_tokens": 5659164672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.9554643630981445, "created_at": "2025-01-16T12:02:56.391559+00:00"} {"global_step": 10795, "acc_step": 0, "speed/wps": 12902.759742195318, "speed/FLOPS": 202655572001060.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051071811467409134, "optim/lr": 0.0029185999735908885, "optim/total_tokens": 5659688960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 2.769672155380249, "created_at": "2025-01-16T12:03:06.554128+00:00"} {"global_step": 10796, "acc_step": 0, "speed/wps": 12905.433440119947, "speed/FLOPS": 202697566101012.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05350799486041069, "optim/lr": 0.002918572130168369, "optim/total_tokens": 5660213248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.8556504249572754, "created_at": "2025-01-16T12:03:16.719331+00:00"} {"global_step": 10797, "acc_step": 0, "speed/wps": 12909.499776045717, "speed/FLOPS": 202761433494458.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05638451129198074, "optim/lr": 0.0029185442821175085, "optim/total_tokens": 5660737536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9056005477905273, "created_at": "2025-01-16T12:03:26.877448+00:00"} {"global_step": 10798, "acc_step": 0, "speed/wps": 12904.723309861974, "speed/FLOPS": 202686412529490.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053437601774930954, "optim/lr": 0.002918516429438398, "optim/total_tokens": 5661261824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 2.8755829334259033, "created_at": "2025-01-16T12:03:37.039715+00:00"} {"global_step": 10799, "acc_step": 0, "speed/wps": 12904.855533313113, "speed/FLOPS": 202688489280484.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0604960061609745, "optim/lr": 0.0029184885721311284, "optim/total_tokens": 5661786112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.9633891582489014, "created_at": "2025-01-16T12:03:47.201447+00:00"} {"global_step": 10800, "acc_step": 0, "speed/wps": 12903.972007357914, "speed/FLOPS": 202674612291266.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.047558803111314774, "optim/lr": 0.00291846071019579, "optim/total_tokens": 5662310400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.89619779586792, "created_at": "2025-01-16T12:03:57.365665+00:00"} {"global_step": 10801, "acc_step": 0, "speed/wps": 12907.87491433512, "speed/FLOPS": 202735912808499.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04936382174491882, "optim/lr": 0.0029184328436324754, "optim/total_tokens": 5662834688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.722827672958374, "created_at": "2025-01-16T12:04:07.527188+00:00"} {"global_step": 10802, "acc_step": 0, "speed/wps": 12904.893834625247, "speed/FLOPS": 202689090855223.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04401126503944397, "optim/lr": 0.002918404972441274, "optim/total_tokens": 5663358976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471481, "loss/out": 2.8648765087127686, "created_at": "2025-01-16T12:04:17.687496+00:00"} {"global_step": 10803, "acc_step": 0, "speed/wps": 12910.811259354914, "speed/FLOPS": 202782032141996.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06436186283826828, "optim/lr": 0.0029183770966222772, "optim/total_tokens": 5663883264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9161150455474854, "created_at": "2025-01-16T12:04:27.843696+00:00"} {"global_step": 10804, "acc_step": 0, "speed/wps": 12905.95953128965, "speed/FLOPS": 202705829085758.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048322755843400955, "optim/lr": 0.002918349216175576, "optim/total_tokens": 5664407552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.954101800918579, "created_at": "2025-01-16T12:04:38.006317+00:00"} {"global_step": 10805, "acc_step": 0, "speed/wps": 12903.507915136, "speed/FLOPS": 202667323085191.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08469640463590622, "optim/lr": 0.002918321331101261, "optim/total_tokens": 5664931840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360518, "loss/out": 2.816908597946167, "created_at": "2025-01-16T12:04:48.167876+00:00"} {"global_step": 10806, "acc_step": 0, "speed/wps": 12899.011751572272, "speed/FLOPS": 202596704658046.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05150477960705757, "optim/lr": 0.0029182934413994237, "optim/total_tokens": 5665456128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.7518157958984375, "created_at": "2025-01-16T12:04:58.334117+00:00"} {"global_step": 10807, "acc_step": 0, "speed/wps": 12902.948808725314, "speed/FLOPS": 202658541550718.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05622624233365059, "optim/lr": 0.002918265547070155, "optim/total_tokens": 5665980416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.8443121910095215, "created_at": "2025-01-16T12:05:08.496139+00:00"} {"global_step": 10808, "acc_step": 0, "speed/wps": 12901.538266861704, "speed/FLOPS": 202636387052463.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04976159334182739, "optim/lr": 0.002918237648113545, "optim/total_tokens": 5666504704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.7538065910339355, "created_at": "2025-01-16T12:05:18.659556+00:00"} {"global_step": 10809, "acc_step": 0, "speed/wps": 12907.749151932241, "speed/FLOPS": 202733937537146.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05687984824180603, "optim/lr": 0.0029182097445296863, "optim/total_tokens": 5667028992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.0028202533721924, "created_at": "2025-01-16T12:05:28.820734+00:00"} {"global_step": 10810, "acc_step": 0, "speed/wps": 12901.711899641745, "speed/FLOPS": 202639114193870.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07003438472747803, "optim/lr": 0.0029181818363186694, "optim/total_tokens": 5667553280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 2.876023292541504, "created_at": "2025-01-16T12:05:38.987632+00:00"} {"global_step": 10811, "acc_step": 0, "speed/wps": 12907.111807733245, "speed/FLOPS": 202723927170698.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04962639883160591, "optim/lr": 0.0029181539234805847, "optim/total_tokens": 5668077568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.913217067718506, "created_at": "2025-01-16T12:05:49.148581+00:00"} {"global_step": 10812, "acc_step": 0, "speed/wps": 12904.831624690985, "speed/FLOPS": 202688113762720.0, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06751968711614609, "optim/lr": 0.002918126006015523, "optim/total_tokens": 5668601856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9817886352539062, "created_at": "2025-01-16T12:05:59.311814+00:00"} {"global_step": 10813, "acc_step": 0, "speed/wps": 12903.963400002498, "speed/FLOPS": 202674477100921.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07441262155771255, "optim/lr": 0.002918098083923577, "optim/total_tokens": 5669126144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 2.866001605987549, "created_at": "2025-01-16T12:06:09.476022+00:00"} {"global_step": 10814, "acc_step": 0, "speed/wps": 12901.919269494529, "speed/FLOPS": 202642371222364.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06085243821144104, "optim/lr": 0.002918070157204837, "optim/total_tokens": 5669650432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.881976842880249, "created_at": "2025-01-16T12:06:19.645291+00:00"} {"global_step": 10815, "acc_step": 0, "speed/wps": 12902.891630858237, "speed/FLOPS": 202657643493745.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04592274874448776, "optim/lr": 0.0029180422258593933, "optim/total_tokens": 5670174720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434219, "loss/out": 2.9922003746032715, "created_at": "2025-01-16T12:06:29.809175+00:00"} {"global_step": 10816, "acc_step": 0, "speed/wps": 12899.732917374924, "speed/FLOPS": 202608031557963.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047788072377443314, "optim/lr": 0.0029180142898873377, "optim/total_tokens": 5670699008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.892606258392334, "created_at": "2025-01-16T12:06:39.977671+00:00"} {"global_step": 10817, "acc_step": 0, "speed/wps": 12900.210551484226, "speed/FLOPS": 202615533457987.84, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046054694801568985, "optim/lr": 0.0029179863492887616, "optim/total_tokens": 5671223296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8357441425323486, "created_at": "2025-01-16T12:06:50.143682+00:00"} {"global_step": 10818, "acc_step": 0, "speed/wps": 12900.364229606123, "speed/FLOPS": 202617947184068.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04846953973174095, "optim/lr": 0.0029179584040637556, "optim/total_tokens": 5671747584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.900852918624878, "created_at": "2025-01-16T12:07:00.310674+00:00"} {"global_step": 10819, "acc_step": 0, "speed/wps": 12889.627833383938, "speed/FLOPS": 202449317328044.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04781917482614517, "optim/lr": 0.0029179304542124116, "optim/total_tokens": 5672271872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.892404794692993, "created_at": "2025-01-16T12:07:10.485087+00:00"} {"global_step": 10820, "acc_step": 0, "speed/wps": 12900.476490593366, "speed/FLOPS": 202619710397135.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057386089116334915, "optim/lr": 0.0029179024997348204, "optim/total_tokens": 5672796160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9426746368408203, "created_at": "2025-01-16T12:07:20.648827+00:00"} {"global_step": 10821, "acc_step": 0, "speed/wps": 12904.166120173913, "speed/FLOPS": 202677661099780.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046725936233997345, "optim/lr": 0.002917874540631073, "optim/total_tokens": 5673320448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317957, "loss/out": 2.8995156288146973, "created_at": "2025-01-16T12:07:30.809973+00:00"} {"global_step": 10822, "acc_step": 0, "speed/wps": 12905.182255493502, "speed/FLOPS": 202693620901290.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06222650408744812, "optim/lr": 0.002917846576901261, "optim/total_tokens": 5673844736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9252991676330566, "created_at": "2025-01-16T12:07:40.975421+00:00"} {"global_step": 10823, "acc_step": 0, "speed/wps": 12904.247594326996, "speed/FLOPS": 202678940763311.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05086198076605797, "optim/lr": 0.002917818608545475, "optim/total_tokens": 5674369024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.993335723876953, "created_at": "2025-01-16T12:07:51.138019+00:00"} {"global_step": 10824, "acc_step": 0, "speed/wps": 12905.65273216652, "speed/FLOPS": 202701010383943.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06766454130411148, "optim/lr": 0.002917790635563807, "optim/total_tokens": 5674893312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 2.8573131561279297, "created_at": "2025-01-16T12:08:01.302507+00:00"} {"global_step": 10825, "acc_step": 0, "speed/wps": 12907.071251305564, "speed/FLOPS": 202723290176271.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056189704686403275, "optim/lr": 0.002917762657956348, "optim/total_tokens": 5675417600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.895388126373291, "created_at": "2025-01-16T12:08:11.461029+00:00"} {"global_step": 10826, "acc_step": 0, "speed/wps": 12899.229081581036, "speed/FLOPS": 202600118124477.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05223485082387924, "optim/lr": 0.002917734675723189, "optim/total_tokens": 5675941888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8191580772399902, "created_at": "2025-01-16T12:08:21.625915+00:00"} {"global_step": 10827, "acc_step": 0, "speed/wps": 12907.551427259828, "speed/FLOPS": 202730831999464.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0535309836268425, "optim/lr": 0.0029177066888644216, "optim/total_tokens": 5676466176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.0122604370117188, "created_at": "2025-01-16T12:08:31.785066+00:00"} {"global_step": 10828, "acc_step": 0, "speed/wps": 12902.602815552747, "speed/FLOPS": 202653107252499.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04933542013168335, "optim/lr": 0.0029176786973801375, "optim/total_tokens": 5676990464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9386074542999268, "created_at": "2025-01-16T12:08:41.947128+00:00"} {"global_step": 10829, "acc_step": 0, "speed/wps": 12905.29836825005, "speed/FLOPS": 202695444611688.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0585719496011734, "optim/lr": 0.002917650701270427, "optim/total_tokens": 5677514752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.8640952110290527, "created_at": "2025-01-16T12:08:52.109669+00:00"} {"global_step": 10830, "acc_step": 0, "speed/wps": 12909.354480518003, "speed/FLOPS": 202759151428538.5, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05632612109184265, "optim/lr": 0.0029176227005353822, "optim/total_tokens": 5678039040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8536198139190674, "created_at": "2025-01-16T12:09:02.267815+00:00"} {"global_step": 10831, "acc_step": 0, "speed/wps": 12906.869242378063, "speed/FLOPS": 202720117348475.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04412119835615158, "optim/lr": 0.0029175946951750945, "optim/total_tokens": 5678563328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344503, "loss/out": 2.948697090148926, "created_at": "2025-01-16T12:09:12.426602+00:00"} {"global_step": 10832, "acc_step": 0, "speed/wps": 12898.089810500078, "speed/FLOPS": 202582224306628.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04525811970233917, "optim/lr": 0.002917566685189655, "optim/total_tokens": 5679087616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.7954368591308594, "created_at": "2025-01-16T12:09:22.593969+00:00"} {"global_step": 10833, "acc_step": 0, "speed/wps": 12911.037499712114, "speed/FLOPS": 202785585557692.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04242734611034393, "optim/lr": 0.002917538670579155, "optim/total_tokens": 5679611904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.0310816764831543, "created_at": "2025-01-16T12:09:32.749823+00:00"} {"global_step": 10834, "acc_step": 0, "speed/wps": 12905.859335188148, "speed/FLOPS": 202704255368300.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046407073736190796, "optim/lr": 0.0029175106513436863, "optim/total_tokens": 5680136192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.9820871353149414, "created_at": "2025-01-16T12:09:42.909336+00:00"} {"global_step": 10835, "acc_step": 0, "speed/wps": 12905.645231460643, "speed/FLOPS": 202700892575050.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051970332860946655, "optim/lr": 0.0029174826274833395, "optim/total_tokens": 5680660480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.9236080646514893, "created_at": "2025-01-16T12:09:53.071259+00:00"} {"global_step": 10836, "acc_step": 0, "speed/wps": 12902.209364762697, "speed/FLOPS": 202646927567183.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04355750232934952, "optim/lr": 0.002917454598998207, "optim/total_tokens": 5681184768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328606, "loss/out": 2.8586859703063965, "created_at": "2025-01-16T12:10:03.235111+00:00"} {"global_step": 10837, "acc_step": 0, "speed/wps": 12904.84799330156, "speed/FLOPS": 202688370854241.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05216902121901512, "optim/lr": 0.00291742656588838, "optim/total_tokens": 5681709056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.875760078430176, "created_at": "2025-01-16T12:10:13.396804+00:00"} {"global_step": 10838, "acc_step": 0, "speed/wps": 12903.70091962873, "speed/FLOPS": 202670354485965.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04841884225606918, "optim/lr": 0.0029173985281539497, "optim/total_tokens": 5682233344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462696, "loss/out": 2.9112348556518555, "created_at": "2025-01-16T12:10:23.558011+00:00"} {"global_step": 10839, "acc_step": 0, "speed/wps": 12906.961126892378, "speed/FLOPS": 202721560521035.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04386405646800995, "optim/lr": 0.0029173704857950074, "optim/total_tokens": 5682757632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.917219638824463, "created_at": "2025-01-16T12:10:33.722523+00:00"} {"global_step": 10840, "acc_step": 0, "speed/wps": 12899.431181285423, "speed/FLOPS": 202603292378048.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04746011272072792, "optim/lr": 0.002917342438811645, "optim/total_tokens": 5683281920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8783626556396484, "created_at": "2025-01-16T12:10:43.887983+00:00"} {"global_step": 10841, "acc_step": 0, "speed/wps": 12904.10394328521, "speed/FLOPS": 202676684526303.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04309920221567154, "optim/lr": 0.002917314387203954, "optim/total_tokens": 5683806208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.8805699348449707, "created_at": "2025-01-16T12:10:54.053933+00:00"} {"global_step": 10842, "acc_step": 0, "speed/wps": 12903.674206633583, "speed/FLOPS": 202669934921669.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05304362252354622, "optim/lr": 0.0029172863309720257, "optim/total_tokens": 5684330496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.866562604904175, "created_at": "2025-01-16T12:11:04.216808+00:00"} {"global_step": 10843, "acc_step": 0, "speed/wps": 12904.918743379867, "speed/FLOPS": 202689482081442.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04178865626454353, "optim/lr": 0.002917258270115952, "optim/total_tokens": 5684854784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372985, "loss/out": 2.8270678520202637, "created_at": "2025-01-16T12:11:14.377321+00:00"} {"global_step": 10844, "acc_step": 0, "speed/wps": 12902.065888020948, "speed/FLOPS": 202644674067796.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05383634567260742, "optim/lr": 0.002917230204635824, "optim/total_tokens": 5685379072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.0297813415527344, "created_at": "2025-01-16T12:11:24.540664+00:00"} {"global_step": 10845, "acc_step": 0, "speed/wps": 12905.300481777193, "speed/FLOPS": 202695477807536.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.060600731521844864, "optim/lr": 0.0029172021345317332, "optim/total_tokens": 5685903360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.833587408065796, "created_at": "2025-01-16T12:11:34.701709+00:00"} {"global_step": 10846, "acc_step": 0, "speed/wps": 12903.0508559456, "speed/FLOPS": 202660144342541.03, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07196429371833801, "optim/lr": 0.002917174059803772, "optim/total_tokens": 5686427648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.747663974761963, "created_at": "2025-01-16T12:11:44.866023+00:00"} {"global_step": 10847, "acc_step": 0, "speed/wps": 12904.987166626186, "speed/FLOPS": 202690556762549.34, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04665952920913696, "optim/lr": 0.002917145980452031, "optim/total_tokens": 5686951936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.9877102375030518, "created_at": "2025-01-16T12:11:55.029299+00:00"} {"global_step": 10848, "acc_step": 0, "speed/wps": 12905.807007454343, "speed/FLOPS": 202703433489334.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05289139598608017, "optim/lr": 0.0029171178964766025, "optim/total_tokens": 5687476224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 2.8436217308044434, "created_at": "2025-01-16T12:12:05.191653+00:00"} {"global_step": 10849, "acc_step": 0, "speed/wps": 12899.875505100596, "speed/FLOPS": 202610271094130.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04871818423271179, "optim/lr": 0.0029170898078775776, "optim/total_tokens": 5688000512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.7829644680023193, "created_at": "2025-01-16T12:12:15.356983+00:00"} {"global_step": 10850, "acc_step": 0, "speed/wps": 12900.38037285609, "speed/FLOPS": 202618200735991.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.049568355083465576, "optim/lr": 0.0029170617146550486, "optim/total_tokens": 5688524800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.792456865310669, "created_at": "2025-01-16T12:12:25.523659+00:00"} {"global_step": 10851, "acc_step": 0, "speed/wps": 12906.699889291664, "speed/FLOPS": 202717457425538.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047863252460956573, "optim/lr": 0.0029170336168091067, "optim/total_tokens": 5689049088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421736, "loss/out": 2.8527426719665527, "created_at": "2025-01-16T12:12:35.684973+00:00"} {"global_step": 10852, "acc_step": 0, "speed/wps": 12901.846260125974, "speed/FLOPS": 202641224509905.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047919344156980515, "optim/lr": 0.002917005514339843, "optim/total_tokens": 5689573376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.788357973098755, "created_at": "2025-01-16T12:12:45.852238+00:00"} {"global_step": 10853, "acc_step": 0, "speed/wps": 12907.144535258758, "speed/FLOPS": 202724441201459.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06074492260813713, "optim/lr": 0.0029169774072473507, "optim/total_tokens": 5690097664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.020927906036377, "created_at": "2025-01-16T12:12:56.014258+00:00"} {"global_step": 10854, "acc_step": 0, "speed/wps": 12911.360932939944, "speed/FLOPS": 202790665520979.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058656610548496246, "optim/lr": 0.002916949295531721, "optim/total_tokens": 5690621952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 2.9047834873199463, "created_at": "2025-01-16T12:13:06.173764+00:00"} {"global_step": 10855, "acc_step": 0, "speed/wps": 12906.331692221696, "speed/FLOPS": 202711674384598.62, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0556313656270504, "optim/lr": 0.0029169211791930444, "optim/total_tokens": 5691146240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8880703449249268, "created_at": "2025-01-16T12:13:16.336049+00:00"} {"global_step": 10856, "acc_step": 0, "speed/wps": 12905.897693349354, "speed/FLOPS": 202704857835931.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04432612285017967, "optim/lr": 0.0029168930582314143, "optim/total_tokens": 5691670528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 2.925313949584961, "created_at": "2025-01-16T12:13:26.495865+00:00"} {"global_step": 10857, "acc_step": 0, "speed/wps": 12903.438132877065, "speed/FLOPS": 202666227058923.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05702689662575722, "optim/lr": 0.002916864932646921, "optim/total_tokens": 5692194816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.875981092453003, "created_at": "2025-01-16T12:13:36.658627+00:00"} {"global_step": 10858, "acc_step": 0, "speed/wps": 12901.15493330479, "speed/FLOPS": 202630366272196.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07055532932281494, "optim/lr": 0.002916836802439657, "optim/total_tokens": 5692719104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.9536900520324707, "created_at": "2025-01-16T12:13:46.824781+00:00"} {"global_step": 10859, "acc_step": 0, "speed/wps": 12901.457706831845, "speed/FLOPS": 202635121746493.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05480637773871422, "optim/lr": 0.0029168086676097144, "optim/total_tokens": 5693243392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 2.874178171157837, "created_at": "2025-01-16T12:13:56.988539+00:00"} {"global_step": 10860, "acc_step": 0, "speed/wps": 12902.969193434703, "speed/FLOPS": 202658861720591.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06340202689170837, "optim/lr": 0.002916780528157184, "optim/total_tokens": 5693767680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9623355865478516, "created_at": "2025-01-16T12:14:07.154202+00:00"} {"global_step": 10861, "acc_step": 0, "speed/wps": 12904.01452963687, "speed/FLOPS": 202675280162088.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06401640176773071, "optim/lr": 0.0029167523840821588, "optim/total_tokens": 5694291968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 3.046586036682129, "created_at": "2025-01-16T12:14:17.315651+00:00"} {"global_step": 10862, "acc_step": 0, "speed/wps": 12906.634287411578, "speed/FLOPS": 202716427057865.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07477288693189621, "optim/lr": 0.00291672423538473, "optim/total_tokens": 5694816256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9589052200317383, "created_at": "2025-01-16T12:14:27.478081+00:00"} {"global_step": 10863, "acc_step": 0, "speed/wps": 12909.190960205022, "speed/FLOPS": 202756583117322.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08305039256811142, "optim/lr": 0.0029166960820649897, "optim/total_tokens": 5695340544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404090, "loss/out": 2.8734099864959717, "created_at": "2025-01-16T12:14:37.635542+00:00"} {"global_step": 10864, "acc_step": 0, "speed/wps": 12904.109925378421, "speed/FLOPS": 202676778483297.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04938944801688194, "optim/lr": 0.002916667924123029, "optim/total_tokens": 5695864832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.7991344928741455, "created_at": "2025-01-16T12:14:47.799537+00:00"} {"global_step": 10865, "acc_step": 0, "speed/wps": 12903.054430892555, "speed/FLOPS": 202660200491995.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061169181019067764, "optim/lr": 0.0029166397615589404, "optim/total_tokens": 5696389120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.8519206047058105, "created_at": "2025-01-16T12:14:57.969615+00:00"} {"global_step": 10866, "acc_step": 0, "speed/wps": 12903.62344999966, "speed/FLOPS": 202669137718986.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04921373724937439, "optim/lr": 0.0029166115943728157, "optim/total_tokens": 5696913408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383767, "loss/out": 2.7662718296051025, "created_at": "2025-01-16T12:15:08.132167+00:00"} {"global_step": 10867, "acc_step": 0, "speed/wps": 12904.431114898405, "speed/FLOPS": 202681823206069.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04744117707014084, "optim/lr": 0.0029165834225647466, "optim/total_tokens": 5697437696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.958094596862793, "created_at": "2025-01-16T12:15:18.293061+00:00"} {"global_step": 10868, "acc_step": 0, "speed/wps": 12902.550196027954, "speed/FLOPS": 202652280790555.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04965177923440933, "optim/lr": 0.002916555246134826, "optim/total_tokens": 5697961984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 2.8831236362457275, "created_at": "2025-01-16T12:15:28.457667+00:00"} {"global_step": 10869, "acc_step": 0, "speed/wps": 12903.00729545889, "speed/FLOPS": 202659460165238.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048590999096632004, "optim/lr": 0.0029165270650831445, "optim/total_tokens": 5698486272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.731070041656494, "created_at": "2025-01-16T12:15:38.619576+00:00"} {"global_step": 10870, "acc_step": 0, "speed/wps": 12902.61006786377, "speed/FLOPS": 202653221160009.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0437592975795269, "optim/lr": 0.002916498879409794, "optim/total_tokens": 5699010560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.9153244495391846, "created_at": "2025-01-16T12:15:48.782632+00:00"} {"global_step": 10871, "acc_step": 0, "speed/wps": 12900.100864551014, "speed/FLOPS": 202613810673976.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04450934752821922, "optim/lr": 0.002916470689114868, "optim/total_tokens": 5699534848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377626, "loss/out": 2.898874521255493, "created_at": "2025-01-16T12:15:58.947408+00:00"} {"global_step": 10872, "acc_step": 0, "speed/wps": 12892.177341282159, "speed/FLOPS": 202489360852976.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.042174939066171646, "optim/lr": 0.0029164424941984573, "optim/total_tokens": 5700059136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.7815370559692383, "created_at": "2025-01-16T12:16:09.122121+00:00"} {"global_step": 10873, "acc_step": 0, "speed/wps": 12902.616582057119, "speed/FLOPS": 202653323474367.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04330234229564667, "optim/lr": 0.002916414294660654, "optim/total_tokens": 5700583424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.959869861602783, "created_at": "2025-01-16T12:16:19.286597+00:00"} {"global_step": 10874, "acc_step": 0, "speed/wps": 12908.39422517191, "speed/FLOPS": 202744069298799.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045118823647499084, "optim/lr": 0.0029163860905015502, "optim/total_tokens": 5701107712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.8564560413360596, "created_at": "2025-01-16T12:16:29.448013+00:00"} {"global_step": 10875, "acc_step": 0, "speed/wps": 12902.620888975198, "speed/FLOPS": 202653391120434.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06327598541975021, "optim/lr": 0.0029163578817212378, "optim/total_tokens": 5701632000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.9354047775268555, "created_at": "2025-01-16T12:16:39.611036+00:00"} {"global_step": 10876, "acc_step": 0, "speed/wps": 12903.68762049721, "speed/FLOPS": 202670145604830.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0567723773419857, "optim/lr": 0.0029163296683198097, "optim/total_tokens": 5702156288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456054, "loss/out": 2.973637580871582, "created_at": "2025-01-16T12:16:49.777869+00:00"} {"global_step": 10877, "acc_step": 0, "speed/wps": 12908.74353642446, "speed/FLOPS": 202749555712024.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048038166016340256, "optim/lr": 0.0029163014502973563, "optim/total_tokens": 5702680576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.9297146797180176, "created_at": "2025-01-16T12:16:59.938512+00:00"} {"global_step": 10878, "acc_step": 0, "speed/wps": 12907.464537005526, "speed/FLOPS": 202729467268620.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0525844506919384, "optim/lr": 0.0029162732276539712, "optim/total_tokens": 5703204864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.956261396408081, "created_at": "2025-01-16T12:17:10.100370+00:00"} {"global_step": 10879, "acc_step": 0, "speed/wps": 12896.794589910898, "speed/FLOPS": 202561881087455.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050543513149023056, "optim/lr": 0.0029162450003897456, "optim/total_tokens": 5703729152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8909857273101807, "created_at": "2025-01-16T12:17:20.267475+00:00"} {"global_step": 10880, "acc_step": 0, "speed/wps": 12895.629346786622, "speed/FLOPS": 202543579342977.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04624491184949875, "optim/lr": 0.0029162167685047717, "optim/total_tokens": 5704253440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9250733852386475, "created_at": "2025-01-16T12:17:30.438124+00:00"} {"global_step": 10881, "acc_step": 0, "speed/wps": 12895.221831950184, "speed/FLOPS": 202537178762486.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052321068942546844, "optim/lr": 0.002916188531999142, "optim/total_tokens": 5704777728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8632442951202393, "created_at": "2025-01-16T12:17:40.606040+00:00"} {"global_step": 10882, "acc_step": 0, "speed/wps": 12899.177839714901, "speed/FLOPS": 202599313300555.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05186651274561882, "optim/lr": 0.0029161602908729487, "optim/total_tokens": 5705302016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288392, "loss/out": 2.821051836013794, "created_at": "2025-01-16T12:17:50.772098+00:00"} {"global_step": 10883, "acc_step": 0, "speed/wps": 12907.084734969643, "speed/FLOPS": 202723501955744.3, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05769436061382294, "optim/lr": 0.0029161320451262835, "optim/total_tokens": 5705826304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9342129230499268, "created_at": "2025-01-16T12:18:00.933268+00:00"} {"global_step": 10884, "acc_step": 0, "speed/wps": 12908.239758975504, "speed/FLOPS": 202741643194925.66, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06116612255573273, "optim/lr": 0.0029161037947592386, "optim/total_tokens": 5706350592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9268369674682617, "created_at": "2025-01-16T12:18:11.091886+00:00"} {"global_step": 10885, "acc_step": 0, "speed/wps": 12901.660777608684, "speed/FLOPS": 202638311252091.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05115801468491554, "optim/lr": 0.002916075539771906, "optim/total_tokens": 5706874880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 2.9079880714416504, "created_at": "2025-01-16T12:18:21.255649+00:00"} {"global_step": 10886, "acc_step": 0, "speed/wps": 12901.887790253359, "speed/FLOPS": 202641876797624.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048444438725709915, "optim/lr": 0.0029160472801643784, "optim/total_tokens": 5707399168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 2.856444835662842, "created_at": "2025-01-16T12:18:31.448701+00:00"} {"global_step": 10887, "acc_step": 0, "speed/wps": 12907.485448565421, "speed/FLOPS": 202729795713404.2, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05466296523809433, "optim/lr": 0.0029160190159367477, "optim/total_tokens": 5707923456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 2.8844997882843018, "created_at": "2025-01-16T12:18:41.608935+00:00"} {"global_step": 10888, "acc_step": 0, "speed/wps": 12905.468524784486, "speed/FLOPS": 202698117153880.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04697751626372337, "optim/lr": 0.0029159907470891065, "optim/total_tokens": 5708447744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.8670570850372314, "created_at": "2025-01-16T12:18:51.769790+00:00"} {"global_step": 10889, "acc_step": 0, "speed/wps": 12903.343600030441, "speed/FLOPS": 202664742290666.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05460433289408684, "optim/lr": 0.0029159624736215467, "optim/total_tokens": 5708972032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.909839153289795, "created_at": "2025-01-16T12:19:01.932276+00:00"} {"global_step": 10890, "acc_step": 0, "speed/wps": 12901.698515905435, "speed/FLOPS": 202638903983900.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05191909521818161, "optim/lr": 0.0029159341955341603, "optim/total_tokens": 5709496320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 2.833366870880127, "created_at": "2025-01-16T12:19:12.101060+00:00"} {"global_step": 10891, "acc_step": 0, "speed/wps": 12899.693932547187, "speed/FLOPS": 202607419247672.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0466763898730278, "optim/lr": 0.0029159059128270395, "optim/total_tokens": 5710020608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.9229140281677246, "created_at": "2025-01-16T12:19:22.268121+00:00"} {"global_step": 10892, "acc_step": 0, "speed/wps": 12902.394175816424, "speed/FLOPS": 202649830278739.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052968062460422516, "optim/lr": 0.0029158776255002778, "optim/total_tokens": 5710544896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.963405132293701, "created_at": "2025-01-16T12:19:32.433163+00:00"} {"global_step": 10893, "acc_step": 0, "speed/wps": 12909.136976143454, "speed/FLOPS": 202755735223453.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05249709635972977, "optim/lr": 0.0029158493335539664, "optim/total_tokens": 5711069184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 2.921996831893921, "created_at": "2025-01-16T12:19:42.593857+00:00"} {"global_step": 10894, "acc_step": 0, "speed/wps": 12908.67632009672, "speed/FLOPS": 202748499987232.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0536104291677475, "optim/lr": 0.0029158210369881967, "optim/total_tokens": 5711593472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.8645427227020264, "created_at": "2025-01-16T12:19:52.752981+00:00"} {"global_step": 10895, "acc_step": 0, "speed/wps": 12905.954720335812, "speed/FLOPS": 202705753523117.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06263332068920135, "optim/lr": 0.0029157927358030634, "optim/total_tokens": 5712117760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 2.9604291915893555, "created_at": "2025-01-16T12:20:02.914205+00:00"} {"global_step": 10896, "acc_step": 0, "speed/wps": 12902.568902146511, "speed/FLOPS": 202652574595852.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06709641963243484, "optim/lr": 0.002915764429998657, "optim/total_tokens": 5712642048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.862825393676758, "created_at": "2025-01-16T12:20:13.077105+00:00"} {"global_step": 10897, "acc_step": 0, "speed/wps": 12886.547420036537, "speed/FLOPS": 202400935203491.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06111765280365944, "optim/lr": 0.0029157361195750707, "optim/total_tokens": 5713166336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8823485374450684, "created_at": "2025-01-16T12:20:23.257312+00:00"} {"global_step": 10898, "acc_step": 0, "speed/wps": 12883.778206919507, "speed/FLOPS": 202357440906191.22, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.047801557928323746, "optim/lr": 0.0029157078045323965, "optim/total_tokens": 5713690624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.9856081008911133, "created_at": "2025-01-16T12:20:33.436508+00:00"} {"global_step": 10899, "acc_step": 0, "speed/wps": 12897.034730915793, "speed/FLOPS": 202565652831925.8, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060519468039274216, "optim/lr": 0.002915679484870727, "optim/total_tokens": 5714214912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296188, "loss/out": 2.8655967712402344, "created_at": "2025-01-16T12:20:43.608754+00:00"} {"global_step": 10900, "acc_step": 0, "speed/wps": 12894.686983518743, "speed/FLOPS": 202528778232912.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.043550245463848114, "optim/lr": 0.0029156511605901545, "optim/total_tokens": 5714739200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.943772315979004, "created_at": "2025-01-16T12:20:53.779432+00:00"} {"global_step": 10901, "acc_step": 0, "speed/wps": 12899.5289791543, "speed/FLOPS": 202604828427967.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05314628407359123, "optim/lr": 0.002915622831690771, "optim/total_tokens": 5715263488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 2.8502774238586426, "created_at": "2025-01-16T12:21:03.950346+00:00"} {"global_step": 10902, "acc_step": 0, "speed/wps": 12894.896336972426, "speed/FLOPS": 202532066416584.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.044366560876369476, "optim/lr": 0.0029155944981726694, "optim/total_tokens": 5715787776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 2.835214376449585, "created_at": "2025-01-16T12:21:14.124083+00:00"} {"global_step": 10903, "acc_step": 0, "speed/wps": 12894.893234409981, "speed/FLOPS": 202532017686577.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04799465835094452, "optim/lr": 0.002915566160035942, "optim/total_tokens": 5716312064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.9278247356414795, "created_at": "2025-01-16T12:21:24.295886+00:00"} {"global_step": 10904, "acc_step": 0, "speed/wps": 12899.281697666394, "speed/FLOPS": 202600944532399.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05247483775019646, "optim/lr": 0.0029155378172806817, "optim/total_tokens": 5716836352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.853713035583496, "created_at": "2025-01-16T12:21:34.460829+00:00"} {"global_step": 10905, "acc_step": 0, "speed/wps": 12896.975266550793, "speed/FLOPS": 202564718862360.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045912567526102066, "optim/lr": 0.00291550946990698, "optim/total_tokens": 5717360640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 2.8777694702148438, "created_at": "2025-01-16T12:21:44.629972+00:00"} {"global_step": 10906, "acc_step": 0, "speed/wps": 12900.226477960183, "speed/FLOPS": 202615783605178.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04423099011182785, "optim/lr": 0.00291548111791493, "optim/total_tokens": 5717884928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.9552459716796875, "created_at": "2025-01-16T12:21:54.794036+00:00"} {"global_step": 10907, "acc_step": 0, "speed/wps": 12899.248414135587, "speed/FLOPS": 202600421768812.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0472874641418457, "optim/lr": 0.0029154527613046246, "optim/total_tokens": 5718409216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.925628900527954, "created_at": "2025-01-16T12:22:04.959184+00:00"} {"global_step": 10908, "acc_step": 0, "speed/wps": 12903.458826725157, "speed/FLOPS": 202666552084244.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04529844596982002, "optim/lr": 0.0029154244000761565, "optim/total_tokens": 5718933504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8773672580718994, "created_at": "2025-01-16T12:22:15.126447+00:00"} {"global_step": 10909, "acc_step": 0, "speed/wps": 12901.448772862672, "speed/FLOPS": 202634981426230.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0418507345020771, "optim/lr": 0.0029153960342296168, "optim/total_tokens": 5719457792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8702759742736816, "created_at": "2025-01-16T12:22:25.289929+00:00"} {"global_step": 10910, "acc_step": 0, "speed/wps": 12904.627928716405, "speed/FLOPS": 202684914437531.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05320369824767113, "optim/lr": 0.0029153676637650987, "optim/total_tokens": 5719982080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 2.831143856048584, "created_at": "2025-01-16T12:22:35.451115+00:00"} {"global_step": 10911, "acc_step": 0, "speed/wps": 12903.80711845427, "speed/FLOPS": 202672022484453.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04117972031235695, "optim/lr": 0.0029153392886826952, "optim/total_tokens": 5720506368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.943326473236084, "created_at": "2025-01-16T12:22:45.614826+00:00"} {"global_step": 10912, "acc_step": 0, "speed/wps": 12903.36855485435, "speed/FLOPS": 202665134240467.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04988490417599678, "optim/lr": 0.0029153109089824986, "optim/total_tokens": 5721030656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9458632469177246, "created_at": "2025-01-16T12:22:55.777399+00:00"} {"global_step": 10913, "acc_step": 0, "speed/wps": 12903.967231013705, "speed/FLOPS": 202674537272216.94, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046482279896736145, "optim/lr": 0.0029152825246646016, "optim/total_tokens": 5721554944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 2.8832974433898926, "created_at": "2025-01-16T12:23:05.944052+00:00"} {"global_step": 10914, "acc_step": 0, "speed/wps": 12896.787546228068, "speed/FLOPS": 202561770456737.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.058488812297582626, "optim/lr": 0.0029152541357290967, "optim/total_tokens": 5722079232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 2.823307514190674, "created_at": "2025-01-16T12:23:16.117510+00:00"} {"global_step": 10915, "acc_step": 0, "speed/wps": 12890.497202092884, "speed/FLOPS": 202462971958256.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.044070567935705185, "optim/lr": 0.0029152257421760767, "optim/total_tokens": 5722603520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.758974075317383, "created_at": "2025-01-16T12:23:26.294323+00:00"} {"global_step": 10916, "acc_step": 0, "speed/wps": 12897.849503726933, "speed/FLOPS": 202578449958539.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040881335735321045, "optim/lr": 0.0029151973440056335, "optim/total_tokens": 5723127808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.8447139263153076, "created_at": "2025-01-16T12:23:36.461690+00:00"} {"global_step": 10917, "acc_step": 0, "speed/wps": 12887.38544740279, "speed/FLOPS": 202414097574849.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.041303474456071854, "optim/lr": 0.002915168941217861, "optim/total_tokens": 5723652096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.732119083404541, "created_at": "2025-01-16T12:23:46.641470+00:00"} {"global_step": 10918, "acc_step": 0, "speed/wps": 12893.811419614785, "speed/FLOPS": 202515026298649.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04517717659473419, "optim/lr": 0.0029151405338128512, "optim/total_tokens": 5724176384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8171262741088867, "created_at": "2025-01-16T12:23:56.817557+00:00"} {"global_step": 10919, "acc_step": 0, "speed/wps": 12886.677353826472, "speed/FLOPS": 202402975992210.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1322161704301834, "optim/lr": 0.002915112121790696, "optim/total_tokens": 5724700672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.702789306640625, "created_at": "2025-01-16T12:24:06.997199+00:00"} {"global_step": 10920, "acc_step": 0, "speed/wps": 12894.134643630181, "speed/FLOPS": 202520102975967.72, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.059361882507801056, "optim/lr": 0.0029150837051514903, "optim/total_tokens": 5725224960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 2.9402709007263184, "created_at": "2025-01-16T12:24:17.170358+00:00"} {"global_step": 10921, "acc_step": 0, "speed/wps": 12893.889438902315, "speed/FLOPS": 202516251698770.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04942711442708969, "optim/lr": 0.0029150552838953245, "optim/total_tokens": 5725749248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.9236183166503906, "created_at": "2025-01-16T12:24:27.340093+00:00"} {"global_step": 10922, "acc_step": 0, "speed/wps": 12888.40623456347, "speed/FLOPS": 202430130439916.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04804337024688721, "optim/lr": 0.002915026858022292, "optim/total_tokens": 5726273536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.7556605339050293, "created_at": "2025-01-16T12:24:37.513774+00:00"} {"global_step": 10923, "acc_step": 0, "speed/wps": 12897.542165763902, "speed/FLOPS": 202573622793503.9, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05089474841952324, "optim/lr": 0.0029149984275324862, "optim/total_tokens": 5726797824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333025, "loss/out": 2.811927318572998, "created_at": "2025-01-16T12:24:47.680036+00:00"} {"global_step": 10924, "acc_step": 0, "speed/wps": 12897.5488143128, "speed/FLOPS": 202573727218100.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06519892811775208, "optim/lr": 0.0029149699924259993, "optim/total_tokens": 5727322112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.9054040908813477, "created_at": "2025-01-16T12:24:57.847268+00:00"} {"global_step": 10925, "acc_step": 0, "speed/wps": 12890.886907904332, "speed/FLOPS": 202469092823536.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05513713136315346, "optim/lr": 0.002914941552702924, "optim/total_tokens": 5727846400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.886380910873413, "created_at": "2025-01-16T12:25:08.028007+00:00"} {"global_step": 10926, "acc_step": 0, "speed/wps": 12898.54924550128, "speed/FLOPS": 202589440364652.03, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05136964097619057, "optim/lr": 0.0029149131083633538, "optim/total_tokens": 5728370688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.0321311950683594, "created_at": "2025-01-16T12:25:18.194442+00:00"} {"global_step": 10927, "acc_step": 0, "speed/wps": 12890.897970305714, "speed/FLOPS": 202469266573751.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055023882538080215, "optim/lr": 0.002914884659407381, "optim/total_tokens": 5728894976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.9645538330078125, "created_at": "2025-01-16T12:25:28.366105+00:00"} {"global_step": 10928, "acc_step": 0, "speed/wps": 12896.165566189688, "speed/FLOPS": 202552001405544.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05407961830496788, "optim/lr": 0.002914856205835098, "optim/total_tokens": 5729419264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8671727180480957, "created_at": "2025-01-16T12:25:38.537823+00:00"} {"global_step": 10929, "acc_step": 0, "speed/wps": 12893.739433997931, "speed/FLOPS": 202513895665618.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06417515873908997, "optim/lr": 0.0029148277476465983, "optim/total_tokens": 5729943552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.7219231128692627, "created_at": "2025-01-16T12:25:48.710421+00:00"} {"global_step": 10930, "acc_step": 0, "speed/wps": 12890.688401894118, "speed/FLOPS": 202465975013870.47, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0706833228468895, "optim/lr": 0.0029147992848419747, "optim/total_tokens": 5730467840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8969035148620605, "created_at": "2025-01-16T12:25:58.883389+00:00"} {"global_step": 10931, "acc_step": 0, "speed/wps": 12890.439822833074, "speed/FLOPS": 202462070738133.34, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05819292366504669, "optim/lr": 0.0029147708174213196, "optim/total_tokens": 5730992128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 3.028968572616577, "created_at": "2025-01-16T12:26:09.056711+00:00"} {"global_step": 10932, "acc_step": 0, "speed/wps": 12888.475419979268, "speed/FLOPS": 202431217091941.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.09096881002187729, "optim/lr": 0.0029147423453847257, "optim/total_tokens": 5731516416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8512659072875977, "created_at": "2025-01-16T12:26:19.233803+00:00"} {"global_step": 10933, "acc_step": 0, "speed/wps": 12900.890361664626, "speed/FLOPS": 202626210811028.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6654651761054993, "optim/lr": 0.0029147138687322873, "optim/total_tokens": 5732040704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 2.884847640991211, "created_at": "2025-01-16T12:26:29.401363+00:00"} {"global_step": 10934, "acc_step": 0, "speed/wps": 12894.815394371488, "speed/FLOPS": 202530795101809.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35594335198402405, "optim/lr": 0.0029146853874640957, "optim/total_tokens": 5732564992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8777356147766113, "created_at": "2025-01-16T12:26:39.571538+00:00"} {"global_step": 10935, "acc_step": 0, "speed/wps": 12900.66909458622, "speed/FLOPS": 202622735507508.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17222699522972107, "optim/lr": 0.002914656901580245, "optim/total_tokens": 5733089280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 2.9081337451934814, "created_at": "2025-01-16T12:26:49.736405+00:00"} {"global_step": 10936, "acc_step": 0, "speed/wps": 12898.07472265262, "speed/FLOPS": 202581987331251.1, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.12018046528100967, "optim/lr": 0.0029146284110808276, "optim/total_tokens": 5733613568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.958407402038574, "created_at": "2025-01-16T12:26:59.905885+00:00"} {"global_step": 10937, "acc_step": 0, "speed/wps": 12892.830317242278, "speed/FLOPS": 202499616737713.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.08097579330205917, "optim/lr": 0.002914599915965936, "optim/total_tokens": 5734137856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8778305053710938, "created_at": "2025-01-16T12:27:10.077489+00:00"} {"global_step": 10938, "acc_step": 0, "speed/wps": 12898.942966075989, "speed/FLOPS": 202595624287306.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07002224028110504, "optim/lr": 0.0029145714162356635, "optim/total_tokens": 5734662144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.964495897293091, "created_at": "2025-01-16T12:27:20.251899+00:00"} {"global_step": 10939, "acc_step": 0, "speed/wps": 12898.20374238256, "speed/FLOPS": 202584013763401.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0700278952717781, "optim/lr": 0.002914542911890104, "optim/total_tokens": 5735186432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 2.9019079208374023, "created_at": "2025-01-16T12:27:30.417429+00:00"} {"global_step": 10940, "acc_step": 0, "speed/wps": 12901.632552090092, "speed/FLOPS": 202637867931536.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.059142425656318665, "optim/lr": 0.002914514402929349, "optim/total_tokens": 5735710720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.9339656829833984, "created_at": "2025-01-16T12:27:40.580990+00:00"} {"global_step": 10941, "acc_step": 0, "speed/wps": 12901.820349615766, "speed/FLOPS": 202640817549738.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.000430941581726, "optim/lr": 0.002914485889353493, "optim/total_tokens": 5736235008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.87471866607666, "created_at": "2025-01-16T12:27:50.743694+00:00"} {"global_step": 10942, "acc_step": 0, "speed/wps": 12907.293569048345, "speed/FLOPS": 202726781981918.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06876132637262344, "optim/lr": 0.0029144573711626276, "optim/total_tokens": 5736759296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9032158851623535, "created_at": "2025-01-16T12:28:00.905869+00:00"} {"global_step": 10943, "acc_step": 0, "speed/wps": 12903.81191006042, "speed/FLOPS": 202672097743212.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08585082739591599, "optim/lr": 0.002914428848356847, "optim/total_tokens": 5737283584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.744311809539795, "created_at": "2025-01-16T12:28:11.070198+00:00"} {"global_step": 10944, "acc_step": 0, "speed/wps": 12900.055383621639, "speed/FLOPS": 202613096333482.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056610263884067535, "optim/lr": 0.002914400320936244, "optim/total_tokens": 5737807872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294366, "loss/out": 2.918642520904541, "created_at": "2025-01-16T12:28:21.234280+00:00"} {"global_step": 10945, "acc_step": 0, "speed/wps": 12905.575674604275, "speed/FLOPS": 202699800089039.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0654492899775505, "optim/lr": 0.002914371788900911, "optim/total_tokens": 5738332160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.05222487449646, "created_at": "2025-01-16T12:28:31.397579+00:00"} {"global_step": 10946, "acc_step": 0, "speed/wps": 12901.773156964566, "speed/FLOPS": 202640076324301.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05837498977780342, "optim/lr": 0.0029143432522509413, "optim/total_tokens": 5738856448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9003472328186035, "created_at": "2025-01-16T12:28:41.560351+00:00"} {"global_step": 10947, "acc_step": 0, "speed/wps": 12902.689684130619, "speed/FLOPS": 202654471642884.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061063770204782486, "optim/lr": 0.0029143147109864293, "optim/total_tokens": 5739380736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.7986278533935547, "created_at": "2025-01-16T12:28:51.725011+00:00"} {"global_step": 10948, "acc_step": 0, "speed/wps": 12904.032889788181, "speed/FLOPS": 202675568533494.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05689653009176254, "optim/lr": 0.002914286165107466, "optim/total_tokens": 5739905024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.809091091156006, "created_at": "2025-01-16T12:29:01.888335+00:00"} {"global_step": 10949, "acc_step": 0, "speed/wps": 12904.744279987797, "speed/FLOPS": 202686741894132.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0705915093421936, "optim/lr": 0.0029142576146141464, "optim/total_tokens": 5740429312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 2.9215497970581055, "created_at": "2025-01-16T12:29:12.048883+00:00"} {"global_step": 10950, "acc_step": 0, "speed/wps": 12901.865518916078, "speed/FLOPS": 202641526995668.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047277841717004776, "optim/lr": 0.0029142290595065624, "optim/total_tokens": 5740953600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9463329315185547, "created_at": "2025-01-16T12:29:22.214559+00:00"} {"global_step": 10951, "acc_step": 0, "speed/wps": 12897.487300427125, "speed/FLOPS": 202572761057996.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07074395567178726, "optim/lr": 0.0029142004997848078, "optim/total_tokens": 5741477888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9028244018554688, "created_at": "2025-01-16T12:29:32.382452+00:00"} {"global_step": 10952, "acc_step": 0, "speed/wps": 12905.65069899862, "speed/FLOPS": 202700978450247.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05884907767176628, "optim/lr": 0.0029141719354489753, "optim/total_tokens": 5742002176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418245, "loss/out": 2.7596821784973145, "created_at": "2025-01-16T12:29:42.546390+00:00"} {"global_step": 10953, "acc_step": 0, "speed/wps": 12904.028193853284, "speed/FLOPS": 202675494777384.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053444214165210724, "optim/lr": 0.002914143366499159, "optim/total_tokens": 5742526464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.993454933166504, "created_at": "2025-01-16T12:29:52.711249+00:00"} {"global_step": 10954, "acc_step": 0, "speed/wps": 12902.507012709722, "speed/FLOPS": 202651602537202.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04454771429300308, "optim/lr": 0.0029141147929354514, "optim/total_tokens": 5743050752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8641343116760254, "created_at": "2025-01-16T12:30:02.874687+00:00"} {"global_step": 10955, "acc_step": 0, "speed/wps": 12903.888963125391, "speed/FLOPS": 202673307967479.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05109395086765289, "optim/lr": 0.0029140862147579456, "optim/total_tokens": 5743575040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 2.819499969482422, "created_at": "2025-01-16T12:30:13.039564+00:00"} {"global_step": 10956, "acc_step": 0, "speed/wps": 12902.322326170215, "speed/FLOPS": 202648701781313.8, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04954847693443298, "optim/lr": 0.0029140576319667353, "optim/total_tokens": 5744099328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.83939790725708, "created_at": "2025-01-16T12:30:23.202076+00:00"} {"global_step": 10957, "acc_step": 0, "speed/wps": 12901.446427609113, "speed/FLOPS": 202634944590800.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051837287843227386, "optim/lr": 0.0029140290445619135, "optim/total_tokens": 5744623616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.817880630493164, "created_at": "2025-01-16T12:30:33.367920+00:00"} {"global_step": 10958, "acc_step": 0, "speed/wps": 12898.78418637316, "speed/FLOPS": 202593130433887.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0517866425216198, "optim/lr": 0.0029140004525435733, "optim/total_tokens": 5745147904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9350640773773193, "created_at": "2025-01-16T12:30:43.533953+00:00"} {"global_step": 10959, "acc_step": 0, "speed/wps": 12898.332040061201, "speed/FLOPS": 202586028854744.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0511862114071846, "optim/lr": 0.0029139718559118085, "optim/total_tokens": 5745672192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 2.897270679473877, "created_at": "2025-01-16T12:30:53.705231+00:00"} {"global_step": 10960, "acc_step": 0, "speed/wps": 12884.881505680882, "speed/FLOPS": 202374769729330.0, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050876863300800323, "optim/lr": 0.002913943254666712, "optim/total_tokens": 5746196480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.907947301864624, "created_at": "2025-01-16T12:31:03.884006+00:00"} {"global_step": 10961, "acc_step": 0, "speed/wps": 12885.647887106546, "speed/FLOPS": 202386806802738.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04857951030135155, "optim/lr": 0.0029139146488083778, "optim/total_tokens": 5746720768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.974221706390381, "created_at": "2025-01-16T12:31:14.092866+00:00"} {"global_step": 10962, "acc_step": 0, "speed/wps": 12894.946205020275, "speed/FLOPS": 202532849662801.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053988829255104065, "optim/lr": 0.002913886038336898, "optim/total_tokens": 5747245056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 2.9262406826019287, "created_at": "2025-01-16T12:31:24.262095+00:00"} {"global_step": 10963, "acc_step": 0, "speed/wps": 12904.105044813428, "speed/FLOPS": 202676701827317.6, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05015638470649719, "optim/lr": 0.0029138574232523665, "optim/total_tokens": 5747769344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.869586944580078, "created_at": "2025-01-16T12:31:34.426305+00:00"} {"global_step": 10964, "acc_step": 0, "speed/wps": 12903.496582639837, "speed/FLOPS": 202667145092766.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0546344593167305, "optim/lr": 0.0029138288035548774, "optim/total_tokens": 5748293632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.959764003753662, "created_at": "2025-01-16T12:31:44.591030+00:00"} {"global_step": 10965, "acc_step": 0, "speed/wps": 12897.605702532328, "speed/FLOPS": 202574620725760.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05664380267262459, "optim/lr": 0.0029138001792445227, "optim/total_tokens": 5748817920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 2.981445789337158, "created_at": "2025-01-16T12:31:54.757270+00:00"} {"global_step": 10966, "acc_step": 0, "speed/wps": 12904.961993726178, "speed/FLOPS": 202690161387563.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04740949347615242, "optim/lr": 0.002913771550321397, "optim/total_tokens": 5749342208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.876157283782959, "created_at": "2025-01-16T12:32:04.923291+00:00"} {"global_step": 10967, "acc_step": 0, "speed/wps": 12900.189365512826, "speed/FLOPS": 202615200703194.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2969039976596832, "optim/lr": 0.002913742916785593, "optim/total_tokens": 5749866496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 2.947035312652588, "created_at": "2025-01-16T12:32:15.090760+00:00"} {"global_step": 10968, "acc_step": 0, "speed/wps": 12899.512754090025, "speed/FLOPS": 202604573591037.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05380187928676605, "optim/lr": 0.0029137142786372044, "optim/total_tokens": 5750390784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.744413375854492, "created_at": "2025-01-16T12:32:25.260861+00:00"} {"global_step": 10969, "acc_step": 0, "speed/wps": 12905.86303359091, "speed/FLOPS": 202704313456797.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04919005557894707, "optim/lr": 0.0029136856358763256, "optim/total_tokens": 5750915072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.827461004257202, "created_at": "2025-01-16T12:32:35.421200+00:00"} {"global_step": 10970, "acc_step": 0, "speed/wps": 12896.705065983766, "speed/FLOPS": 202560474991161.25, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04586220532655716, "optim/lr": 0.002913656988503048, "optim/total_tokens": 5751439360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 2.9183011054992676, "created_at": "2025-01-16T12:32:45.590668+00:00"} {"global_step": 10971, "acc_step": 0, "speed/wps": 12906.80305996584, "speed/FLOPS": 202719077862747.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050682540982961655, "optim/lr": 0.002913628336517466, "optim/total_tokens": 5751963648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8995442390441895, "created_at": "2025-01-16T12:32:55.749848+00:00"} {"global_step": 10972, "acc_step": 0, "speed/wps": 12907.010528409692, "speed/FLOPS": 202722336439749.75, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04677485674619675, "optim/lr": 0.0029135996799196737, "optim/total_tokens": 5752487936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 2.9328510761260986, "created_at": "2025-01-16T12:33:05.910240+00:00"} {"global_step": 10973, "acc_step": 0, "speed/wps": 12904.740552644122, "speed/FLOPS": 202686683351077.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044939637184143066, "optim/lr": 0.0029135710187097636, "optim/total_tokens": 5753012224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.799088954925537, "created_at": "2025-01-16T12:33:16.076576+00:00"} {"global_step": 10974, "acc_step": 0, "speed/wps": 12902.153104211728, "speed/FLOPS": 202646043917920.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04708520695567131, "optim/lr": 0.0029135423528878304, "optim/total_tokens": 5753536512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.886239767074585, "created_at": "2025-01-16T12:33:26.242515+00:00"} {"global_step": 10975, "acc_step": 0, "speed/wps": 12898.771330055652, "speed/FLOPS": 202592928507754.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046815045177936554, "optim/lr": 0.0029135136824539666, "optim/total_tokens": 5754060800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.8084917068481445, "created_at": "2025-01-16T12:33:36.407946+00:00"} {"global_step": 10976, "acc_step": 0, "speed/wps": 12903.116337410922, "speed/FLOPS": 202661172818934.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05096373334527016, "optim/lr": 0.0029134850074082662, "optim/total_tokens": 5754585088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.895400047302246, "created_at": "2025-01-16T12:33:46.572853+00:00"} {"global_step": 10977, "acc_step": 0, "speed/wps": 12896.258647900555, "speed/FLOPS": 202553463381721.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04695224389433861, "optim/lr": 0.0029134563277508226, "optim/total_tokens": 5755109376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.768350839614868, "created_at": "2025-01-16T12:33:56.740232+00:00"} {"global_step": 10978, "acc_step": 0, "speed/wps": 12901.767050484737, "speed/FLOPS": 202639980413644.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04652271047234535, "optim/lr": 0.0029134276434817294, "optim/total_tokens": 5755633664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 2.934138774871826, "created_at": "2025-01-16T12:34:06.905464+00:00"} {"global_step": 10979, "acc_step": 0, "speed/wps": 12890.312939491318, "speed/FLOPS": 202460077860895.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053791217505931854, "optim/lr": 0.00291339895460108, "optim/total_tokens": 5756157952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.9358267784118652, "created_at": "2025-01-16T12:34:17.077775+00:00"} {"global_step": 10980, "acc_step": 0, "speed/wps": 12891.889899386812, "speed/FLOPS": 202484846183023.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04862544313073158, "optim/lr": 0.002913370261108968, "optim/total_tokens": 5756682240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.850433349609375, "created_at": "2025-01-16T12:34:27.251357+00:00"} {"global_step": 10981, "acc_step": 0, "speed/wps": 12884.312059937543, "speed/FLOPS": 202365825801430.4, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05558878555893898, "optim/lr": 0.002913341563005488, "optim/total_tokens": 5757206528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 2.910789728164673, "created_at": "2025-01-16T12:34:37.431141+00:00"} {"global_step": 10982, "acc_step": 0, "speed/wps": 12884.653936266997, "speed/FLOPS": 202371195438971.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.04593675956130028, "optim/lr": 0.0029133128602907322, "optim/total_tokens": 5757730816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 2.9326112270355225, "created_at": "2025-01-16T12:34:47.608584+00:00"} {"global_step": 10983, "acc_step": 0, "speed/wps": 12893.734098218192, "speed/FLOPS": 202513811859865.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05925694853067398, "optim/lr": 0.0029132841529647953, "optim/total_tokens": 5758255104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9791760444641113, "created_at": "2025-01-16T12:34:57.781286+00:00"} {"global_step": 10984, "acc_step": 0, "speed/wps": 12889.734618798226, "speed/FLOPS": 202450994539714.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04994292929768562, "optim/lr": 0.0029132554410277707, "optim/total_tokens": 5758779392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.9557547569274902, "created_at": "2025-01-16T12:35:07.956211+00:00"} {"global_step": 10985, "acc_step": 0, "speed/wps": 12893.837984292986, "speed/FLOPS": 202515443533424.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0506390780210495, "optim/lr": 0.002913226724479751, "optim/total_tokens": 5759303680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 2.8513426780700684, "created_at": "2025-01-16T12:35:18.128901+00:00"} {"global_step": 10986, "acc_step": 0, "speed/wps": 12903.292622406987, "speed/FLOPS": 202663941617039.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.04635510966181755, "optim/lr": 0.002913198003320832, "optim/total_tokens": 5759827968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.825026512145996, "created_at": "2025-01-16T12:35:28.293270+00:00"} {"global_step": 10987, "acc_step": 0, "speed/wps": 12903.138785783254, "speed/FLOPS": 202661525401469.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049027618020772934, "optim/lr": 0.0029131692775511055, "optim/total_tokens": 5760352256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 2.949615955352783, "created_at": "2025-01-16T12:35:38.464897+00:00"} {"global_step": 10988, "acc_step": 0, "speed/wps": 12898.301279309362, "speed/FLOPS": 202585545714866.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05614475533366203, "optim/lr": 0.002913140547170666, "optim/total_tokens": 5760876544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.960425615310669, "created_at": "2025-01-16T12:35:48.630659+00:00"} {"global_step": 10989, "acc_step": 0, "speed/wps": 12897.920184180412, "speed/FLOPS": 202579560092186.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05400075018405914, "optim/lr": 0.0029131118121796078, "optim/total_tokens": 5761400832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.899385929107666, "created_at": "2025-01-16T12:35:58.801127+00:00"} {"global_step": 10990, "acc_step": 0, "speed/wps": 12891.912321703658, "speed/FLOPS": 202485198356321.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.048481784760951996, "optim/lr": 0.0029130830725780234, "optim/total_tokens": 5761925120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9193923473358154, "created_at": "2025-01-16T12:36:08.974329+00:00"} {"global_step": 10991, "acc_step": 0, "speed/wps": 12892.530733417283, "speed/FLOPS": 202494911362070.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06215855851769447, "optim/lr": 0.002913054328366008, "optim/total_tokens": 5762449408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.8738601207733154, "created_at": "2025-01-16T12:36:19.147568+00:00"} {"global_step": 10992, "acc_step": 0, "speed/wps": 12898.565625367944, "speed/FLOPS": 202589697632966.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.043637845665216446, "optim/lr": 0.002913025579543654, "optim/total_tokens": 5762973696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9066014289855957, "created_at": "2025-01-16T12:36:29.313073+00:00"} {"global_step": 10993, "acc_step": 0, "speed/wps": 12890.995675430406, "speed/FLOPS": 202470801166995.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05810348317027092, "optim/lr": 0.002912996826111056, "optim/total_tokens": 5763497984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8653063774108887, "created_at": "2025-01-16T12:36:39.485258+00:00"} {"global_step": 10994, "acc_step": 0, "speed/wps": 12892.596319558352, "speed/FLOPS": 202495941482540.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0485546700656414, "optim/lr": 0.0029129680680683077, "optim/total_tokens": 5764022272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 3.001272201538086, "created_at": "2025-01-16T12:36:49.656719+00:00"} {"global_step": 10995, "acc_step": 0, "speed/wps": 12887.417343300132, "speed/FLOPS": 202414598543746.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.058452557772397995, "optim/lr": 0.002912939305415502, "optim/total_tokens": 5764546560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.7931032180786133, "created_at": "2025-01-16T12:36:59.832230+00:00"} {"global_step": 10996, "acc_step": 0, "speed/wps": 12895.654471474365, "speed/FLOPS": 202543973960722.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07503559440374374, "optim/lr": 0.002912910538152735, "optim/total_tokens": 5765070848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.950381278991699, "created_at": "2025-01-16T12:37:10.000069+00:00"} {"global_step": 10997, "acc_step": 0, "speed/wps": 12897.330207707846, "speed/FLOPS": 202570293700972.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0572131872177124, "optim/lr": 0.002912881766280099, "optim/total_tokens": 5765595136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346134, "loss/out": 2.9812307357788086, "created_at": "2025-01-16T12:37:20.167662+00:00"} {"global_step": 10998, "acc_step": 0, "speed/wps": 12890.656090725994, "speed/FLOPS": 202465467522574.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.002, "optim/grad_norm": 0.05319013446569443, "optim/lr": 0.0029128529897976866, "optim/total_tokens": 5766119424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.693230628967285, "created_at": "2025-01-16T12:37:30.343096+00:00"} {"global_step": 10999, "acc_step": 0, "speed/wps": 12899.837124791753, "speed/FLOPS": 202609668278640.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047166742384433746, "optim/lr": 0.0029128242087055945, "optim/total_tokens": 5766643712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 2.8871655464172363, "created_at": "2025-01-16T12:37:40.511333+00:00"} {"global_step": 11000, "acc_step": 0, "speed/wps": 12896.824955835666, "speed/FLOPS": 202562358026031.25, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04869913309812546, "optim/lr": 0.0029127954230039146, "optim/total_tokens": 5767168000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 2.708582878112793, "created_at": "2025-01-16T12:37:50.678355+00:00"} {"global_step": 11001, "acc_step": 0, "speed/wps": 12593.353785937321, "speed/FLOPS": 197795926289690.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.251, "optim/grad_norm": 0.045740630477666855, "optim/lr": 0.0029127666326927417, "optim/total_tokens": 5767692288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283142, "loss/out": 2.909930944442749, "created_at": "2025-01-16T12:38:01.093031+00:00"} {"global_step": 11002, "acc_step": 0, "speed/wps": 12899.449935112212, "speed/FLOPS": 202603586932668.94, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048840224742889404, "optim/lr": 0.0029127378377721692, "optim/total_tokens": 5768216576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.8808910846710205, "created_at": "2025-01-16T12:38:11.260080+00:00"} {"global_step": 11003, "acc_step": 0, "speed/wps": 12891.859256742755, "speed/FLOPS": 202484364898191.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04878964647650719, "optim/lr": 0.0029127090382422916, "optim/total_tokens": 5768740864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8065757751464844, "created_at": "2025-01-16T12:38:21.432881+00:00"} {"global_step": 11004, "acc_step": 0, "speed/wps": 12903.519196780517, "speed/FLOPS": 202667500278921.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053065910935401917, "optim/lr": 0.0029126802341032025, "optim/total_tokens": 5769265152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 2.934642791748047, "created_at": "2025-01-16T12:38:31.597130+00:00"} {"global_step": 11005, "acc_step": 0, "speed/wps": 12903.90577656792, "speed/FLOPS": 202673572045698.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047966424375772476, "optim/lr": 0.0029126514253549956, "optim/total_tokens": 5769789440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.899372100830078, "created_at": "2025-01-16T12:38:41.763062+00:00"} {"global_step": 11006, "acc_step": 0, "speed/wps": 12896.523101686764, "speed/FLOPS": 202557616991831.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05683790147304535, "optim/lr": 0.002912622611997766, "optim/total_tokens": 5770313728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.830822467803955, "created_at": "2025-01-16T12:38:51.930031+00:00"} {"global_step": 11007, "acc_step": 0, "speed/wps": 12901.971432053022, "speed/FLOPS": 202643190507024.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046141814440488815, "optim/lr": 0.0029125937940316063, "optim/total_tokens": 5770838016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.9476394653320312, "created_at": "2025-01-16T12:39:02.099687+00:00"} {"global_step": 11008, "acc_step": 0, "speed/wps": 12905.928409866718, "speed/FLOPS": 202705340281045.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05493525415658951, "optim/lr": 0.0029125649714566116, "optim/total_tokens": 5771362304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.920764446258545, "created_at": "2025-01-16T12:39:12.261599+00:00"} {"global_step": 11009, "acc_step": 0, "speed/wps": 12887.841985162113, "speed/FLOPS": 202421268127707.84, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05000174045562744, "optim/lr": 0.002912536144272875, "optim/total_tokens": 5771886592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8348093032836914, "created_at": "2025-01-16T12:39:22.437969+00:00"} {"global_step": 11010, "acc_step": 0, "speed/wps": 12888.98814486364, "speed/FLOPS": 202439270140808.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.057008471339941025, "optim/lr": 0.002912507312480492, "optim/total_tokens": 5772410880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 2.9266319274902344, "created_at": "2025-01-16T12:39:32.612421+00:00"} {"global_step": 11011, "acc_step": 0, "speed/wps": 12891.164197462225, "speed/FLOPS": 202473448037079.22, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04283437132835388, "optim/lr": 0.002912478476079555, "optim/total_tokens": 5772935168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.8792357444763184, "created_at": "2025-01-16T12:39:42.787032+00:00"} {"global_step": 11012, "acc_step": 0, "speed/wps": 12888.429024818597, "speed/FLOPS": 202430488392189.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08172480762004852, "optim/lr": 0.0029124496350701593, "optim/total_tokens": 5773459456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.8214969635009766, "created_at": "2025-01-16T12:39:52.961610+00:00"} {"global_step": 11013, "acc_step": 0, "speed/wps": 12889.316034255431, "speed/FLOPS": 202444420094274.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06449336558580399, "optim/lr": 0.002912420789452398, "optim/total_tokens": 5773983744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486049, "loss/out": 2.872269630432129, "created_at": "2025-01-16T12:40:03.135026+00:00"} {"global_step": 11014, "acc_step": 0, "speed/wps": 12889.861997213677, "speed/FLOPS": 202452995192765.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.053705837577581406, "optim/lr": 0.0029123919392263662, "optim/total_tokens": 5774508032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 2.9722142219543457, "created_at": "2025-01-16T12:40:13.307386+00:00"} {"global_step": 11015, "acc_step": 0, "speed/wps": 12890.367033055434, "speed/FLOPS": 202460927474651.88, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08115149289369583, "optim/lr": 0.0029123630843921576, "optim/total_tokens": 5775032320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 2.869931697845459, "created_at": "2025-01-16T12:40:23.484906+00:00"} {"global_step": 11016, "acc_step": 0, "speed/wps": 12892.955623741002, "speed/FLOPS": 202501584848465.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05215016379952431, "optim/lr": 0.002912334224949866, "optim/total_tokens": 5775556608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 2.8096423149108887, "created_at": "2025-01-16T12:40:33.655978+00:00"} {"global_step": 11017, "acc_step": 0, "speed/wps": 12893.918184580527, "speed/FLOPS": 202516703189147.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.07787435501813889, "optim/lr": 0.002912305360899586, "optim/total_tokens": 5776080896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 2.9770421981811523, "created_at": "2025-01-16T12:40:43.826612+00:00"} {"global_step": 11018, "acc_step": 0, "speed/wps": 12892.224544066641, "speed/FLOPS": 202490102237570.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05663993954658508, "optim/lr": 0.0029122764922414114, "optim/total_tokens": 5776605184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.822800636291504, "created_at": "2025-01-16T12:40:53.997911+00:00"} {"global_step": 11019, "acc_step": 0, "speed/wps": 12897.761569132776, "speed/FLOPS": 202577068824903.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05764536187052727, "optim/lr": 0.0029122476189754366, "optim/total_tokens": 5777129472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 2.895667314529419, "created_at": "2025-01-16T12:41:04.165148+00:00"} {"global_step": 11020, "acc_step": 0, "speed/wps": 12889.241493587135, "speed/FLOPS": 202443249330648.9, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05798288807272911, "optim/lr": 0.0029122187411017566, "optim/total_tokens": 5777653760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 266264, "loss/out": 2.9775047302246094, "created_at": "2025-01-16T12:41:14.342780+00:00"} {"global_step": 11021, "acc_step": 0, "speed/wps": 12890.538195181301, "speed/FLOPS": 202463615811040.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04482826963067055, "optim/lr": 0.002912189858620464, "optim/total_tokens": 5778178048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.8267786502838135, "created_at": "2025-01-16T12:41:24.517047+00:00"} {"global_step": 11022, "acc_step": 0, "speed/wps": 12905.349080075142, "speed/FLOPS": 202696241110586.4, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04975299537181854, "optim/lr": 0.0029121609715316546, "optim/total_tokens": 5778702336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.90556263923645, "created_at": "2025-01-16T12:41:34.679572+00:00"} {"global_step": 11023, "acc_step": 0, "speed/wps": 12905.81149639155, "speed/FLOPS": 202703503994262.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047468483448028564, "optim/lr": 0.0029121320798354217, "optim/total_tokens": 5779226624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 2.798772096633911, "created_at": "2025-01-16T12:41:44.840713+00:00"} {"global_step": 11024, "acc_step": 0, "speed/wps": 12893.661551188488, "speed/FLOPS": 202512672409073.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07412420958280563, "optim/lr": 0.00291210318353186, "optim/total_tokens": 5779750912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.869418144226074, "created_at": "2025-01-16T12:41:55.011460+00:00"} {"global_step": 11025, "acc_step": 0, "speed/wps": 12898.386199767796, "speed/FLOPS": 202586879507359.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.13310930132865906, "optim/lr": 0.0029120742826210634, "optim/total_tokens": 5780275200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.8666210174560547, "created_at": "2025-01-16T12:42:05.179201+00:00"} {"global_step": 11026, "acc_step": 0, "speed/wps": 12892.322230328446, "speed/FLOPS": 202491636534547.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06712105870246887, "optim/lr": 0.0029120453771031263, "optim/total_tokens": 5780799488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 2.9462649822235107, "created_at": "2025-01-16T12:42:15.350789+00:00"} {"global_step": 11027, "acc_step": 0, "speed/wps": 12892.129830609663, "speed/FLOPS": 202488614632578.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08792869001626968, "optim/lr": 0.0029120164669781427, "optim/total_tokens": 5781323776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9055771827697754, "created_at": "2025-01-16T12:42:25.525509+00:00"} {"global_step": 11028, "acc_step": 0, "speed/wps": 12888.111387224366, "speed/FLOPS": 202425499457292.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04801669716835022, "optim/lr": 0.002911987552246208, "optim/total_tokens": 5781848064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.7191085815429688, "created_at": "2025-01-16T12:42:35.704716+00:00"} {"global_step": 11029, "acc_step": 0, "speed/wps": 12892.40333886308, "speed/FLOPS": 202492910455538.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05842006951570511, "optim/lr": 0.0029119586329074154, "optim/total_tokens": 5782372352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 2.7953238487243652, "created_at": "2025-01-16T12:42:45.881572+00:00"} {"global_step": 11030, "acc_step": 0, "speed/wps": 12893.939824842586, "speed/FLOPS": 202517043079199.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.048933759331703186, "optim/lr": 0.00291192970896186, "optim/total_tokens": 5782896640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.9217286109924316, "created_at": "2025-01-16T12:42:56.053170+00:00"} {"global_step": 11031, "acc_step": 0, "speed/wps": 12902.697648650324, "speed/FLOPS": 202654596736611.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05509357899427414, "optim/lr": 0.002911900780409636, "optim/total_tokens": 5783420928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.950800895690918, "created_at": "2025-01-16T12:43:06.217397+00:00"} {"global_step": 11032, "acc_step": 0, "speed/wps": 12893.898119058262, "speed/FLOPS": 202516388032546.9, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.058378126472234726, "optim/lr": 0.0029118718472508375, "optim/total_tokens": 5783945216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9133987426757812, "created_at": "2025-01-16T12:43:16.391307+00:00"} {"global_step": 11033, "acc_step": 0, "speed/wps": 12897.430500715871, "speed/FLOPS": 202571868940480.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050255563110113144, "optim/lr": 0.002911842909485559, "optim/total_tokens": 5784469504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.733635187149048, "created_at": "2025-01-16T12:43:26.557575+00:00"} {"global_step": 11034, "acc_step": 0, "speed/wps": 12900.168937558301, "speed/FLOPS": 202614879854097.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04483744129538536, "optim/lr": 0.002911813967113895, "optim/total_tokens": 5784993792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.908092498779297, "created_at": "2025-01-16T12:43:36.726421+00:00"} {"global_step": 11035, "acc_step": 0, "speed/wps": 12903.252751279684, "speed/FLOPS": 202663315386196.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.059018027037382126, "optim/lr": 0.00291178502013594, "optim/total_tokens": 5785518080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 2.970046043395996, "created_at": "2025-01-16T12:43:46.888260+00:00"} {"global_step": 11036, "acc_step": 0, "speed/wps": 12899.817572914462, "speed/FLOPS": 202609361189540.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04697604849934578, "optim/lr": 0.0029117560685517885, "optim/total_tokens": 5786042368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.029200315475464, "created_at": "2025-01-16T12:43:57.052594+00:00"} {"global_step": 11037, "acc_step": 0, "speed/wps": 12901.47113382785, "speed/FLOPS": 202635332635916.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04973239824175835, "optim/lr": 0.002911727112361535, "optim/total_tokens": 5786566656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.90651535987854, "created_at": "2025-01-16T12:44:07.216855+00:00"} {"global_step": 11038, "acc_step": 0, "speed/wps": 12900.654719684555, "speed/FLOPS": 202622509729925.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044191453605890274, "optim/lr": 0.002911698151565273, "optim/total_tokens": 5787090944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 2.813778877258301, "created_at": "2025-01-16T12:44:17.384442+00:00"} {"global_step": 11039, "acc_step": 0, "speed/wps": 12909.991288307161, "speed/FLOPS": 202769153369932.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05608135834336281, "optim/lr": 0.0029116691861630984, "optim/total_tokens": 5787615232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.8851540088653564, "created_at": "2025-01-16T12:44:27.544318+00:00"} {"global_step": 11040, "acc_step": 0, "speed/wps": 12899.474419685454, "speed/FLOPS": 202603971496537.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0613466240465641, "optim/lr": 0.0029116402161551047, "optim/total_tokens": 5788139520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 2.8844656944274902, "created_at": "2025-01-16T12:44:37.711004+00:00"} {"global_step": 11041, "acc_step": 0, "speed/wps": 12906.338650463915, "speed/FLOPS": 202711783673354.28, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057147007435560226, "optim/lr": 0.0029116112415413873, "optim/total_tokens": 5788663808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 2.835427761077881, "created_at": "2025-01-16T12:44:47.871956+00:00"} {"global_step": 11042, "acc_step": 0, "speed/wps": 12905.341313458672, "speed/FLOPS": 202696119125202.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05169863626360893, "optim/lr": 0.0029115822623220398, "optim/total_tokens": 5789188096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8632349967956543, "created_at": "2025-01-16T12:44:58.031944+00:00"} {"global_step": 11043, "acc_step": 0, "speed/wps": 12903.271000426024, "speed/FLOPS": 202663602014116.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04558992013335228, "optim/lr": 0.002911553278497157, "optim/total_tokens": 5789712384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.8305206298828125, "created_at": "2025-01-16T12:45:08.194778+00:00"} {"global_step": 11044, "acc_step": 0, "speed/wps": 12900.46194347609, "speed/FLOPS": 202619481914668.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05071330815553665, "optim/lr": 0.0029115242900668342, "optim/total_tokens": 5790236672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8349428176879883, "created_at": "2025-01-16T12:45:18.359917+00:00"} {"global_step": 11045, "acc_step": 0, "speed/wps": 12903.786811548587, "speed/FLOPS": 202671703536594.94, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04311615601181984, "optim/lr": 0.0029114952970311653, "optim/total_tokens": 5790760960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.9188716411590576, "created_at": "2025-01-16T12:45:28.521127+00:00"} {"global_step": 11046, "acc_step": 0, "speed/wps": 12902.17946907613, "speed/FLOPS": 202646458014345.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04803166165947914, "optim/lr": 0.0029114662993902454, "optim/total_tokens": 5791285248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.867640972137451, "created_at": "2025-01-16T12:45:38.683641+00:00"} {"global_step": 11047, "acc_step": 0, "speed/wps": 12902.808245652015, "speed/FLOPS": 202656333814498.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04616705700755119, "optim/lr": 0.002911437297144168, "optim/total_tokens": 5791809536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.8970839977264404, "created_at": "2025-01-16T12:45:48.846800+00:00"} {"global_step": 11048, "acc_step": 0, "speed/wps": 12908.766367875742, "speed/FLOPS": 202749914311340.34, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055828824639320374, "optim/lr": 0.002911408290293029, "optim/total_tokens": 5792333824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.7617740631103516, "created_at": "2025-01-16T12:45:59.009588+00:00"} {"global_step": 11049, "acc_step": 0, "speed/wps": 12899.755986543703, "speed/FLOPS": 202608393890959.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05387838929891586, "optim/lr": 0.0029113792788369223, "optim/total_tokens": 5792858112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8032777309417725, "created_at": "2025-01-16T12:46:09.180602+00:00"} {"global_step": 11050, "acc_step": 0, "speed/wps": 12894.699648572108, "speed/FLOPS": 202528977154978.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04437999427318573, "optim/lr": 0.0029113502627759425, "optim/total_tokens": 5793382400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9895968437194824, "created_at": "2025-01-16T12:46:19.353115+00:00"} {"global_step": 11051, "acc_step": 0, "speed/wps": 12891.880662175474, "speed/FLOPS": 202484701099926.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04063982889056206, "optim/lr": 0.002911321242110185, "optim/total_tokens": 5793906688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 2.867408037185669, "created_at": "2025-01-16T12:46:29.524864+00:00"} {"global_step": 11052, "acc_step": 0, "speed/wps": 12906.941498964286, "speed/FLOPS": 202721252237453.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05070493742823601, "optim/lr": 0.0029112922168397434, "optim/total_tokens": 5794430976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.771253824234009, "created_at": "2025-01-16T12:46:39.687375+00:00"} {"global_step": 11053, "acc_step": 0, "speed/wps": 12898.874710942426, "speed/FLOPS": 202594552246641.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06419558823108673, "optim/lr": 0.0029112631869647135, "optim/total_tokens": 5794955264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397147, "loss/out": 2.8622970581054688, "created_at": "2025-01-16T12:46:49.855429+00:00"} {"global_step": 11054, "acc_step": 0, "speed/wps": 12897.891178339685, "speed/FLOPS": 202579104515600.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047603122889995575, "optim/lr": 0.0029112341524851894, "optim/total_tokens": 5795479552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 2.960228204727173, "created_at": "2025-01-16T12:47:00.021265+00:00"} {"global_step": 11055, "acc_step": 0, "speed/wps": 12898.013597372948, "speed/FLOPS": 202581027274738.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057259272783994675, "optim/lr": 0.0029112051134012663, "optim/total_tokens": 5796003840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 2.833005905151367, "created_at": "2025-01-16T12:47:10.188130+00:00"} {"global_step": 11056, "acc_step": 0, "speed/wps": 12901.536544856297, "speed/FLOPS": 202636360006002.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052166473120450974, "optim/lr": 0.002911176069713038, "optim/total_tokens": 5796528128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420398, "loss/out": 2.8464136123657227, "created_at": "2025-01-16T12:47:20.353524+00:00"} {"global_step": 11057, "acc_step": 0, "speed/wps": 12907.958784132987, "speed/FLOPS": 202737230098924.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.12418430298566818, "optim/lr": 0.0029111470214206008, "optim/total_tokens": 5797052416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8411593437194824, "created_at": "2025-01-16T12:47:30.511380+00:00"} {"global_step": 11058, "acc_step": 0, "speed/wps": 12903.135364578127, "speed/FLOPS": 202661471666741.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09495355188846588, "optim/lr": 0.002911117968524048, "optim/total_tokens": 5797576704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.908996343612671, "created_at": "2025-01-16T12:47:40.674202+00:00"} {"global_step": 11059, "acc_step": 0, "speed/wps": 12896.27397848133, "speed/FLOPS": 202553704169559.16, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04821040853857994, "optim/lr": 0.002911088911023474, "optim/total_tokens": 5798100992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 263306, "loss/out": 2.723335027694702, "created_at": "2025-01-16T12:47:50.847384+00:00"} {"global_step": 11060, "acc_step": 0, "speed/wps": 12900.36491071518, "speed/FLOPS": 202617957881821.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07094834744930267, "optim/lr": 0.002911059848918976, "optim/total_tokens": 5798625280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.983642339706421, "created_at": "2025-01-16T12:48:01.012614+00:00"} {"global_step": 11061, "acc_step": 0, "speed/wps": 12902.846588044613, "speed/FLOPS": 202656936034462.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05669640377163887, "optim/lr": 0.002911030782210647, "optim/total_tokens": 5799149568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421239, "loss/out": 2.9012186527252197, "created_at": "2025-01-16T12:48:11.178381+00:00"} {"global_step": 11062, "acc_step": 0, "speed/wps": 12895.996647492704, "speed/FLOPS": 202549348305290.1, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06725680828094482, "optim/lr": 0.0029110017108985816, "optim/total_tokens": 5799673856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.842048168182373, "created_at": "2025-01-16T12:48:21.346688+00:00"} {"global_step": 11063, "acc_step": 0, "speed/wps": 12893.46486738088, "speed/FLOPS": 202509583219603.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06999220699071884, "optim/lr": 0.0029109726349828757, "optim/total_tokens": 5800198144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.7439448833465576, "created_at": "2025-01-16T12:48:31.515932+00:00"} {"global_step": 11064, "acc_step": 0, "speed/wps": 12904.610648370917, "speed/FLOPS": 202684643025960.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055160071700811386, "optim/lr": 0.0029109435544636235, "optim/total_tokens": 5800722432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.976013660430908, "created_at": "2025-01-16T12:48:41.678473+00:00"} {"global_step": 11065, "acc_step": 0, "speed/wps": 12905.629795750185, "speed/FLOPS": 202700650136006.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06334207952022552, "optim/lr": 0.0029109144693409204, "optim/total_tokens": 5801246720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.819833755493164, "created_at": "2025-01-16T12:48:51.844611+00:00"} {"global_step": 11066, "acc_step": 0, "speed/wps": 12910.215383605508, "speed/FLOPS": 202772673094532.6, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04697539657354355, "optim/lr": 0.002910885379614861, "optim/total_tokens": 5801771008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.7694711685180664, "created_at": "2025-01-16T12:49:02.000796+00:00"} {"global_step": 11067, "acc_step": 0, "speed/wps": 12905.573307731669, "speed/FLOPS": 202699762914053.25, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22864766418933868, "optim/lr": 0.00291085628528554, "optim/total_tokens": 5802295296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 2.948831796646118, "created_at": "2025-01-16T12:49:12.161183+00:00"} {"global_step": 11068, "acc_step": 0, "speed/wps": 12912.269281575363, "speed/FLOPS": 202804932384500.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08120157569646835, "optim/lr": 0.0029108271863530523, "optim/total_tokens": 5802819584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.979832410812378, "created_at": "2025-01-16T12:49:22.316284+00:00"} {"global_step": 11069, "acc_step": 0, "speed/wps": 12910.68983138741, "speed/FLOPS": 202780124948906.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08103121817111969, "optim/lr": 0.0029107980828174933, "optim/total_tokens": 5803343872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 2.85062313079834, "created_at": "2025-01-16T12:49:32.473252+00:00"} {"global_step": 11070, "acc_step": 0, "speed/wps": 12903.416147214995, "speed/FLOPS": 202665881743889.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06119966134428978, "optim/lr": 0.0029107689746789575, "optim/total_tokens": 5803868160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.7889981269836426, "created_at": "2025-01-16T12:49:42.634901+00:00"} {"global_step": 11071, "acc_step": 0, "speed/wps": 12908.479887143774, "speed/FLOPS": 202745414737779.0, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.09207845479249954, "optim/lr": 0.0029107398619375403, "optim/total_tokens": 5804392448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8756463527679443, "created_at": "2025-01-16T12:49:52.793520+00:00"} {"global_step": 11072, "acc_step": 0, "speed/wps": 12914.002916501504, "speed/FLOPS": 202832161503279.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09113115072250366, "optim/lr": 0.0029107107445933366, "optim/total_tokens": 5804916736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.9130594730377197, "created_at": "2025-01-16T12:50:02.956458+00:00"} {"global_step": 11073, "acc_step": 0, "speed/wps": 12908.602262177546, "speed/FLOPS": 202747336805845.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06585197150707245, "optim/lr": 0.0029106816226464414, "optim/total_tokens": 5805441024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8604564666748047, "created_at": "2025-01-16T12:50:13.117421+00:00"} {"global_step": 11074, "acc_step": 0, "speed/wps": 12900.150938403547, "speed/FLOPS": 202614597152638.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07566312700510025, "optim/lr": 0.0029106524960969492, "optim/total_tokens": 5805965312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 2.8062620162963867, "created_at": "2025-01-16T12:50:23.283984+00:00"} {"global_step": 11075, "acc_step": 0, "speed/wps": 12905.648700151405, "speed/FLOPS": 202700947055605.25, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058446966111660004, "optim/lr": 0.0029106233649449555, "optim/total_tokens": 5806489600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.936175584793091, "created_at": "2025-01-16T12:50:33.446613+00:00"} {"global_step": 11076, "acc_step": 0, "speed/wps": 12911.461107878087, "speed/FLOPS": 202792238906037.97, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.10977715998888016, "optim/lr": 0.0029105942291905557, "optim/total_tokens": 5807013888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.0042850971221924, "created_at": "2025-01-16T12:50:43.603723+00:00"} {"global_step": 11077, "acc_step": 0, "speed/wps": 12909.25131878299, "speed/FLOPS": 202757531131731.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.10247990489006042, "optim/lr": 0.002910565088833844, "optim/total_tokens": 5807538176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9016871452331543, "created_at": "2025-01-16T12:50:53.763625+00:00"} {"global_step": 11078, "acc_step": 0, "speed/wps": 12907.498342384097, "speed/FLOPS": 202729998228544.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06091732904314995, "optim/lr": 0.002910535943874916, "optim/total_tokens": 5808062464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.722318410873413, "created_at": "2025-01-16T12:51:03.931110+00:00"} {"global_step": 11079, "acc_step": 0, "speed/wps": 12897.87105080497, "speed/FLOPS": 202578788385009.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06731773167848587, "optim/lr": 0.0029105067943138666, "optim/total_tokens": 5808586752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 2.8107523918151855, "created_at": "2025-01-16T12:51:14.097248+00:00"} {"global_step": 11080, "acc_step": 0, "speed/wps": 12899.309379688046, "speed/FLOPS": 202601379316589.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06118708848953247, "optim/lr": 0.002910477640150791, "optim/total_tokens": 5809111040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.965060234069824, "created_at": "2025-01-16T12:51:24.263575+00:00"} {"global_step": 11081, "acc_step": 0, "speed/wps": 12909.694535987588, "speed/FLOPS": 202764492466973.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06428239494562149, "optim/lr": 0.002910448481385784, "optim/total_tokens": 5809635328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8022990226745605, "created_at": "2025-01-16T12:51:34.424005+00:00"} {"global_step": 11082, "acc_step": 0, "speed/wps": 12904.228303577751, "speed/FLOPS": 202678637775586.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04606702923774719, "optim/lr": 0.002910419318018942, "optim/total_tokens": 5810159616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.932407855987549, "created_at": "2025-01-16T12:51:44.586007+00:00"} {"global_step": 11083, "acc_step": 0, "speed/wps": 12906.727017255938, "speed/FLOPS": 202717883507494.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08001662790775299, "optim/lr": 0.002910390150050358, "optim/total_tokens": 5810683904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 2.779808521270752, "created_at": "2025-01-16T12:51:54.747267+00:00"} {"global_step": 11084, "acc_step": 0, "speed/wps": 12904.302714789857, "speed/FLOPS": 202679806505924.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05209605023264885, "optim/lr": 0.0029103609774801292, "optim/total_tokens": 5811208192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.904008626937866, "created_at": "2025-01-16T12:52:04.908067+00:00"} {"global_step": 11085, "acc_step": 0, "speed/wps": 12905.00388826417, "speed/FLOPS": 202690819398851.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07298917323350906, "optim/lr": 0.00291033180030835, "optim/total_tokens": 5811732480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.9425806999206543, "created_at": "2025-01-16T12:52:15.068436+00:00"} {"global_step": 11086, "acc_step": 0, "speed/wps": 12904.319119921305, "speed/FLOPS": 202680064171056.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05303032696247101, "optim/lr": 0.002910302618535115, "optim/total_tokens": 5812256768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.934248208999634, "created_at": "2025-01-16T12:52:25.233707+00:00"} {"global_step": 11087, "acc_step": 0, "speed/wps": 12905.58769359677, "speed/FLOPS": 202699988863832.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07244233042001724, "optim/lr": 0.00291027343216052, "optim/total_tokens": 5812781056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.9217138290405273, "created_at": "2025-01-16T12:52:35.395370+00:00"} {"global_step": 11088, "acc_step": 0, "speed/wps": 12906.453893965077, "speed/FLOPS": 202713593730901.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0504472441971302, "optim/lr": 0.0029102442411846605, "optim/total_tokens": 5813305344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9331624507904053, "created_at": "2025-01-16T12:52:45.562823+00:00"} {"global_step": 11089, "acc_step": 0, "speed/wps": 12911.775436366872, "speed/FLOPS": 202797175866886.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053520746529102325, "optim/lr": 0.0029102150456076317, "optim/total_tokens": 5813829632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.847541093826294, "created_at": "2025-01-16T12:52:55.717707+00:00"} {"global_step": 11090, "acc_step": 0, "speed/wps": 12899.611709527146, "speed/FLOPS": 202606127822156.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.06307034194469452, "optim/lr": 0.002910185845429528, "optim/total_tokens": 5814353920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9193387031555176, "created_at": "2025-01-16T12:53:05.882511+00:00"} {"global_step": 11091, "acc_step": 0, "speed/wps": 12890.712725700996, "speed/FLOPS": 202466357052681.88, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04871797561645508, "optim/lr": 0.002910156640650446, "optim/total_tokens": 5814878208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8122894763946533, "created_at": "2025-01-16T12:53:16.055543+00:00"} {"global_step": 11092, "acc_step": 0, "speed/wps": 12887.862171102073, "speed/FLOPS": 202421585175633.3, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05773613229393959, "optim/lr": 0.0029101274312704796, "optim/total_tokens": 5815402496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476549, "loss/out": 2.8637423515319824, "created_at": "2025-01-16T12:53:26.230380+00:00"} {"global_step": 11093, "acc_step": 0, "speed/wps": 12889.308301090241, "speed/FLOPS": 202444298634288.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04428868368268013, "optim/lr": 0.0029100982172897247, "optim/total_tokens": 5815926784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.958483934402466, "created_at": "2025-01-16T12:53:36.404273+00:00"} {"global_step": 11094, "acc_step": 0, "speed/wps": 12897.581989787765, "speed/FLOPS": 202574248284522.62, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0493779219686985, "optim/lr": 0.0029100689987082767, "optim/total_tokens": 5816451072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8751909732818604, "created_at": "2025-01-16T12:53:46.570676+00:00"} {"global_step": 11095, "acc_step": 0, "speed/wps": 12885.961569973177, "speed/FLOPS": 202391733623204.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0506068654358387, "optim/lr": 0.0029100397755262315, "optim/total_tokens": 5816975360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411310, "loss/out": 2.8062212467193604, "created_at": "2025-01-16T12:53:56.746770+00:00"} {"global_step": 11096, "acc_step": 0, "speed/wps": 12890.369624580313, "speed/FLOPS": 202460968178111.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06827647984027863, "optim/lr": 0.002910010547743683, "optim/total_tokens": 5817499648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 3.0039658546447754, "created_at": "2025-01-16T12:54:06.919549+00:00"} {"global_step": 11097, "acc_step": 0, "speed/wps": 12897.276221155156, "speed/FLOPS": 202569445767977.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05870020017027855, "optim/lr": 0.002909981315360728, "optim/total_tokens": 5818023936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8848297595977783, "created_at": "2025-01-16T12:54:17.085987+00:00"} {"global_step": 11098, "acc_step": 0, "speed/wps": 12900.443282488708, "speed/FLOPS": 202619188818219.12, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05445995554327965, "optim/lr": 0.002909952078377461, "optim/total_tokens": 5818548224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.7808139324188232, "created_at": "2025-01-16T12:54:27.249773+00:00"} {"global_step": 11099, "acc_step": 0, "speed/wps": 12905.837405182692, "speed/FLOPS": 202703910927429.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055438317358493805, "optim/lr": 0.0029099228367939775, "optim/total_tokens": 5819072512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8444018363952637, "created_at": "2025-01-16T12:54:37.411163+00:00"} {"global_step": 11100, "acc_step": 0, "speed/wps": 12915.121962883488, "speed/FLOPS": 202849737664439.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04581941291689873, "optim/lr": 0.0029098935906103737, "optim/total_tokens": 5819596800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.8399288654327393, "created_at": "2025-01-16T12:54:47.563838+00:00"} {"global_step": 11101, "acc_step": 0, "speed/wps": 12903.61706847733, "speed/FLOPS": 202669037488409.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046251967549324036, "optim/lr": 0.002909864339826744, "optim/total_tokens": 5820121088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.882180690765381, "created_at": "2025-01-16T12:54:57.727787+00:00"} {"global_step": 11102, "acc_step": 0, "speed/wps": 12899.270754468942, "speed/FLOPS": 202600772654446.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04176511988043785, "optim/lr": 0.0029098350844431837, "optim/total_tokens": 5820645376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8032073974609375, "created_at": "2025-01-16T12:55:07.893927+00:00"} {"global_step": 11103, "acc_step": 0, "speed/wps": 12896.221755627133, "speed/FLOPS": 202552883937872.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04833948612213135, "optim/lr": 0.0029098058244597896, "optim/total_tokens": 5821169664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.923379421234131, "created_at": "2025-01-16T12:55:18.072282+00:00"} {"global_step": 11104, "acc_step": 0, "speed/wps": 12894.582236683675, "speed/FLOPS": 202527133039930.16, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04539249837398529, "optim/lr": 0.0029097765598766557, "optim/total_tokens": 5821693952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.8604092597961426, "created_at": "2025-01-16T12:55:28.241861+00:00"} {"global_step": 11105, "acc_step": 0, "speed/wps": 12896.437274531181, "speed/FLOPS": 202556268958414.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03921010345220566, "optim/lr": 0.0029097472906938785, "optim/total_tokens": 5822218240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.7875895500183105, "created_at": "2025-01-16T12:55:38.414217+00:00"} {"global_step": 11106, "acc_step": 0, "speed/wps": 12891.72955026863, "speed/FLOPS": 202482327679779.97, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049507904797792435, "optim/lr": 0.002909718016911553, "optim/total_tokens": 5822742528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9671616554260254, "created_at": "2025-01-16T12:55:48.585052+00:00"} {"global_step": 11107, "acc_step": 0, "speed/wps": 12897.828238040707, "speed/FLOPS": 202578115951715.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0454130545258522, "optim/lr": 0.0029096887385297748, "optim/total_tokens": 5823266816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.819509267807007, "created_at": "2025-01-16T12:55:58.754007+00:00"} {"global_step": 11108, "acc_step": 0, "speed/wps": 12906.265172088852, "speed/FLOPS": 202710629594504.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.046025365591049194, "optim/lr": 0.0029096594555486394, "optim/total_tokens": 5823791104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.8509693145751953, "created_at": "2025-01-16T12:56:08.913117+00:00"} {"global_step": 11109, "acc_step": 0, "speed/wps": 12904.278213067166, "speed/FLOPS": 202679421672700.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08694592863321304, "optim/lr": 0.0029096301679682425, "optim/total_tokens": 5824315392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.7547383308410645, "created_at": "2025-01-16T12:56:19.083321+00:00"} {"global_step": 11110, "acc_step": 0, "speed/wps": 12900.4078550229, "speed/FLOPS": 202618632381186.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05431869626045227, "optim/lr": 0.0029096008757886795, "optim/total_tokens": 5824839680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 2.882248878479004, "created_at": "2025-01-16T12:56:29.251138+00:00"} {"global_step": 11111, "acc_step": 0, "speed/wps": 12901.854651761465, "speed/FLOPS": 202641356312072.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05952472984790802, "optim/lr": 0.0029095715790100458, "optim/total_tokens": 5825363968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8701276779174805, "created_at": "2025-01-16T12:56:39.413905+00:00"} {"global_step": 11112, "acc_step": 0, "speed/wps": 12906.873133114923, "speed/FLOPS": 202720178457844.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.049192894250154495, "optim/lr": 0.002909542277632438, "optim/total_tokens": 5825888256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466943, "loss/out": 2.8635358810424805, "created_at": "2025-01-16T12:56:49.572667+00:00"} {"global_step": 11113, "acc_step": 0, "speed/wps": 12897.845824099815, "speed/FLOPS": 202578392164939.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04674559459090233, "optim/lr": 0.00290951297165595, "optim/total_tokens": 5826412544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8994312286376953, "created_at": "2025-01-16T12:56:59.738586+00:00"} {"global_step": 11114, "acc_step": 0, "speed/wps": 12905.210739136723, "speed/FLOPS": 202694068276048.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052385810762643814, "optim/lr": 0.002909483661080679, "optim/total_tokens": 5826936832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 2.7991647720336914, "created_at": "2025-01-16T12:57:09.899691+00:00"} {"global_step": 11115, "acc_step": 0, "speed/wps": 12905.8171368586, "speed/FLOPS": 202703592585548.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04580582678318024, "optim/lr": 0.002909454345906719, "optim/total_tokens": 5827461120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.863612174987793, "created_at": "2025-01-16T12:57:20.060887+00:00"} {"global_step": 11116, "acc_step": 0, "speed/wps": 12908.167639393288, "speed/FLOPS": 202740510457784.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048758722841739655, "optim/lr": 0.0029094250261341675, "optim/total_tokens": 5827985408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.813746929168701, "created_at": "2025-01-16T12:57:30.219869+00:00"} {"global_step": 11117, "acc_step": 0, "speed/wps": 12903.769412778933, "speed/FLOPS": 202671430265009.1, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04101844131946564, "optim/lr": 0.0029093957017631194, "optim/total_tokens": 5828509696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473347, "loss/out": 2.8280277252197266, "created_at": "2025-01-16T12:57:40.389038+00:00"} {"global_step": 11118, "acc_step": 0, "speed/wps": 12904.418214191484, "speed/FLOPS": 202681620582739.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05221685767173767, "optim/lr": 0.00290936637279367, "optim/total_tokens": 5829033984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 2.804790496826172, "created_at": "2025-01-16T12:57:50.556546+00:00"} {"global_step": 11119, "acc_step": 0, "speed/wps": 12905.401122445313, "speed/FLOPS": 202697058507523.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06052301451563835, "optim/lr": 0.0029093370392259147, "optim/total_tokens": 5829558272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.8154313564300537, "created_at": "2025-01-16T12:58:00.718147+00:00"} {"global_step": 11120, "acc_step": 0, "speed/wps": 12902.3207352507, "speed/FLOPS": 202648676793736.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04909196123480797, "optim/lr": 0.00290930770105995, "optim/total_tokens": 5830082560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 3.0208165645599365, "created_at": "2025-01-16T12:58:10.881236+00:00"} {"global_step": 11121, "acc_step": 0, "speed/wps": 12905.415955254462, "speed/FLOPS": 202697291477173.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04896226525306702, "optim/lr": 0.0029092783582958717, "optim/total_tokens": 5830606848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8919386863708496, "created_at": "2025-01-16T12:58:21.042380+00:00"} {"global_step": 11122, "acc_step": 0, "speed/wps": 12913.06094658312, "speed/FLOPS": 202817366571309.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04824963957071304, "optim/lr": 0.002909249010933775, "optim/total_tokens": 5831131136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.965118408203125, "created_at": "2025-01-16T12:58:31.200280+00:00"} {"global_step": 11123, "acc_step": 0, "speed/wps": 12902.273379256603, "speed/FLOPS": 202647933002775.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05880455672740936, "optim/lr": 0.0029092196589737556, "optim/total_tokens": 5831655424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.948761224746704, "created_at": "2025-01-16T12:58:41.365411+00:00"} {"global_step": 11124, "acc_step": 0, "speed/wps": 12907.819282708628, "speed/FLOPS": 202735039037359.4, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04713689535856247, "optim/lr": 0.0029091903024159095, "optim/total_tokens": 5832179712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.787153482437134, "created_at": "2025-01-16T12:58:51.525415+00:00"} {"global_step": 11125, "acc_step": 0, "speed/wps": 12907.990206583134, "speed/FLOPS": 202737723631682.56, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045167628675699234, "optim/lr": 0.0029091609412603327, "optim/total_tokens": 5832704000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 2.811514377593994, "created_at": "2025-01-16T12:59:01.688027+00:00"} {"global_step": 11126, "acc_step": 0, "speed/wps": 12902.931191003472, "speed/FLOPS": 202658264840188.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0543820895254612, "optim/lr": 0.0029091315755071213, "optim/total_tokens": 5833228288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8485395908355713, "created_at": "2025-01-16T12:59:11.858806+00:00"} {"global_step": 11127, "acc_step": 0, "speed/wps": 12900.87827810014, "speed/FLOPS": 202626021022043.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06656806170940399, "optim/lr": 0.0029091022051563704, "optim/total_tokens": 5833752576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.837707996368408, "created_at": "2025-01-16T12:59:22.025822+00:00"} {"global_step": 11128, "acc_step": 0, "speed/wps": 12903.452143688637, "speed/FLOPS": 202666447117972.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.09059075266122818, "optim/lr": 0.002909072830208175, "optim/total_tokens": 5834276864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.785560131072998, "created_at": "2025-01-16T12:59:32.190251+00:00"} {"global_step": 11129, "acc_step": 0, "speed/wps": 12903.65983331569, "speed/FLOPS": 202669709168960.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058188099414110184, "optim/lr": 0.0029090434506626327, "optim/total_tokens": 5834801152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.9742918014526367, "created_at": "2025-01-16T12:59:42.351818+00:00"} {"global_step": 11130, "acc_step": 0, "speed/wps": 12898.033933394025, "speed/FLOPS": 202581346679894.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0629640743136406, "optim/lr": 0.002909014066519839, "optim/total_tokens": 5835325440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397090, "loss/out": 2.8692080974578857, "created_at": "2025-01-16T12:59:52.518034+00:00"} {"global_step": 11131, "acc_step": 0, "speed/wps": 12897.381590910913, "speed/FLOPS": 202571100744784.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0706259235739708, "optim/lr": 0.002908984677779889, "optim/total_tokens": 5835849728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.9708096981048584, "created_at": "2025-01-16T13:00:02.691066+00:00"} {"global_step": 11132, "acc_step": 0, "speed/wps": 12897.74859074353, "speed/FLOPS": 202576864981466.6, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05694298818707466, "optim/lr": 0.0029089552844428785, "optim/total_tokens": 5836374016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481564, "loss/out": 2.9441304206848145, "created_at": "2025-01-16T13:00:12.861552+00:00"} {"global_step": 11133, "acc_step": 0, "speed/wps": 12899.957583682784, "speed/FLOPS": 202611560251052.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0576699934899807, "optim/lr": 0.0029089258865089046, "optim/total_tokens": 5836898304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 2.9882326126098633, "created_at": "2025-01-16T13:00:23.029445+00:00"} {"global_step": 11134, "acc_step": 0, "speed/wps": 12893.471733795757, "speed/FLOPS": 202509691066084.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04515053704380989, "optim/lr": 0.0029088964839780623, "optim/total_tokens": 5837422592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505601, "loss/out": 2.8348379135131836, "created_at": "2025-01-16T13:00:33.204472+00:00"} {"global_step": 11135, "acc_step": 0, "speed/wps": 12906.902741616417, "speed/FLOPS": 202720643500046.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059780899435281754, "optim/lr": 0.0029088670768504473, "optim/total_tokens": 5837946880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.031716823577881, "created_at": "2025-01-16T13:00:43.365738+00:00"} {"global_step": 11136, "acc_step": 0, "speed/wps": 12902.277069690252, "speed/FLOPS": 202647990966107.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05126620829105377, "optim/lr": 0.0029088376651261565, "optim/total_tokens": 5838471168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342387, "loss/out": 2.800372362136841, "created_at": "2025-01-16T13:00:53.531535+00:00"} {"global_step": 11137, "acc_step": 0, "speed/wps": 12896.645356480842, "speed/FLOPS": 202559537171368.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045698877424001694, "optim/lr": 0.002908808248805285, "optim/total_tokens": 5838995456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8605830669403076, "created_at": "2025-01-16T13:01:03.701699+00:00"} {"global_step": 11138, "acc_step": 0, "speed/wps": 12893.303263640324, "speed/FLOPS": 202507045010791.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0483904667198658, "optim/lr": 0.002908778827887929, "optim/total_tokens": 5839519744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.7915754318237305, "created_at": "2025-01-16T13:01:13.873545+00:00"} {"global_step": 11139, "acc_step": 0, "speed/wps": 12900.437435051454, "speed/FLOPS": 202619096976181.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04580124840140343, "optim/lr": 0.0029087494023741856, "optim/total_tokens": 5840044032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 2.7789177894592285, "created_at": "2025-01-16T13:01:24.045701+00:00"} {"global_step": 11140, "acc_step": 0, "speed/wps": 12904.690796953768, "speed/FLOPS": 202685901869589.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046826306730508804, "optim/lr": 0.002908719972264149, "optim/total_tokens": 5840568320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.842146635055542, "created_at": "2025-01-16T13:01:34.206336+00:00"} {"global_step": 11141, "acc_step": 0, "speed/wps": 12896.617703618942, "speed/FLOPS": 202559102845173.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.04007735475897789, "optim/lr": 0.0029086905375579164, "optim/total_tokens": 5841092608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8566880226135254, "created_at": "2025-01-16T13:01:44.373798+00:00"} {"global_step": 11142, "acc_step": 0, "speed/wps": 12890.303900525463, "speed/FLOPS": 202459935891515.47, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0024, "optim/grad_norm": 0.047641366720199585, "optim/lr": 0.002908661098255583, "optim/total_tokens": 5841616896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350240, "loss/out": 2.8958072662353516, "created_at": "2025-01-16T13:01:54.547432+00:00"} {"global_step": 11143, "acc_step": 0, "speed/wps": 12893.328422969098, "speed/FLOPS": 202507440172622.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05077333003282547, "optim/lr": 0.002908631654357245, "optim/total_tokens": 5842141184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8640756607055664, "created_at": "2025-01-16T13:02:04.717116+00:00"} {"global_step": 11144, "acc_step": 0, "speed/wps": 12902.259865427048, "speed/FLOPS": 202647720749512.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044367700815200806, "optim/lr": 0.0029086022058629995, "optim/total_tokens": 5842665472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.960063934326172, "created_at": "2025-01-16T13:02:14.882488+00:00"} {"global_step": 11145, "acc_step": 0, "speed/wps": 12900.551300803254, "speed/FLOPS": 202620885394281.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06358514726161957, "optim/lr": 0.002908572752772942, "optim/total_tokens": 5843189760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406246, "loss/out": 2.8712100982666016, "created_at": "2025-01-16T13:02:25.046587+00:00"} {"global_step": 11146, "acc_step": 0, "speed/wps": 12907.77082443905, "speed/FLOPS": 202734277933647.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058931149542331696, "optim/lr": 0.002908543295087168, "optim/total_tokens": 5843714048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.7723796367645264, "created_at": "2025-01-16T13:02:35.211318+00:00"} {"global_step": 11147, "acc_step": 0, "speed/wps": 12905.062984117758, "speed/FLOPS": 202691747580438.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04821847006678581, "optim/lr": 0.0029085138328057737, "optim/total_tokens": 5844238336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.865821599960327, "created_at": "2025-01-16T13:02:45.379587+00:00"} {"global_step": 11148, "acc_step": 0, "speed/wps": 12912.217737031287, "speed/FLOPS": 202804122806606.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05509749799966812, "optim/lr": 0.0029084843659288565, "optim/total_tokens": 5844762624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.8599281311035156, "created_at": "2025-01-16T13:02:55.536421+00:00"} {"global_step": 11149, "acc_step": 0, "speed/wps": 12905.52289591819, "speed/FLOPS": 202698971127249.47, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04500048607587814, "optim/lr": 0.002908454894456511, "optim/total_tokens": 5845286912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 2.7972638607025146, "created_at": "2025-01-16T13:03:05.702578+00:00"} {"global_step": 11150, "acc_step": 0, "speed/wps": 12901.809779406474, "speed/FLOPS": 202640651530076.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.043930839747190475, "optim/lr": 0.002908425418388834, "optim/total_tokens": 5845811200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.87772274017334, "created_at": "2025-01-16T13:03:15.866768+00:00"} {"global_step": 11151, "acc_step": 0, "speed/wps": 12908.626161093436, "speed/FLOPS": 202747712171159.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05330866575241089, "optim/lr": 0.002908395937725922, "optim/total_tokens": 5846335488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.7825493812561035, "created_at": "2025-01-16T13:03:26.024071+00:00"} {"global_step": 11152, "acc_step": 0, "speed/wps": 12910.356459853687, "speed/FLOPS": 202774888890869.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05886496230959892, "optim/lr": 0.0029083664524678704, "optim/total_tokens": 5846859776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.905763626098633, "created_at": "2025-01-16T13:03:36.182343+00:00"} {"global_step": 11153, "acc_step": 0, "speed/wps": 12903.971495057263, "speed/FLOPS": 202674604244880.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0529475063085556, "optim/lr": 0.0029083369626147766, "optim/total_tokens": 5847384064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8313374519348145, "created_at": "2025-01-16T13:03:46.345811+00:00"} {"global_step": 11154, "acc_step": 0, "speed/wps": 12911.49990089857, "speed/FLOPS": 202792848203731.88, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06231596693396568, "optim/lr": 0.002908307468166735, "optim/total_tokens": 5847908352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.785714864730835, "created_at": "2025-01-16T13:03:56.501055+00:00"} {"global_step": 11155, "acc_step": 0, "speed/wps": 12901.111331217902, "speed/FLOPS": 202629681441506.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046530831605196, "optim/lr": 0.0029082779691238436, "optim/total_tokens": 5848432640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296188, "loss/out": 2.8775830268859863, "created_at": "2025-01-16T13:04:06.694157+00:00"} {"global_step": 11156, "acc_step": 0, "speed/wps": 12901.984299481577, "speed/FLOPS": 202643392607671.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056539181619882584, "optim/lr": 0.0029082484654861978, "optim/total_tokens": 5848956928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 2.8812167644500732, "created_at": "2025-01-16T13:04:16.863203+00:00"} {"global_step": 11157, "acc_step": 0, "speed/wps": 12906.721468317977, "speed/FLOPS": 202717796353799.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06035381183028221, "optim/lr": 0.0029082189572538944, "optim/total_tokens": 5849481216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.899326801300049, "created_at": "2025-01-16T13:04:27.024163+00:00"} {"global_step": 11158, "acc_step": 0, "speed/wps": 12903.16307142326, "speed/FLOPS": 202661906840817.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0462326854467392, "optim/lr": 0.002908189444427029, "optim/total_tokens": 5850005504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.9123239517211914, "created_at": "2025-01-16T13:04:37.186001+00:00"} {"global_step": 11159, "acc_step": 0, "speed/wps": 12903.05281258409, "speed/FLOPS": 202660175074237.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054979804903268814, "optim/lr": 0.0029081599270056975, "optim/total_tokens": 5850529792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.883711099624634, "created_at": "2025-01-16T13:04:47.347866+00:00"} {"global_step": 11160, "acc_step": 0, "speed/wps": 12901.043675185001, "speed/FLOPS": 202628618810540.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05332358926534653, "optim/lr": 0.002908130404989998, "optim/total_tokens": 5851054080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.9190990924835205, "created_at": "2025-01-16T13:04:57.512271+00:00"} {"global_step": 11161, "acc_step": 0, "speed/wps": 12905.06778257336, "speed/FLOPS": 202691822946777.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048689115792512894, "optim/lr": 0.002908100878380025, "optim/total_tokens": 5851578368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 2.8572702407836914, "created_at": "2025-01-16T13:05:07.677047+00:00"} {"global_step": 11162, "acc_step": 0, "speed/wps": 12903.924668819924, "speed/FLOPS": 202673868774477.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06338829547166824, "optim/lr": 0.002908071347175875, "optim/total_tokens": 5852102656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9546027183532715, "created_at": "2025-01-16T13:05:17.838603+00:00"} {"global_step": 11163, "acc_step": 0, "speed/wps": 12902.517349852114, "speed/FLOPS": 202651764896228.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05290042236447334, "optim/lr": 0.002908041811377646, "optim/total_tokens": 5852626944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.749509572982788, "created_at": "2025-01-16T13:05:28.000842+00:00"} {"global_step": 11164, "acc_step": 0, "speed/wps": 12904.01050456106, "speed/FLOPS": 202675216942742.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0474482886493206, "optim/lr": 0.002908012270985433, "optim/total_tokens": 5853151232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.9395079612731934, "created_at": "2025-01-16T13:05:38.165013+00:00"} {"global_step": 11165, "acc_step": 0, "speed/wps": 12901.019923779848, "speed/FLOPS": 202628245762084.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052318476140499115, "optim/lr": 0.002907982725999332, "optim/total_tokens": 5853675520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.0194454193115234, "created_at": "2025-01-16T13:05:48.332969+00:00"} {"global_step": 11166, "acc_step": 0, "speed/wps": 12903.634011350172, "speed/FLOPS": 202669303599509.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06222639977931976, "optim/lr": 0.0029079531764194407, "optim/total_tokens": 5854199808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306453, "loss/out": 2.8666088581085205, "created_at": "2025-01-16T13:05:58.494730+00:00"} {"global_step": 11167, "acc_step": 0, "speed/wps": 12905.00326701508, "speed/FLOPS": 202690809641281.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057855743914842606, "optim/lr": 0.0029079236222458544, "optim/total_tokens": 5854724096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8981940746307373, "created_at": "2025-01-16T13:06:08.655147+00:00"} {"global_step": 11168, "acc_step": 0, "speed/wps": 12903.173966808048, "speed/FLOPS": 202662077967807.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051190346479415894, "optim/lr": 0.00290789406347867, "optim/total_tokens": 5855248384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.869180917739868, "created_at": "2025-01-16T13:06:18.819076+00:00"} {"global_step": 11169, "acc_step": 0, "speed/wps": 12898.439971009035, "speed/FLOPS": 202587724058592.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04788631945848465, "optim/lr": 0.0029078645001179844, "optim/total_tokens": 5855772672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 2.9289894104003906, "created_at": "2025-01-16T13:06:28.984998+00:00"} {"global_step": 11170, "acc_step": 0, "speed/wps": 12913.012008203446, "speed/FLOPS": 202816597926808.12, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0487644299864769, "optim/lr": 0.002907834932163893, "optim/total_tokens": 5856296960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 2.866638660430908, "created_at": "2025-01-16T13:06:39.140832+00:00"} {"global_step": 11171, "acc_step": 0, "speed/wps": 12899.984474024977, "speed/FLOPS": 202611982600827.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05491148307919502, "optim/lr": 0.0029078053596164927, "optim/total_tokens": 5856821248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8802084922790527, "created_at": "2025-01-16T13:06:49.306637+00:00"} {"global_step": 11172, "acc_step": 0, "speed/wps": 12888.541208011584, "speed/FLOPS": 202432250383391.66, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0443386547267437, "optim/lr": 0.002907775782475881, "optim/total_tokens": 5857345536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.932284355163574, "created_at": "2025-01-16T13:06:59.480824+00:00"} {"global_step": 11173, "acc_step": 0, "speed/wps": 12896.959892937655, "speed/FLOPS": 202564477398641.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05635958909988403, "optim/lr": 0.002907746200742153, "optim/total_tokens": 5857869824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.964099884033203, "created_at": "2025-01-16T13:07:09.647492+00:00"} {"global_step": 11174, "acc_step": 0, "speed/wps": 12905.901409526543, "speed/FLOPS": 202704916203600.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050493866205215454, "optim/lr": 0.0029077166144154056, "optim/total_tokens": 5858394112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 2.8657312393188477, "created_at": "2025-01-16T13:07:19.832668+00:00"} {"global_step": 11175, "acc_step": 0, "speed/wps": 12907.4242174838, "speed/FLOPS": 202728833995127.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06691158562898636, "optim/lr": 0.0029076870234957353, "optim/total_tokens": 5858918400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.880187749862671, "created_at": "2025-01-16T13:07:29.993163+00:00"} {"global_step": 11176, "acc_step": 0, "speed/wps": 12894.599855241964, "speed/FLOPS": 202527409763598.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05628411844372749, "optim/lr": 0.0029076574279832392, "optim/total_tokens": 5859442688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.8091678619384766, "created_at": "2025-01-16T13:07:40.161639+00:00"} {"global_step": 11177, "acc_step": 0, "speed/wps": 12901.690867223779, "speed/FLOPS": 202638783850844.5, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05222201347351074, "optim/lr": 0.0029076278278780134, "optim/total_tokens": 5859966976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.9677650928497314, "created_at": "2025-01-16T13:07:50.324887+00:00"} {"global_step": 11178, "acc_step": 0, "speed/wps": 12901.435818984151, "speed/FLOPS": 202634777967768.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05552971735596657, "optim/lr": 0.0029075982231801544, "optim/total_tokens": 5860491264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 2.8765952587127686, "created_at": "2025-01-16T13:08:00.488651+00:00"} {"global_step": 11179, "acc_step": 0, "speed/wps": 12906.234517880555, "speed/FLOPS": 202710148128039.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04558819904923439, "optim/lr": 0.0029075686138897586, "optim/total_tokens": 5861015552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8837618827819824, "created_at": "2025-01-16T13:08:10.649499+00:00"} {"global_step": 11180, "acc_step": 0, "speed/wps": 12896.478188413334, "speed/FLOPS": 202556911567155.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06021245941519737, "optim/lr": 0.002907539000006924, "optim/total_tokens": 5861539840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 2.924302577972412, "created_at": "2025-01-16T13:08:20.817739+00:00"} {"global_step": 11181, "acc_step": 0, "speed/wps": 12898.089744304492, "speed/FLOPS": 202582223266935.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04618959128856659, "optim/lr": 0.0029075093815317454, "optim/total_tokens": 5862064128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.898517370223999, "created_at": "2025-01-16T13:08:30.983392+00:00"} {"global_step": 11182, "acc_step": 0, "speed/wps": 12896.38945238055, "speed/FLOPS": 202555517845824.2, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045166339725255966, "optim/lr": 0.00290747975846432, "optim/total_tokens": 5862588416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.7370083332061768, "created_at": "2025-01-16T13:08:41.150347+00:00"} {"global_step": 11183, "acc_step": 0, "speed/wps": 12904.30359745941, "speed/FLOPS": 202679820369462.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048301804810762405, "optim/lr": 0.002907450130804745, "optim/total_tokens": 5863112704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 2.8416666984558105, "created_at": "2025-01-16T13:08:51.314618+00:00"} {"global_step": 11184, "acc_step": 0, "speed/wps": 12903.208493824315, "speed/FLOPS": 202662620262043.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04293356090784073, "optim/lr": 0.0029074204985531167, "optim/total_tokens": 5863636992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.8095905780792236, "created_at": "2025-01-16T13:09:01.476265+00:00"} {"global_step": 11185, "acc_step": 0, "speed/wps": 12902.157281894139, "speed/FLOPS": 202646109534163.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04822452738881111, "optim/lr": 0.002907390861709532, "optim/total_tokens": 5864161280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8159008026123047, "created_at": "2025-01-16T13:09:11.643608+00:00"} {"global_step": 11186, "acc_step": 0, "speed/wps": 12901.697164050263, "speed/FLOPS": 202638882751157.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050167813897132874, "optim/lr": 0.002907361220274087, "optim/total_tokens": 5864685568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 3.018986225128174, "created_at": "2025-01-16T13:09:21.809576+00:00"} {"global_step": 11187, "acc_step": 0, "speed/wps": 12899.337823225014, "speed/FLOPS": 202601826061422.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05002312734723091, "optim/lr": 0.002907331574246879, "optim/total_tokens": 5865209856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 2.812169313430786, "created_at": "2025-01-16T13:09:31.976457+00:00"} {"global_step": 11188, "acc_step": 0, "speed/wps": 12904.510120606628, "speed/FLOPS": 202683064099282.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04362430423498154, "optim/lr": 0.0029073019236280045, "optim/total_tokens": 5865734144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.788698673248291, "created_at": "2025-01-16T13:09:42.139768+00:00"} {"global_step": 11189, "acc_step": 0, "speed/wps": 12899.077832824772, "speed/FLOPS": 202597742554922.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05126815661787987, "optim/lr": 0.00290727226841756, "optim/total_tokens": 5866258432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.8546950817108154, "created_at": "2025-01-16T13:09:52.304510+00:00"} {"global_step": 11190, "acc_step": 0, "speed/wps": 12900.418761127576, "speed/FLOPS": 202618803676546.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.055283889174461365, "optim/lr": 0.0029072426086156427, "optim/total_tokens": 5866782720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9047374725341797, "created_at": "2025-01-16T13:10:02.468462+00:00"} {"global_step": 11191, "acc_step": 0, "speed/wps": 12901.179663242357, "speed/FLOPS": 202630754689848.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04817938804626465, "optim/lr": 0.0029072129442223498, "optim/total_tokens": 5867307008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.9122891426086426, "created_at": "2025-01-16T13:10:12.634603+00:00"} {"global_step": 11192, "acc_step": 0, "speed/wps": 12901.304635375984, "speed/FLOPS": 202632717548936.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05204342305660248, "optim/lr": 0.0029071832752377764, "optim/total_tokens": 5867831296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.7416162490844727, "created_at": "2025-01-16T13:10:22.797728+00:00"} {"global_step": 11193, "acc_step": 0, "speed/wps": 12903.669583128047, "speed/FLOPS": 202669862303161.62, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06729527562856674, "optim/lr": 0.002907153601662021, "optim/total_tokens": 5868355584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.840031623840332, "created_at": "2025-01-16T13:10:32.959161+00:00"} {"global_step": 11194, "acc_step": 0, "speed/wps": 12900.421304666079, "speed/FLOPS": 202618843626314.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048696935176849365, "optim/lr": 0.00290712392349518, "optim/total_tokens": 5868879872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8525424003601074, "created_at": "2025-01-16T13:10:43.124148+00:00"} {"global_step": 11195, "acc_step": 0, "speed/wps": 12894.190234147345, "speed/FLOPS": 202520976101429.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05692312493920326, "optim/lr": 0.002907094240737349, "optim/total_tokens": 5869404160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.7753114700317383, "created_at": "2025-01-16T13:10:53.292858+00:00"} {"global_step": 11196, "acc_step": 0, "speed/wps": 12904.21336464535, "speed/FLOPS": 202678403139124.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05644763261079788, "optim/lr": 0.002907064553388627, "optim/total_tokens": 5869928448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470430, "loss/out": 2.9928784370422363, "created_at": "2025-01-16T13:11:03.454827+00:00"} {"global_step": 11197, "acc_step": 0, "speed/wps": 12895.930217308656, "speed/FLOPS": 202548304927964.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05119685083627701, "optim/lr": 0.0029070348614491093, "optim/total_tokens": 5870452736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.885697841644287, "created_at": "2025-01-16T13:11:13.623415+00:00"} {"global_step": 11198, "acc_step": 0, "speed/wps": 12898.615034750088, "speed/FLOPS": 202590473675208.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0508972592651844, "optim/lr": 0.002907005164918893, "optim/total_tokens": 5870977024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9168190956115723, "created_at": "2025-01-16T13:11:23.788836+00:00"} {"global_step": 11199, "acc_step": 0, "speed/wps": 12899.26940903389, "speed/FLOPS": 202600751522540.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0477728471159935, "optim/lr": 0.0029069754637980755, "optim/total_tokens": 5871501312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.8232288360595703, "created_at": "2025-01-16T13:11:33.954336+00:00"} {"global_step": 11200, "acc_step": 0, "speed/wps": 12901.20519746656, "speed/FLOPS": 202631155739927.2, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05991436168551445, "optim/lr": 0.0029069457580867533, "optim/total_tokens": 5872025600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.894819736480713, "created_at": "2025-01-16T13:11:44.120114+00:00"} {"global_step": 11201, "acc_step": 0, "speed/wps": 12904.203169133723, "speed/FLOPS": 202678243004604.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05906488746404648, "optim/lr": 0.002906916047785023, "optim/total_tokens": 5872549888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.833195209503174, "created_at": "2025-01-16T13:11:54.290025+00:00"} {"global_step": 11202, "acc_step": 0, "speed/wps": 12903.144975188015, "speed/FLOPS": 202661622614576.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049997132271528244, "optim/lr": 0.0029068863328929827, "optim/total_tokens": 5873074176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.8850393295288086, "created_at": "2025-01-16T13:12:04.452795+00:00"} {"global_step": 11203, "acc_step": 0, "speed/wps": 12899.826405304519, "speed/FLOPS": 202609499914363.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057966116815805435, "optim/lr": 0.0029068566134107282, "optim/total_tokens": 5873598464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8780224323272705, "created_at": "2025-01-16T13:12:14.617951+00:00"} {"global_step": 11204, "acc_step": 0, "speed/wps": 12905.866537903448, "speed/FLOPS": 202704368496841.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04670599848031998, "optim/lr": 0.002906826889338357, "optim/total_tokens": 5874122752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8652162551879883, "created_at": "2025-01-16T13:12:24.782405+00:00"} {"global_step": 11205, "acc_step": 0, "speed/wps": 12907.34141764922, "speed/FLOPS": 202727533509946.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047777753323316574, "optim/lr": 0.0029067971606759656, "optim/total_tokens": 5874647040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.927835702896118, "created_at": "2025-01-16T13:12:34.940626+00:00"} {"global_step": 11206, "acc_step": 0, "speed/wps": 12898.14890096891, "speed/FLOPS": 202583152403640.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05530289188027382, "optim/lr": 0.002906767427423652, "optim/total_tokens": 5875171328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.8493995666503906, "created_at": "2025-01-16T13:12:45.106256+00:00"} {"global_step": 11207, "acc_step": 0, "speed/wps": 12899.637736495815, "speed/FLOPS": 202606536611464.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042975813150405884, "optim/lr": 0.002906737689581512, "optim/total_tokens": 5875695616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 2.9578921794891357, "created_at": "2025-01-16T13:12:55.271945+00:00"} {"global_step": 11208, "acc_step": 0, "speed/wps": 12902.314021456385, "speed/FLOPS": 202648571344371.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04329656437039375, "optim/lr": 0.0029067079471496436, "optim/total_tokens": 5876219904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8198657035827637, "created_at": "2025-01-16T13:13:05.434526+00:00"} {"global_step": 11209, "acc_step": 0, "speed/wps": 12901.109131548976, "speed/FLOPS": 202629646892683.12, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.044442787766456604, "optim/lr": 0.002906678200128143, "optim/total_tokens": 5876744192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.9215946197509766, "created_at": "2025-01-16T13:13:15.599534+00:00"} {"global_step": 11210, "acc_step": 0, "speed/wps": 12904.075087099014, "speed/FLOPS": 202676231300246.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04877300560474396, "optim/lr": 0.0029066484485171075, "optim/total_tokens": 5877268480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.892979621887207, "created_at": "2025-01-16T13:13:25.763539+00:00"} {"global_step": 11211, "acc_step": 0, "speed/wps": 12900.207292645227, "speed/FLOPS": 202615482273443.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04584171622991562, "optim/lr": 0.0029066186923166347, "optim/total_tokens": 5877792768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9291343688964844, "created_at": "2025-01-16T13:13:35.929170+00:00"} {"global_step": 11212, "acc_step": 0, "speed/wps": 12905.582325520947, "speed/FLOPS": 202699904550825.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05355530604720116, "optim/lr": 0.0029065889315268213, "optim/total_tokens": 5878317056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.0057783126831055, "created_at": "2025-01-16T13:13:46.091061+00:00"} {"global_step": 11213, "acc_step": 0, "speed/wps": 12896.525329336542, "speed/FLOPS": 202557651980132.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07956861704587936, "optim/lr": 0.0029065591661477648, "optim/total_tokens": 5878841344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.0171170234680176, "created_at": "2025-01-16T13:13:56.263108+00:00"} {"global_step": 11214, "acc_step": 0, "speed/wps": 12901.198568800153, "speed/FLOPS": 202631051627612.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05133332684636116, "optim/lr": 0.0029065293961795615, "optim/total_tokens": 5879365632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.94230055809021, "created_at": "2025-01-16T13:14:06.427401+00:00"} {"global_step": 11215, "acc_step": 0, "speed/wps": 12902.504612659235, "speed/FLOPS": 202651564841111.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050860047340393066, "optim/lr": 0.002906499621622309, "optim/total_tokens": 5879889920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489248, "loss/out": 2.763472557067871, "created_at": "2025-01-16T13:14:16.589891+00:00"} {"global_step": 11216, "acc_step": 0, "speed/wps": 12894.127770602532, "speed/FLOPS": 202519995025623.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046788107603788376, "optim/lr": 0.0029064698424761046, "optim/total_tokens": 5880414208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.9459900856018066, "created_at": "2025-01-16T13:14:26.761492+00:00"} {"global_step": 11217, "acc_step": 0, "speed/wps": 12900.609396855774, "speed/FLOPS": 202621797872618.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04671990126371384, "optim/lr": 0.002906440058741045, "optim/total_tokens": 5880938496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.7869224548339844, "created_at": "2025-01-16T13:14:36.925576+00:00"} {"global_step": 11218, "acc_step": 0, "speed/wps": 12899.13423798614, "speed/FLOPS": 202598628475490.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047044701874256134, "optim/lr": 0.002906410270417228, "optim/total_tokens": 5881462784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.923691749572754, "created_at": "2025-01-16T13:14:47.092266+00:00"} {"global_step": 11219, "acc_step": 0, "speed/wps": 12905.195619241156, "speed/FLOPS": 202693830797311.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05770350620150566, "optim/lr": 0.00290638047750475, "optim/total_tokens": 5881987072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.868290901184082, "created_at": "2025-01-16T13:14:57.257347+00:00"} {"global_step": 11220, "acc_step": 0, "speed/wps": 12903.08074738964, "speed/FLOPS": 202660613828744.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08022384345531464, "optim/lr": 0.002906350680003709, "optim/total_tokens": 5882511360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9212474822998047, "created_at": "2025-01-16T13:15:07.419199+00:00"} {"global_step": 11221, "acc_step": 0, "speed/wps": 12905.888665651393, "speed/FLOPS": 202704716043529.72, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06025972589850426, "optim/lr": 0.002906320877914202, "optim/total_tokens": 5883035648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.8637635707855225, "created_at": "2025-01-16T13:15:17.580997+00:00"} {"global_step": 11222, "acc_step": 0, "speed/wps": 12903.612921792583, "speed/FLOPS": 202668972359027.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0543086938560009, "optim/lr": 0.002906291071236326, "optim/total_tokens": 5883559936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.9147353172302246, "created_at": "2025-01-16T13:15:27.742507+00:00"} {"global_step": 11223, "acc_step": 0, "speed/wps": 12905.004057480599, "speed/FLOPS": 202690822056628.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05903296917676926, "optim/lr": 0.002906261259970178, "optim/total_tokens": 5884084224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485284, "loss/out": 2.787198066711426, "created_at": "2025-01-16T13:15:37.904751+00:00"} {"global_step": 11224, "acc_step": 0, "speed/wps": 12905.530014234562, "speed/FLOPS": 202699082930189.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06064041703939438, "optim/lr": 0.0029062314441158556, "optim/total_tokens": 5884608512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.012239456176758, "created_at": "2025-01-16T13:15:48.072469+00:00"} {"global_step": 11225, "acc_step": 0, "speed/wps": 12904.17939307863, "speed/FLOPS": 202677869568987.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05547189339995384, "optim/lr": 0.0029062016236734564, "optim/total_tokens": 5885132800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.877810478210449, "created_at": "2025-01-16T13:15:58.233340+00:00"} {"global_step": 11226, "acc_step": 0, "speed/wps": 12901.331849469498, "speed/FLOPS": 202633144983671.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0495641827583313, "optim/lr": 0.0029061717986430774, "optim/total_tokens": 5885657088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.830374240875244, "created_at": "2025-01-16T13:16:08.398405+00:00"} {"global_step": 11227, "acc_step": 0, "speed/wps": 12900.760743024022, "speed/FLOPS": 202624174972162.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07521311193704605, "optim/lr": 0.0029061419690248162, "optim/total_tokens": 5886181376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8705925941467285, "created_at": "2025-01-16T13:16:18.564575+00:00"} {"global_step": 11228, "acc_step": 0, "speed/wps": 12910.190976668657, "speed/FLOPS": 202772289750050.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07040394842624664, "optim/lr": 0.002906112134818769, "optim/total_tokens": 5886705664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440101, "loss/out": 2.8427047729492188, "created_at": "2025-01-16T13:16:28.721260+00:00"} {"global_step": 11229, "acc_step": 0, "speed/wps": 12901.028944667702, "speed/FLOPS": 202628387447524.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05462340638041496, "optim/lr": 0.0029060822960250347, "optim/total_tokens": 5887229952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 2.875674247741699, "created_at": "2025-01-16T13:16:38.887940+00:00"} {"global_step": 11230, "acc_step": 0, "speed/wps": 12897.89476934056, "speed/FLOPS": 202579160917203.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06051100417971611, "optim/lr": 0.0029060524526437095, "optim/total_tokens": 5887754240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.869882822036743, "created_at": "2025-01-16T13:16:49.053835+00:00"} {"global_step": 11231, "acc_step": 0, "speed/wps": 12904.114097229145, "speed/FLOPS": 202676844007945.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046232372522354126, "optim/lr": 0.002906022604674891, "optim/total_tokens": 5888278528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428623, "loss/out": 2.917464256286621, "created_at": "2025-01-16T13:16:59.214869+00:00"} {"global_step": 11232, "acc_step": 0, "speed/wps": 12903.423393342113, "speed/FLOPS": 202665995554273.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04865938052535057, "optim/lr": 0.0029059927521186773, "optim/total_tokens": 5888802816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.8867733478546143, "created_at": "2025-01-16T13:17:09.379804+00:00"} {"global_step": 11233, "acc_step": 0, "speed/wps": 12905.405646602334, "speed/FLOPS": 202697129565626.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04775340482592583, "optim/lr": 0.002905962894975165, "optim/total_tokens": 5889327104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367082, "loss/out": 2.871772527694702, "created_at": "2025-01-16T13:17:19.542071+00:00"} {"global_step": 11234, "acc_step": 0, "speed/wps": 12904.471220489273, "speed/FLOPS": 202682453119484.84, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05113748461008072, "optim/lr": 0.0029059330332444518, "optim/total_tokens": 5889851392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9523091316223145, "created_at": "2025-01-16T13:17:29.706497+00:00"} {"global_step": 11235, "acc_step": 0, "speed/wps": 12908.794799852969, "speed/FLOPS": 202750360874612.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06345131993293762, "optim/lr": 0.0029059031669266353, "optim/total_tokens": 5890375680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.014151096343994, "created_at": "2025-01-16T13:17:39.866414+00:00"} {"global_step": 11236, "acc_step": 0, "speed/wps": 12905.519534986837, "speed/FLOPS": 202698918339204.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04757820442318916, "optim/lr": 0.0029058732960218128, "optim/total_tokens": 5890899968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 2.810641050338745, "created_at": "2025-01-16T13:17:50.028730+00:00"} {"global_step": 11237, "acc_step": 0, "speed/wps": 12906.533006112957, "speed/FLOPS": 202714836295895.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0628308653831482, "optim/lr": 0.0029058434205300816, "optim/total_tokens": 5891424256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9278934001922607, "created_at": "2025-01-16T13:18:00.189811+00:00"} {"global_step": 11238, "acc_step": 0, "speed/wps": 12903.101733471174, "speed/FLOPS": 202660943443993.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051674239337444305, "optim/lr": 0.002905813540451539, "optim/total_tokens": 5891948544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9401278495788574, "created_at": "2025-01-16T13:18:10.354028+00:00"} {"global_step": 11239, "acc_step": 0, "speed/wps": 12899.846476276027, "speed/FLOPS": 202609815156551.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04521029070019722, "optim/lr": 0.002905783655786283, "optim/total_tokens": 5892472832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.865757465362549, "created_at": "2025-01-16T13:18:20.520423+00:00"} {"global_step": 11240, "acc_step": 0, "speed/wps": 12903.632776214303, "speed/FLOPS": 202669284200003.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048377253115177155, "optim/lr": 0.0029057537665344107, "optim/total_tokens": 5892997120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8222994804382324, "created_at": "2025-01-16T13:18:30.683723+00:00"} {"global_step": 11241, "acc_step": 0, "speed/wps": 12905.20253129045, "speed/FLOPS": 202693939360543.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0597008541226387, "optim/lr": 0.0029057238726960203, "optim/total_tokens": 5893521408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.857978343963623, "created_at": "2025-01-16T13:18:40.845869+00:00"} {"global_step": 11242, "acc_step": 0, "speed/wps": 12904.554695819566, "speed/FLOPS": 202683764214254.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05522449314594269, "optim/lr": 0.002905693974271208, "optim/total_tokens": 5894045696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.848360538482666, "created_at": "2025-01-16T13:18:51.009534+00:00"} {"global_step": 11243, "acc_step": 0, "speed/wps": 12906.052026568894, "speed/FLOPS": 202707281851220.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05142873898148537, "optim/lr": 0.0029056640712600734, "optim/total_tokens": 5894569984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.7543745040893555, "created_at": "2025-01-16T13:19:01.171692+00:00"} {"global_step": 11244, "acc_step": 0, "speed/wps": 12903.49644540543, "speed/FLOPS": 202667142937311.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05612725764513016, "optim/lr": 0.002905634163662712, "optim/total_tokens": 5895094272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 2.8188657760620117, "created_at": "2025-01-16T13:19:11.336158+00:00"} {"global_step": 11245, "acc_step": 0, "speed/wps": 12903.533658514756, "speed/FLOPS": 202667727420329.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04923398047685623, "optim/lr": 0.002905604251479222, "optim/total_tokens": 5895618560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.9145607948303223, "created_at": "2025-01-16T13:19:21.500529+00:00"} {"global_step": 11246, "acc_step": 0, "speed/wps": 12909.282327165074, "speed/FLOPS": 202758018160981.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040865834802389145, "optim/lr": 0.002905574334709702, "optim/total_tokens": 5896142848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.8202404975891113, "created_at": "2025-01-16T13:19:31.660185+00:00"} {"global_step": 11247, "acc_step": 0, "speed/wps": 12900.8743353957, "speed/FLOPS": 202625959096452.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04786909371614456, "optim/lr": 0.0029055444133542483, "optim/total_tokens": 5896667136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.844290256500244, "created_at": "2025-01-16T13:19:41.824649+00:00"} {"global_step": 11248, "acc_step": 0, "speed/wps": 12905.448484834342, "speed/FLOPS": 202697802398926.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043865494430065155, "optim/lr": 0.002905514487412959, "optim/total_tokens": 5897191424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8667421340942383, "created_at": "2025-01-16T13:19:51.988079+00:00"} {"global_step": 11249, "acc_step": 0, "speed/wps": 12902.335810559553, "speed/FLOPS": 202648913572177.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05047052353620529, "optim/lr": 0.002905484556885932, "optim/total_tokens": 5897715712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8517372608184814, "created_at": "2025-01-16T13:20:02.153513+00:00"} {"global_step": 11250, "acc_step": 0, "speed/wps": 12903.783544986936, "speed/FLOPS": 202671652230755.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0483773872256279, "optim/lr": 0.002905454621773265, "optim/total_tokens": 5898240000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.923234462738037, "created_at": "2025-01-16T13:20:12.314847+00:00"} {"global_step": 11251, "acc_step": 0, "speed/wps": 12903.297794563088, "speed/FLOPS": 202664022852857.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05658411979675293, "optim/lr": 0.0029054246820750555, "optim/total_tokens": 5898764288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453382, "loss/out": 2.9801831245422363, "created_at": "2025-01-16T13:20:22.478720+00:00"} {"global_step": 11252, "acc_step": 0, "speed/wps": 12904.209596196852, "speed/FLOPS": 202678343950462.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055491853505373, "optim/lr": 0.0029053947377914004, "optim/total_tokens": 5899288576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8098556995391846, "created_at": "2025-01-16T13:20:32.649213+00:00"} {"global_step": 11253, "acc_step": 0, "speed/wps": 12902.26755263393, "speed/FLOPS": 202647841487659.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05199594050645828, "optim/lr": 0.002905364788922398, "optim/total_tokens": 5899812864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.8901844024658203, "created_at": "2025-01-16T13:20:42.811629+00:00"} {"global_step": 11254, "acc_step": 0, "speed/wps": 12906.467635547735, "speed/FLOPS": 202713809561339.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04561278969049454, "optim/lr": 0.0029053348354681472, "optim/total_tokens": 5900337152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 2.7935843467712402, "created_at": "2025-01-16T13:20:52.971449+00:00"} {"global_step": 11255, "acc_step": 0, "speed/wps": 12910.802519695977, "speed/FLOPS": 202781894873642.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049623213708400726, "optim/lr": 0.002905304877428744, "optim/total_tokens": 5900861440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.8949358463287354, "created_at": "2025-01-16T13:21:03.128045+00:00"} {"global_step": 11256, "acc_step": 0, "speed/wps": 12908.759404647506, "speed/FLOPS": 202749804944272.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04681253060698509, "optim/lr": 0.002905274914804287, "optim/total_tokens": 5901385728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 2.8720510005950928, "created_at": "2025-01-16T13:21:13.286615+00:00"} {"global_step": 11257, "acc_step": 0, "speed/wps": 12908.36926037676, "speed/FLOPS": 202743677192385.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04617703706026077, "optim/lr": 0.002905244947594873, "optim/total_tokens": 5901910016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.961552143096924, "created_at": "2025-01-16T13:21:23.450983+00:00"} {"global_step": 11258, "acc_step": 0, "speed/wps": 12907.660231133843, "speed/FLOPS": 202732540913817.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053208958357572556, "optim/lr": 0.0029052149758006015, "optim/total_tokens": 5902434304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.786067247390747, "created_at": "2025-01-16T13:21:33.610584+00:00"} {"global_step": 11259, "acc_step": 0, "speed/wps": 12902.686354247453, "speed/FLOPS": 202654419342493.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06258898228406906, "optim/lr": 0.002905184999421569, "optim/total_tokens": 5902958592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.809706211090088, "created_at": "2025-01-16T13:21:43.772598+00:00"} {"global_step": 11260, "acc_step": 0, "speed/wps": 12910.547735235323, "speed/FLOPS": 202777893133577.84, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06619153916835785, "optim/lr": 0.0029051550184578732, "optim/total_tokens": 5903482880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.9307732582092285, "created_at": "2025-01-16T13:21:53.930887+00:00"} {"global_step": 11261, "acc_step": 0, "speed/wps": 12911.59011837464, "speed/FLOPS": 202794265193165.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05360981076955795, "optim/lr": 0.0029051250329096123, "optim/total_tokens": 5904007168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8183112144470215, "created_at": "2025-01-16T13:22:04.085904+00:00"} {"global_step": 11262, "acc_step": 0, "speed/wps": 12905.043799797746, "speed/FLOPS": 202691446264330.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0597546249628067, "optim/lr": 0.002905095042776885, "optim/total_tokens": 5904531456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8456873893737793, "created_at": "2025-01-16T13:22:14.247396+00:00"} {"global_step": 11263, "acc_step": 0, "speed/wps": 12906.72817959807, "speed/FLOPS": 202717901763675.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05169496685266495, "optim/lr": 0.0029050650480597876, "optim/total_tokens": 5905055744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 2.7531802654266357, "created_at": "2025-01-16T13:22:24.409907+00:00"} {"global_step": 11264, "acc_step": 0, "speed/wps": 12909.360924554785, "speed/FLOPS": 202759252640991.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059960637241601944, "optim/lr": 0.0029050350487584186, "optim/total_tokens": 5905580032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.939992904663086, "created_at": "2025-01-16T13:22:34.569171+00:00"} {"global_step": 11265, "acc_step": 0, "speed/wps": 12902.56765301992, "speed/FLOPS": 202652554976602.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05833979323506355, "optim/lr": 0.0029050050448728762, "optim/total_tokens": 5906104320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 2.955994129180908, "created_at": "2025-01-16T13:22:44.731273+00:00"} {"global_step": 11266, "acc_step": 0, "speed/wps": 12899.152947205936, "speed/FLOPS": 202598922329496.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05422559380531311, "optim/lr": 0.002904975036403258, "optim/total_tokens": 5906628608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8795828819274902, "created_at": "2025-01-16T13:22:54.896318+00:00"} {"global_step": 11267, "acc_step": 0, "speed/wps": 12907.573951440798, "speed/FLOPS": 202731185772677.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0677674412727356, "optim/lr": 0.0029049450233496624, "optim/total_tokens": 5907152896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.728959083557129, "created_at": "2025-01-16T13:23:05.056027+00:00"} {"global_step": 11268, "acc_step": 0, "speed/wps": 12899.693625134098, "speed/FLOPS": 202607414419327.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060881875455379486, "optim/lr": 0.0029049150057121865, "optim/total_tokens": 5907677184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294596, "loss/out": 2.9829659461975098, "created_at": "2025-01-16T13:23:15.221433+00:00"} {"global_step": 11269, "acc_step": 0, "speed/wps": 12899.364316096027, "speed/FLOPS": 202602242168366.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05112725868821144, "optim/lr": 0.002904884983490929, "optim/total_tokens": 5908201472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.8959169387817383, "created_at": "2025-01-16T13:23:25.398126+00:00"} {"global_step": 11270, "acc_step": 0, "speed/wps": 12903.810082150987, "speed/FLOPS": 202672069033382.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05050303414463997, "optim/lr": 0.0029048549566859866, "optim/total_tokens": 5908725760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.0082216262817383, "created_at": "2025-01-16T13:23:35.559639+00:00"} {"global_step": 11271, "acc_step": 0, "speed/wps": 12899.629671706803, "speed/FLOPS": 202606409942870.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0487230159342289, "optim/lr": 0.0029048249252974583, "optim/total_tokens": 5909250048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 2.90242862701416, "created_at": "2025-01-16T13:23:45.723964+00:00"} {"global_step": 11272, "acc_step": 0, "speed/wps": 12905.415486625217, "speed/FLOPS": 202697284116707.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05141829326748848, "optim/lr": 0.0029047948893254423, "optim/total_tokens": 5909774336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9416110515594482, "created_at": "2025-01-16T13:23:55.884217+00:00"} {"global_step": 11273, "acc_step": 0, "speed/wps": 12907.046200482018, "speed/FLOPS": 202722896718664.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06815075874328613, "optim/lr": 0.0029047648487700362, "optim/total_tokens": 5910298624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 2.8999791145324707, "created_at": "2025-01-16T13:24:06.042819+00:00"} {"global_step": 11274, "acc_step": 0, "speed/wps": 12904.000015949387, "speed/FLOPS": 202675052204683.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04774933308362961, "optim/lr": 0.002904734803631338, "optim/total_tokens": 5910822912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.8756370544433594, "created_at": "2025-01-16T13:24:16.205879+00:00"} {"global_step": 11275, "acc_step": 0, "speed/wps": 12907.344689563735, "speed/FLOPS": 202727584899859.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05733652412891388, "optim/lr": 0.002904704753909446, "optim/total_tokens": 5911347200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9932713508605957, "created_at": "2025-01-16T13:24:26.364471+00:00"} {"global_step": 11276, "acc_step": 0, "speed/wps": 12901.707470326857, "speed/FLOPS": 202639044625393.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04968211054801941, "optim/lr": 0.002904674699604457, "optim/total_tokens": 5911871488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.908456325531006, "created_at": "2025-01-16T13:24:36.529254+00:00"} {"global_step": 11277, "acc_step": 0, "speed/wps": 12906.309547568122, "speed/FLOPS": 202711326572384.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049323663115501404, "optim/lr": 0.002904644640716471, "optim/total_tokens": 5912395776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9280219078063965, "created_at": "2025-01-16T13:24:46.692163+00:00"} {"global_step": 11278, "acc_step": 0, "speed/wps": 12902.928060914368, "speed/FLOPS": 202658215677837.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04453301802277565, "optim/lr": 0.002904614577245585, "optim/total_tokens": 5912920064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396559, "loss/out": 2.8001928329467773, "created_at": "2025-01-16T13:24:56.856411+00:00"} {"global_step": 11279, "acc_step": 0, "speed/wps": 12904.34684841538, "speed/FLOPS": 202680499685159.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052686892449855804, "optim/lr": 0.002904584509191897, "optim/total_tokens": 5913444352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8387670516967773, "created_at": "2025-01-16T13:25:07.018187+00:00"} {"global_step": 11280, "acc_step": 0, "speed/wps": 12903.789462895205, "speed/FLOPS": 202671745179636.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056074146181344986, "optim/lr": 0.0029045544365555055, "optim/total_tokens": 5913968640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.8101439476013184, "created_at": "2025-01-16T13:25:17.179362+00:00"} {"global_step": 11281, "acc_step": 0, "speed/wps": 12898.805034350327, "speed/FLOPS": 202593457880017.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07811344414949417, "optim/lr": 0.002904524359336508, "optim/total_tokens": 5914492928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.930969476699829, "created_at": "2025-01-16T13:25:27.345611+00:00"} {"global_step": 11282, "acc_step": 0, "speed/wps": 12909.539085418182, "speed/FLOPS": 202762050902169.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05167024955153465, "optim/lr": 0.0029044942775350033, "optim/total_tokens": 5915017216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.976574182510376, "created_at": "2025-01-16T13:25:37.502454+00:00"} {"global_step": 11283, "acc_step": 0, "speed/wps": 12898.561393267924, "speed/FLOPS": 202589631162019.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059202875941991806, "optim/lr": 0.00290446419115109, "optim/total_tokens": 5915541504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486334, "loss/out": 2.930459499359131, "created_at": "2025-01-16T13:25:47.668355+00:00"} {"global_step": 11284, "acc_step": 0, "speed/wps": 12902.813504956628, "speed/FLOPS": 202656416419104.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05316450819373131, "optim/lr": 0.0029044341001848648, "optim/total_tokens": 5916065792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.844825267791748, "created_at": "2025-01-16T13:25:57.832429+00:00"} {"global_step": 11285, "acc_step": 0, "speed/wps": 12904.684087792317, "speed/FLOPS": 202685796492989.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049831416457891464, "optim/lr": 0.002904404004636427, "optim/total_tokens": 5916590080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468245, "loss/out": 2.861583948135376, "created_at": "2025-01-16T13:26:07.994768+00:00"} {"global_step": 11286, "acc_step": 0, "speed/wps": 12904.663275236917, "speed/FLOPS": 202685469603207.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049799803644418716, "optim/lr": 0.002904373904505874, "optim/total_tokens": 5917114368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.759751796722412, "created_at": "2025-01-16T13:26:18.156693+00:00"} {"global_step": 11287, "acc_step": 0, "speed/wps": 12906.767079119838, "speed/FLOPS": 202718512734118.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06028255447745323, "optim/lr": 0.0029043437997933047, "optim/total_tokens": 5917638656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.914353370666504, "created_at": "2025-01-16T13:26:28.323373+00:00"} {"global_step": 11288, "acc_step": 0, "speed/wps": 12903.491494321226, "speed/FLOPS": 202667065173730.84, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057667359709739685, "optim/lr": 0.002904313690498817, "optim/total_tokens": 5918162944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438233, "loss/out": 2.8445475101470947, "created_at": "2025-01-16T13:26:38.485700+00:00"} {"global_step": 11289, "acc_step": 0, "speed/wps": 12900.762510941942, "speed/FLOPS": 202624202739743.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054295994341373444, "optim/lr": 0.0029042835766225094, "optim/total_tokens": 5918687232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8329720497131348, "created_at": "2025-01-16T13:26:48.651761+00:00"} {"global_step": 11290, "acc_step": 0, "speed/wps": 12900.689045195903, "speed/FLOPS": 202623048859248.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05144792050123215, "optim/lr": 0.0029042534581644797, "optim/total_tokens": 5919211520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.82470703125, "created_at": "2025-01-16T13:26:58.823219+00:00"} {"global_step": 11291, "acc_step": 0, "speed/wps": 12907.091724800448, "speed/FLOPS": 202723611740642.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0582282617688179, "optim/lr": 0.0029042233351248264, "optim/total_tokens": 5919735808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466943, "loss/out": 2.8722641468048096, "created_at": "2025-01-16T13:27:08.984595+00:00"} {"global_step": 11292, "acc_step": 0, "speed/wps": 12903.208925620213, "speed/FLOPS": 202662627043991.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06807754933834076, "optim/lr": 0.0029041932075036486, "optim/total_tokens": 5920260096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.800631284713745, "created_at": "2025-01-16T13:27:19.148021+00:00"} {"global_step": 11293, "acc_step": 0, "speed/wps": 12905.838019411189, "speed/FLOPS": 202703920574732.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04876890033483505, "optim/lr": 0.002904163075301043, "optim/total_tokens": 5920784384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.0350265502929688, "created_at": "2025-01-16T13:27:29.310854+00:00"} {"global_step": 11294, "acc_step": 0, "speed/wps": 12908.2951598073, "speed/FLOPS": 202742513341117.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04945157840847969, "optim/lr": 0.002904132938517109, "optim/total_tokens": 5921308672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.916012763977051, "created_at": "2025-01-16T13:27:39.469393+00:00"} {"global_step": 11295, "acc_step": 0, "speed/wps": 12905.949536581447, "speed/FLOPS": 202705672105131.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04813683405518532, "optim/lr": 0.0029041027971519443, "optim/total_tokens": 5921832960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8440732955932617, "created_at": "2025-01-16T13:27:49.628825+00:00"} {"global_step": 11296, "acc_step": 0, "speed/wps": 12907.865052803541, "speed/FLOPS": 202735757919594.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046839967370033264, "optim/lr": 0.0029040726512056477, "optim/total_tokens": 5922357248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 2.852846384048462, "created_at": "2025-01-16T13:27:59.787583+00:00"} {"global_step": 11297, "acc_step": 0, "speed/wps": 12906.83148566255, "speed/FLOPS": 202719524327375.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053218889981508255, "optim/lr": 0.0029040425006783177, "optim/total_tokens": 5922881536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.9196577072143555, "created_at": "2025-01-16T13:28:09.946948+00:00"} {"global_step": 11298, "acc_step": 0, "speed/wps": 12903.194978035814, "speed/FLOPS": 202662407978012.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04780489578843117, "optim/lr": 0.0029040123455700523, "optim/total_tokens": 5923405824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422470, "loss/out": 2.862424850463867, "created_at": "2025-01-16T13:28:20.112536+00:00"} {"global_step": 11299, "acc_step": 0, "speed/wps": 12904.787335104817, "speed/FLOPS": 202687418133908.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04966694489121437, "optim/lr": 0.0029039821858809502, "optim/total_tokens": 5923930112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.856682062149048, "created_at": "2025-01-16T13:28:30.273304+00:00"} {"global_step": 11300, "acc_step": 0, "speed/wps": 12903.588701894294, "speed/FLOPS": 202668591952243.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05461934208869934, "optim/lr": 0.0029039520216111096, "optim/total_tokens": 5924454400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8500194549560547, "created_at": "2025-01-16T13:28:40.437976+00:00"} {"global_step": 11301, "acc_step": 0, "speed/wps": 12903.602141607791, "speed/FLOPS": 202668803041411.62, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055271752178668976, "optim/lr": 0.0029039218527606285, "optim/total_tokens": 5924978688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 2.9113054275512695, "created_at": "2025-01-16T13:28:50.599771+00:00"} {"global_step": 11302, "acc_step": 0, "speed/wps": 12902.799594897817, "speed/FLOPS": 202656197942516.75, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056399133056402206, "optim/lr": 0.0029038916793296065, "optim/total_tokens": 5925502976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.920513153076172, "created_at": "2025-01-16T13:29:00.766208+00:00"} {"global_step": 11303, "acc_step": 0, "speed/wps": 12902.92477940913, "speed/FLOPS": 202658164137288.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05470491573214531, "optim/lr": 0.00290386150131814, "optim/total_tokens": 5926027264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8792827129364014, "created_at": "2025-01-16T13:29:10.930491+00:00"} {"global_step": 11304, "acc_step": 0, "speed/wps": 12904.636416265412, "speed/FLOPS": 202685047746151.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04725423455238342, "optim/lr": 0.00290383131872633, "optim/total_tokens": 5926551552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.919936180114746, "created_at": "2025-01-16T13:29:21.091277+00:00"} {"global_step": 11305, "acc_step": 0, "speed/wps": 12902.142903675782, "speed/FLOPS": 202645883704486.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06543908268213272, "optim/lr": 0.0029038011315542732, "optim/total_tokens": 5927075840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.93969988822937, "created_at": "2025-01-16T13:29:31.255716+00:00"} {"global_step": 11306, "acc_step": 0, "speed/wps": 12896.350539755807, "speed/FLOPS": 202554906669581.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0610223226249218, "optim/lr": 0.0029037709398020686, "optim/total_tokens": 5927600128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 2.8870561122894287, "created_at": "2025-01-16T13:29:41.424156+00:00"} {"global_step": 11307, "acc_step": 0, "speed/wps": 12906.285701021341, "speed/FLOPS": 202710952029598.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05454009771347046, "optim/lr": 0.002903740743469815, "optim/total_tokens": 5928124416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.838334798812866, "created_at": "2025-01-16T13:29:51.584636+00:00"} {"global_step": 11308, "acc_step": 0, "speed/wps": 12902.891537405638, "speed/FLOPS": 202657642025944.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05524000898003578, "optim/lr": 0.0029037105425576103, "optim/total_tokens": 5928648704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.7655959129333496, "created_at": "2025-01-16T13:30:01.746453+00:00"} {"global_step": 11309, "acc_step": 0, "speed/wps": 12902.253642569409, "speed/FLOPS": 202647623010981.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050129689276218414, "optim/lr": 0.0029036803370655535, "optim/total_tokens": 5929172992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9449212551116943, "created_at": "2025-01-16T13:30:11.910175+00:00"} {"global_step": 11310, "acc_step": 0, "speed/wps": 12896.062118826865, "speed/FLOPS": 202550376622560.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06129604950547218, "optim/lr": 0.002903650126993743, "optim/total_tokens": 5929697280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.867166519165039, "created_at": "2025-01-16T13:30:22.077424+00:00"} {"global_step": 11311, "acc_step": 0, "speed/wps": 12907.842614027817, "speed/FLOPS": 202735405487787.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059752970933914185, "optim/lr": 0.002903619912342278, "optim/total_tokens": 5930221568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 2.8464181423187256, "created_at": "2025-01-16T13:30:32.239785+00:00"} {"global_step": 11312, "acc_step": 0, "speed/wps": 12898.555378489209, "speed/FLOPS": 202589536691655.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10474550724029541, "optim/lr": 0.0029035896931112557, "optim/total_tokens": 5930745856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.9543120861053467, "created_at": "2025-01-16T13:30:42.405030+00:00"} {"global_step": 11313, "acc_step": 0, "speed/wps": 12901.620952046856, "speed/FLOPS": 202637685736917.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06587178260087967, "optim/lr": 0.002903559469300776, "optim/total_tokens": 5931270144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8739914894104004, "created_at": "2025-01-16T13:30:52.573811+00:00"} {"global_step": 11314, "acc_step": 0, "speed/wps": 12906.775253446674, "speed/FLOPS": 202718641123153.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07697516679763794, "optim/lr": 0.0029035292409109368, "optim/total_tokens": 5931794432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 2.901230573654175, "created_at": "2025-01-16T13:31:02.736041+00:00"} {"global_step": 11315, "acc_step": 0, "speed/wps": 12905.482196814059, "speed/FLOPS": 202698331891892.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06943759322166443, "optim/lr": 0.0029034990079418368, "optim/total_tokens": 5932318720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.932401657104492, "created_at": "2025-01-16T13:31:12.895843+00:00"} {"global_step": 11316, "acc_step": 0, "speed/wps": 12905.565291140601, "speed/FLOPS": 202699637002474.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06106773018836975, "optim/lr": 0.002903468770393575, "optim/total_tokens": 5932843008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.7672924995422363, "created_at": "2025-01-16T13:31:23.055823+00:00"} {"global_step": 11317, "acc_step": 0, "speed/wps": 12907.181256857402, "speed/FLOPS": 202725017964626.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06548592448234558, "optim/lr": 0.0029034385282662497, "optim/total_tokens": 5933367296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470104, "loss/out": 2.918252944946289, "created_at": "2025-01-16T13:31:33.214973+00:00"} {"global_step": 11318, "acc_step": 0, "speed/wps": 12897.942441898807, "speed/FLOPS": 202579909680239.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04803469404578209, "optim/lr": 0.0029034082815599595, "optim/total_tokens": 5933891584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8468010425567627, "created_at": "2025-01-16T13:31:43.383218+00:00"} {"global_step": 11319, "acc_step": 0, "speed/wps": 12900.362801169307, "speed/FLOPS": 202617924748505.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06740875542163849, "optim/lr": 0.002903378030274804, "optim/total_tokens": 5934415872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 2.8866004943847656, "created_at": "2025-01-16T13:31:53.552689+00:00"} {"global_step": 11320, "acc_step": 0, "speed/wps": 12900.134964816554, "speed/FLOPS": 202614346265505.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0555238351225853, "optim/lr": 0.00290334777441088, "optim/total_tokens": 5934940160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8461689949035645, "created_at": "2025-01-16T13:32:03.719420+00:00"} {"global_step": 11321, "acc_step": 0, "speed/wps": 12896.421261687647, "speed/FLOPS": 202556017454702.72, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07640562206506729, "optim/lr": 0.002903317513968288, "optim/total_tokens": 5935464448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9440550804138184, "created_at": "2025-01-16T13:32:13.888183+00:00"} {"global_step": 11322, "acc_step": 0, "speed/wps": 12898.380304546314, "speed/FLOPS": 202586786914805.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05390789732336998, "optim/lr": 0.002903287248947126, "optim/total_tokens": 5935988736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 2.907230854034424, "created_at": "2025-01-16T13:32:24.056407+00:00"} {"global_step": 11323, "acc_step": 0, "speed/wps": 12901.461710209656, "speed/FLOPS": 202635184625042.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05292802304029465, "optim/lr": 0.002903256979347493, "optim/total_tokens": 5936513024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9107840061187744, "created_at": "2025-01-16T13:32:34.219476+00:00"} {"global_step": 11324, "acc_step": 0, "speed/wps": 12902.608207173833, "speed/FLOPS": 202653191935317.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04985035955905914, "optim/lr": 0.0029032267051694875, "optim/total_tokens": 5937037312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.791412115097046, "created_at": "2025-01-16T13:32:44.382818+00:00"} {"global_step": 11325, "acc_step": 0, "speed/wps": 12898.711495022333, "speed/FLOPS": 202591988716334.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051718417555093765, "optim/lr": 0.002903196426413209, "optim/total_tokens": 5937561600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.906543254852295, "created_at": "2025-01-16T13:32:54.551797+00:00"} {"global_step": 11326, "acc_step": 0, "speed/wps": 12901.908416442537, "speed/FLOPS": 202642200760269.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04874292388558388, "optim/lr": 0.002903166143078755, "optim/total_tokens": 5938085888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8918445110321045, "created_at": "2025-01-16T13:33:04.714457+00:00"} {"global_step": 11327, "acc_step": 0, "speed/wps": 12900.76579016489, "speed/FLOPS": 202624254244445.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05369795113801956, "optim/lr": 0.002903135855166225, "optim/total_tokens": 5938610176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 2.797884464263916, "created_at": "2025-01-16T13:33:14.879841+00:00"} {"global_step": 11328, "acc_step": 0, "speed/wps": 12901.854971105102, "speed/FLOPS": 202641361327803.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042793698608875275, "optim/lr": 0.0029031055626757177, "optim/total_tokens": 5939134464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.659853935241699, "created_at": "2025-01-16T13:33:25.049078+00:00"} {"global_step": 11329, "acc_step": 0, "speed/wps": 12903.410294716308, "speed/FLOPS": 202665789822355.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055286020040512085, "optim/lr": 0.0029030752656073323, "optim/total_tokens": 5939658752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.814117431640625, "created_at": "2025-01-16T13:33:35.211763+00:00"} {"global_step": 11330, "acc_step": 0, "speed/wps": 12902.087219726449, "speed/FLOPS": 202645009111544.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.062015634030103683, "optim/lr": 0.0029030449639611674, "optim/total_tokens": 5940183040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 2.789584159851074, "created_at": "2025-01-16T13:33:45.375426+00:00"} {"global_step": 11331, "acc_step": 0, "speed/wps": 12903.346468872027, "speed/FLOPS": 202664787349765.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05349966138601303, "optim/lr": 0.0029030146577373223, "optim/total_tokens": 5940707328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9303781986236572, "created_at": "2025-01-16T13:33:55.542082+00:00"} {"global_step": 11332, "acc_step": 0, "speed/wps": 12907.05596369838, "speed/FLOPS": 202723050063393.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050966475158929825, "optim/lr": 0.0029029843469358945, "optim/total_tokens": 5941231616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 2.887669086456299, "created_at": "2025-01-16T13:34:05.702344+00:00"} {"global_step": 11333, "acc_step": 0, "speed/wps": 12894.628970331089, "speed/FLOPS": 202527867056081.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05948276445269585, "optim/lr": 0.002902954031556984, "optim/total_tokens": 5941755904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8444862365722656, "created_at": "2025-01-16T13:34:15.873055+00:00"} {"global_step": 11334, "acc_step": 0, "speed/wps": 12898.982775106344, "speed/FLOPS": 202596249542831.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05705777183175087, "optim/lr": 0.0029029237116006902, "optim/total_tokens": 5942280192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9529595375061035, "created_at": "2025-01-16T13:34:26.040887+00:00"} {"global_step": 11335, "acc_step": 0, "speed/wps": 12898.14943172054, "speed/FLOPS": 202583160739823.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05729418620467186, "optim/lr": 0.002902893387067111, "optim/total_tokens": 5942804480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.7509279251098633, "created_at": "2025-01-16T13:34:36.207993+00:00"} {"global_step": 11336, "acc_step": 0, "speed/wps": 12898.957136185258, "speed/FLOPS": 202595846848344.4, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0508921779692173, "optim/lr": 0.0029028630579563458, "optim/total_tokens": 5943328768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.754896640777588, "created_at": "2025-01-16T13:34:46.372972+00:00"} {"global_step": 11337, "acc_step": 0, "speed/wps": 12902.91888240409, "speed/FLOPS": 202658071516720.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04686642438173294, "optim/lr": 0.0029028327242684934, "optim/total_tokens": 5943853056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.760338068008423, "created_at": "2025-01-16T13:34:56.539346+00:00"} {"global_step": 11338, "acc_step": 0, "speed/wps": 12904.803625505572, "speed/FLOPS": 202687673997038.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.06748271733522415, "optim/lr": 0.0029028023860036526, "optim/total_tokens": 5944377344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 2.8722023963928223, "created_at": "2025-01-16T13:35:06.700918+00:00"} {"global_step": 11339, "acc_step": 0, "speed/wps": 12899.15930303286, "speed/FLOPS": 202599022156492.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06716907024383545, "optim/lr": 0.0029027720431619233, "optim/total_tokens": 5944901632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.870481491088867, "created_at": "2025-01-16T13:35:16.868006+00:00"} {"global_step": 11340, "acc_step": 0, "speed/wps": 12907.398463313975, "speed/FLOPS": 202728429490499.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050175633281469345, "optim/lr": 0.0029027416957434033, "optim/total_tokens": 5945425920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347078, "loss/out": 2.9795477390289307, "created_at": "2025-01-16T13:35:27.029113+00:00"} {"global_step": 11341, "acc_step": 0, "speed/wps": 12902.450693968334, "speed/FLOPS": 202650717973979.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047069232910871506, "optim/lr": 0.002902711343748192, "optim/total_tokens": 5945950208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.7812373638153076, "created_at": "2025-01-16T13:35:37.191575+00:00"} {"global_step": 11342, "acc_step": 0, "speed/wps": 12891.30814623857, "speed/FLOPS": 202475708950418.06, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05446476489305496, "optim/lr": 0.0029026809871763887, "optim/total_tokens": 5946474496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9171037673950195, "created_at": "2025-01-16T13:35:47.366202+00:00"} {"global_step": 11343, "acc_step": 0, "speed/wps": 12903.961270349322, "speed/FLOPS": 202674443651792.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047291431576013565, "optim/lr": 0.0029026506260280923, "optim/total_tokens": 5946998784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 2.8565850257873535, "created_at": "2025-01-16T13:35:57.527371+00:00"} {"global_step": 11344, "acc_step": 0, "speed/wps": 12906.800348197608, "speed/FLOPS": 202719035270701.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04590332508087158, "optim/lr": 0.002902620260303402, "optim/total_tokens": 5947523072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.775999069213867, "created_at": "2025-01-16T13:36:07.686342+00:00"} {"global_step": 11345, "acc_step": 0, "speed/wps": 12896.808708027202, "speed/FLOPS": 202562102831872.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05324619263410568, "optim/lr": 0.0029025898900024165, "optim/total_tokens": 5948047360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 2.901909828186035, "created_at": "2025-01-16T13:36:17.855437+00:00"} {"global_step": 11346, "acc_step": 0, "speed/wps": 12906.052124801377, "speed/FLOPS": 202707283394096.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04038355499505997, "optim/lr": 0.0029025595151252353, "optim/total_tokens": 5948571648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9698023796081543, "created_at": "2025-01-16T13:36:28.015403+00:00"} {"global_step": 11347, "acc_step": 0, "speed/wps": 12902.353807531867, "speed/FLOPS": 202649196239358.47, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04842449352145195, "optim/lr": 0.0029025291356719572, "optim/total_tokens": 5949095936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 2.8688547611236572, "created_at": "2025-01-16T13:36:38.178113+00:00"} {"global_step": 11348, "acc_step": 0, "speed/wps": 12898.676889219472, "speed/FLOPS": 202591445184647.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0478510782122612, "optim/lr": 0.0029024987516426816, "optim/total_tokens": 5949620224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9268112182617188, "created_at": "2025-01-16T13:36:48.344118+00:00"} {"global_step": 11349, "acc_step": 0, "speed/wps": 12903.449583575224, "speed/FLOPS": 202666406907873.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04511036351323128, "optim/lr": 0.0029024683630375072, "optim/total_tokens": 5950144512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.7862982749938965, "created_at": "2025-01-16T13:36:58.507303+00:00"} {"global_step": 11350, "acc_step": 0, "speed/wps": 12904.613870387453, "speed/FLOPS": 202684693632157.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04696747660636902, "optim/lr": 0.002902437969856534, "optim/total_tokens": 5950668800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.8956711292266846, "created_at": "2025-01-16T13:37:08.669145+00:00"} {"global_step": 11351, "acc_step": 0, "speed/wps": 12905.076526303483, "speed/FLOPS": 202691960279074.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04139956086874008, "optim/lr": 0.0029024075720998598, "optim/total_tokens": 5951193088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 2.8616201877593994, "created_at": "2025-01-16T13:37:18.829497+00:00"} {"global_step": 11352, "acc_step": 0, "speed/wps": 12901.672821243837, "speed/FLOPS": 202638500413931.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058074045926332474, "optim/lr": 0.002902377169767585, "optim/total_tokens": 5951717376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.915018081665039, "created_at": "2025-01-16T13:37:28.995096+00:00"} {"global_step": 11353, "acc_step": 0, "speed/wps": 12898.482189673734, "speed/FLOPS": 202588387160736.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056854408234357834, "optim/lr": 0.002902346762859808, "optim/total_tokens": 5952241664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 2.7772297859191895, "created_at": "2025-01-16T13:37:39.167863+00:00"} {"global_step": 11354, "acc_step": 0, "speed/wps": 12904.203421152095, "speed/FLOPS": 202678246962899.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04926391318440437, "optim/lr": 0.002902316351376629, "optim/total_tokens": 5952765952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.0320334434509277, "created_at": "2025-01-16T13:37:49.328839+00:00"} {"global_step": 11355, "acc_step": 0, "speed/wps": 12902.847288342798, "speed/FLOPS": 202656947033607.84, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049667391926050186, "optim/lr": 0.002902285935318146, "optim/total_tokens": 5953290240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.7547407150268555, "created_at": "2025-01-16T13:37:59.492455+00:00"} {"global_step": 11356, "acc_step": 0, "speed/wps": 12904.146526775601, "speed/FLOPS": 202677353358535.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047205064445734024, "optim/lr": 0.002902255514684459, "optim/total_tokens": 5953814528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9723265171051025, "created_at": "2025-01-16T13:38:09.655169+00:00"} {"global_step": 11357, "acc_step": 0, "speed/wps": 12900.298011013496, "speed/FLOPS": 202616907130077.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05410168692469597, "optim/lr": 0.0029022250894756674, "optim/total_tokens": 5954338816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9442138671875, "created_at": "2025-01-16T13:38:19.821795+00:00"} {"global_step": 11358, "acc_step": 0, "speed/wps": 12897.171276239731, "speed/FLOPS": 202567797463871.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052860043942928314, "optim/lr": 0.0029021946596918696, "optim/total_tokens": 5954863104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.915828227996826, "created_at": "2025-01-16T13:38:29.990022+00:00"} {"global_step": 11359, "acc_step": 0, "speed/wps": 12903.527215562071, "speed/FLOPS": 202667626224904.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05154132843017578, "optim/lr": 0.002902164225333166, "optim/total_tokens": 5955387392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.830230712890625, "created_at": "2025-01-16T13:38:40.152883+00:00"} {"global_step": 11360, "acc_step": 0, "speed/wps": 12903.919001611768, "speed/FLOPS": 202673779763185.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06093843653798103, "optim/lr": 0.0029021337863996545, "optim/total_tokens": 5955911680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.0238099098205566, "created_at": "2025-01-16T13:38:50.316486+00:00"} {"global_step": 11361, "acc_step": 0, "speed/wps": 12900.84528788476, "speed/FLOPS": 202625502865377.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.040498286485672, "optim/lr": 0.0029021033428914362, "optim/total_tokens": 5956435968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 2.8892455101013184, "created_at": "2025-01-16T13:39:00.485244+00:00"} {"global_step": 11362, "acc_step": 0, "speed/wps": 12903.268010950636, "speed/FLOPS": 202663555060297.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05069537088274956, "optim/lr": 0.0029020728948086086, "optim/total_tokens": 5956960256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 2.9488296508789062, "created_at": "2025-01-16T13:39:10.646937+00:00"} {"global_step": 11363, "acc_step": 0, "speed/wps": 12901.503948657979, "speed/FLOPS": 202635848037915.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04777388274669647, "optim/lr": 0.0029020424421512723, "optim/total_tokens": 5957484544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.946819543838501, "created_at": "2025-01-16T13:39:20.809883+00:00"} {"global_step": 11364, "acc_step": 0, "speed/wps": 12899.937597498525, "speed/FLOPS": 202611246340564.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05301417037844658, "optim/lr": 0.0029020119849195265, "optim/total_tokens": 5958008832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 2.8590927124023438, "created_at": "2025-01-16T13:39:30.976939+00:00"} {"global_step": 11365, "acc_step": 0, "speed/wps": 12895.044643797959, "speed/FLOPS": 202534395779073.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05189788341522217, "optim/lr": 0.0029019815231134706, "optim/total_tokens": 5958533120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8963406085968018, "created_at": "2025-01-16T13:39:41.145010+00:00"} {"global_step": 11366, "acc_step": 0, "speed/wps": 12904.311853856858, "speed/FLOPS": 202679950047530.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04661343991756439, "optim/lr": 0.0029019510567332026, "optim/total_tokens": 5959057408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.931257963180542, "created_at": "2025-01-16T13:39:51.305772+00:00"} {"global_step": 11367, "acc_step": 0, "speed/wps": 12898.719874305225, "speed/FLOPS": 202592120324486.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04611274600028992, "optim/lr": 0.002901920585778824, "optim/total_tokens": 5959581696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 2.9036765098571777, "created_at": "2025-01-16T13:40:01.471762+00:00"} {"global_step": 11368, "acc_step": 0, "speed/wps": 12904.922354861439, "speed/FLOPS": 202689538804723.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055291999131441116, "optim/lr": 0.0029018901102504328, "optim/total_tokens": 5960105984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 2.8811964988708496, "created_at": "2025-01-16T13:40:11.632043+00:00"} {"global_step": 11369, "acc_step": 0, "speed/wps": 12905.948323488847, "speed/FLOPS": 202705653051845.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04856180399656296, "optim/lr": 0.002901859630148129, "optim/total_tokens": 5960630272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8222556114196777, "created_at": "2025-01-16T13:40:21.795115+00:00"} {"global_step": 11370, "acc_step": 0, "speed/wps": 12902.63196206177, "speed/FLOPS": 202653565038474.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042263396084308624, "optim/lr": 0.002901829145472012, "optim/total_tokens": 5961154560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.8434793949127197, "created_at": "2025-01-16T13:40:31.961255+00:00"} {"global_step": 11371, "acc_step": 0, "speed/wps": 12899.51711805251, "speed/FLOPS": 202604642133065.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05077318474650383, "optim/lr": 0.002901798656222181, "optim/total_tokens": 5961678848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8372104167938232, "created_at": "2025-01-16T13:40:42.127579+00:00"} {"global_step": 11372, "acc_step": 0, "speed/wps": 12897.999163375975, "speed/FLOPS": 202580800568982.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05036310851573944, "optim/lr": 0.002901768162398736, "optim/total_tokens": 5962203136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.9069888591766357, "created_at": "2025-01-16T13:40:52.293467+00:00"} {"global_step": 11373, "acc_step": 0, "speed/wps": 12901.896385377906, "speed/FLOPS": 202642011795866.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05794920399785042, "optim/lr": 0.0029017376640017756, "optim/total_tokens": 5962727424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.9679551124572754, "created_at": "2025-01-16T13:41:02.456987+00:00"} {"global_step": 11374, "acc_step": 0, "speed/wps": 12901.79060002652, "speed/FLOPS": 202640350291558.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05626150965690613, "optim/lr": 0.0029017071610314, "optim/total_tokens": 5963251712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.932183265686035, "created_at": "2025-01-16T13:41:12.621039+00:00"} {"global_step": 11375, "acc_step": 0, "speed/wps": 12900.358104357772, "speed/FLOPS": 202617850978625.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04991603642702103, "optim/lr": 0.002901676653487709, "optim/total_tokens": 5963776000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.879007339477539, "created_at": "2025-01-16T13:41:22.786870+00:00"} {"global_step": 11376, "acc_step": 0, "speed/wps": 12900.117802921068, "speed/FLOPS": 202614076714353.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05945075303316116, "optim/lr": 0.0029016461413708017, "optim/total_tokens": 5964300288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.8754525184631348, "created_at": "2025-01-16T13:41:32.953863+00:00"} {"global_step": 11377, "acc_step": 0, "speed/wps": 12902.352457905057, "speed/FLOPS": 202649175041614.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05348476395010948, "optim/lr": 0.002901615624680777, "optim/total_tokens": 5964824576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.878302574157715, "created_at": "2025-01-16T13:41:43.116337+00:00"} {"global_step": 11378, "acc_step": 0, "speed/wps": 12898.77562731387, "speed/FLOPS": 202592996002099.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.041293736547231674, "optim/lr": 0.002901585103417736, "optim/total_tokens": 5965348864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.9404265880584717, "created_at": "2025-01-16T13:41:53.281680+00:00"} {"global_step": 11379, "acc_step": 0, "speed/wps": 12899.381140213329, "speed/FLOPS": 202602506414247.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04803559184074402, "optim/lr": 0.0029015545775817764, "optim/total_tokens": 5965873152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8937082290649414, "created_at": "2025-01-16T13:42:03.449809+00:00"} {"global_step": 11380, "acc_step": 0, "speed/wps": 12902.57743548559, "speed/FLOPS": 202652708623668.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04728962108492851, "optim/lr": 0.002901524047173, "optim/total_tokens": 5966397440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 2.826202392578125, "created_at": "2025-01-16T13:42:13.615005+00:00"} {"global_step": 11381, "acc_step": 0, "speed/wps": 12899.415417566885, "speed/FLOPS": 202603044787187.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04325202479958534, "optim/lr": 0.0029014935121915043, "optim/total_tokens": 5966921728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8180341720581055, "created_at": "2025-01-16T13:42:23.779649+00:00"} {"global_step": 11382, "acc_step": 0, "speed/wps": 12901.085501711961, "speed/FLOPS": 202629275753622.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04244893789291382, "optim/lr": 0.0029014629726373904, "optim/total_tokens": 5967446016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.80753231048584, "created_at": "2025-01-16T13:42:33.943873+00:00"} {"global_step": 11383, "acc_step": 0, "speed/wps": 12900.938576707726, "speed/FLOPS": 202626968094534.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04948452115058899, "optim/lr": 0.002901432428510757, "optim/total_tokens": 5967970304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343977, "loss/out": 2.797645330429077, "created_at": "2025-01-16T13:42:44.107530+00:00"} {"global_step": 11384, "acc_step": 0, "speed/wps": 12900.755931215097, "speed/FLOPS": 202624099396091.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05279921740293503, "optim/lr": 0.0029014018798117045, "optim/total_tokens": 5968494592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.8043150901794434, "created_at": "2025-01-16T13:42:54.271562+00:00"} {"global_step": 11385, "acc_step": 0, "speed/wps": 12898.455776082426, "speed/FLOPS": 202587972298988.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04787181317806244, "optim/lr": 0.002901371326540332, "optim/total_tokens": 5969018880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 2.904270887374878, "created_at": "2025-01-16T13:43:04.439435+00:00"} {"global_step": 11386, "acc_step": 0, "speed/wps": 12906.246952346104, "speed/FLOPS": 202710343428407.66, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05242439731955528, "optim/lr": 0.0029013407686967396, "optim/total_tokens": 5969543168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.874925136566162, "created_at": "2025-01-16T13:43:14.606963+00:00"} {"global_step": 11387, "acc_step": 0, "speed/wps": 12904.168953885901, "speed/FLOPS": 202677705607121.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05455056205391884, "optim/lr": 0.002901310206281026, "optim/total_tokens": 5970067456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.902108907699585, "created_at": "2025-01-16T13:43:24.768001+00:00"} {"global_step": 11388, "acc_step": 0, "speed/wps": 12899.896588355114, "speed/FLOPS": 202610602235614.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05390813946723938, "optim/lr": 0.0029012796392932924, "optim/total_tokens": 5970591744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366944, "loss/out": 2.9399571418762207, "created_at": "2025-01-16T13:43:34.933337+00:00"} {"global_step": 11389, "acc_step": 0, "speed/wps": 12900.81052044604, "speed/FLOPS": 202624956794976.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04551422968506813, "optim/lr": 0.002901249067733638, "optim/total_tokens": 5971116032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505601, "loss/out": 3.0029234886169434, "created_at": "2025-01-16T13:43:45.097180+00:00"} {"global_step": 11390, "acc_step": 0, "speed/wps": 12901.772356250147, "speed/FLOPS": 202640063747981.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058672986924648285, "optim/lr": 0.0029012184916021623, "optim/total_tokens": 5971640320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8135933876037598, "created_at": "2025-01-16T13:43:55.261950+00:00"} {"global_step": 11391, "acc_step": 0, "speed/wps": 12905.620782628346, "speed/FLOPS": 202700508572542.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059489600360393524, "optim/lr": 0.0029011879108989647, "optim/total_tokens": 5972164608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.8140647411346436, "created_at": "2025-01-16T13:44:05.422091+00:00"} {"global_step": 11392, "acc_step": 0, "speed/wps": 12897.420291098826, "speed/FLOPS": 202571708584415.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05182887986302376, "optim/lr": 0.002901157325624146, "optim/total_tokens": 5972688896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8999428749084473, "created_at": "2025-01-16T13:44:15.588496+00:00"} {"global_step": 11393, "acc_step": 0, "speed/wps": 12902.40751959613, "speed/FLOPS": 202650039861135.84, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05113797262310982, "optim/lr": 0.002901126735777805, "optim/total_tokens": 5973213184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 2.8535337448120117, "created_at": "2025-01-16T13:44:25.757565+00:00"} {"global_step": 11394, "acc_step": 0, "speed/wps": 12899.774798068838, "speed/FLOPS": 202608689351811.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047959182411432266, "optim/lr": 0.002901096141360042, "optim/total_tokens": 5973737472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.915609836578369, "created_at": "2025-01-16T13:44:35.926315+00:00"} {"global_step": 11395, "acc_step": 0, "speed/wps": 12905.282590188297, "speed/FLOPS": 202695196795546.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053317841142416, "optim/lr": 0.0029010655423709567, "optim/total_tokens": 5974261760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8211164474487305, "created_at": "2025-01-16T13:44:46.089421+00:00"} {"global_step": 11396, "acc_step": 0, "speed/wps": 12903.952615682674, "speed/FLOPS": 202674307718359.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05242663994431496, "optim/lr": 0.0029010349388106493, "optim/total_tokens": 5974786048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.8683066368103027, "created_at": "2025-01-16T13:44:56.250544+00:00"} {"global_step": 11397, "acc_step": 0, "speed/wps": 12905.167293051774, "speed/FLOPS": 202693385895582.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04555438458919525, "optim/lr": 0.002901004330679219, "optim/total_tokens": 5975310336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.9308958053588867, "created_at": "2025-01-16T13:45:06.418575+00:00"} {"global_step": 11398, "acc_step": 0, "speed/wps": 12900.43395970147, "speed/FLOPS": 202619042391034.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04824452847242355, "optim/lr": 0.0029009737179767664, "optim/total_tokens": 5975834624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 2.93815279006958, "created_at": "2025-01-16T13:45:16.585760+00:00"} {"global_step": 11399, "acc_step": 0, "speed/wps": 12900.576115933158, "speed/FLOPS": 202621275149996.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05052082613110542, "optim/lr": 0.00290094310070339, "optim/total_tokens": 5976358912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.922576904296875, "created_at": "2025-01-16T13:45:26.750552+00:00"} {"global_step": 11400, "acc_step": 0, "speed/wps": 12900.212061476584, "speed/FLOPS": 202615557174492.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06244971975684166, "optim/lr": 0.0029009124788591917, "optim/total_tokens": 5976883200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.075782537460327, "created_at": "2025-01-16T13:45:36.919767+00:00"} {"global_step": 11401, "acc_step": 0, "speed/wps": 12902.098110928822, "speed/FLOPS": 202645180172843.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04443051293492317, "optim/lr": 0.0029008818524442702, "optim/total_tokens": 5977407488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 2.7734451293945312, "created_at": "2025-01-16T13:45:47.082998+00:00"} {"global_step": 11402, "acc_step": 0, "speed/wps": 12902.12803935688, "speed/FLOPS": 202645650239932.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04566475749015808, "optim/lr": 0.0029008512214587254, "optim/total_tokens": 5977931776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.984219551086426, "created_at": "2025-01-16T13:45:57.251076+00:00"} {"global_step": 11403, "acc_step": 0, "speed/wps": 12894.937993684694, "speed/FLOPS": 202532720692492.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04659944400191307, "optim/lr": 0.0029008205859026577, "optim/total_tokens": 5978456064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.8163058757781982, "created_at": "2025-01-16T13:46:07.429190+00:00"} {"global_step": 11404, "acc_step": 0, "speed/wps": 12904.398191701752, "speed/FLOPS": 202681306102024.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043448884040117264, "optim/lr": 0.002900789945776166, "optim/total_tokens": 5978980352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313142, "loss/out": 2.7937307357788086, "created_at": "2025-01-16T13:46:17.589771+00:00"} {"global_step": 11405, "acc_step": 0, "speed/wps": 12905.719424689243, "speed/FLOPS": 202702057881657.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04557885229587555, "optim/lr": 0.002900759301079352, "optim/total_tokens": 5979504640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8065414428710938, "created_at": "2025-01-16T13:46:27.755074+00:00"} {"global_step": 11406, "acc_step": 0, "speed/wps": 12905.963550480403, "speed/FLOPS": 202705892212671.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04222473129630089, "optim/lr": 0.002900728651812314, "optim/total_tokens": 5980028928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 2.826849937438965, "created_at": "2025-01-16T13:46:37.917878+00:00"} {"global_step": 11407, "acc_step": 0, "speed/wps": 12903.482467623124, "speed/FLOPS": 202666923397033.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05634254589676857, "optim/lr": 0.0029006979979751533, "optim/total_tokens": 5980553216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.9113574028015137, "created_at": "2025-01-16T13:46:48.087906+00:00"} {"global_step": 11408, "acc_step": 0, "speed/wps": 12904.885324254734, "speed/FLOPS": 202688957188159.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05658378824591637, "optim/lr": 0.0029006673395679692, "optim/total_tokens": 5981077504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8497605323791504, "created_at": "2025-01-16T13:46:58.253592+00:00"} {"global_step": 11409, "acc_step": 0, "speed/wps": 12892.760893728306, "speed/FLOPS": 202498526346028.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.053422294557094574, "optim/lr": 0.0029006366765908617, "optim/total_tokens": 5981601792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.876251220703125, "created_at": "2025-01-16T13:47:08.429581+00:00"} {"global_step": 11410, "acc_step": 0, "speed/wps": 12900.62524508714, "speed/FLOPS": 202622046790870.3, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0766276866197586, "optim/lr": 0.002900606009043931, "optim/total_tokens": 5982126080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 2.8433213233947754, "created_at": "2025-01-16T13:47:18.595657+00:00"} {"global_step": 11411, "acc_step": 0, "speed/wps": 12902.232695914043, "speed/FLOPS": 202647294014975.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06374761462211609, "optim/lr": 0.002900575336927277, "optim/total_tokens": 5982650368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.884261131286621, "created_at": "2025-01-16T13:47:28.757975+00:00"} {"global_step": 11412, "acc_step": 0, "speed/wps": 12901.18753601795, "speed/FLOPS": 202630878342607.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0510888546705246, "optim/lr": 0.002900544660241001, "optim/total_tokens": 5983174656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 3.0246901512145996, "created_at": "2025-01-16T13:47:38.921262+00:00"} {"global_step": 11413, "acc_step": 0, "speed/wps": 12900.974159643025, "speed/FLOPS": 202627526973429.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0846516564488411, "optim/lr": 0.0029005139789852006, "optim/total_tokens": 5983698944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.79331374168396, "created_at": "2025-01-16T13:47:49.085625+00:00"} {"global_step": 11414, "acc_step": 0, "speed/wps": 12895.417305080671, "speed/FLOPS": 202540248936609.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06372979283332825, "optim/lr": 0.0029004832931599786, "optim/total_tokens": 5984223232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.805265426635742, "created_at": "2025-01-16T13:47:59.253951+00:00"} {"global_step": 11415, "acc_step": 0, "speed/wps": 12902.226994707293, "speed/FLOPS": 202647204469689.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.059862278401851654, "optim/lr": 0.002900452602765433, "optim/total_tokens": 5984747520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 2.8102784156799316, "created_at": "2025-01-16T13:48:09.419810+00:00"} {"global_step": 11416, "acc_step": 0, "speed/wps": 12897.721400038941, "speed/FLOPS": 202576437914087.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05407065153121948, "optim/lr": 0.0029004219078016653, "optim/total_tokens": 5985271808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8208162784576416, "created_at": "2025-01-16T13:48:19.585761+00:00"} {"global_step": 11417, "acc_step": 0, "speed/wps": 12893.607100403777, "speed/FLOPS": 202511817184676.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06178134307265282, "optim/lr": 0.002900391208268775, "optim/total_tokens": 5985796096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.776960849761963, "created_at": "2025-01-16T13:48:29.754926+00:00"} {"global_step": 11418, "acc_step": 0, "speed/wps": 12891.713786473649, "speed/FLOPS": 202482080087718.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05223479121923447, "optim/lr": 0.0029003605041668626, "optim/total_tokens": 5986320384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.830220937728882, "created_at": "2025-01-16T13:48:39.928262+00:00"} {"global_step": 11419, "acc_step": 0, "speed/wps": 12888.846497679779, "speed/FLOPS": 202437045377142.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0496298149228096, "optim/lr": 0.0029003297954960277, "optim/total_tokens": 5986844672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.832658290863037, "created_at": "2025-01-16T13:48:50.104988+00:00"} {"global_step": 11420, "acc_step": 0, "speed/wps": 12889.577765639639, "speed/FLOPS": 202448530945319.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05861258879303932, "optim/lr": 0.0029002990822563713, "optim/total_tokens": 5987368960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.898646831512451, "created_at": "2025-01-16T13:49:00.280594+00:00"} {"global_step": 11421, "acc_step": 0, "speed/wps": 12894.131364233468, "speed/FLOPS": 202520051468536.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04266538843512535, "optim/lr": 0.0029002683644479927, "optim/total_tokens": 5987893248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.9341042041778564, "created_at": "2025-01-16T13:49:10.452255+00:00"} {"global_step": 11422, "acc_step": 0, "speed/wps": 12901.324111340662, "speed/FLOPS": 202633023445724.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04545166343450546, "optim/lr": 0.002900237642070993, "optim/total_tokens": 5988417536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8091814517974854, "created_at": "2025-01-16T13:49:20.615903+00:00"} {"global_step": 11423, "acc_step": 0, "speed/wps": 12901.292190326845, "speed/FLOPS": 202632522082338.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04230300337076187, "optim/lr": 0.002900206915125472, "optim/total_tokens": 5988941824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 2.8782825469970703, "created_at": "2025-01-16T13:49:30.779704+00:00"} {"global_step": 11424, "acc_step": 0, "speed/wps": 12900.455342770878, "speed/FLOPS": 202619378241523.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04604477435350418, "optim/lr": 0.0029001761836115296, "optim/total_tokens": 5989466112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 2.8122329711914062, "created_at": "2025-01-16T13:49:40.946981+00:00"} {"global_step": 11425, "acc_step": 0, "speed/wps": 12897.660761540079, "speed/FLOPS": 202575485503136.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05007442831993103, "optim/lr": 0.0029001454475292665, "optim/total_tokens": 5989990400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9101505279541016, "created_at": "2025-01-16T13:49:51.115253+00:00"} {"global_step": 11426, "acc_step": 0, "speed/wps": 12894.780354609897, "speed/FLOPS": 202530244754204.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04309762641787529, "optim/lr": 0.0029001147068787834, "optim/total_tokens": 5990514688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.9645843505859375, "created_at": "2025-01-16T13:50:01.284186+00:00"} {"global_step": 11427, "acc_step": 0, "speed/wps": 12900.25623216121, "speed/FLOPS": 202616250935791.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048921868205070496, "optim/lr": 0.0029000839616601807, "optim/total_tokens": 5991038976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.990659713745117, "created_at": "2025-01-16T13:50:11.450878+00:00"} {"global_step": 11428, "acc_step": 0, "speed/wps": 12901.156156140085, "speed/FLOPS": 202630385478505.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.039198290556669235, "optim/lr": 0.002900053211873557, "optim/total_tokens": 5991563264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.7618846893310547, "created_at": "2025-01-16T13:50:21.615274+00:00"} {"global_step": 11429, "acc_step": 0, "speed/wps": 12900.586619082416, "speed/FLOPS": 202621440116388.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04758524149656296, "optim/lr": 0.0029000224575190144, "optim/total_tokens": 5992087552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 2.9998202323913574, "created_at": "2025-01-16T13:50:31.784180+00:00"} {"global_step": 11430, "acc_step": 0, "speed/wps": 12897.778909152612, "speed/FLOPS": 202577341173743.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045327458530664444, "optim/lr": 0.0028999916985966525, "optim/total_tokens": 5992611840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 2.780897855758667, "created_at": "2025-01-16T13:50:41.951907+00:00"} {"global_step": 11431, "acc_step": 0, "speed/wps": 12903.646832405719, "speed/FLOPS": 202669504971804.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05469053238630295, "optim/lr": 0.0028999609351065716, "optim/total_tokens": 5993136128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.7888669967651367, "created_at": "2025-01-16T13:50:52.113177+00:00"} {"global_step": 11432, "acc_step": 0, "speed/wps": 12900.81243619799, "speed/FLOPS": 202624986884493.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04789208620786667, "optim/lr": 0.0028999301670488723, "optim/total_tokens": 5993660416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314573, "loss/out": 2.939438581466675, "created_at": "2025-01-16T13:51:02.278832+00:00"} {"global_step": 11433, "acc_step": 0, "speed/wps": 12902.296492997784, "speed/FLOPS": 202648296035842.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06911200284957886, "optim/lr": 0.002899899394423655, "optim/total_tokens": 5994184704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.859302043914795, "created_at": "2025-01-16T13:51:12.443295+00:00"} {"global_step": 11434, "acc_step": 0, "speed/wps": 12897.992178661661, "speed/FLOPS": 202580690864445.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05964605510234833, "optim/lr": 0.00289986861723102, "optim/total_tokens": 5994708992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.839682102203369, "created_at": "2025-01-16T13:51:22.609212+00:00"} {"global_step": 11435, "acc_step": 0, "speed/wps": 12903.311937568293, "speed/FLOPS": 202664244988189.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0530400276184082, "optim/lr": 0.0028998378354710684, "optim/total_tokens": 5995233280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 2.9976389408111572, "created_at": "2025-01-16T13:51:32.773472+00:00"} {"global_step": 11436, "acc_step": 0, "speed/wps": 12897.023492517708, "speed/FLOPS": 202565476317440.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06327138096094131, "optim/lr": 0.002899807049143899, "optim/total_tokens": 5995757568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.9992692470550537, "created_at": "2025-01-16T13:51:42.948154+00:00"} {"global_step": 11437, "acc_step": 0, "speed/wps": 12901.672320952937, "speed/FLOPS": 202638492556175.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052422989159822464, "optim/lr": 0.0028997762582496133, "optim/total_tokens": 5996281856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9328882694244385, "created_at": "2025-01-16T13:51:53.114399+00:00"} {"global_step": 11438, "acc_step": 0, "speed/wps": 12904.99007169853, "speed/FLOPS": 202690602390702.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04853401705622673, "optim/lr": 0.002899745462788313, "optim/total_tokens": 5996806144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.84519624710083, "created_at": "2025-01-16T13:52:03.275063+00:00"} {"global_step": 11439, "acc_step": 0, "speed/wps": 12900.161963500655, "speed/FLOPS": 202614770316938.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04977283626794815, "optim/lr": 0.002899714662760096, "optim/total_tokens": 5997330432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8733184337615967, "created_at": "2025-01-16T13:52:13.442529+00:00"} {"global_step": 11440, "acc_step": 0, "speed/wps": 12898.263226351115, "speed/FLOPS": 202584948040867.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04636450484395027, "optim/lr": 0.002899683858165064, "optim/total_tokens": 5997854720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471773, "loss/out": 2.8518195152282715, "created_at": "2025-01-16T13:52:23.608120+00:00"} {"global_step": 11441, "acc_step": 0, "speed/wps": 12895.923508981708, "speed/FLOPS": 202548199564472.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04857942834496498, "optim/lr": 0.0028996530490033178, "optim/total_tokens": 5998379008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8985517024993896, "created_at": "2025-01-16T13:52:33.778534+00:00"} {"global_step": 11442, "acc_step": 0, "speed/wps": 12898.741120470295, "speed/FLOPS": 202592454024704.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04361778497695923, "optim/lr": 0.002899622235274958, "optim/total_tokens": 5998903296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.8905272483825684, "created_at": "2025-01-16T13:52:43.943729+00:00"} {"global_step": 11443, "acc_step": 0, "speed/wps": 12904.77707952169, "speed/FLOPS": 202687257055883.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044252797961235046, "optim/lr": 0.0028995914169800845, "optim/total_tokens": 5999427584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 2.8931374549865723, "created_at": "2025-01-16T13:52:54.109023+00:00"} {"global_step": 11444, "acc_step": 0, "speed/wps": 12900.592801308869, "speed/FLOPS": 202621537216750.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04498409107327461, "optim/lr": 0.0028995605941187987, "optim/total_tokens": 5999951872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.896352529525757, "created_at": "2025-01-16T13:53:04.273112+00:00"} {"global_step": 11445, "acc_step": 0, "speed/wps": 12900.626486742487, "speed/FLOPS": 202622066292773.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049691010266542435, "optim/lr": 0.0028995297666912, "optim/total_tokens": 6000476160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9113681316375732, "created_at": "2025-01-16T13:53:14.438151+00:00"} {"global_step": 11446, "acc_step": 0, "speed/wps": 12901.368050742678, "speed/FLOPS": 202633713574412.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0649823248386383, "optim/lr": 0.00289949893469739, "optim/total_tokens": 6001000448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 2.9601874351501465, "created_at": "2025-01-16T13:53:24.602906+00:00"} {"global_step": 11447, "acc_step": 0, "speed/wps": 12902.945856070799, "speed/FLOPS": 202658495175222.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.03963455557823181, "optim/lr": 0.0028994680981374687, "optim/total_tokens": 6001524736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.838489532470703, "created_at": "2025-01-16T13:53:34.764902+00:00"} {"global_step": 11448, "acc_step": 0, "speed/wps": 12899.61569993596, "speed/FLOPS": 202606190497010.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056315742433071136, "optim/lr": 0.0028994372570115374, "optim/total_tokens": 6002049024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.9461309909820557, "created_at": "2025-01-16T13:53:44.930521+00:00"} {"global_step": 11449, "acc_step": 0, "speed/wps": 12902.992142871231, "speed/FLOPS": 202659222173027.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05503414198756218, "optim/lr": 0.0028994064113196955, "optim/total_tokens": 6002573312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.8556365966796875, "created_at": "2025-01-16T13:53:55.092247+00:00"} {"global_step": 11450, "acc_step": 0, "speed/wps": 12902.216099738864, "speed/FLOPS": 202647033349239.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07358911633491516, "optim/lr": 0.002899375561062045, "optim/total_tokens": 6003097600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.725484848022461, "created_at": "2025-01-16T13:54:05.255077+00:00"} {"global_step": 11451, "acc_step": 0, "speed/wps": 12897.13499805444, "speed/FLOPS": 202567227665119.75, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07420556992292404, "optim/lr": 0.0028993447062386855, "optim/total_tokens": 6003621888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.962012767791748, "created_at": "2025-01-16T13:54:15.425026+00:00"} {"global_step": 11452, "acc_step": 0, "speed/wps": 12900.778901160571, "speed/FLOPS": 202624460170649.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06341806799173355, "optim/lr": 0.0028993138468497185, "optim/total_tokens": 6004146176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8451805114746094, "created_at": "2025-01-16T13:54:25.588593+00:00"} {"global_step": 11453, "acc_step": 0, "speed/wps": 12899.708232001401, "speed/FLOPS": 202607643840250.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07160470634698868, "optim/lr": 0.0028992829828952443, "optim/total_tokens": 6004670464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8183159828186035, "created_at": "2025-01-16T13:54:35.755449+00:00"} {"global_step": 11454, "acc_step": 0, "speed/wps": 12900.528648581223, "speed/FLOPS": 202620529610007.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05058320239186287, "optim/lr": 0.002899252114375363, "optim/total_tokens": 6005194752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 2.8392136096954346, "created_at": "2025-01-16T13:54:45.921480+00:00"} {"global_step": 11455, "acc_step": 0, "speed/wps": 12902.482476405214, "speed/FLOPS": 202651217160823.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04796673730015755, "optim/lr": 0.0028992212412901766, "optim/total_tokens": 6005719040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.876796245574951, "created_at": "2025-01-16T13:54:56.086714+00:00"} {"global_step": 11456, "acc_step": 0, "speed/wps": 12904.706889503499, "speed/FLOPS": 202686154625196.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052375391125679016, "optim/lr": 0.002899190363639785, "optim/total_tokens": 6006243328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.790743589401245, "created_at": "2025-01-16T13:55:06.248396+00:00"} {"global_step": 11457, "acc_step": 0, "speed/wps": 12905.335530217038, "speed/FLOPS": 202696028291445.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05062704533338547, "optim/lr": 0.002899159481424289, "optim/total_tokens": 6006767616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.928469181060791, "created_at": "2025-01-16T13:55:16.408481+00:00"} {"global_step": 11458, "acc_step": 0, "speed/wps": 12899.084070353047, "speed/FLOPS": 202597840523875.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05602366849780083, "optim/lr": 0.002899128594643789, "optim/total_tokens": 6007291904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.941046714782715, "created_at": "2025-01-16T13:55:26.576532+00:00"} {"global_step": 11459, "acc_step": 0, "speed/wps": 12904.43328020348, "speed/FLOPS": 202681857215161.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061238788068294525, "optim/lr": 0.0028990977032983868, "optim/total_tokens": 6007816192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 2.914454221725464, "created_at": "2025-01-16T13:55:36.739194+00:00"} {"global_step": 11460, "acc_step": 0, "speed/wps": 12900.10447924866, "speed/FLOPS": 202613867447769.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055102258920669556, "optim/lr": 0.002899066807388182, "optim/total_tokens": 6008340480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9236981868743896, "created_at": "2025-01-16T13:55:46.904912+00:00"} {"global_step": 11461, "acc_step": 0, "speed/wps": 12904.950201913547, "speed/FLOPS": 202689976180942.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05167759209871292, "optim/lr": 0.0028990359069132766, "optim/total_tokens": 6008864768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.9268856048583984, "created_at": "2025-01-16T13:55:57.065371+00:00"} {"global_step": 11462, "acc_step": 0, "speed/wps": 12900.368886217739, "speed/FLOPS": 202618020322552.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05606318265199661, "optim/lr": 0.0028990050018737707, "optim/total_tokens": 6009389056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.9510557651519775, "created_at": "2025-01-16T13:56:07.232762+00:00"} {"global_step": 11463, "acc_step": 0, "speed/wps": 12903.567936645526, "speed/FLOPS": 202668265805476.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052992988377809525, "optim/lr": 0.0028989740922697655, "optim/total_tokens": 6009913344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.8579108715057373, "created_at": "2025-01-16T13:56:17.395629+00:00"} {"global_step": 11464, "acc_step": 0, "speed/wps": 12904.635021196991, "speed/FLOPS": 202685025834685.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05052623897790909, "optim/lr": 0.0028989431781013607, "optim/total_tokens": 6010437632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.7154574394226074, "created_at": "2025-01-16T13:56:27.557745+00:00"} {"global_step": 11465, "acc_step": 0, "speed/wps": 12896.947437359608, "speed/FLOPS": 202564281766672.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043318863958120346, "optim/lr": 0.002898912259368659, "optim/total_tokens": 6010961920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.840768814086914, "created_at": "2025-01-16T13:56:37.726893+00:00"} {"global_step": 11466, "acc_step": 0, "speed/wps": 12901.845766918139, "speed/FLOPS": 202641216763398.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05032212659716606, "optim/lr": 0.00289888133607176, "optim/total_tokens": 6011486208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.0114662647247314, "created_at": "2025-01-16T13:56:47.889798+00:00"} {"global_step": 11467, "acc_step": 0, "speed/wps": 12903.365452945794, "speed/FLOPS": 202665085520730.7, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04808548837900162, "optim/lr": 0.002898850408210765, "optim/total_tokens": 6012010496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 2.8714218139648438, "created_at": "2025-01-16T13:56:58.059314+00:00"} {"global_step": 11468, "acc_step": 0, "speed/wps": 12899.039683069797, "speed/FLOPS": 202597143360597.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051601700484752655, "optim/lr": 0.002898819475785775, "optim/total_tokens": 6012534784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.9738571643829346, "created_at": "2025-01-16T13:57:08.227829+00:00"} {"global_step": 11469, "acc_step": 0, "speed/wps": 12902.444033267571, "speed/FLOPS": 202650613358520.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05799734219908714, "optim/lr": 0.00289878853879689, "optim/total_tokens": 6013059072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.8911843299865723, "created_at": "2025-01-16T13:57:18.393048+00:00"} {"global_step": 11470, "acc_step": 0, "speed/wps": 12898.233711850653, "speed/FLOPS": 202584484475080.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07955840975046158, "optim/lr": 0.0028987575972442125, "optim/total_tokens": 6013583360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288392, "loss/out": 2.7841310501098633, "created_at": "2025-01-16T13:57:28.558755+00:00"} {"global_step": 11471, "acc_step": 0, "speed/wps": 12889.049905856604, "speed/FLOPS": 202440240182071.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052870262414216995, "optim/lr": 0.0028987266511278425, "optim/total_tokens": 6014107648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.863521099090576, "created_at": "2025-01-16T13:57:38.735372+00:00"} {"global_step": 11472, "acc_step": 0, "speed/wps": 12900.565488636626, "speed/FLOPS": 202621108233701.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0748138278722763, "optim/lr": 0.002898695700447881, "optim/total_tokens": 6014631936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.8428566455841064, "created_at": "2025-01-16T13:57:48.899512+00:00"} {"global_step": 11473, "acc_step": 0, "speed/wps": 12906.012154955928, "speed/FLOPS": 202706655612749.5, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052141979336738586, "optim/lr": 0.002898664745204429, "optim/total_tokens": 6015156224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.8793749809265137, "created_at": "2025-01-16T13:57:59.059021+00:00"} {"global_step": 11474, "acc_step": 0, "speed/wps": 12899.760674640214, "speed/FLOPS": 202608467523957.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11087080836296082, "optim/lr": 0.002898633785397588, "optim/total_tokens": 6015680512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.916572093963623, "created_at": "2025-01-16T13:58:09.223845+00:00"} {"global_step": 11475, "acc_step": 0, "speed/wps": 12907.79259646146, "speed/FLOPS": 202734619893176.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.12956155836582184, "optim/lr": 0.002898602821027458, "optim/total_tokens": 6016204800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.845893383026123, "created_at": "2025-01-16T13:58:19.383085+00:00"} {"global_step": 11476, "acc_step": 0, "speed/wps": 12908.092075472638, "speed/FLOPS": 202739323622574.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08544415235519409, "optim/lr": 0.0028985718520941413, "optim/total_tokens": 6016729088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.791524887084961, "created_at": "2025-01-16T13:58:29.544581+00:00"} {"global_step": 11477, "acc_step": 0, "speed/wps": 12906.318250372406, "speed/FLOPS": 202711463261884.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08326795697212219, "optim/lr": 0.002898540878597738, "optim/total_tokens": 6017253376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9239468574523926, "created_at": "2025-01-16T13:58:39.703850+00:00"} {"global_step": 11478, "acc_step": 0, "speed/wps": 12899.5331007504, "speed/FLOPS": 202604893163298.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05604556202888489, "optim/lr": 0.0028985099005383493, "optim/total_tokens": 6017777664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 2.8764891624450684, "created_at": "2025-01-16T13:58:49.868374+00:00"} {"global_step": 11479, "acc_step": 0, "speed/wps": 12900.630532175523, "speed/FLOPS": 202622129831858.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06360924988985062, "optim/lr": 0.002898478917916077, "optim/total_tokens": 6018301952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.9427390098571777, "created_at": "2025-01-16T13:59:00.034405+00:00"} {"global_step": 11480, "acc_step": 0, "speed/wps": 12905.435174999877, "speed/FLOPS": 202697593349686.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04636204242706299, "optim/lr": 0.0028984479307310206, "optim/total_tokens": 6018826240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472295, "loss/out": 2.9661242961883545, "created_at": "2025-01-16T13:59:10.198989+00:00"} {"global_step": 11481, "acc_step": 0, "speed/wps": 12898.82894665526, "speed/FLOPS": 202593833455624.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07352573424577713, "optim/lr": 0.002898416938983283, "optim/total_tokens": 6019350528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.917123317718506, "created_at": "2025-01-16T13:59:20.364842+00:00"} {"global_step": 11482, "acc_step": 0, "speed/wps": 12899.672821578593, "speed/FLOPS": 202607087670901.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05020967125892639, "optim/lr": 0.0028983859426729647, "optim/total_tokens": 6019874816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.886178970336914, "created_at": "2025-01-16T13:59:30.531515+00:00"} {"global_step": 11483, "acc_step": 0, "speed/wps": 12902.283410853768, "speed/FLOPS": 202648090562793.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06816490739583969, "optim/lr": 0.002898354941800166, "optim/total_tokens": 6020399104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411602, "loss/out": 2.92972993850708, "created_at": "2025-01-16T13:59:40.702201+00:00"} {"global_step": 11484, "acc_step": 0, "speed/wps": 12897.61353075939, "speed/FLOPS": 202574743678823.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061687108129262924, "optim/lr": 0.0028983239363649885, "optim/total_tokens": 6020923392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9964656829833984, "created_at": "2025-01-16T13:59:50.869591+00:00"} {"global_step": 11485, "acc_step": 0, "speed/wps": 12895.344365735304, "speed/FLOPS": 202539103323961.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060119304805994034, "optim/lr": 0.002898292926367534, "optim/total_tokens": 6021447680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.8683981895446777, "created_at": "2025-01-16T14:00:01.038502+00:00"} {"global_step": 11486, "acc_step": 0, "speed/wps": 12898.400249259608, "speed/FLOPS": 202587100173934.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06066583842039108, "optim/lr": 0.0028982619118079036, "optim/total_tokens": 6021971968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 2.8482041358947754, "created_at": "2025-01-16T14:00:11.204001+00:00"} {"global_step": 11487, "acc_step": 0, "speed/wps": 12902.08894542725, "speed/FLOPS": 202645036216046.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06163063272833824, "optim/lr": 0.002898230892686198, "optim/total_tokens": 6022496256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.89029598236084, "created_at": "2025-01-16T14:00:21.369232+00:00"} {"global_step": 11488, "acc_step": 0, "speed/wps": 12904.275611214016, "speed/FLOPS": 202679380807021.62, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07019098103046417, "optim/lr": 0.002898199869002518, "optim/total_tokens": 6023020544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.7974400520324707, "created_at": "2025-01-16T14:00:31.530116+00:00"} {"global_step": 11489, "acc_step": 0, "speed/wps": 12905.006137778044, "speed/FLOPS": 202690854730558.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05534130707383156, "optim/lr": 0.0028981688407569654, "optim/total_tokens": 6023544832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 2.8151373863220215, "created_at": "2025-01-16T14:00:41.693034+00:00"} {"global_step": 11490, "acc_step": 0, "speed/wps": 12908.94477637271, "speed/FLOPS": 202752716461944.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08394607156515121, "optim/lr": 0.0028981378079496415, "optim/total_tokens": 6024069120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474572, "loss/out": 2.9212188720703125, "created_at": "2025-01-16T14:00:51.850336+00:00"} {"global_step": 11491, "acc_step": 0, "speed/wps": 12904.512224409658, "speed/FLOPS": 202683097142399.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05025892332196236, "optim/lr": 0.0028981067705806476, "optim/total_tokens": 6024593408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.883349895477295, "created_at": "2025-01-16T14:01:02.016705+00:00"} {"global_step": 11492, "acc_step": 0, "speed/wps": 12902.27453015123, "speed/FLOPS": 202647951079157.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07828065752983093, "optim/lr": 0.002898075728650085, "optim/total_tokens": 6025117696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.8699750900268555, "created_at": "2025-01-16T14:01:12.182130+00:00"} {"global_step": 11493, "acc_step": 0, "speed/wps": 12900.17537006305, "speed/FLOPS": 202614980885423.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04118559882044792, "optim/lr": 0.0028980446821580547, "optim/total_tokens": 6025641984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 2.929222583770752, "created_at": "2025-01-16T14:01:22.347224+00:00"} {"global_step": 11494, "acc_step": 0, "speed/wps": 12901.81348731612, "speed/FLOPS": 202640709767892.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055512044578790665, "optim/lr": 0.002898013631104658, "optim/total_tokens": 6026166272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.945863962173462, "created_at": "2025-01-16T14:01:32.515278+00:00"} {"global_step": 11495, "acc_step": 0, "speed/wps": 12906.145190691022, "speed/FLOPS": 202708745121779.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05803034082055092, "optim/lr": 0.002897982575489996, "optim/total_tokens": 6026690560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.902573347091675, "created_at": "2025-01-16T14:01:42.674596+00:00"} {"global_step": 11496, "acc_step": 0, "speed/wps": 12898.016656519196, "speed/FLOPS": 202581075322834.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.047027453780174255, "optim/lr": 0.00289795151531417, "optim/total_tokens": 6027214848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8342132568359375, "created_at": "2025-01-16T14:01:52.840252+00:00"} {"global_step": 11497, "acc_step": 0, "speed/wps": 12904.999237772432, "speed/FLOPS": 202690746356488.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054807256907224655, "optim/lr": 0.002897920450577282, "optim/total_tokens": 6027739136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8347392082214355, "created_at": "2025-01-16T14:02:03.001797+00:00"} {"global_step": 11498, "acc_step": 0, "speed/wps": 12905.850309910824, "speed/FLOPS": 202704113613917.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04929376393556595, "optim/lr": 0.0028978893812794337, "optim/total_tokens": 6028263424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416139, "loss/out": 2.9239754676818848, "created_at": "2025-01-16T14:02:13.166925+00:00"} {"global_step": 11499, "acc_step": 0, "speed/wps": 12902.576195820437, "speed/FLOPS": 202652689153024.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0646560862660408, "optim/lr": 0.0028978583074207253, "optim/total_tokens": 6028787712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 2.8560986518859863, "created_at": "2025-01-16T14:02:23.334411+00:00"} {"global_step": 11500, "acc_step": 0, "speed/wps": 12906.006036178152, "speed/FLOPS": 202706559508936.47, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053203798830509186, "optim/lr": 0.0028978272290012584, "optim/total_tokens": 6029312000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8891804218292236, "created_at": "2025-01-16T14:02:33.496258+00:00"} {"global_step": 11501, "acc_step": 0, "speed/wps": 12899.317745560285, "speed/FLOPS": 202601510714108.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0774020329117775, "optim/lr": 0.002897796146021135, "optim/total_tokens": 6029836288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8495888710021973, "created_at": "2025-01-16T14:02:43.663457+00:00"} {"global_step": 11502, "acc_step": 0, "speed/wps": 12904.641398993164, "speed/FLOPS": 202685126006738.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05054231360554695, "optim/lr": 0.002897765058480456, "optim/total_tokens": 6030360576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.856126308441162, "created_at": "2025-01-16T14:02:53.824450+00:00"} {"global_step": 11503, "acc_step": 0, "speed/wps": 12897.182816279297, "speed/FLOPS": 202567978716050.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05625545233488083, "optim/lr": 0.002897733966379323, "optim/total_tokens": 6030884864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.8371880054473877, "created_at": "2025-01-16T14:03:03.993792+00:00"} {"global_step": 11504, "acc_step": 0, "speed/wps": 12897.796075463573, "speed/FLOPS": 202577610794245.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04290684685111046, "optim/lr": 0.0028977028697178375, "optim/total_tokens": 6031409152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.8631720542907715, "created_at": "2025-01-16T14:03:14.162866+00:00"} {"global_step": 11505, "acc_step": 0, "speed/wps": 12904.276043081338, "speed/FLOPS": 202679387590091.34, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06664404273033142, "optim/lr": 0.0028976717684961, "optim/total_tokens": 6031933440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.710930824279785, "created_at": "2025-01-16T14:03:24.324226+00:00"} {"global_step": 11506, "acc_step": 0, "speed/wps": 12905.883068911056, "speed/FLOPS": 202704628139032.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03951817378401756, "optim/lr": 0.0028976406627142137, "optim/total_tokens": 6032457728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.941617012023926, "created_at": "2025-01-16T14:03:34.486263+00:00"} {"global_step": 11507, "acc_step": 0, "speed/wps": 12898.359670722502, "speed/FLOPS": 202586462832248.28, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06567493081092834, "optim/lr": 0.0028976095523722796, "optim/total_tokens": 6032982016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.6855053901672363, "created_at": "2025-01-16T14:03:44.652689+00:00"} {"global_step": 11508, "acc_step": 0, "speed/wps": 12901.0435025246, "speed/FLOPS": 202628616098671.6, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04583513364195824, "optim/lr": 0.0028975784374703985, "optim/total_tokens": 6033506304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8863511085510254, "created_at": "2025-01-16T14:03:54.819364+00:00"} {"global_step": 11509, "acc_step": 0, "speed/wps": 12902.874389467815, "speed/FLOPS": 202657372694017.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042843107134103775, "optim/lr": 0.0028975473180086722, "optim/total_tokens": 6034030592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.972470998764038, "created_at": "2025-01-16T14:04:04.983097+00:00"} {"global_step": 11510, "acc_step": 0, "speed/wps": 12904.473402373087, "speed/FLOPS": 202682487388968.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.03999582305550575, "optim/lr": 0.0028975161939872023, "optim/total_tokens": 6034554880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9233880043029785, "created_at": "2025-01-16T14:04:15.144114+00:00"} {"global_step": 11511, "acc_step": 0, "speed/wps": 12902.765751432411, "speed/FLOPS": 202655666384386.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07172726839780807, "optim/lr": 0.0028974850654060903, "optim/total_tokens": 6035079168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 2.762653350830078, "created_at": "2025-01-16T14:04:25.311170+00:00"} {"global_step": 11512, "acc_step": 0, "speed/wps": 12902.229637133198, "speed/FLOPS": 202647245972619.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07005691528320312, "optim/lr": 0.0028974539322654384, "optim/total_tokens": 6035603456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 2.8134617805480957, "created_at": "2025-01-16T14:04:35.475833+00:00"} {"global_step": 11513, "acc_step": 0, "speed/wps": 12903.861492153443, "speed/FLOPS": 202672876498116.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04916137829422951, "optim/lr": 0.002897422794565347, "optim/total_tokens": 6036127744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.896838426589966, "created_at": "2025-01-16T14:04:45.640553+00:00"} {"global_step": 11514, "acc_step": 0, "speed/wps": 12910.34689652946, "speed/FLOPS": 202774738685720.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0498049259185791, "optim/lr": 0.0028973916523059185, "optim/total_tokens": 6036652032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9789624214172363, "created_at": "2025-01-16T14:04:55.796592+00:00"} {"global_step": 11515, "acc_step": 0, "speed/wps": 12903.72243183946, "speed/FLOPS": 202670692364796.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0520729199051857, "optim/lr": 0.0028973605054872544, "optim/total_tokens": 6037176320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 2.9257853031158447, "created_at": "2025-01-16T14:05:05.958205+00:00"} {"global_step": 11516, "acc_step": 0, "speed/wps": 12899.369129231498, "speed/FLOPS": 202602317765273.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05239398404955864, "optim/lr": 0.002897329354109456, "optim/total_tokens": 6037700608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.816100835800171, "created_at": "2025-01-16T14:05:16.131636+00:00"} {"global_step": 11517, "acc_step": 0, "speed/wps": 12907.655130060402, "speed/FLOPS": 202732460794449.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05011455714702606, "optim/lr": 0.002897298198172625, "optim/total_tokens": 6038224896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 2.9120571613311768, "created_at": "2025-01-16T14:05:26.293719+00:00"} {"global_step": 11518, "acc_step": 0, "speed/wps": 12903.425777171833, "speed/FLOPS": 202666032995594.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050548408180475235, "optim/lr": 0.002897267037676863, "optim/total_tokens": 6038749184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.889633893966675, "created_at": "2025-01-16T14:05:36.455420+00:00"} {"global_step": 11519, "acc_step": 0, "speed/wps": 12901.557646550395, "speed/FLOPS": 202636691437104.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04549354314804077, "optim/lr": 0.0028972358726222732, "optim/total_tokens": 6039273472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8327393531799316, "created_at": "2025-01-16T14:05:46.621934+00:00"} {"global_step": 11520, "acc_step": 0, "speed/wps": 12908.248014490937, "speed/FLOPS": 202741772859139.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04779410734772682, "optim/lr": 0.002897204703008955, "optim/total_tokens": 6039797760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.871584415435791, "created_at": "2025-01-16T14:05:56.780274+00:00"} {"global_step": 11521, "acc_step": 0, "speed/wps": 12900.138214159493, "speed/FLOPS": 202614397300901.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.041524261236190796, "optim/lr": 0.002897173528837011, "optim/total_tokens": 6040322048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8938536643981934, "created_at": "2025-01-16T14:06:06.946536+00:00"} {"global_step": 11522, "acc_step": 0, "speed/wps": 12901.468598154132, "speed/FLOPS": 202635292809675.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04422735050320625, "optim/lr": 0.002897142350106543, "optim/total_tokens": 6040846336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9114553928375244, "created_at": "2025-01-16T14:06:17.112826+00:00"} {"global_step": 11523, "acc_step": 0, "speed/wps": 12897.649529142831, "speed/FLOPS": 202575309082902.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04458363726735115, "optim/lr": 0.0028971111668176525, "optim/total_tokens": 6041370624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.940676689147949, "created_at": "2025-01-16T14:06:27.278912+00:00"} {"global_step": 11524, "acc_step": 0, "speed/wps": 12899.420957859882, "speed/FLOPS": 202603131805102.16, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04234399273991585, "optim/lr": 0.0028970799789704424, "optim/total_tokens": 6041894912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.8980369567871094, "created_at": "2025-01-16T14:06:37.444797+00:00"} {"global_step": 11525, "acc_step": 0, "speed/wps": 12901.632164160086, "speed/FLOPS": 202637861838562.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057598087936639786, "optim/lr": 0.0028970487865650125, "optim/total_tokens": 6042419200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.851733684539795, "created_at": "2025-01-16T14:06:47.608324+00:00"} {"global_step": 11526, "acc_step": 0, "speed/wps": 12908.635730156553, "speed/FLOPS": 202747862466445.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05023779720067978, "optim/lr": 0.0028970175896014657, "optim/total_tokens": 6042943488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 2.8944973945617676, "created_at": "2025-01-16T14:06:57.766003+00:00"} {"global_step": 11527, "acc_step": 0, "speed/wps": 12903.365415088796, "speed/FLOPS": 202665084926134.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0455930158495903, "optim/lr": 0.002896986388079904, "optim/total_tokens": 6043467776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373078, "loss/out": 2.736661195755005, "created_at": "2025-01-16T14:07:07.930121+00:00"} {"global_step": 11528, "acc_step": 0, "speed/wps": 12905.714236390044, "speed/FLOPS": 202701976392289.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04794979840517044, "optim/lr": 0.0028969551820004287, "optim/total_tokens": 6043992064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475100, "loss/out": 2.800690174102783, "created_at": "2025-01-16T14:07:18.090219+00:00"} {"global_step": 11529, "acc_step": 0, "speed/wps": 12901.211561229245, "speed/FLOPS": 202631255691564.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05490020290017128, "optim/lr": 0.0028969239713631414, "optim/total_tokens": 6044516352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.892491579055786, "created_at": "2025-01-16T14:07:28.255368+00:00"} {"global_step": 11530, "acc_step": 0, "speed/wps": 12907.746979595073, "speed/FLOPS": 202733903417606.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056335337460041046, "optim/lr": 0.002896892756168145, "optim/total_tokens": 6045040640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.794081926345825, "created_at": "2025-01-16T14:07:38.416224+00:00"} {"global_step": 11531, "acc_step": 0, "speed/wps": 12903.511611008144, "speed/FLOPS": 202667381133942.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05251898244023323, "optim/lr": 0.00289686153641554, "optim/total_tokens": 6045564928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9203484058380127, "created_at": "2025-01-16T14:07:48.586003+00:00"} {"global_step": 11532, "acc_step": 0, "speed/wps": 12904.791760618204, "speed/FLOPS": 202687487642678.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0463222861289978, "optim/lr": 0.0028968303121054296, "optim/total_tokens": 6046089216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.8531863689422607, "created_at": "2025-01-16T14:07:58.746436+00:00"} {"global_step": 11533, "acc_step": 0, "speed/wps": 12901.131212238135, "speed/FLOPS": 202629993700248.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04766116291284561, "optim/lr": 0.002896799083237914, "optim/total_tokens": 6046613504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.861689567565918, "created_at": "2025-01-16T14:08:08.910203+00:00"} {"global_step": 11534, "acc_step": 0, "speed/wps": 12903.692855673813, "speed/FLOPS": 202670227830472.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04754924774169922, "optim/lr": 0.002896767849813097, "optim/total_tokens": 6047137792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9249379634857178, "created_at": "2025-01-16T14:08:19.071552+00:00"} {"global_step": 11535, "acc_step": 0, "speed/wps": 12903.4852016584, "speed/FLOPS": 202666966338814.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0661056637763977, "optim/lr": 0.0028967366118310793, "optim/total_tokens": 6047662080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 2.9306232929229736, "created_at": "2025-01-16T14:08:29.232771+00:00"} {"global_step": 11536, "acc_step": 0, "speed/wps": 12903.504647532496, "speed/FLOPS": 202667271762988.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05294833332300186, "optim/lr": 0.0028967053692919627, "optim/total_tokens": 6048186368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8620717525482178, "created_at": "2025-01-16T14:08:39.395857+00:00"} {"global_step": 11537, "acc_step": 0, "speed/wps": 12885.37460169887, "speed/FLOPS": 202382514479877.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07397744059562683, "optim/lr": 0.00289667412219585, "optim/total_tokens": 6048710656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372174, "loss/out": 2.8018975257873535, "created_at": "2025-01-16T14:08:49.571753+00:00"} {"global_step": 11538, "acc_step": 0, "speed/wps": 12891.89772421976, "speed/FLOPS": 202484969082777.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0758843719959259, "optim/lr": 0.0028966428705428423, "optim/total_tokens": 6049234944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.8562920093536377, "created_at": "2025-01-16T14:08:59.745334+00:00"} {"global_step": 11539, "acc_step": 0, "speed/wps": 12891.577755818338, "speed/FLOPS": 202479943539350.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0026, "optim/grad_norm": 0.04540680721402168, "optim/lr": 0.0028966116143330422, "optim/total_tokens": 6049759232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.9434163570404053, "created_at": "2025-01-16T14:09:09.916328+00:00"} {"global_step": 11540, "acc_step": 0, "speed/wps": 12893.47644213126, "speed/FLOPS": 202509765016963.53, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06782443821430206, "optim/lr": 0.002896580353566551, "optim/total_tokens": 6050283520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.887112855911255, "created_at": "2025-01-16T14:09:20.090968+00:00"} {"global_step": 11541, "acc_step": 0, "speed/wps": 12894.674606761151, "speed/FLOPS": 202528583838926.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06433174014091492, "optim/lr": 0.0028965490882434716, "optim/total_tokens": 6050807808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382664, "loss/out": 2.89547061920166, "created_at": "2025-01-16T14:09:30.259414+00:00"} {"global_step": 11542, "acc_step": 0, "speed/wps": 12898.713678505414, "speed/FLOPS": 202592023010936.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04239543154835701, "optim/lr": 0.002896517818363905, "optim/total_tokens": 6051332096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.83547306060791, "created_at": "2025-01-16T14:09:40.425729+00:00"} {"global_step": 11543, "acc_step": 0, "speed/wps": 12889.208448357811, "speed/FLOPS": 202442730309913.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04809460788965225, "optim/lr": 0.0028964865439279545, "optim/total_tokens": 6051856384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.873692512512207, "created_at": "2025-01-16T14:09:50.601603+00:00"} {"global_step": 11544, "acc_step": 0, "speed/wps": 12896.954178661981, "speed/FLOPS": 202564387648089.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0471692718565464, "optim/lr": 0.002896455264935721, "optim/total_tokens": 6052380672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290845, "loss/out": 2.8460187911987305, "created_at": "2025-01-16T14:10:00.769251+00:00"} {"global_step": 11545, "acc_step": 0, "speed/wps": 12907.87717313703, "speed/FLOPS": 202735948286087.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04974750801920891, "optim/lr": 0.002896423981387307, "optim/total_tokens": 6052904960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 2.881972312927246, "created_at": "2025-01-16T14:10:10.929870+00:00"} {"global_step": 11546, "acc_step": 0, "speed/wps": 12902.82674787151, "speed/FLOPS": 202656624417280.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050453465431928635, "optim/lr": 0.0028963926932828138, "optim/total_tokens": 6053429248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.846734046936035, "created_at": "2025-01-16T14:10:21.091966+00:00"} {"global_step": 11547, "acc_step": 0, "speed/wps": 12910.768251677315, "speed/FLOPS": 202781356647320.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06082361191511154, "optim/lr": 0.002896361400622345, "optim/total_tokens": 6053953536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457361, "loss/out": 2.8650641441345215, "created_at": "2025-01-16T14:10:31.247932+00:00"} {"global_step": 11548, "acc_step": 0, "speed/wps": 12900.281779284722, "speed/FLOPS": 202616652188471.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04326077178120613, "optim/lr": 0.002896330103406002, "optim/total_tokens": 6054477824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9195899963378906, "created_at": "2025-01-16T14:10:41.413513+00:00"} {"global_step": 11549, "acc_step": 0, "speed/wps": 12904.03248041705, "speed/FLOPS": 202675562103758.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06600339710712433, "optim/lr": 0.0028962988016338863, "optim/total_tokens": 6055002112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 2.8115105628967285, "created_at": "2025-01-16T14:10:51.574566+00:00"} {"global_step": 11550, "acc_step": 0, "speed/wps": 12908.807074672333, "speed/FLOPS": 202750553667517.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06284104287624359, "optim/lr": 0.0028962674953061005, "optim/total_tokens": 6055526400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.9065871238708496, "created_at": "2025-01-16T14:11:01.735669+00:00"} {"global_step": 11551, "acc_step": 0, "speed/wps": 12903.522885560762, "speed/FLOPS": 202667558216283.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041998617351055145, "optim/lr": 0.002896236184422747, "optim/total_tokens": 6056050688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421661, "loss/out": 2.890613555908203, "created_at": "2025-01-16T14:11:11.903585+00:00"} {"global_step": 11552, "acc_step": 0, "speed/wps": 12904.040751848588, "speed/FLOPS": 202675692017957.0, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062189437448978424, "optim/lr": 0.0028962048689839272, "optim/total_tokens": 6056574976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9460911750793457, "created_at": "2025-01-16T14:11:22.065430+00:00"} {"global_step": 11553, "acc_step": 0, "speed/wps": 12907.279967756771, "speed/FLOPS": 202726568354944.1, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05486997216939926, "optim/lr": 0.002896173548989744, "optim/total_tokens": 6057099264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.867640733718872, "created_at": "2025-01-16T14:11:32.225516+00:00"} {"global_step": 11554, "acc_step": 0, "speed/wps": 12903.873815565645, "speed/FLOPS": 202673070054238.88, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08039924502372742, "optim/lr": 0.0028961422244403, "optim/total_tokens": 6057623552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366652, "loss/out": 2.916078567504883, "created_at": "2025-01-16T14:11:42.386836+00:00"} {"global_step": 11555, "acc_step": 0, "speed/wps": 12900.907954892498, "speed/FLOPS": 202626487136847.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047444868832826614, "optim/lr": 0.0028961108953356966, "optim/total_tokens": 6058147840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.7731785774230957, "created_at": "2025-01-16T14:11:52.554773+00:00"} {"global_step": 11556, "acc_step": 0, "speed/wps": 12906.968696598364, "speed/FLOPS": 202721679413669.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0657007023692131, "optim/lr": 0.0028960795616760357, "optim/total_tokens": 6058672128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8211593627929688, "created_at": "2025-01-16T14:12:02.713474+00:00"} {"global_step": 11557, "acc_step": 0, "speed/wps": 12901.246871297313, "speed/FLOPS": 202631810284705.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0505513995885849, "optim/lr": 0.0028960482234614204, "optim/total_tokens": 6059196416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.8564534187316895, "created_at": "2025-01-16T14:12:12.877536+00:00"} {"global_step": 11558, "acc_step": 0, "speed/wps": 12895.679650614637, "speed/FLOPS": 202544369433720.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04906304180622101, "optim/lr": 0.0028960168806919525, "optim/total_tokens": 6059720704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8541831970214844, "created_at": "2025-01-16T14:12:23.045495+00:00"} {"global_step": 11559, "acc_step": 0, "speed/wps": 12904.208804645468, "speed/FLOPS": 202678331518060.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0453018844127655, "optim/lr": 0.0028959855333677343, "optim/total_tokens": 6060244992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346784, "loss/out": 2.810854911804199, "created_at": "2025-01-16T14:12:33.206387+00:00"} {"global_step": 11560, "acc_step": 0, "speed/wps": 12902.669817233256, "speed/FLOPS": 202654159605961.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05002811178565025, "optim/lr": 0.002895954181488868, "optim/total_tokens": 6060769280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9020962715148926, "created_at": "2025-01-16T14:12:43.370488+00:00"} {"global_step": 11561, "acc_step": 0, "speed/wps": 12901.641468573067, "speed/FLOPS": 202638007977154.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051062535494565964, "optim/lr": 0.0028959228250554562, "optim/total_tokens": 6061293568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.9083685874938965, "created_at": "2025-01-16T14:12:53.534528+00:00"} {"global_step": 11562, "acc_step": 0, "speed/wps": 12902.61025121214, "speed/FLOPS": 202653224039747.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04354643076658249, "optim/lr": 0.002895891464067601, "optim/total_tokens": 6061817856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288730, "loss/out": 2.8844974040985107, "created_at": "2025-01-16T14:13:03.696414+00:00"} {"global_step": 11563, "acc_step": 0, "speed/wps": 12903.865162199432, "speed/FLOPS": 202672934141231.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04906218871474266, "optim/lr": 0.002895860098525405, "optim/total_tokens": 6062342144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8060832023620605, "created_at": "2025-01-16T14:13:13.862380+00:00"} {"global_step": 11564, "acc_step": 0, "speed/wps": 12901.158958964605, "speed/FLOPS": 202630429500715.62, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.043978817760944366, "optim/lr": 0.0028958287284289695, "optim/total_tokens": 6062866432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361335, "loss/out": 2.9411540031433105, "created_at": "2025-01-16T14:13:24.026221+00:00"} {"global_step": 11565, "acc_step": 0, "speed/wps": 12897.157477622643, "speed/FLOPS": 202567580737628.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05151433125138283, "optim/lr": 0.0028957973537783984, "optim/total_tokens": 6063390720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.8557915687561035, "created_at": "2025-01-16T14:13:34.192890+00:00"} {"global_step": 11566, "acc_step": 0, "speed/wps": 12904.398673272546, "speed/FLOPS": 202681313665755.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047164514660835266, "optim/lr": 0.0028957659745737926, "optim/total_tokens": 6063915008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.929748773574829, "created_at": "2025-01-16T14:13:44.358509+00:00"} {"global_step": 11567, "acc_step": 0, "speed/wps": 12904.135055909697, "speed/FLOPS": 202677173192824.06, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0473334901034832, "optim/lr": 0.0028957345908152557, "optim/total_tokens": 6064439296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403804, "loss/out": 2.9393858909606934, "created_at": "2025-01-16T14:13:54.519905+00:00"} {"global_step": 11568, "acc_step": 0, "speed/wps": 12903.651431050197, "speed/FLOPS": 202669577199835.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04051348567008972, "optim/lr": 0.0028957032025028894, "optim/total_tokens": 6064963584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.043513298034668, "created_at": "2025-01-16T14:14:04.682839+00:00"} {"global_step": 11569, "acc_step": 0, "speed/wps": 12902.829761984027, "speed/FLOPS": 202656671758059.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04726585000753403, "optim/lr": 0.0028956718096367962, "optim/total_tokens": 6065487872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.986912250518799, "created_at": "2025-01-16T14:14:14.845005+00:00"} {"global_step": 11570, "acc_step": 0, "speed/wps": 12901.291017137097, "speed/FLOPS": 202632503655781.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.039779361337423325, "optim/lr": 0.002895640412217079, "optim/total_tokens": 6066012160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 2.861135721206665, "created_at": "2025-01-16T14:14:25.009188+00:00"} {"global_step": 11571, "acc_step": 0, "speed/wps": 12905.215893896726, "speed/FLOPS": 202694149238637.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05461730435490608, "optim/lr": 0.002895609010243839, "optim/total_tokens": 6066536448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8950071334838867, "created_at": "2025-01-16T14:14:35.172443+00:00"} {"global_step": 11572, "acc_step": 0, "speed/wps": 12900.94577511414, "speed/FLOPS": 202627081155398.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052083879709243774, "optim/lr": 0.00289557760371718, "optim/total_tokens": 6067060736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 3.0368847846984863, "created_at": "2025-01-16T14:14:45.335834+00:00"} {"global_step": 11573, "acc_step": 0, "speed/wps": 12908.582186446194, "speed/FLOPS": 202747021488897.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0631546676158905, "optim/lr": 0.002895546192637204, "optim/total_tokens": 6067585024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.8939449787139893, "created_at": "2025-01-16T14:14:55.506319+00:00"} {"global_step": 11574, "acc_step": 0, "speed/wps": 12899.600502107716, "speed/FLOPS": 202605951794234.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057597815990448, "optim/lr": 0.0028955147770040133, "optim/total_tokens": 6068109312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.7424683570861816, "created_at": "2025-01-16T14:15:05.672367+00:00"} {"global_step": 11575, "acc_step": 0, "speed/wps": 12898.704736514423, "speed/FLOPS": 202591882564680.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05004210025072098, "optim/lr": 0.0028954833568177107, "optim/total_tokens": 6068633600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379498, "loss/out": 2.8607301712036133, "created_at": "2025-01-16T14:15:15.843723+00:00"} {"global_step": 11576, "acc_step": 0, "speed/wps": 12906.47860753009, "speed/FLOPS": 202713981891399.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.049257829785346985, "optim/lr": 0.0028954519320783987, "optim/total_tokens": 6069157888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.8290958404541016, "created_at": "2025-01-16T14:15:26.003777+00:00"} {"global_step": 11577, "acc_step": 0, "speed/wps": 12905.25627535931, "speed/FLOPS": 202694783484997.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04809791222214699, "optim/lr": 0.002895420502786179, "optim/total_tokens": 6069682176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467961, "loss/out": 2.882439613342285, "created_at": "2025-01-16T14:15:36.163824+00:00"} {"global_step": 11578, "acc_step": 0, "speed/wps": 12905.897751340637, "speed/FLOPS": 202704858746764.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05146178975701332, "optim/lr": 0.0028953890689411557, "optim/total_tokens": 6070206464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9473676681518555, "created_at": "2025-01-16T14:15:46.325927+00:00"} {"global_step": 11579, "acc_step": 0, "speed/wps": 12902.015453928176, "speed/FLOPS": 202643881931065.94, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04615257680416107, "optim/lr": 0.00289535763054343, "optim/total_tokens": 6070730752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.884547472000122, "created_at": "2025-01-16T14:15:56.491869+00:00"} {"global_step": 11580, "acc_step": 0, "speed/wps": 12907.83997285385, "speed/FLOPS": 202735364004521.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04738452285528183, "optim/lr": 0.0028953261875931053, "optim/total_tokens": 6071255040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419583, "loss/out": 2.810150623321533, "created_at": "2025-01-16T14:16:06.655895+00:00"} {"global_step": 11581, "acc_step": 0, "speed/wps": 12901.920432153645, "speed/FLOPS": 202642389483523.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051811326295137405, "optim/lr": 0.0028952947400902834, "optim/total_tokens": 6071779328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8362905979156494, "created_at": "2025-01-16T14:16:16.821581+00:00"} {"global_step": 11582, "acc_step": 0, "speed/wps": 12906.983226434184, "speed/FLOPS": 202721907624707.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04573702812194824, "optim/lr": 0.002895263288035067, "optim/total_tokens": 6072303616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 2.803460121154785, "created_at": "2025-01-16T14:16:26.984407+00:00"} {"global_step": 11583, "acc_step": 0, "speed/wps": 12908.92653840714, "speed/FLOPS": 202752430009633.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06486901640892029, "optim/lr": 0.00289523183142756, "optim/total_tokens": 6072827904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.07480788230896, "created_at": "2025-01-16T14:16:37.144144+00:00"} {"global_step": 11584, "acc_step": 0, "speed/wps": 12905.203023571594, "speed/FLOPS": 202693947092495.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06496534496545792, "optim/lr": 0.002895200370267864, "optim/total_tokens": 6073352192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.031508445739746, "created_at": "2025-01-16T14:16:47.305964+00:00"} {"global_step": 11585, "acc_step": 0, "speed/wps": 12902.059071588978, "speed/FLOPS": 202644567006365.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047299575060606, "optim/lr": 0.002895168904556081, "optim/total_tokens": 6073876480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 2.9040427207946777, "created_at": "2025-01-16T14:16:57.471190+00:00"} {"global_step": 11586, "acc_step": 0, "speed/wps": 12905.986554359744, "speed/FLOPS": 202706253520207.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06382813304662704, "optim/lr": 0.002895137434292315, "optim/total_tokens": 6074400768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.869750738143921, "created_at": "2025-01-16T14:17:07.631603+00:00"} {"global_step": 11587, "acc_step": 0, "speed/wps": 12911.589325917548, "speed/FLOPS": 202794252746537.78, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048944734036922455, "optim/lr": 0.0028951059594766676, "optim/total_tokens": 6074925056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9327878952026367, "created_at": "2025-01-16T14:17:17.792706+00:00"} {"global_step": 11588, "acc_step": 0, "speed/wps": 12905.336411844495, "speed/FLOPS": 202696042138616.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052116770297288895, "optim/lr": 0.002895074480109243, "optim/total_tokens": 6075449344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 2.8783891201019287, "created_at": "2025-01-16T14:17:27.952626+00:00"} {"global_step": 11589, "acc_step": 0, "speed/wps": 12907.08671413858, "speed/FLOPS": 202723533041312.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051151301711797714, "optim/lr": 0.0028950429961901417, "optim/total_tokens": 6075973632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.831388473510742, "created_at": "2025-01-16T14:17:38.112764+00:00"} {"global_step": 11590, "acc_step": 0, "speed/wps": 12908.81643443019, "speed/FLOPS": 202750700675376.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04486563056707382, "optim/lr": 0.0028950115077194683, "optim/total_tokens": 6076497920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 2.9563636779785156, "created_at": "2025-01-16T14:17:48.270526+00:00"} {"global_step": 11591, "acc_step": 0, "speed/wps": 12901.445050970508, "speed/FLOPS": 202634922968799.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045096494257450104, "optim/lr": 0.0028949800146973246, "optim/total_tokens": 6077022208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.991764545440674, "created_at": "2025-01-16T14:17:58.433582+00:00"} {"global_step": 11592, "acc_step": 0, "speed/wps": 12910.2644950145, "speed/FLOPS": 202773444456696.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04425089433789253, "optim/lr": 0.0028949485171238143, "optim/total_tokens": 6077546496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.829547166824341, "created_at": "2025-01-16T14:18:08.590905+00:00"} {"global_step": 11593, "acc_step": 0, "speed/wps": 12907.161043498481, "speed/FLOPS": 202724700486048.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04256346449255943, "optim/lr": 0.002894917014999039, "optim/total_tokens": 6078070784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.966150999069214, "created_at": "2025-01-16T14:18:18.752705+00:00"} {"global_step": 11594, "acc_step": 0, "speed/wps": 12899.284400364735, "speed/FLOPS": 202600986981991.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.03972181677818298, "optim/lr": 0.002894885508323101, "optim/total_tokens": 6078595072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9176368713378906, "created_at": "2025-01-16T14:18:28.919378+00:00"} {"global_step": 11595, "acc_step": 0, "speed/wps": 12903.116527871594, "speed/FLOPS": 202661175810381.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.042840395122766495, "optim/lr": 0.002894853997096105, "optim/total_tokens": 6079119360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416139, "loss/out": 2.8853774070739746, "created_at": "2025-01-16T14:18:39.084073+00:00"} {"global_step": 11596, "acc_step": 0, "speed/wps": 12904.32906122645, "speed/FLOPS": 202680220312914.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04866327717900276, "optim/lr": 0.002894822481318153, "optim/total_tokens": 6079643648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 2.736438751220703, "created_at": "2025-01-16T14:18:49.247053+00:00"} {"global_step": 11597, "acc_step": 0, "speed/wps": 12897.275035694625, "speed/FLOPS": 202569427148690.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046559784561395645, "optim/lr": 0.002894790960989347, "optim/total_tokens": 6080167936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 2.920449733734131, "created_at": "2025-01-16T14:18:59.415543+00:00"} {"global_step": 11598, "acc_step": 0, "speed/wps": 12907.38842731258, "speed/FLOPS": 202728271861307.12, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04815421625971794, "optim/lr": 0.002894759436109791, "optim/total_tokens": 6080692224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8481149673461914, "created_at": "2025-01-16T14:19:09.574031+00:00"} {"global_step": 11599, "acc_step": 0, "speed/wps": 12899.856558489524, "speed/FLOPS": 202609973511568.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05367497354745865, "optim/lr": 0.002894727906679587, "optim/total_tokens": 6081216512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.809908151626587, "created_at": "2025-01-16T14:19:19.739575+00:00"} {"global_step": 11600, "acc_step": 0, "speed/wps": 12906.210384007663, "speed/FLOPS": 202709769072402.28, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05475124716758728, "optim/lr": 0.002894696372698839, "optim/total_tokens": 6081740800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 2.9802489280700684, "created_at": "2025-01-16T14:19:29.899043+00:00"} {"global_step": 11601, "acc_step": 0, "speed/wps": 12904.519764028853, "speed/FLOPS": 202683215562479.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04981425032019615, "optim/lr": 0.002894664834167648, "optim/total_tokens": 6082265088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.9582228660583496, "created_at": "2025-01-16T14:19:40.060657+00:00"} {"global_step": 11602, "acc_step": 0, "speed/wps": 12903.46473723729, "speed/FLOPS": 202666644916959.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04657936468720436, "optim/lr": 0.0028946332910861194, "optim/total_tokens": 6082789376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.800419569015503, "created_at": "2025-01-16T14:19:50.224900+00:00"} {"global_step": 11603, "acc_step": 0, "speed/wps": 12905.667631827439, "speed/FLOPS": 202701244403592.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05776551365852356, "optim/lr": 0.0028946017434543538, "optim/total_tokens": 6083313664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283675, "loss/out": 2.889479160308838, "created_at": "2025-01-16T14:20:00.388142+00:00"} {"global_step": 11604, "acc_step": 0, "speed/wps": 12894.85783033206, "speed/FLOPS": 202531461616883.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0533747561275959, "optim/lr": 0.0028945701912724553, "optim/total_tokens": 6083837952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8033175468444824, "created_at": "2025-01-16T14:20:10.559681+00:00"} {"global_step": 11605, "acc_step": 0, "speed/wps": 12892.52335070736, "speed/FLOPS": 202494795406466.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05975694954395294, "optim/lr": 0.0028945386345405263, "optim/total_tokens": 6084362240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9070582389831543, "created_at": "2025-01-16T14:20:20.730064+00:00"} {"global_step": 11606, "acc_step": 0, "speed/wps": 12894.047753175471, "speed/FLOPS": 202518738241976.53, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04378385841846466, "optim/lr": 0.0028945070732586706, "optim/total_tokens": 6084886528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282898, "loss/out": 2.8747119903564453, "created_at": "2025-01-16T14:20:30.902169+00:00"} {"global_step": 11607, "acc_step": 0, "speed/wps": 12896.863209493426, "speed/FLOPS": 202562958852292.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0530424565076828, "optim/lr": 0.0028944755074269903, "optim/total_tokens": 6085410816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8452651500701904, "created_at": "2025-01-16T14:20:41.069918+00:00"} {"global_step": 11608, "acc_step": 0, "speed/wps": 12900.966995500086, "speed/FLOPS": 202627414450719.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05489271506667137, "optim/lr": 0.0028944439370455887, "optim/total_tokens": 6085935104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.7976884841918945, "created_at": "2025-01-16T14:20:51.233910+00:00"} {"global_step": 11609, "acc_step": 0, "speed/wps": 12902.428460904854, "speed/FLOPS": 202650368773165.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043017853051424026, "optim/lr": 0.0028944123621145695, "optim/total_tokens": 6086459392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341330, "loss/out": 2.8572933673858643, "created_at": "2025-01-16T14:21:01.397265+00:00"} {"global_step": 11610, "acc_step": 0, "speed/wps": 12902.147657379011, "speed/FLOPS": 202645958367927.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04421503469347954, "optim/lr": 0.0028943807826340346, "optim/total_tokens": 6086983680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.9510133266448975, "created_at": "2025-01-16T14:21:11.561547+00:00"} {"global_step": 11611, "acc_step": 0, "speed/wps": 12900.285696795183, "speed/FLOPS": 202616713718356.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04865537956357002, "optim/lr": 0.0028943491986040872, "optim/total_tokens": 6087507968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 2.948413848876953, "created_at": "2025-01-16T14:21:21.728314+00:00"} {"global_step": 11612, "acc_step": 0, "speed/wps": 12894.719980051686, "speed/FLOPS": 202529296488804.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06109481677412987, "optim/lr": 0.0028943176100248314, "optim/total_tokens": 6088032256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8345158100128174, "created_at": "2025-01-16T14:21:31.899331+00:00"} {"global_step": 11613, "acc_step": 0, "speed/wps": 12892.28786311355, "speed/FLOPS": 202491096750211.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04473115876317024, "optim/lr": 0.0028942860168963687, "optim/total_tokens": 6088556544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.817598342895508, "created_at": "2025-01-16T14:21:42.075628+00:00"} {"global_step": 11614, "acc_step": 0, "speed/wps": 12897.513288092925, "speed/FLOPS": 202573169229999.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05366602540016174, "optim/lr": 0.0028942544192188035, "optim/total_tokens": 6089080832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.90116024017334, "created_at": "2025-01-16T14:21:52.243207+00:00"} {"global_step": 11615, "acc_step": 0, "speed/wps": 12895.908792517906, "speed/FLOPS": 202547968422185.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07429896295070648, "optim/lr": 0.0028942228169922387, "optim/total_tokens": 6089605120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 2.788827419281006, "created_at": "2025-01-16T14:22:02.410501+00:00"} {"global_step": 11616, "acc_step": 0, "speed/wps": 12907.344480038046, "speed/FLOPS": 202727581608970.7, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05296165868639946, "optim/lr": 0.0028941912102167767, "optim/total_tokens": 6090129408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9231197834014893, "created_at": "2025-01-16T14:22:12.572617+00:00"} {"global_step": 11617, "acc_step": 0, "speed/wps": 12902.020251300519, "speed/FLOPS": 202643957280390.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.062196534126996994, "optim/lr": 0.002894159598892521, "optim/total_tokens": 6090653696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339277, "loss/out": 2.806459903717041, "created_at": "2025-01-16T14:22:22.739100+00:00"} {"global_step": 11618, "acc_step": 0, "speed/wps": 12900.66231392361, "speed/FLOPS": 202622629007885.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05487242341041565, "optim/lr": 0.0028941279830195744, "optim/total_tokens": 6091177984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9151968955993652, "created_at": "2025-01-16T14:22:32.910001+00:00"} {"global_step": 11619, "acc_step": 0, "speed/wps": 12900.677374714405, "speed/FLOPS": 202622865558300.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04946429282426834, "optim/lr": 0.0028940963625980407, "optim/total_tokens": 6091702272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.925224781036377, "created_at": "2025-01-16T14:22:43.074590+00:00"} {"global_step": 11620, "acc_step": 0, "speed/wps": 12899.854382896208, "speed/FLOPS": 202609939340886.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04893190786242485, "optim/lr": 0.0028940647376280225, "optim/total_tokens": 6092226560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 2.798055648803711, "created_at": "2025-01-16T14:22:53.239710+00:00"} {"global_step": 11621, "acc_step": 0, "speed/wps": 12904.082246412192, "speed/FLOPS": 202676343747097.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051195546984672546, "optim/lr": 0.0028940331081096235, "optim/total_tokens": 6092750848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8678691387176514, "created_at": "2025-01-16T14:23:03.401432+00:00"} {"global_step": 11622, "acc_step": 0, "speed/wps": 12897.173713309045, "speed/FLOPS": 202567835741393.66, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05419277399778366, "optim/lr": 0.0028940014740429466, "optim/total_tokens": 6093275136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.825697183609009, "created_at": "2025-01-16T14:23:13.569514+00:00"} {"global_step": 11623, "acc_step": 0, "speed/wps": 12904.605123734482, "speed/FLOPS": 202684556253953.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06515313684940338, "optim/lr": 0.0028939698354280953, "optim/total_tokens": 6093799424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 2.7067251205444336, "created_at": "2025-01-16T14:23:23.735248+00:00"} {"global_step": 11624, "acc_step": 0, "speed/wps": 12892.221392007239, "speed/FLOPS": 202490052730146.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04578026011586189, "optim/lr": 0.002893938192265173, "optim/total_tokens": 6094323712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.9191982746124268, "created_at": "2025-01-16T14:23:33.909483+00:00"} {"global_step": 11625, "acc_step": 0, "speed/wps": 12893.989500026435, "speed/FLOPS": 202517823296222.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04957886412739754, "optim/lr": 0.002893906544554282, "optim/total_tokens": 6094848000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8072028160095215, "created_at": "2025-01-16T14:23:44.081042+00:00"} {"global_step": 11626, "acc_step": 0, "speed/wps": 12870.079424439513, "speed/FLOPS": 202142282703243.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05124356597661972, "optim/lr": 0.0028938748922955264, "optim/total_tokens": 6095372288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.955045223236084, "created_at": "2025-01-16T14:23:54.274861+00:00"} {"global_step": 11627, "acc_step": 0, "speed/wps": 12885.476765162139, "speed/FLOPS": 202384119097455.38, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0025, "optim/grad_norm": 0.04474283754825592, "optim/lr": 0.002893843235489008, "optim/total_tokens": 6095896576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369048, "loss/out": 2.8518009185791016, "created_at": "2025-01-16T14:24:04.451929+00:00"} {"global_step": 11628, "acc_step": 0, "speed/wps": 12896.074914204699, "speed/FLOPS": 202550577591551.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.044334691017866135, "optim/lr": 0.002893811574134832, "optim/total_tokens": 6096420864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.7653794288635254, "created_at": "2025-01-16T14:24:14.619726+00:00"} {"global_step": 11629, "acc_step": 0, "speed/wps": 12897.09989486442, "speed/FLOPS": 202566676321283.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05374422296881676, "optim/lr": 0.0028937799082331015, "optim/total_tokens": 6096945152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.877943277359009, "created_at": "2025-01-16T14:24:24.787470+00:00"} {"global_step": 11630, "acc_step": 0, "speed/wps": 12888.782891374463, "speed/FLOPS": 202436046352713.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05098247900605202, "optim/lr": 0.002893748237783919, "optim/total_tokens": 6097469440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 2.8216140270233154, "created_at": "2025-01-16T14:24:34.960847+00:00"} {"global_step": 11631, "acc_step": 0, "speed/wps": 12893.201635125737, "speed/FLOPS": 202505448795317.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0538165457546711, "optim/lr": 0.0028937165627873876, "optim/total_tokens": 6097993728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8272433280944824, "created_at": "2025-01-16T14:24:45.130993+00:00"} {"global_step": 11632, "acc_step": 0, "speed/wps": 12883.3248971615, "speed/FLOPS": 202350321053528.84, "speed/curr_iter_time": 1.2891, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0450577549636364, "optim/lr": 0.0028936848832436116, "optim/total_tokens": 6098518016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9358863830566406, "created_at": "2025-01-16T14:24:55.319617+00:00"} {"global_step": 11633, "acc_step": 0, "speed/wps": 12890.208766650927, "speed/FLOPS": 202458441683288.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.0621185265481472, "optim/lr": 0.0028936531991526934, "optim/total_tokens": 6099042304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.7237162590026855, "created_at": "2025-01-16T14:25:05.494897+00:00"} {"global_step": 11634, "acc_step": 0, "speed/wps": 12891.411328411546, "speed/FLOPS": 202477329568231.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04290815442800522, "optim/lr": 0.002893621510514737, "optim/total_tokens": 6099566592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 2.814194679260254, "created_at": "2025-01-16T14:25:15.667435+00:00"} {"global_step": 11635, "acc_step": 0, "speed/wps": 12897.148793077751, "speed/FLOPS": 202567444334917.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05985689535737038, "optim/lr": 0.0028935898173298464, "optim/total_tokens": 6100090880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 2.916581630706787, "created_at": "2025-01-16T14:25:25.834393+00:00"} {"global_step": 11636, "acc_step": 0, "speed/wps": 12901.480991505441, "speed/FLOPS": 202635487464288.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04984600841999054, "optim/lr": 0.0028935581195981236, "optim/total_tokens": 6100615168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.893540143966675, "created_at": "2025-01-16T14:25:35.997567+00:00"} {"global_step": 11637, "acc_step": 0, "speed/wps": 12896.864760066643, "speed/FLOPS": 202562983206175.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049354538321495056, "optim/lr": 0.0028935264173196727, "optim/total_tokens": 6101139456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 2.8877248764038086, "created_at": "2025-01-16T14:25:46.167204+00:00"} {"global_step": 11638, "acc_step": 0, "speed/wps": 12894.360943752603, "speed/FLOPS": 202523657330359.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04839133098721504, "optim/lr": 0.0028934947104945975, "optim/total_tokens": 6101663744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9003446102142334, "created_at": "2025-01-16T14:25:56.335888+00:00"} {"global_step": 11639, "acc_step": 0, "speed/wps": 12906.411556818142, "speed/FLOPS": 202712928767831.25, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06634820997714996, "optim/lr": 0.0028934629991230003, "optim/total_tokens": 6102188032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8263347148895264, "created_at": "2025-01-16T14:26:06.503016+00:00"} {"global_step": 11640, "acc_step": 0, "speed/wps": 12898.644392585507, "speed/FLOPS": 202590934780355.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06198984384536743, "optim/lr": 0.002893431283204986, "optim/total_tokens": 6102712320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.7649898529052734, "created_at": "2025-01-16T14:26:16.671692+00:00"} {"global_step": 11641, "acc_step": 0, "speed/wps": 12898.293392324453, "speed/FLOPS": 202585421838930.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054048094898462296, "optim/lr": 0.002893399562740657, "optim/total_tokens": 6103236608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.716965913772583, "created_at": "2025-01-16T14:26:26.840971+00:00"} {"global_step": 11642, "acc_step": 0, "speed/wps": 12901.835564502651, "speed/FLOPS": 202641056520444.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055834099650382996, "optim/lr": 0.002893367837730118, "optim/total_tokens": 6103760896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.9823899269104004, "created_at": "2025-01-16T14:26:37.004594+00:00"} {"global_step": 11643, "acc_step": 0, "speed/wps": 12899.100069711587, "speed/FLOPS": 202598091815786.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05403762683272362, "optim/lr": 0.0028933361081734713, "optim/total_tokens": 6104285184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.893134593963623, "created_at": "2025-01-16T14:26:47.170219+00:00"} {"global_step": 11644, "acc_step": 0, "speed/wps": 12902.12372212259, "speed/FLOPS": 202645582431835.28, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05261205509305, "optim/lr": 0.0028933043740708205, "optim/total_tokens": 6104809472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.774226427078247, "created_at": "2025-01-16T14:26:57.333575+00:00"} {"global_step": 11645, "acc_step": 0, "speed/wps": 12903.273238688678, "speed/FLOPS": 202663637169107.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04462762176990509, "optim/lr": 0.0028932726354222696, "optim/total_tokens": 6105333760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 2.9014015197753906, "created_at": "2025-01-16T14:27:07.496313+00:00"} {"global_step": 11646, "acc_step": 0, "speed/wps": 12891.759186165853, "speed/FLOPS": 202482793152269.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05321861803531647, "optim/lr": 0.002893240892227922, "optim/total_tokens": 6105858048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.821949005126953, "created_at": "2025-01-16T14:27:17.671215+00:00"} {"global_step": 11647, "acc_step": 0, "speed/wps": 12888.26812443809, "speed/FLOPS": 202427961230613.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058118339627981186, "optim/lr": 0.0028932091444878816, "optim/total_tokens": 6106382336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.91253399848938, "created_at": "2025-01-16T14:27:27.847518+00:00"} {"global_step": 11648, "acc_step": 0, "speed/wps": 12894.239967965877, "speed/FLOPS": 202521757239390.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.058371711522340775, "optim/lr": 0.002893177392202252, "optim/total_tokens": 6106906624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 2.8836727142333984, "created_at": "2025-01-16T14:27:38.017064+00:00"} {"global_step": 11649, "acc_step": 0, "speed/wps": 12900.760914494302, "speed/FLOPS": 202624177665339.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051038216799497604, "optim/lr": 0.0028931456353711357, "optim/total_tokens": 6107430912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.793224334716797, "created_at": "2025-01-16T14:27:48.180539+00:00"} {"global_step": 11650, "acc_step": 0, "speed/wps": 12901.691583954658, "speed/FLOPS": 202638795108087.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04425691440701485, "optim/lr": 0.002893113873994637, "optim/total_tokens": 6107955200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8509626388549805, "created_at": "2025-01-16T14:27:58.345007+00:00"} {"global_step": 11651, "acc_step": 0, "speed/wps": 12904.45084641566, "speed/FLOPS": 202682133116661.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06053905189037323, "optim/lr": 0.0028930821080728605, "optim/total_tokens": 6108479488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 2.865525960922241, "created_at": "2025-01-16T14:28:08.506955+00:00"} {"global_step": 11652, "acc_step": 0, "speed/wps": 12903.770593518107, "speed/FLOPS": 202671448810140.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04929255321621895, "optim/lr": 0.002893050337605908, "optim/total_tokens": 6109003776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.765476942062378, "created_at": "2025-01-16T14:28:18.670302+00:00"} {"global_step": 11653, "acc_step": 0, "speed/wps": 12898.934329961756, "speed/FLOPS": 202595488645265.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05890177562832832, "optim/lr": 0.0028930185625938845, "optim/total_tokens": 6109528064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.871502637863159, "created_at": "2025-01-16T14:28:28.837201+00:00"} {"global_step": 11654, "acc_step": 0, "speed/wps": 12898.252877110668, "speed/FLOPS": 202584785491824.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.059872809797525406, "optim/lr": 0.0028929867830368937, "optim/total_tokens": 6110052352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402512, "loss/out": 2.842205047607422, "created_at": "2025-01-16T14:28:39.005334+00:00"} {"global_step": 11655, "acc_step": 0, "speed/wps": 12901.4597079291, "speed/FLOPS": 202635153176475.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05482763797044754, "optim/lr": 0.002892954998935038, "optim/total_tokens": 6110576640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.775418996810913, "created_at": "2025-01-16T14:28:49.173245+00:00"} {"global_step": 11656, "acc_step": 0, "speed/wps": 12894.899682921108, "speed/FLOPS": 202532118969305.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05894118547439575, "optim/lr": 0.002892923210288423, "optim/total_tokens": 6111100928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.896601915359497, "created_at": "2025-01-16T14:28:59.344496+00:00"} {"global_step": 11657, "acc_step": 0, "speed/wps": 12895.614667624111, "speed/FLOPS": 202543348786558.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05365152657032013, "optim/lr": 0.002892891417097151, "optim/total_tokens": 6111625216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.869699001312256, "created_at": "2025-01-16T14:29:09.519905+00:00"} {"global_step": 11658, "acc_step": 0, "speed/wps": 12902.133304021016, "speed/FLOPS": 202645732928716.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06906388700008392, "optim/lr": 0.0028928596193613255, "optim/total_tokens": 6112149504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.876673698425293, "created_at": "2025-01-16T14:29:19.685328+00:00"} {"global_step": 11659, "acc_step": 0, "speed/wps": 12902.226878790692, "speed/FLOPS": 202647202649059.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043687816709280014, "optim/lr": 0.0028928278170810514, "optim/total_tokens": 6112673792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9133660793304443, "created_at": "2025-01-16T14:29:29.849903+00:00"} {"global_step": 11660, "acc_step": 0, "speed/wps": 12900.352643668297, "speed/FLOPS": 202617765210993.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07594650238752365, "optim/lr": 0.0028927960102564324, "optim/total_tokens": 6113198080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 2.8982131481170654, "created_at": "2025-01-16T14:29:40.017234+00:00"} {"global_step": 11661, "acc_step": 0, "speed/wps": 12906.998416805869, "speed/FLOPS": 202722146210368.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06301688402891159, "optim/lr": 0.0028927641988875713, "optim/total_tokens": 6113722368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.95888614654541, "created_at": "2025-01-16T14:29:50.178191+00:00"} {"global_step": 11662, "acc_step": 0, "speed/wps": 12907.118457865872, "speed/FLOPS": 202724031620169.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059370990842580795, "optim/lr": 0.002892732382974572, "optim/total_tokens": 6114246656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420398, "loss/out": 2.7807071208953857, "created_at": "2025-01-16T14:30:00.337336+00:00"} {"global_step": 11663, "acc_step": 0, "speed/wps": 12907.270198282786, "speed/FLOPS": 202726414911930.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08860377967357635, "optim/lr": 0.0028927005625175393, "optim/total_tokens": 6114770944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8981645107269287, "created_at": "2025-01-16T14:30:10.505246+00:00"} {"global_step": 11664, "acc_step": 0, "speed/wps": 12903.956570917133, "speed/FLOPS": 202674369840751.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05047331377863884, "optim/lr": 0.002892668737516576, "optim/total_tokens": 6115295232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 2.8936400413513184, "created_at": "2025-01-16T14:30:20.667198+00:00"} {"global_step": 11665, "acc_step": 0, "speed/wps": 12904.912448141591, "speed/FLOPS": 202689383206074.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04830499738454819, "optim/lr": 0.002892636907971787, "optim/total_tokens": 6115819520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.847151756286621, "created_at": "2025-01-16T14:30:30.828979+00:00"} {"global_step": 11666, "acc_step": 0, "speed/wps": 12901.173193074506, "speed/FLOPS": 202630653066971.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05244796350598335, "optim/lr": 0.002892605073883275, "optim/total_tokens": 6116343808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.7541041374206543, "created_at": "2025-01-16T14:30:40.992127+00:00"} {"global_step": 11667, "acc_step": 0, "speed/wps": 12901.610907287282, "speed/FLOPS": 202637527970165.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04880828782916069, "optim/lr": 0.0028925732352511444, "optim/total_tokens": 6116868096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393414, "loss/out": 2.9064908027648926, "created_at": "2025-01-16T14:30:51.155200+00:00"} {"global_step": 11668, "acc_step": 0, "speed/wps": 12900.116210180293, "speed/FLOPS": 202614051698171.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05434674024581909, "optim/lr": 0.002892541392075499, "optim/total_tokens": 6117392384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.1874217987060547, "created_at": "2025-01-16T14:31:01.325446+00:00"} {"global_step": 11669, "acc_step": 0, "speed/wps": 12903.990833547681, "speed/FLOPS": 202674907982446.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04191594570875168, "optim/lr": 0.002892509544356443, "optim/total_tokens": 6117916672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.839918613433838, "created_at": "2025-01-16T14:31:11.487722+00:00"} {"global_step": 11670, "acc_step": 0, "speed/wps": 12899.812572628316, "speed/FLOPS": 202609282653175.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.045892827212810516, "optim/lr": 0.0028924776920940796, "optim/total_tokens": 6118440960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 2.7449707984924316, "created_at": "2025-01-16T14:31:21.652863+00:00"} {"global_step": 11671, "acc_step": 0, "speed/wps": 12903.617968799248, "speed/FLOPS": 202669051629202.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04781229421496391, "optim/lr": 0.0028924458352885137, "optim/total_tokens": 6118965248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.8640260696411133, "created_at": "2025-01-16T14:31:31.814235+00:00"} {"global_step": 11672, "acc_step": 0, "speed/wps": 12908.414939655533, "speed/FLOPS": 202744394648229.0, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042384885251522064, "optim/lr": 0.002892413973939849, "optim/total_tokens": 6119489536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 2.9401164054870605, "created_at": "2025-01-16T14:31:41.971893+00:00"} {"global_step": 11673, "acc_step": 0, "speed/wps": 12899.937365747594, "speed/FLOPS": 202611242700597.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045143984258174896, "optim/lr": 0.0028923821080481882, "optim/total_tokens": 6120013824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9458513259887695, "created_at": "2025-01-16T14:31:52.138965+00:00"} {"global_step": 11674, "acc_step": 0, "speed/wps": 12902.642048629878, "speed/FLOPS": 202653723461887.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04455455392599106, "optim/lr": 0.002892350237613637, "optim/total_tokens": 6120538112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 2.859206199645996, "created_at": "2025-01-16T14:32:02.301352+00:00"} {"global_step": 11675, "acc_step": 0, "speed/wps": 12905.703505949625, "speed/FLOPS": 202701807855977.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04755787178874016, "optim/lr": 0.0028923183626362987, "optim/total_tokens": 6121062400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375761, "loss/out": 2.768425941467285, "created_at": "2025-01-16T14:32:12.466516+00:00"} {"global_step": 11676, "acc_step": 0, "speed/wps": 12905.372273474713, "speed/FLOPS": 202696605394797.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046130090951919556, "optim/lr": 0.002892286483116276, "optim/total_tokens": 6121586688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.921215534210205, "created_at": "2025-01-16T14:32:22.629398+00:00"} {"global_step": 11677, "acc_step": 0, "speed/wps": 12899.561518080329, "speed/FLOPS": 202605339496514.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0465247668325901, "optim/lr": 0.002892254599053675, "optim/total_tokens": 6122110976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486049, "loss/out": 2.8943703174591064, "created_at": "2025-01-16T14:32:32.794875+00:00"} {"global_step": 11678, "acc_step": 0, "speed/wps": 12902.761528404972, "speed/FLOPS": 202655600055937.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051357824355363846, "optim/lr": 0.002892222710448599, "optim/total_tokens": 6122635264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8329198360443115, "created_at": "2025-01-16T14:32:42.960222+00:00"} {"global_step": 11679, "acc_step": 0, "speed/wps": 12899.08001406739, "speed/FLOPS": 202597776814335.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.08223208039999008, "optim/lr": 0.0028921908173011514, "optim/total_tokens": 6123159552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8558433055877686, "created_at": "2025-01-16T14:32:53.126977+00:00"} {"global_step": 11680, "acc_step": 0, "speed/wps": 12909.027802807665, "speed/FLOPS": 202754020506195.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07598140090703964, "optim/lr": 0.0028921589196114375, "optim/total_tokens": 6123683840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.8403525352478027, "created_at": "2025-01-16T14:33:03.285931+00:00"} {"global_step": 11681, "acc_step": 0, "speed/wps": 12901.951221998052, "speed/FLOPS": 202642873080339.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07844146341085434, "optim/lr": 0.00289212701737956, "optim/total_tokens": 6124208128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8831944465637207, "created_at": "2025-01-16T14:33:13.451684+00:00"} {"global_step": 11682, "acc_step": 0, "speed/wps": 12896.518458487122, "speed/FLOPS": 202557544064000.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07837814837694168, "optim/lr": 0.002892095110605624, "optim/total_tokens": 6124732416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8628134727478027, "created_at": "2025-01-16T14:33:23.626593+00:00"} {"global_step": 11683, "acc_step": 0, "speed/wps": 12900.912416770716, "speed/FLOPS": 202626557216776.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05934278294444084, "optim/lr": 0.002892063199289733, "optim/total_tokens": 6125256704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 2.938676118850708, "created_at": "2025-01-16T14:33:33.797981+00:00"} {"global_step": 11684, "acc_step": 0, "speed/wps": 12902.106724075606, "speed/FLOPS": 202645315454149.62, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053960252553224564, "optim/lr": 0.0028920312834319916, "optim/total_tokens": 6125780992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 2.896101474761963, "created_at": "2025-01-16T14:33:43.965938+00:00"} {"global_step": 11685, "acc_step": 0, "speed/wps": 12907.888151067638, "speed/FLOPS": 202736120709572.72, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06480852514505386, "optim/lr": 0.002891999363032503, "optim/total_tokens": 6126305280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 2.716156005859375, "created_at": "2025-01-16T14:33:54.124168+00:00"} {"global_step": 11686, "acc_step": 0, "speed/wps": 12909.211917618819, "speed/FLOPS": 202756912282304.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058701734989881516, "optim/lr": 0.0028919674380913727, "optim/total_tokens": 6126829568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 2.8536813259124756, "created_at": "2025-01-16T14:34:04.284643+00:00"} {"global_step": 11687, "acc_step": 0, "speed/wps": 12899.180894684387, "speed/FLOPS": 202599361283049.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048853930085897446, "optim/lr": 0.0028919355086087038, "optim/total_tokens": 6127353856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9401774406433105, "created_at": "2025-01-16T14:34:14.451915+00:00"} {"global_step": 11688, "acc_step": 0, "speed/wps": 12903.858554461785, "speed/FLOPS": 202672830357631.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06769351661205292, "optim/lr": 0.0028919035745846007, "optim/total_tokens": 6127878144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.8075413703918457, "created_at": "2025-01-16T14:34:24.613551+00:00"} {"global_step": 11689, "acc_step": 0, "speed/wps": 12908.175814310358, "speed/FLOPS": 202740638856090.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047947436571121216, "optim/lr": 0.002891871636019168, "optim/total_tokens": 6128402432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.941577434539795, "created_at": "2025-01-16T14:34:34.775144+00:00"} {"global_step": 11690, "acc_step": 0, "speed/wps": 12905.555023670042, "speed/FLOPS": 202699475737740.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05797603353857994, "optim/lr": 0.00289183969291251, "optim/total_tokens": 6128926720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.918285846710205, "created_at": "2025-01-16T14:34:44.937791+00:00"} {"global_step": 11691, "acc_step": 0, "speed/wps": 12899.141811519065, "speed/FLOPS": 202598747428232.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046897877007722855, "optim/lr": 0.0028918077452647303, "optim/total_tokens": 6129451008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 2.8252663612365723, "created_at": "2025-01-16T14:34:55.113060+00:00"} {"global_step": 11692, "acc_step": 0, "speed/wps": 12902.633180445495, "speed/FLOPS": 202653584174865.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.049244169145822525, "optim/lr": 0.002891775793075933, "optim/total_tokens": 6129975296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.9339239597320557, "created_at": "2025-01-16T14:35:05.275159+00:00"} {"global_step": 11693, "acc_step": 0, "speed/wps": 12903.24320320871, "speed/FLOPS": 202663165420621.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04827076196670532, "optim/lr": 0.0028917438363462227, "optim/total_tokens": 6130499584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.937089681625366, "created_at": "2025-01-16T14:35:15.438165+00:00"} {"global_step": 11694, "acc_step": 0, "speed/wps": 12892.067742494793, "speed/FLOPS": 202487639453416.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05013100430369377, "optim/lr": 0.0028917118750757045, "optim/total_tokens": 6131023872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379498, "loss/out": 2.8410286903381348, "created_at": "2025-01-16T14:35:25.610805+00:00"} {"global_step": 11695, "acc_step": 0, "speed/wps": 12897.425158334941, "speed/FLOPS": 202571785031047.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05500762164592743, "optim/lr": 0.0028916799092644813, "optim/total_tokens": 6131548160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.74198055267334, "created_at": "2025-01-16T14:35:35.779395+00:00"} {"global_step": 11696, "acc_step": 0, "speed/wps": 12905.525067506807, "speed/FLOPS": 202699005235032.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04360978677868843, "optim/lr": 0.0028916479389126584, "optim/total_tokens": 6132072448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9269402027130127, "created_at": "2025-01-16T14:35:45.943353+00:00"} {"global_step": 11697, "acc_step": 0, "speed/wps": 12900.574128117709, "speed/FLOPS": 202621243928623.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06303370743989944, "optim/lr": 0.002891615964020339, "optim/total_tokens": 6132596736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.8653125762939453, "created_at": "2025-01-16T14:35:56.113987+00:00"} {"global_step": 11698, "acc_step": 0, "speed/wps": 12904.387343944589, "speed/FLOPS": 202681135723092.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05609409138560295, "optim/lr": 0.0028915839845876285, "optim/total_tokens": 6133121024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9819560050964355, "created_at": "2025-01-16T14:36:06.275538+00:00"} {"global_step": 11699, "acc_step": 0, "speed/wps": 12904.891852578832, "speed/FLOPS": 202689059724460.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052031245082616806, "optim/lr": 0.002891552000614631, "optim/total_tokens": 6133645312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389901, "loss/out": 2.911412239074707, "created_at": "2025-01-16T14:36:16.437680+00:00"} {"global_step": 11700, "acc_step": 0, "speed/wps": 12899.677921074226, "speed/FLOPS": 202607167765487.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05743223801255226, "optim/lr": 0.0028915200121014508, "optim/total_tokens": 6134169600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9366631507873535, "created_at": "2025-01-16T14:36:26.602877+00:00"} {"global_step": 11701, "acc_step": 0, "speed/wps": 12898.3462691159, "speed/FLOPS": 202586252341600.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048865918070077896, "optim/lr": 0.0028914880190481916, "optim/total_tokens": 6134693888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9663469791412354, "created_at": "2025-01-16T14:36:36.775723+00:00"} {"global_step": 11702, "acc_step": 0, "speed/wps": 12893.427518606159, "speed/FLOPS": 202508996605773.84, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05296601727604866, "optim/lr": 0.002891456021454959, "optim/total_tokens": 6135218176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462167, "loss/out": 2.869093894958496, "created_at": "2025-01-16T14:36:46.945474+00:00"} {"global_step": 11703, "acc_step": 0, "speed/wps": 12889.145563261021, "speed/FLOPS": 202441742613055.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047739770263433456, "optim/lr": 0.0028914240193218567, "optim/total_tokens": 6135742464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 2.8545312881469727, "created_at": "2025-01-16T14:36:57.121663+00:00"} {"global_step": 11704, "acc_step": 0, "speed/wps": 12894.178065079073, "speed/FLOPS": 202520784969489.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056388452649116516, "optim/lr": 0.0028913920126489886, "optim/total_tokens": 6136266752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.847358226776123, "created_at": "2025-01-16T14:37:07.301777+00:00"} {"global_step": 11705, "acc_step": 0, "speed/wps": 12895.524240797133, "speed/FLOPS": 202541928508981.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06704584509134293, "optim/lr": 0.0028913600014364603, "optim/total_tokens": 6136791040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.904771089553833, "created_at": "2025-01-16T14:37:17.469473+00:00"} {"global_step": 11706, "acc_step": 0, "speed/wps": 12907.26104437129, "speed/FLOPS": 202726271137171.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062183037400245667, "optim/lr": 0.0028913279856843752, "optim/total_tokens": 6137315328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8273909091949463, "created_at": "2025-01-16T14:37:27.628948+00:00"} {"global_step": 11707, "acc_step": 0, "speed/wps": 12900.332838361046, "speed/FLOPS": 202617454141428.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054095059633255005, "optim/lr": 0.0028912959653928385, "optim/total_tokens": 6137839616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.038544178009033, "created_at": "2025-01-16T14:37:37.793710+00:00"} {"global_step": 11708, "acc_step": 0, "speed/wps": 12900.844063925828, "speed/FLOPS": 202625483641420.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06785521656274796, "optim/lr": 0.002891263940561954, "optim/total_tokens": 6138363904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 2.988901138305664, "created_at": "2025-01-16T14:37:47.958419+00:00"} {"global_step": 11709, "acc_step": 0, "speed/wps": 12901.203066358437, "speed/FLOPS": 202631122267945.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06355487555265427, "optim/lr": 0.002891231911191827, "optim/total_tokens": 6138888192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7338056564331055, "created_at": "2025-01-16T14:37:58.124885+00:00"} {"global_step": 11710, "acc_step": 0, "speed/wps": 12894.208769545448, "speed/FLOPS": 202521267225326.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05915346369147301, "optim/lr": 0.0028911998772825614, "optim/total_tokens": 6139412480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.881274461746216, "created_at": "2025-01-16T14:38:08.296734+00:00"} {"global_step": 11711, "acc_step": 0, "speed/wps": 12882.684747996966, "speed/FLOPS": 202340266631242.7, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08302013576030731, "optim/lr": 0.002891167838834262, "optim/total_tokens": 6139936768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 2.9181532859802246, "created_at": "2025-01-16T14:38:18.475932+00:00"} {"global_step": 11712, "acc_step": 0, "speed/wps": 12888.355390242103, "speed/FLOPS": 202429331859981.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0668388307094574, "optim/lr": 0.002891135795847033, "optim/total_tokens": 6140461056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.865828514099121, "created_at": "2025-01-16T14:38:28.650691+00:00"} {"global_step": 11713, "acc_step": 0, "speed/wps": 12893.225136872621, "speed/FLOPS": 202505817922547.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06317144632339478, "optim/lr": 0.0028911037483209793, "optim/total_tokens": 6140985344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8910458087921143, "created_at": "2025-01-16T14:38:38.821406+00:00"} {"global_step": 11714, "acc_step": 0, "speed/wps": 12899.217841722882, "speed/FLOPS": 202599941587059.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06968306005001068, "optim/lr": 0.002891071696256205, "optim/total_tokens": 6141509632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 2.68557071685791, "created_at": "2025-01-16T14:38:48.986145+00:00"} {"global_step": 11715, "acc_step": 0, "speed/wps": 12909.645335364346, "speed/FLOPS": 202763719703576.78, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055442918092012405, "optim/lr": 0.0028910396396528157, "optim/total_tokens": 6142033920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.906588315963745, "created_at": "2025-01-16T14:38:59.145254+00:00"} {"global_step": 11716, "acc_step": 0, "speed/wps": 12898.143833416032, "speed/FLOPS": 202583072810758.6, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06203613430261612, "optim/lr": 0.002891007578510914, "optim/total_tokens": 6142558208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431014, "loss/out": 2.9004039764404297, "created_at": "2025-01-16T14:39:09.310902+00:00"} {"global_step": 11717, "acc_step": 0, "speed/wps": 12900.975605954129, "speed/FLOPS": 202627549689732.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054104048758745193, "optim/lr": 0.0028909755128306075, "optim/total_tokens": 6143082496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 2.8232951164245605, "created_at": "2025-01-16T14:39:19.474662+00:00"} {"global_step": 11718, "acc_step": 0, "speed/wps": 12900.673798719437, "speed/FLOPS": 202622809392385.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04657286778092384, "optim/lr": 0.002890943442611998, "optim/total_tokens": 6143606784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.911625862121582, "created_at": "2025-01-16T14:39:29.639624+00:00"} {"global_step": 11719, "acc_step": 0, "speed/wps": 12908.32760563642, "speed/FLOPS": 202743022947449.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04829473793506622, "optim/lr": 0.002890911367855191, "optim/total_tokens": 6144131072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.8237802982330322, "created_at": "2025-01-16T14:39:39.797546+00:00"} {"global_step": 11720, "acc_step": 0, "speed/wps": 12903.162431423489, "speed/FLOPS": 202661896788742.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050965551286935806, "optim/lr": 0.0028908792885602923, "optim/total_tokens": 6144655360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505844, "loss/out": 2.8834986686706543, "created_at": "2025-01-16T14:39:49.961521+00:00"} {"global_step": 11721, "acc_step": 0, "speed/wps": 12905.908799281424, "speed/FLOPS": 202705032269855.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05175555497407913, "optim/lr": 0.002890847204727405, "optim/total_tokens": 6145179648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8791749477386475, "created_at": "2025-01-16T14:40:00.121045+00:00"} {"global_step": 11722, "acc_step": 0, "speed/wps": 12906.231874997524, "speed/FLOPS": 202710106617929.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04524875059723854, "optim/lr": 0.0028908151163566345, "optim/total_tokens": 6145703936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.937673330307007, "created_at": "2025-01-16T14:40:10.280405+00:00"} {"global_step": 11723, "acc_step": 0, "speed/wps": 12897.95901644299, "speed/FLOPS": 202580170006231.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.049596503376960754, "optim/lr": 0.0028907830234480856, "optim/total_tokens": 6146228224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.849198818206787, "created_at": "2025-01-16T14:40:20.446413+00:00"} {"global_step": 11724, "acc_step": 0, "speed/wps": 12883.397759568143, "speed/FLOPS": 202351465457748.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.003, "optim/grad_norm": 0.0452812984585762, "optim/lr": 0.002890750926001863, "optim/total_tokens": 6146752512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.786815643310547, "created_at": "2025-01-16T14:40:30.623742+00:00"} {"global_step": 11725, "acc_step": 0, "speed/wps": 12896.204018009712, "speed/FLOPS": 202552605344216.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04922812059521675, "optim/lr": 0.0028907188240180712, "optim/total_tokens": 6147276800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.865656614303589, "created_at": "2025-01-16T14:40:40.791870+00:00"} {"global_step": 11726, "acc_step": 0, "speed/wps": 12896.762898126035, "speed/FLOPS": 202561383324425.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05355725809931755, "optim/lr": 0.002890686717496814, "optim/total_tokens": 6147801088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.864539623260498, "created_at": "2025-01-16T14:40:50.962586+00:00"} {"global_step": 11727, "acc_step": 0, "speed/wps": 12889.472345103808, "speed/FLOPS": 202446875170941.5, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05090167373418808, "optim/lr": 0.002890654606438198, "optim/total_tokens": 6148325376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.786433219909668, "created_at": "2025-01-16T14:41:01.134994+00:00"} {"global_step": 11728, "acc_step": 0, "speed/wps": 12891.416346984415, "speed/FLOPS": 202477408391814.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05013120546936989, "optim/lr": 0.0028906224908423275, "optim/total_tokens": 6148849664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.754516124725342, "created_at": "2025-01-16T14:41:11.305888+00:00"} {"global_step": 11729, "acc_step": 0, "speed/wps": 12890.66629908237, "speed/FLOPS": 202465627858839.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05591215938329697, "optim/lr": 0.0028905903707093067, "optim/total_tokens": 6149373952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.969601631164551, "created_at": "2025-01-16T14:41:21.477360+00:00"} {"global_step": 11730, "acc_step": 0, "speed/wps": 12879.125531417209, "speed/FLOPS": 202284364243981.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06640470027923584, "optim/lr": 0.0028905582460392407, "optim/total_tokens": 6149898240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.8508901596069336, "created_at": "2025-01-16T14:41:31.669674+00:00"} {"global_step": 11731, "acc_step": 0, "speed/wps": 12895.950460498565, "speed/FLOPS": 202548622875079.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06264746189117432, "optim/lr": 0.002890526116832233, "optim/total_tokens": 6150422528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284736, "loss/out": 2.841386556625366, "created_at": "2025-01-16T14:41:41.837657+00:00"} {"global_step": 11732, "acc_step": 0, "speed/wps": 12893.600198407483, "speed/FLOPS": 202511708779340.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05393407493829727, "optim/lr": 0.0028904939830883907, "optim/total_tokens": 6150946816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.9011807441711426, "created_at": "2025-01-16T14:41:52.008783+00:00"} {"global_step": 11733, "acc_step": 0, "speed/wps": 12907.899203600267, "speed/FLOPS": 202736294304785.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0521208681166172, "optim/lr": 0.002890461844807818, "optim/total_tokens": 6151471104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.790537118911743, "created_at": "2025-01-16T14:42:02.166775+00:00"} {"global_step": 11734, "acc_step": 0, "speed/wps": 12902.076317897308, "speed/FLOPS": 202644837883337.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05100857838988304, "optim/lr": 0.0028904297019906188, "optim/total_tokens": 6151995392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.953313112258911, "created_at": "2025-01-16T14:42:12.329185+00:00"} {"global_step": 11735, "acc_step": 0, "speed/wps": 12900.156693350535, "speed/FLOPS": 202614687541989.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04790567234158516, "optim/lr": 0.0028903975546368983, "optim/total_tokens": 6152519680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.736846685409546, "created_at": "2025-01-16T14:42:22.493350+00:00"} {"global_step": 11736, "acc_step": 0, "speed/wps": 12883.946970804867, "speed/FLOPS": 202360091574916.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0027, "optim/grad_norm": 0.05153755471110344, "optim/lr": 0.0028903654027467615, "optim/total_tokens": 6153043968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419583, "loss/out": 2.9853243827819824, "created_at": "2025-01-16T14:42:32.670498+00:00"} {"global_step": 11737, "acc_step": 0, "speed/wps": 12882.012172919862, "speed/FLOPS": 202329702915441.78, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.05914513021707535, "optim/lr": 0.002890333246320314, "optim/total_tokens": 6153568256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.8172719478607178, "created_at": "2025-01-16T14:42:42.853271+00:00"} {"global_step": 11738, "acc_step": 0, "speed/wps": 12891.156097214014, "speed/FLOPS": 202473320811550.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051549673080444336, "optim/lr": 0.0028903010853576594, "optim/total_tokens": 6154092544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.7112655639648438, "created_at": "2025-01-16T14:42:53.026762+00:00"} {"global_step": 11739, "acc_step": 0, "speed/wps": 12901.48901366738, "speed/FLOPS": 202635613463365.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04242962598800659, "optim/lr": 0.0028902689198589035, "optim/total_tokens": 6154616832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.701570510864258, "created_at": "2025-01-16T14:43:03.191356+00:00"} {"global_step": 11740, "acc_step": 0, "speed/wps": 12899.513280224604, "speed/FLOPS": 202604581854704.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04554238170385361, "optim/lr": 0.0028902367498241516, "optim/total_tokens": 6155141120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 2.910000801086426, "created_at": "2025-01-16T14:43:13.356450+00:00"} {"global_step": 11741, "acc_step": 0, "speed/wps": 12901.344686055827, "speed/FLOPS": 202633346599899.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04770283401012421, "optim/lr": 0.0028902045752535084, "optim/total_tokens": 6155665408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.954939842224121, "created_at": "2025-01-16T14:43:23.519755+00:00"} {"global_step": 11742, "acc_step": 0, "speed/wps": 12903.141717232245, "speed/FLOPS": 202661571443904.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.042845144867897034, "optim/lr": 0.0028901723961470773, "optim/total_tokens": 6156189696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8703036308288574, "created_at": "2025-01-16T14:43:33.682036+00:00"} {"global_step": 11743, "acc_step": 0, "speed/wps": 12900.681502957093, "speed/FLOPS": 202622930398024.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048182640224695206, "optim/lr": 0.0028901402125049657, "optim/total_tokens": 6156713984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8511855602264404, "created_at": "2025-01-16T14:43:43.848969+00:00"} {"global_step": 11744, "acc_step": 0, "speed/wps": 12896.974602345656, "speed/FLOPS": 202564708430106.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047305818647146225, "optim/lr": 0.0028901080243272773, "optim/total_tokens": 6157238272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.861802816390991, "created_at": "2025-01-16T14:43:54.017039+00:00"} {"global_step": 11745, "acc_step": 0, "speed/wps": 12902.006399725547, "speed/FLOPS": 202643739722371.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045339733362197876, "optim/lr": 0.002890075831614117, "optim/total_tokens": 6157762560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412610, "loss/out": 2.8326549530029297, "created_at": "2025-01-16T14:44:04.179704+00:00"} {"global_step": 11746, "acc_step": 0, "speed/wps": 12884.233797933812, "speed/FLOPS": 202364596589118.53, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06204916536808014, "optim/lr": 0.0028900436343655906, "optim/total_tokens": 6158286848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.0243446826934814, "created_at": "2025-01-16T14:44:14.358696+00:00"} {"global_step": 11747, "acc_step": 0, "speed/wps": 12895.053519276576, "speed/FLOPS": 202534535180661.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041796449571847916, "optim/lr": 0.002890011432581803, "optim/total_tokens": 6158811136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.9776909351348877, "created_at": "2025-01-16T14:44:24.528289+00:00"} {"global_step": 11748, "acc_step": 0, "speed/wps": 12901.09031613197, "speed/FLOPS": 202629351370703.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06559459865093231, "optim/lr": 0.0028899792262628583, "optim/total_tokens": 6159335424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.860888957977295, "created_at": "2025-01-16T14:44:34.691808+00:00"} {"global_step": 11749, "acc_step": 0, "speed/wps": 12901.50929323579, "speed/FLOPS": 202635931981854.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052662864327430725, "optim/lr": 0.0028899470154088625, "optim/total_tokens": 6159859712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8249826431274414, "created_at": "2025-01-16T14:44:44.857835+00:00"} {"global_step": 11750, "acc_step": 0, "speed/wps": 12890.83055359417, "speed/FLOPS": 202468207701656.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054358575493097305, "optim/lr": 0.0028899148000199205, "optim/total_tokens": 6160384000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9478366374969482, "created_at": "2025-01-16T14:44:55.032059+00:00"} {"global_step": 11751, "acc_step": 0, "speed/wps": 12902.853413589293, "speed/FLOPS": 202657043239021.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06005176529288292, "optim/lr": 0.0028898825800961372, "optim/total_tokens": 6160908288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 2.9795613288879395, "created_at": "2025-01-16T14:45:05.194084+00:00"} {"global_step": 11752, "acc_step": 0, "speed/wps": 12907.14933289499, "speed/FLOPS": 202724516554928.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04900802671909332, "optim/lr": 0.002889850355637618, "optim/total_tokens": 6161432576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.966888427734375, "created_at": "2025-01-16T14:45:15.355547+00:00"} {"global_step": 11753, "acc_step": 0, "speed/wps": 12902.363686666768, "speed/FLOPS": 202649351404747.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059428151696920395, "optim/lr": 0.0028898181266444685, "optim/total_tokens": 6161956864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.921602249145508, "created_at": "2025-01-16T14:45:25.519055+00:00"} {"global_step": 11754, "acc_step": 0, "speed/wps": 12901.718045099427, "speed/FLOPS": 202639210716727.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05241404101252556, "optim/lr": 0.0028897858931167925, "optim/total_tokens": 6162481152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.8689308166503906, "created_at": "2025-01-16T14:45:35.691494+00:00"} {"global_step": 11755, "acc_step": 0, "speed/wps": 12900.70194554296, "speed/FLOPS": 202623251476926.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07941218465566635, "optim/lr": 0.002889753655054696, "optim/total_tokens": 6163005440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9211268424987793, "created_at": "2025-01-16T14:45:45.858612+00:00"} {"global_step": 11756, "acc_step": 0, "speed/wps": 12900.602312347095, "speed/FLOPS": 202621686600675.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06169349327683449, "optim/lr": 0.002889721412458284, "optim/total_tokens": 6163529728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418774, "loss/out": 2.8377885818481445, "created_at": "2025-01-16T14:45:56.023078+00:00"} {"global_step": 11757, "acc_step": 0, "speed/wps": 12900.284736632422, "speed/FLOPS": 202616698637681.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.09819388389587402, "optim/lr": 0.002889689165327663, "optim/total_tokens": 6164054016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8398427963256836, "created_at": "2025-01-16T14:46:06.187029+00:00"} {"global_step": 11758, "acc_step": 0, "speed/wps": 12902.34657206874, "speed/FLOPS": 202649082596467.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06273623555898666, "optim/lr": 0.0028896569136629357, "optim/total_tokens": 6164578304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.979309320449829, "created_at": "2025-01-16T14:46:16.350454+00:00"} {"global_step": 11759, "acc_step": 0, "speed/wps": 12895.967311162183, "speed/FLOPS": 202548887537906.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.073664590716362, "optim/lr": 0.0028896246574642093, "optim/total_tokens": 6165102592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467256, "loss/out": 2.8901803493499756, "created_at": "2025-01-16T14:46:26.520214+00:00"} {"global_step": 11760, "acc_step": 0, "speed/wps": 12892.363107475676, "speed/FLOPS": 202492278566315.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07714501023292542, "optim/lr": 0.0028895923967315884, "optim/total_tokens": 6165626880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477322, "loss/out": 2.9068922996520996, "created_at": "2025-01-16T14:46:36.696640+00:00"} {"global_step": 11761, "acc_step": 0, "speed/wps": 12907.182187270037, "speed/FLOPS": 202725032578035.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06171979755163193, "optim/lr": 0.002889560131465178, "optim/total_tokens": 6166151168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.806392192840576, "created_at": "2025-01-16T14:46:46.858467+00:00"} {"global_step": 11762, "acc_step": 0, "speed/wps": 12898.02467673756, "speed/FLOPS": 202581201291384.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.10648005455732346, "optim/lr": 0.002889527861665084, "optim/total_tokens": 6166675456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400880, "loss/out": 2.943737030029297, "created_at": "2025-01-16T14:46:57.024479+00:00"} {"global_step": 11763, "acc_step": 0, "speed/wps": 12890.4018646537, "speed/FLOPS": 202461474552766.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04724133759737015, "optim/lr": 0.002889495587331411, "optim/total_tokens": 6167199744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.8857524394989014, "created_at": "2025-01-16T14:47:07.197615+00:00"} {"global_step": 11764, "acc_step": 0, "speed/wps": 12898.742119415345, "speed/FLOPS": 202592469714508.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05836985632777214, "optim/lr": 0.0028894633084642646, "optim/total_tokens": 6167724032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.899657726287842, "created_at": "2025-01-16T14:47:17.365396+00:00"} {"global_step": 11765, "acc_step": 0, "speed/wps": 12900.695063147241, "speed/FLOPS": 202623143379444.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05394197255373001, "optim/lr": 0.0028894310250637506, "optim/total_tokens": 6168248320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 2.8593595027923584, "created_at": "2025-01-16T14:47:27.530599+00:00"} {"global_step": 11766, "acc_step": 0, "speed/wps": 12901.240362034878, "speed/FLOPS": 202631708047793.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0459476113319397, "optim/lr": 0.0028893987371299734, "optim/total_tokens": 6168772608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8896894454956055, "created_at": "2025-01-16T14:47:37.697171+00:00"} {"global_step": 11767, "acc_step": 0, "speed/wps": 12901.138901648032, "speed/FLOPS": 202630114472997.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06551841646432877, "optim/lr": 0.002889366444663039, "optim/total_tokens": 6169296896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 2.9058094024658203, "created_at": "2025-01-16T14:47:47.860480+00:00"} {"global_step": 11768, "acc_step": 0, "speed/wps": 12903.618141528566, "speed/FLOPS": 202669054342153.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05067061632871628, "optim/lr": 0.0028893341476630527, "optim/total_tokens": 6169821184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 2.74838924407959, "created_at": "2025-01-16T14:47:58.022077+00:00"} {"global_step": 11769, "acc_step": 0, "speed/wps": 12902.951718794788, "speed/FLOPS": 202658587257358.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04409165307879448, "optim/lr": 0.00288930184613012, "optim/total_tokens": 6170345472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.9161314964294434, "created_at": "2025-01-16T14:48:08.184674+00:00"} {"global_step": 11770, "acc_step": 0, "speed/wps": 12902.806990564435, "speed/FLOPS": 202656314101623.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046246398240327835, "optim/lr": 0.002889269540064345, "optim/total_tokens": 6170869760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 2.866023540496826, "created_at": "2025-01-16T14:48:18.347268+00:00"} {"global_step": 11771, "acc_step": 0, "speed/wps": 12900.793502207871, "speed/FLOPS": 202624689500160.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054520685225725174, "optim/lr": 0.0028892372294658346, "optim/total_tokens": 6171394048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.84535813331604, "created_at": "2025-01-16T14:48:28.513516+00:00"} {"global_step": 11772, "acc_step": 0, "speed/wps": 12906.816371438288, "speed/FLOPS": 202719286937714.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048312004655599594, "optim/lr": 0.002889204914334694, "optim/total_tokens": 6171918336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.7940595149993896, "created_at": "2025-01-16T14:48:38.672428+00:00"} {"global_step": 11773, "acc_step": 0, "speed/wps": 12897.259861109917, "speed/FLOPS": 202569188810985.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.056897591799497604, "optim/lr": 0.0028891725946710287, "optim/total_tokens": 6172442624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337405, "loss/out": 2.814258098602295, "created_at": "2025-01-16T14:48:48.839973+00:00"} {"global_step": 11774, "acc_step": 0, "speed/wps": 12899.556954273532, "speed/FLOPS": 202605267815657.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04482709616422653, "optim/lr": 0.0028891402704749438, "optim/total_tokens": 6172966912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.867130756378174, "created_at": "2025-01-16T14:48:59.006137+00:00"} {"global_step": 11775, "acc_step": 0, "speed/wps": 12908.805895379326, "speed/FLOPS": 202750535145100.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04732644557952881, "optim/lr": 0.002889107941746544, "optim/total_tokens": 6173491200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.935635566711426, "created_at": "2025-01-16T14:49:09.166563+00:00"} {"global_step": 11776, "acc_step": 0, "speed/wps": 12901.98546924908, "speed/FLOPS": 202643410980477.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05460654944181442, "optim/lr": 0.0028890756084859366, "optim/total_tokens": 6174015488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.0504567623138428, "created_at": "2025-01-16T14:49:19.329187+00:00"} {"global_step": 11777, "acc_step": 0, "speed/wps": 12904.174596466839, "speed/FLOPS": 202677794231608.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044526390731334686, "optim/lr": 0.0028890432706932256, "optim/total_tokens": 6174539776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.7091684341430664, "created_at": "2025-01-16T14:49:29.490203+00:00"} {"global_step": 11778, "acc_step": 0, "speed/wps": 12901.502742314407, "speed/FLOPS": 202635829090632.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07987594604492188, "optim/lr": 0.0028890109283685166, "optim/total_tokens": 6175064064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392005, "loss/out": 2.895373821258545, "created_at": "2025-01-16T14:49:39.653873+00:00"} {"global_step": 11779, "acc_step": 0, "speed/wps": 12902.494089826434, "speed/FLOPS": 202651399565562.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05665263533592224, "optim/lr": 0.0028889785815119163, "optim/total_tokens": 6175588352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.909135341644287, "created_at": "2025-01-16T14:49:49.819535+00:00"} {"global_step": 11780, "acc_step": 0, "speed/wps": 12893.510934987002, "speed/FLOPS": 202510306774661.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06198639050126076, "optim/lr": 0.0028889462301235295, "optim/total_tokens": 6176112640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.918618679046631, "created_at": "2025-01-16T14:49:59.989943+00:00"} {"global_step": 11781, "acc_step": 0, "speed/wps": 12903.920849665321, "speed/FLOPS": 202673808789405.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062265776097774506, "optim/lr": 0.002888913874203461, "optim/total_tokens": 6176636928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373078, "loss/out": 2.945624351501465, "created_at": "2025-01-16T14:50:10.151065+00:00"} {"global_step": 11782, "acc_step": 0, "speed/wps": 12903.24595843676, "speed/FLOPS": 202663208695263.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06181968003511429, "optim/lr": 0.002888881513751817, "optim/total_tokens": 6177161216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 2.9247679710388184, "created_at": "2025-01-16T14:50:20.315367+00:00"} {"global_step": 11783, "acc_step": 0, "speed/wps": 12907.11498956872, "speed/FLOPS": 202723977145796.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05778534337878227, "optim/lr": 0.0028888491487687037, "optim/total_tokens": 6177685504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369775, "loss/out": 2.8733997344970703, "created_at": "2025-01-16T14:50:30.476251+00:00"} {"global_step": 11784, "acc_step": 0, "speed/wps": 12905.598570591254, "speed/FLOPS": 202700159701977.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050053276121616364, "optim/lr": 0.0028888167792542256, "optim/total_tokens": 6178209792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.941671133041382, "created_at": "2025-01-16T14:50:40.636133+00:00"} {"global_step": 11785, "acc_step": 0, "speed/wps": 12905.049185175678, "speed/FLOPS": 202691530849091.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05338279530405998, "optim/lr": 0.0028887844052084893, "optim/total_tokens": 6178734080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.920397996902466, "created_at": "2025-01-16T14:50:50.796313+00:00"} {"global_step": 11786, "acc_step": 0, "speed/wps": 12901.129729230881, "speed/FLOPS": 202629970407581.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04503506049513817, "optim/lr": 0.0028887520266315998, "optim/total_tokens": 6179258368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.8306145668029785, "created_at": "2025-01-16T14:51:00.961423+00:00"} {"global_step": 11787, "acc_step": 0, "speed/wps": 12899.143600273925, "speed/FLOPS": 202598775523085.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045194171369075775, "optim/lr": 0.0028887196435236633, "optim/total_tokens": 6179782656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.908268451690674, "created_at": "2025-01-16T14:51:11.126225+00:00"} {"global_step": 11788, "acc_step": 0, "speed/wps": 12905.34051703487, "speed/FLOPS": 202696106616272.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05252908542752266, "optim/lr": 0.0028886872558847847, "optim/total_tokens": 6180306944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.784540891647339, "created_at": "2025-01-16T14:51:21.292373+00:00"} {"global_step": 11789, "acc_step": 0, "speed/wps": 12904.201961102139, "speed/FLOPS": 202678224030808.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047404319047927856, "optim/lr": 0.00288865486371507, "optim/total_tokens": 6180831232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.784433603286743, "created_at": "2025-01-16T14:51:31.455663+00:00"} {"global_step": 11790, "acc_step": 0, "speed/wps": 12902.950396250857, "speed/FLOPS": 202658566484989.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06618975847959518, "optim/lr": 0.002888622467014625, "optim/total_tokens": 6181355520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 2.955798625946045, "created_at": "2025-01-16T14:51:41.618341+00:00"} {"global_step": 11791, "acc_step": 0, "speed/wps": 12882.904236986458, "speed/FLOPS": 202343714007432.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05570825934410095, "optim/lr": 0.0028885900657835556, "optim/total_tokens": 6181879808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.958282470703125, "created_at": "2025-01-16T14:51:51.827201+00:00"} {"global_step": 11792, "acc_step": 0, "speed/wps": 12908.19650441466, "speed/FLOPS": 202740963822609.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0621308907866478, "optim/lr": 0.0028885576600219667, "optim/total_tokens": 6182404096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 2.9070892333984375, "created_at": "2025-01-16T14:52:01.985056+00:00"} {"global_step": 11793, "acc_step": 0, "speed/wps": 12905.17778592783, "speed/FLOPS": 202693550700620.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05426807701587677, "optim/lr": 0.0028885252497299653, "optim/total_tokens": 6182928384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.039255142211914, "created_at": "2025-01-16T14:52:12.149546+00:00"} {"global_step": 11794, "acc_step": 0, "speed/wps": 12901.207360506545, "speed/FLOPS": 202631189713442.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05597976595163345, "optim/lr": 0.0028884928349076563, "optim/total_tokens": 6183452672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.7899255752563477, "created_at": "2025-01-16T14:52:22.314119+00:00"} {"global_step": 11795, "acc_step": 0, "speed/wps": 12902.328184779068, "speed/FLOPS": 202648793798816.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04063231870532036, "optim/lr": 0.0028884604155551452, "optim/total_tokens": 6183976960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 2.9102957248687744, "created_at": "2025-01-16T14:52:32.476714+00:00"} {"global_step": 11796, "acc_step": 0, "speed/wps": 12904.2334883159, "speed/FLOPS": 202678719209023.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06896598637104034, "optim/lr": 0.002888427991672538, "optim/total_tokens": 6184501248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9235124588012695, "created_at": "2025-01-16T14:52:42.642807+00:00"} {"global_step": 11797, "acc_step": 0, "speed/wps": 12899.094773237359, "speed/FLOPS": 202598008627380.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05833989009261131, "optim/lr": 0.002888395563259941, "optim/total_tokens": 6185025536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396909, "loss/out": 2.79072904586792, "created_at": "2025-01-16T14:52:52.807647+00:00"} {"global_step": 11798, "acc_step": 0, "speed/wps": 12902.109373549052, "speed/FLOPS": 202645357067770.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056055910885334015, "optim/lr": 0.0028883631303174596, "optim/total_tokens": 6185549824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.808804512023926, "created_at": "2025-01-16T14:53:02.972653+00:00"} {"global_step": 11799, "acc_step": 0, "speed/wps": 12902.24631377537, "speed/FLOPS": 202647507902200.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06142249330878258, "optim/lr": 0.0028883306928451998, "optim/total_tokens": 6186074112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.9272944927215576, "created_at": "2025-01-16T14:53:13.135837+00:00"} {"global_step": 11800, "acc_step": 0, "speed/wps": 12902.07275532027, "speed/FLOPS": 202644781928169.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045699696987867355, "optim/lr": 0.0028882982508432672, "optim/total_tokens": 6186598400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8426458835601807, "created_at": "2025-01-16T14:53:23.304576+00:00"} {"global_step": 11801, "acc_step": 0, "speed/wps": 12907.195646307495, "speed/FLOPS": 202725243970713.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0655687153339386, "optim/lr": 0.0028882658043117675, "optim/total_tokens": 6187122688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 2.8866987228393555, "created_at": "2025-01-16T14:53:33.464649+00:00"} {"global_step": 11802, "acc_step": 0, "speed/wps": 12907.42798781008, "speed/FLOPS": 202728893213282.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05025212839245796, "optim/lr": 0.0028882333532508075, "optim/total_tokens": 6187646976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.7073447704315186, "created_at": "2025-01-16T14:53:43.623954+00:00"} {"global_step": 11803, "acc_step": 0, "speed/wps": 12902.815008460826, "speed/FLOPS": 202656440033704.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04883110523223877, "optim/lr": 0.002888200897660492, "optim/total_tokens": 6188171264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.946110725402832, "created_at": "2025-01-16T14:53:53.785908+00:00"} {"global_step": 11804, "acc_step": 0, "speed/wps": 12898.777412420603, "speed/FLOPS": 202593024039654.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05426976457238197, "optim/lr": 0.002888168437540927, "optim/total_tokens": 6188695552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.916347026824951, "created_at": "2025-01-16T14:54:03.950988+00:00"} {"global_step": 11805, "acc_step": 0, "speed/wps": 12901.745425337116, "speed/FLOPS": 202639640760985.16, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04835154861211777, "optim/lr": 0.002888135972892219, "optim/total_tokens": 6189219840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 2.937765598297119, "created_at": "2025-01-16T14:54:14.115841+00:00"} {"global_step": 11806, "acc_step": 0, "speed/wps": 12906.536483566246, "speed/FLOPS": 202714890914077.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04639523848891258, "optim/lr": 0.0028881035037144737, "optim/total_tokens": 6189744128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.813378095626831, "created_at": "2025-01-16T14:54:24.274852+00:00"} {"global_step": 11807, "acc_step": 0, "speed/wps": 12900.509637378598, "speed/FLOPS": 202620231012946.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04631641507148743, "optim/lr": 0.0028880710300077965, "optim/total_tokens": 6190268416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.794379949569702, "created_at": "2025-01-16T14:54:34.439129+00:00"} {"global_step": 11808, "acc_step": 0, "speed/wps": 12899.474537917238, "speed/FLOPS": 202603973353529.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04512915387749672, "optim/lr": 0.0028880385517722944, "optim/total_tokens": 6190792704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 2.8820550441741943, "created_at": "2025-01-16T14:54:44.603578+00:00"} {"global_step": 11809, "acc_step": 0, "speed/wps": 12901.053840869896, "speed/FLOPS": 202628778476590.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05268550291657448, "optim/lr": 0.002888006069008072, "optim/total_tokens": 6191316992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8279941082000732, "created_at": "2025-01-16T14:54:54.771715+00:00"} {"global_step": 11810, "acc_step": 0, "speed/wps": 12904.175676709885, "speed/FLOPS": 202677811198309.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043068550527095795, "optim/lr": 0.002887973581715237, "optim/total_tokens": 6191841280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306453, "loss/out": 2.9454967975616455, "created_at": "2025-01-16T14:55:04.936003+00:00"} {"global_step": 11811, "acc_step": 0, "speed/wps": 12901.164015878085, "speed/FLOPS": 202630508926490.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06490974128246307, "optim/lr": 0.0028879410898938933, "optim/total_tokens": 6192365568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9271652698516846, "created_at": "2025-01-16T14:55:15.099609+00:00"} {"global_step": 11812, "acc_step": 0, "speed/wps": 12901.139664439886, "speed/FLOPS": 202630126453691.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053700946271419525, "optim/lr": 0.002887908593544149, "optim/total_tokens": 6192889856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.7735936641693115, "created_at": "2025-01-16T14:55:25.263038+00:00"} {"global_step": 11813, "acc_step": 0, "speed/wps": 12894.595331569451, "speed/FLOPS": 202527338713105.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.060660652816295624, "optim/lr": 0.002887876092666109, "optim/total_tokens": 6193414144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 2.8512892723083496, "created_at": "2025-01-16T14:55:35.432556+00:00"} {"global_step": 11814, "acc_step": 0, "speed/wps": 12907.246164705799, "speed/FLOPS": 202726037431578.3, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05897139757871628, "optim/lr": 0.0028878435872598794, "optim/total_tokens": 6193938432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.883392572402954, "created_at": "2025-01-16T14:55:45.593070+00:00"} {"global_step": 11815, "acc_step": 0, "speed/wps": 12901.166382316273, "speed/FLOPS": 202630546094654.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061561468988657, "optim/lr": 0.0028878110773255658, "optim/total_tokens": 6194462720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478894, "loss/out": 2.962000608444214, "created_at": "2025-01-16T14:55:55.760548+00:00"} {"global_step": 11816, "acc_step": 0, "speed/wps": 12899.564945666883, "speed/FLOPS": 202605393331471.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07556667178869247, "optim/lr": 0.002887778562863276, "optim/total_tokens": 6194987008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497150, "loss/out": 2.8058717250823975, "created_at": "2025-01-16T14:56:05.927101+00:00"} {"global_step": 11817, "acc_step": 0, "speed/wps": 12903.67140389915, "speed/FLOPS": 202669890900873.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061551533639431, "optim/lr": 0.0028877460438731144, "optim/total_tokens": 6195511296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.8148255348205566, "created_at": "2025-01-16T14:56:16.091252+00:00"} {"global_step": 11818, "acc_step": 0, "speed/wps": 12903.083419739747, "speed/FLOPS": 202660655801675.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07436598092317581, "optim/lr": 0.0028877135203551874, "optim/total_tokens": 6196035584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369341, "loss/out": 2.810647964477539, "created_at": "2025-01-16T14:56:26.259929+00:00"} {"global_step": 11819, "acc_step": 0, "speed/wps": 12905.602453461423, "speed/FLOPS": 202700220687789.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0641307607293129, "optim/lr": 0.0028876809923096014, "optim/total_tokens": 6196559872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.782282829284668, "created_at": "2025-01-16T14:56:36.421666+00:00"} {"global_step": 11820, "acc_step": 0, "speed/wps": 12900.306883082994, "speed/FLOPS": 202617046478119.97, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06585153192281723, "optim/lr": 0.002887648459736463, "optim/total_tokens": 6197084160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9369540214538574, "created_at": "2025-01-16T14:56:46.586317+00:00"} {"global_step": 11821, "acc_step": 0, "speed/wps": 12901.120924621478, "speed/FLOPS": 202629832119092.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045416973531246185, "optim/lr": 0.0028876159226358766, "optim/total_tokens": 6197608448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 2.9037723541259766, "created_at": "2025-01-16T14:56:56.756500+00:00"} {"global_step": 11822, "acc_step": 0, "speed/wps": 12905.6524090845, "speed/FLOPS": 202701005309496.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058842550963163376, "optim/lr": 0.0028875833810079504, "optim/total_tokens": 6198132736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.9388113021850586, "created_at": "2025-01-16T14:57:06.923643+00:00"} {"global_step": 11823, "acc_step": 0, "speed/wps": 12901.727535325024, "speed/FLOPS": 202639359773761.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04611029475927353, "optim/lr": 0.00288755083485279, "optim/total_tokens": 6198657024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.6917901039123535, "created_at": "2025-01-16T14:57:17.087504+00:00"} {"global_step": 11824, "acc_step": 0, "speed/wps": 12899.418741032727, "speed/FLOPS": 202603096986785.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.043349489569664, "optim/lr": 0.002887518284170501, "optim/total_tokens": 6199181312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 2.870105028152466, "created_at": "2025-01-16T14:57:27.252309+00:00"} {"global_step": 11825, "acc_step": 0, "speed/wps": 12898.652920734288, "speed/FLOPS": 202591068726651.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04766229912638664, "optim/lr": 0.00288748572896119, "optim/total_tokens": 6199705600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8831746578216553, "created_at": "2025-01-16T14:57:37.420855+00:00"} {"global_step": 11826, "acc_step": 0, "speed/wps": 12896.586721776486, "speed/FLOPS": 202558616232763.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.04208654910326004, "optim/lr": 0.002887453169224963, "optim/total_tokens": 6200229888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8982934951782227, "created_at": "2025-01-16T14:57:47.590796+00:00"} {"global_step": 11827, "acc_step": 0, "speed/wps": 12902.60851354296, "speed/FLOPS": 202653196747265.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04433664679527283, "optim/lr": 0.0028874206049619264, "optim/total_tokens": 6200754176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 2.879744052886963, "created_at": "2025-01-16T14:57:57.754015+00:00"} {"global_step": 11828, "acc_step": 0, "speed/wps": 12906.257664753255, "speed/FLOPS": 202710511681482.34, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04134932905435562, "optim/lr": 0.002887388036172186, "optim/total_tokens": 6201278464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.903198719024658, "created_at": "2025-01-16T14:58:07.915025+00:00"} {"global_step": 11829, "acc_step": 0, "speed/wps": 12899.653739592564, "speed/FLOPS": 202606787962089.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054326679557561874, "optim/lr": 0.0028873554628558494, "optim/total_tokens": 6201802752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338225, "loss/out": 2.888725757598877, "created_at": "2025-01-16T14:58:18.080597+00:00"} {"global_step": 11830, "acc_step": 0, "speed/wps": 12898.65719427025, "speed/FLOPS": 202591135848405.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05428745225071907, "optim/lr": 0.0028873228850130215, "optim/total_tokens": 6202327040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.866994857788086, "created_at": "2025-01-16T14:58:28.245883+00:00"} {"global_step": 11831, "acc_step": 0, "speed/wps": 12901.655749873706, "speed/FLOPS": 202638232284605.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07058867067098618, "optim/lr": 0.002887290302643809, "optim/total_tokens": 6202851328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.784904956817627, "created_at": "2025-01-16T14:58:38.408841+00:00"} {"global_step": 11832, "acc_step": 0, "speed/wps": 12904.839195449249, "speed/FLOPS": 202688232671881.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05337250605225563, "optim/lr": 0.0028872577157483185, "optim/total_tokens": 6203375616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 2.9266839027404785, "created_at": "2025-01-16T14:58:48.572923+00:00"} {"global_step": 11833, "acc_step": 0, "speed/wps": 12904.580855147698, "speed/FLOPS": 202684175082449.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04985063150525093, "optim/lr": 0.002887225124326656, "optim/total_tokens": 6203899904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.9502153396606445, "created_at": "2025-01-16T14:58:58.734515+00:00"} {"global_step": 11834, "acc_step": 0, "speed/wps": 12905.552459171819, "speed/FLOPS": 202699435458771.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0527985543012619, "optim/lr": 0.0028871925283789276, "optim/total_tokens": 6204424192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.9687676429748535, "created_at": "2025-01-16T14:59:08.897490+00:00"} {"global_step": 11835, "acc_step": 0, "speed/wps": 12905.018243276474, "speed/FLOPS": 202691044864045.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042296409606933594, "optim/lr": 0.0028871599279052403, "optim/total_tokens": 6204948480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.82224702835083, "created_at": "2025-01-16T14:59:19.057765+00:00"} {"global_step": 11836, "acc_step": 0, "speed/wps": 12906.94621240774, "speed/FLOPS": 202721326268559.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05502764508128166, "optim/lr": 0.0028871273229057006, "optim/total_tokens": 6205472768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.00272274017334, "created_at": "2025-01-16T14:59:29.218258+00:00"} {"global_step": 11837, "acc_step": 0, "speed/wps": 12905.50016942916, "speed/FLOPS": 202698614176509.78, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03917324170470238, "optim/lr": 0.0028870947133804135, "optim/total_tokens": 6205997056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330191, "loss/out": 2.9271321296691895, "created_at": "2025-01-16T14:59:39.382616+00:00"} {"global_step": 11838, "acc_step": 0, "speed/wps": 12904.406759407666, "speed/FLOPS": 202681440669619.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051166802644729614, "optim/lr": 0.0028870620993294866, "optim/total_tokens": 6206521344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9527125358581543, "created_at": "2025-01-16T14:59:49.546430+00:00"} {"global_step": 11839, "acc_step": 0, "speed/wps": 12906.043138310642, "speed/FLOPS": 202707142248910.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05210661143064499, "optim/lr": 0.002887029480753026, "optim/total_tokens": 6207045632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.763773202896118, "created_at": "2025-01-16T14:59:59.706798+00:00"} {"global_step": 11840, "acc_step": 0, "speed/wps": 12904.1296748442, "speed/FLOPS": 202677088675796.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0507817342877388, "optim/lr": 0.0028869968576511387, "optim/total_tokens": 6207569920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 2.7556276321411133, "created_at": "2025-01-16T15:00:09.869964+00:00"} {"global_step": 11841, "acc_step": 0, "speed/wps": 12904.858376794335, "speed/FLOPS": 202688533941264.34, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.040937744081020355, "optim/lr": 0.0028869642300239297, "optim/total_tokens": 6208094208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9665720462799072, "created_at": "2025-01-16T15:00:20.033245+00:00"} {"global_step": 11842, "acc_step": 0, "speed/wps": 12905.18969767631, "speed/FLOPS": 202693737790998.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04667327553033829, "optim/lr": 0.002886931597871507, "optim/total_tokens": 6208618496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 2.9939382076263428, "created_at": "2025-01-16T15:00:30.194540+00:00"} {"global_step": 11843, "acc_step": 0, "speed/wps": 12899.64620565836, "speed/FLOPS": 202606669631300.2, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04379377141594887, "optim/lr": 0.002886898961193976, "optim/total_tokens": 6209142784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8461785316467285, "created_at": "2025-01-16T15:00:40.359014+00:00"} {"global_step": 11844, "acc_step": 0, "speed/wps": 12903.8093923963, "speed/FLOPS": 202672058199837.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04556378722190857, "optim/lr": 0.0028868663199914436, "optim/total_tokens": 6209667072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9280314445495605, "created_at": "2025-01-16T15:00:50.527314+00:00"} {"global_step": 11845, "acc_step": 0, "speed/wps": 12904.721930157406, "speed/FLOPS": 202686390859334.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04919853433966637, "optim/lr": 0.0028868336742640167, "optim/total_tokens": 6210191360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331823, "loss/out": 2.802506923675537, "created_at": "2025-01-16T15:01:00.689707+00:00"} {"global_step": 11846, "acc_step": 0, "speed/wps": 12903.647245301629, "speed/FLOPS": 202669511456902.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05556056275963783, "optim/lr": 0.002886801024011801, "optim/total_tokens": 6210715648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.890017509460449, "created_at": "2025-01-16T15:01:10.851058+00:00"} {"global_step": 11847, "acc_step": 0, "speed/wps": 12903.956789797961, "speed/FLOPS": 202674373278575.94, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05458235740661621, "optim/lr": 0.0028867683692349036, "optim/total_tokens": 6211239936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.832033634185791, "created_at": "2025-01-16T15:01:21.016455+00:00"} {"global_step": 11848, "acc_step": 0, "speed/wps": 12902.19459846443, "speed/FLOPS": 202646695642178.2, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04673386737704277, "optim/lr": 0.0028867357099334307, "optim/total_tokens": 6211764224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 2.847769021987915, "created_at": "2025-01-16T15:01:31.179069+00:00"} {"global_step": 11849, "acc_step": 0, "speed/wps": 12906.529731059105, "speed/FLOPS": 202714784856674.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04897027835249901, "optim/lr": 0.0028867030461074894, "optim/total_tokens": 6212288512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8001489639282227, "created_at": "2025-01-16T15:01:41.338089+00:00"} {"global_step": 11850, "acc_step": 0, "speed/wps": 12904.435743682598, "speed/FLOPS": 202681895907485.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04937669634819031, "optim/lr": 0.0028866703777571854, "optim/total_tokens": 6212812800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.8283560276031494, "created_at": "2025-01-16T15:01:51.500447+00:00"} {"global_step": 11851, "acc_step": 0, "speed/wps": 12906.165366545503, "speed/FLOPS": 202709062011298.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04240132123231888, "optim/lr": 0.0028866377048826263, "optim/total_tokens": 6213337088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.9512219429016113, "created_at": "2025-01-16T15:02:01.662716+00:00"} {"global_step": 11852, "acc_step": 0, "speed/wps": 12907.50662062702, "speed/FLOPS": 202730128249725.06, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046290431171655655, "optim/lr": 0.002886605027483918, "optim/total_tokens": 6213861376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.849931478500366, "created_at": "2025-01-16T15:02:11.820951+00:00"} {"global_step": 11853, "acc_step": 0, "speed/wps": 12905.85496219946, "speed/FLOPS": 202704186684503.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053793683648109436, "optim/lr": 0.002886572345561167, "optim/total_tokens": 6214385664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.976773500442505, "created_at": "2025-01-16T15:02:21.981281+00:00"} {"global_step": 11854, "acc_step": 0, "speed/wps": 12909.699966694554, "speed/FLOPS": 202764577763689.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05276407673954964, "optim/lr": 0.0028865396591144805, "optim/total_tokens": 6214909952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.8681693077087402, "created_at": "2025-01-16T15:02:32.139060+00:00"} {"global_step": 11855, "acc_step": 0, "speed/wps": 12901.114650827214, "speed/FLOPS": 202629733580532.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05365592986345291, "optim/lr": 0.0028865069681439644, "optim/total_tokens": 6215434240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.7753982543945312, "created_at": "2025-01-16T15:02:42.302796+00:00"} {"global_step": 11856, "acc_step": 0, "speed/wps": 12895.953591929918, "speed/FLOPS": 202548672058512.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060712166130542755, "optim/lr": 0.0028864742726497266, "optim/total_tokens": 6215958528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.866959571838379, "created_at": "2025-01-16T15:02:52.472146+00:00"} {"global_step": 11857, "acc_step": 0, "speed/wps": 12906.842751813427, "speed/FLOPS": 202719701277756.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05028292536735535, "optim/lr": 0.0028864415726318725, "optim/total_tokens": 6216482816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 2.916712760925293, "created_at": "2025-01-16T15:03:02.630928+00:00"} {"global_step": 11858, "acc_step": 0, "speed/wps": 12903.809367550935, "speed/FLOPS": 202672057809607.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049704477190971375, "optim/lr": 0.0028864088680905097, "optim/total_tokens": 6217007104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.848083257675171, "created_at": "2025-01-16T15:03:12.795829+00:00"} {"global_step": 11859, "acc_step": 0, "speed/wps": 12904.287542594704, "speed/FLOPS": 202679568205750.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059342917054891586, "optim/lr": 0.002886376159025744, "optim/total_tokens": 6217531392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.850663423538208, "created_at": "2025-01-16T15:03:22.956702+00:00"} {"global_step": 11860, "acc_step": 0, "speed/wps": 12903.920758564078, "speed/FLOPS": 202673807358535.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05486778914928436, "optim/lr": 0.002886343445437683, "optim/total_tokens": 6218055680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.9525809288024902, "created_at": "2025-01-16T15:03:33.122428+00:00"} {"global_step": 11861, "acc_step": 0, "speed/wps": 12903.099770902934, "speed/FLOPS": 202660912619162.16, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06546078622341156, "optim/lr": 0.0028863107273264326, "optim/total_tokens": 6218579968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.92480731010437, "created_at": "2025-01-16T15:03:43.289387+00:00"} {"global_step": 11862, "acc_step": 0, "speed/wps": 12904.231182278269, "speed/FLOPS": 202678682989533.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048744574189186096, "optim/lr": 0.0028862780046921004, "optim/total_tokens": 6219104256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 2.7942113876342773, "created_at": "2025-01-16T15:03:53.453092+00:00"} {"global_step": 11863, "acc_step": 0, "speed/wps": 12900.793334284404, "speed/FLOPS": 202624686862692.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06781845539808273, "optim/lr": 0.002886245277534793, "optim/total_tokens": 6219628544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 2.963696241378784, "created_at": "2025-01-16T15:04:03.617029+00:00"} {"global_step": 11864, "acc_step": 0, "speed/wps": 12908.010938712225, "speed/FLOPS": 202738049258258.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.053821414709091187, "optim/lr": 0.0028862125458546163, "optim/total_tokens": 6220152832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471481, "loss/out": 2.925539255142212, "created_at": "2025-01-16T15:04:13.778492+00:00"} {"global_step": 11865, "acc_step": 0, "speed/wps": 12903.707214868147, "speed/FLOPS": 202670453361351.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08506406098604202, "optim/lr": 0.0028861798096516777, "optim/total_tokens": 6220677120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.907094717025757, "created_at": "2025-01-16T15:04:23.948920+00:00"} {"global_step": 11866, "acc_step": 0, "speed/wps": 12899.673653955995, "speed/FLOPS": 202607100744532.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06070558354258537, "optim/lr": 0.002886147068926085, "optim/total_tokens": 6221201408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.858577251434326, "created_at": "2025-01-16T15:04:34.116313+00:00"} {"global_step": 11867, "acc_step": 0, "speed/wps": 12904.730622537769, "speed/FLOPS": 202686527385112.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05690344050526619, "optim/lr": 0.002886114323677943, "optim/total_tokens": 6221725696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.7911195755004883, "created_at": "2025-01-16T15:04:44.278455+00:00"} {"global_step": 11868, "acc_step": 0, "speed/wps": 12910.93659463783, "speed/FLOPS": 202784000704843.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06475993245840073, "optim/lr": 0.00288608157390736, "optim/total_tokens": 6222249984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.80171537399292, "created_at": "2025-01-16T15:04:54.434724+00:00"} {"global_step": 11869, "acc_step": 0, "speed/wps": 12902.684544400487, "speed/FLOPS": 202654390916360.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05175863206386566, "optim/lr": 0.0028860488196144423, "optim/total_tokens": 6222774272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8236961364746094, "created_at": "2025-01-16T15:05:04.606426+00:00"} {"global_step": 11870, "acc_step": 0, "speed/wps": 12901.239875969215, "speed/FLOPS": 202631700413464.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05347779020667076, "optim/lr": 0.0028860160607992974, "optim/total_tokens": 6223298560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 2.822786808013916, "created_at": "2025-01-16T15:05:14.770706+00:00"} {"global_step": 11871, "acc_step": 0, "speed/wps": 12893.3318649619, "speed/FLOPS": 202507494233848.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048859547823667526, "optim/lr": 0.002885983297462031, "optim/total_tokens": 6223822848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.842164993286133, "created_at": "2025-01-16T15:05:24.942509+00:00"} {"global_step": 11872, "acc_step": 0, "speed/wps": 12891.937312635044, "speed/FLOPS": 202485590873239.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043878838419914246, "optim/lr": 0.0028859505296027514, "optim/total_tokens": 6224347136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.856201648712158, "created_at": "2025-01-16T15:05:35.114419+00:00"} {"global_step": 11873, "acc_step": 0, "speed/wps": 12900.882372176704, "speed/FLOPS": 202626085325141.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044429145753383636, "optim/lr": 0.0028859177572215645, "optim/total_tokens": 6224871424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476257, "loss/out": 2.872860908508301, "created_at": "2025-01-16T15:05:45.284680+00:00"} {"global_step": 11874, "acc_step": 0, "speed/wps": 12894.954084367802, "speed/FLOPS": 202532973418781.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051862604916095734, "optim/lr": 0.0028858849803185776, "optim/total_tokens": 6225395712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 2.9574978351593018, "created_at": "2025-01-16T15:05:55.457826+00:00"} {"global_step": 11875, "acc_step": 0, "speed/wps": 12907.779317306653, "speed/FLOPS": 202734411325803.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04700411856174469, "optim/lr": 0.0028858521988938974, "optim/total_tokens": 6225920000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.8756837844848633, "created_at": "2025-01-16T15:06:05.617915+00:00"} {"global_step": 11876, "acc_step": 0, "speed/wps": 12898.86929406816, "speed/FLOPS": 202594467167187.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051572285592556, "optim/lr": 0.0028858194129476314, "optim/total_tokens": 6226444288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.7802276611328125, "created_at": "2025-01-16T15:06:15.783676+00:00"} {"global_step": 11877, "acc_step": 0, "speed/wps": 12891.758377246704, "speed/FLOPS": 202482780447083.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060084596276283264, "optim/lr": 0.002885786622479886, "optim/total_tokens": 6226968576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.7716526985168457, "created_at": "2025-01-16T15:06:25.956637+00:00"} {"global_step": 11878, "acc_step": 0, "speed/wps": 12905.330789548512, "speed/FLOPS": 202695953832731.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048238229006528854, "optim/lr": 0.002885753827490768, "optim/total_tokens": 6227492864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.9452085494995117, "created_at": "2025-01-16T15:06:36.123161+00:00"} {"global_step": 11879, "acc_step": 0, "speed/wps": 12905.360832381226, "speed/FLOPS": 202696425696702.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04382805898785591, "optim/lr": 0.0028857210279803857, "optim/total_tokens": 6228017152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 2.9654250144958496, "created_at": "2025-01-16T15:06:46.286112+00:00"} {"global_step": 11880, "acc_step": 0, "speed/wps": 12904.86343069185, "speed/FLOPS": 202688613319669.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06295379251241684, "optim/lr": 0.0028856882239488442, "optim/total_tokens": 6228541440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9267115592956543, "created_at": "2025-01-16T15:06:56.453836+00:00"} {"global_step": 11881, "acc_step": 0, "speed/wps": 12901.804764565559, "speed/FLOPS": 202640572765108.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05162747576832771, "optim/lr": 0.0028856554153962525, "optim/total_tokens": 6229065728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8435893058776855, "created_at": "2025-01-16T15:07:06.618912+00:00"} {"global_step": 11882, "acc_step": 0, "speed/wps": 12902.894667477021, "speed/FLOPS": 202657691188016.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06497848778963089, "optim/lr": 0.002885622602322716, "optim/total_tokens": 6229590016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.839583158493042, "created_at": "2025-01-16T15:07:16.783497+00:00"} {"global_step": 11883, "acc_step": 0, "speed/wps": 12903.063794139913, "speed/FLOPS": 202660347554661.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06406552344560623, "optim/lr": 0.0028855897847283422, "optim/total_tokens": 6230114304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.9064698219299316, "created_at": "2025-01-16T15:07:26.945744+00:00"} {"global_step": 11884, "acc_step": 0, "speed/wps": 12904.483539149924, "speed/FLOPS": 202682646600977.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05642083287239075, "optim/lr": 0.002885556962613239, "optim/total_tokens": 6230638592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 2.8659234046936035, "created_at": "2025-01-16T15:07:37.107595+00:00"} {"global_step": 11885, "acc_step": 0, "speed/wps": 12900.55680659653, "speed/FLOPS": 202620971870330.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05966109782457352, "optim/lr": 0.002885524135977513, "optim/total_tokens": 6231162880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.8233280181884766, "created_at": "2025-01-16T15:07:47.271524+00:00"} {"global_step": 11886, "acc_step": 0, "speed/wps": 12902.881406678584, "speed/FLOPS": 202657482908954.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05841565132141113, "optim/lr": 0.002885491304821271, "optim/total_tokens": 6231687168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.939438819885254, "created_at": "2025-01-16T15:07:57.436270+00:00"} {"global_step": 11887, "acc_step": 0, "speed/wps": 12899.710947884456, "speed/FLOPS": 202607686496926.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06372245401144028, "optim/lr": 0.0028854584691446203, "optim/total_tokens": 6232211456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 2.8716025352478027, "created_at": "2025-01-16T15:08:07.604402+00:00"} {"global_step": 11888, "acc_step": 0, "speed/wps": 12901.907659473594, "speed/FLOPS": 202642188871031.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04762447252869606, "optim/lr": 0.002885425628947668, "optim/total_tokens": 6232735744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.7630271911621094, "created_at": "2025-01-16T15:08:17.768751+00:00"} {"global_step": 11889, "acc_step": 0, "speed/wps": 12905.438159562485, "speed/FLOPS": 202697640226343.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059363991022109985, "optim/lr": 0.002885392784230521, "optim/total_tokens": 6233260032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375761, "loss/out": 2.8441638946533203, "created_at": "2025-01-16T15:08:27.928764+00:00"} {"global_step": 11890, "acc_step": 0, "speed/wps": 12903.156708101178, "speed/FLOPS": 202661806896100.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06249884143471718, "optim/lr": 0.0028853599349932873, "optim/total_tokens": 6233784320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.840973377227783, "created_at": "2025-01-16T15:08:38.090513+00:00"} {"global_step": 11891, "acc_step": 0, "speed/wps": 12898.238345646656, "speed/FLOPS": 202584557255213.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06459823250770569, "optim/lr": 0.002885327081236073, "optim/total_tokens": 6234308608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8889572620391846, "created_at": "2025-01-16T15:08:48.259432+00:00"} {"global_step": 11892, "acc_step": 0, "speed/wps": 12893.531822620575, "speed/FLOPS": 202510634843649.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.053848572075366974, "optim/lr": 0.002885294222958986, "optim/total_tokens": 6234832896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428359, "loss/out": 2.892691135406494, "created_at": "2025-01-16T15:08:58.429151+00:00"} {"global_step": 11893, "acc_step": 0, "speed/wps": 12887.950334266496, "speed/FLOPS": 202422969899279.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07437727600336075, "optim/lr": 0.0028852613601621333, "optim/total_tokens": 6235357184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 2.903954267501831, "created_at": "2025-01-16T15:09:08.606231+00:00"} {"global_step": 11894, "acc_step": 0, "speed/wps": 12895.229111412344, "speed/FLOPS": 202537293096442.94, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06572864204645157, "optim/lr": 0.0028852284928456225, "optim/total_tokens": 6235881472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.7693989276885986, "created_at": "2025-01-16T15:09:18.774722+00:00"} {"global_step": 11895, "acc_step": 0, "speed/wps": 12904.007040300294, "speed/FLOPS": 202675162531766.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06578429788351059, "optim/lr": 0.0028851956210095603, "optim/total_tokens": 6236405760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 2.9290642738342285, "created_at": "2025-01-16T15:09:28.935701+00:00"} {"global_step": 11896, "acc_step": 0, "speed/wps": 12903.802149384983, "speed/FLOPS": 202671944438391.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07373357564210892, "optim/lr": 0.0028851627446540543, "optim/total_tokens": 6236930048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.894200563430786, "created_at": "2025-01-16T15:09:39.099987+00:00"} {"global_step": 11897, "acc_step": 0, "speed/wps": 12903.556934119426, "speed/FLOPS": 202668092995685.25, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05148552358150482, "optim/lr": 0.002885129863779211, "optim/total_tokens": 6237454336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.776228189468384, "created_at": "2025-01-16T15:09:49.265415+00:00"} {"global_step": 11898, "acc_step": 0, "speed/wps": 12905.456083534933, "speed/FLOPS": 202697921746960.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1012822613120079, "optim/lr": 0.002885096978385139, "optim/total_tokens": 6237978624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 2.8277299404144287, "created_at": "2025-01-16T15:09:59.425538+00:00"} {"global_step": 11899, "acc_step": 0, "speed/wps": 12912.288885432301, "speed/FLOPS": 202805240290012.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04180402681231499, "optim/lr": 0.0028850640884719443, "optim/total_tokens": 6238502912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.9028124809265137, "created_at": "2025-01-16T15:10:09.581167+00:00"} {"global_step": 11900, "acc_step": 0, "speed/wps": 12899.325152605743, "speed/FLOPS": 202601627051935.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08104468137025833, "optim/lr": 0.0028850311940397354, "optim/total_tokens": 6239027200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9281909465789795, "created_at": "2025-01-16T15:10:19.753104+00:00"} {"global_step": 11901, "acc_step": 0, "speed/wps": 12907.596223549455, "speed/FLOPS": 202731535586749.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04661588743329048, "optim/lr": 0.0028849982950886186, "optim/total_tokens": 6239551488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468765, "loss/out": 2.863328456878662, "created_at": "2025-01-16T15:10:29.911222+00:00"} {"global_step": 11902, "acc_step": 0, "speed/wps": 12902.485115387157, "speed/FLOPS": 202651258609661.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06663721799850464, "optim/lr": 0.0028849653916187016, "optim/total_tokens": 6240075776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.9961671829223633, "created_at": "2025-01-16T15:10:40.073429+00:00"} {"global_step": 11903, "acc_step": 0, "speed/wps": 12904.049806540412, "speed/FLOPS": 202675834234334.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050967924296855927, "optim/lr": 0.002884932483630092, "optim/total_tokens": 6240600064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 2.8673624992370605, "created_at": "2025-01-16T15:10:50.234499+00:00"} {"global_step": 11904, "acc_step": 0, "speed/wps": 12905.580858058316, "speed/FLOPS": 202699881502308.47, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06491027772426605, "optim/lr": 0.0028848995711228966, "optim/total_tokens": 6241124352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.875060558319092, "created_at": "2025-01-16T15:11:00.397818+00:00"} {"global_step": 11905, "acc_step": 0, "speed/wps": 12903.88214358382, "speed/FLOPS": 202673200857208.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.063433438539505, "optim/lr": 0.0028848666540972233, "optim/total_tokens": 6241648640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8843741416931152, "created_at": "2025-01-16T15:11:10.562466+00:00"} {"global_step": 11906, "acc_step": 0, "speed/wps": 12902.145715208719, "speed/FLOPS": 202645927863474.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047804173082113266, "optim/lr": 0.00288483373255318, "optim/total_tokens": 6242172928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 2.906230926513672, "created_at": "2025-01-16T15:11:20.730619+00:00"} {"global_step": 11907, "acc_step": 0, "speed/wps": 12898.392624584192, "speed/FLOPS": 202586980417929.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05125068873167038, "optim/lr": 0.0028848008064908726, "optim/total_tokens": 6242697216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.8869853019714355, "created_at": "2025-01-16T15:11:30.896454+00:00"} {"global_step": 11908, "acc_step": 0, "speed/wps": 12903.604404829905, "speed/FLOPS": 202668838588424.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04825961962342262, "optim/lr": 0.0028847678759104096, "optim/total_tokens": 6243221504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.9121222496032715, "created_at": "2025-01-16T15:11:41.058192+00:00"} {"global_step": 11909, "acc_step": 0, "speed/wps": 12901.822181692518, "speed/FLOPS": 202640846325021.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05396965518593788, "optim/lr": 0.0028847349408118983, "optim/total_tokens": 6243745792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.9155995845794678, "created_at": "2025-01-16T15:11:51.220840+00:00"} {"global_step": 11910, "acc_step": 0, "speed/wps": 12903.00768465456, "speed/FLOPS": 202659466278091.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055136870592832565, "optim/lr": 0.002884702001195446, "optim/total_tokens": 6244270080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.8665990829467773, "created_at": "2025-01-16T15:12:01.383665+00:00"} {"global_step": 11911, "acc_step": 0, "speed/wps": 12907.051686962723, "speed/FLOPS": 202722982891382.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053221024572849274, "optim/lr": 0.0028846690570611603, "optim/total_tokens": 6244794368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8744354248046875, "created_at": "2025-01-16T15:12:11.544726+00:00"} {"global_step": 11912, "acc_step": 0, "speed/wps": 12901.077717756778, "speed/FLOPS": 202629153495909.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06059623509645462, "optim/lr": 0.002884636108409149, "optim/total_tokens": 6245318656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8365654945373535, "created_at": "2025-01-16T15:12:21.711032+00:00"} {"global_step": 11913, "acc_step": 0, "speed/wps": 12895.055257274418, "speed/FLOPS": 202534562478306.2, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05782610923051834, "optim/lr": 0.0028846031552395195, "optim/total_tokens": 6245842944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9572324752807617, "created_at": "2025-01-16T15:12:31.881316+00:00"} {"global_step": 11914, "acc_step": 0, "speed/wps": 12907.649596897809, "speed/FLOPS": 202732373888527.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043429985642433167, "optim/lr": 0.002884570197552378, "optim/total_tokens": 6246367232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 2.928748607635498, "created_at": "2025-01-16T15:12:42.040797+00:00"} {"global_step": 11915, "acc_step": 0, "speed/wps": 12900.088687871004, "speed/FLOPS": 202613619422484.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04399890452623367, "optim/lr": 0.002884537235347834, "optim/total_tokens": 6246891520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.859182834625244, "created_at": "2025-01-16T15:12:52.209321+00:00"} {"global_step": 11916, "acc_step": 0, "speed/wps": 12900.03144666813, "speed/FLOPS": 202612720370735.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04056611657142639, "optim/lr": 0.0028845042686259934, "optim/total_tokens": 6247415808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290845, "loss/out": 2.9424962997436523, "created_at": "2025-01-16T15:13:02.373473+00:00"} {"global_step": 11917, "acc_step": 0, "speed/wps": 12904.681311834433, "speed/FLOPS": 202685752892756.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04531204327940941, "optim/lr": 0.0028844712973869653, "optim/total_tokens": 6247940096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.8778188228607178, "created_at": "2025-01-16T15:13:12.541668+00:00"} {"global_step": 11918, "acc_step": 0, "speed/wps": 12903.052937979322, "speed/FLOPS": 202660177043741.66, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043831098824739456, "optim/lr": 0.0028844383216308556, "optim/total_tokens": 6248464384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.900996208190918, "created_at": "2025-01-16T15:13:22.704380+00:00"} {"global_step": 11919, "acc_step": 0, "speed/wps": 12906.466717074114, "speed/FLOPS": 202713795135449.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04256178066134453, "optim/lr": 0.0028844053413577733, "optim/total_tokens": 6248988672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 2.795351982116699, "created_at": "2025-01-16T15:13:32.863434+00:00"} {"global_step": 11920, "acc_step": 0, "speed/wps": 12897.350450565511, "speed/FLOPS": 202570611642868.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050526391714811325, "optim/lr": 0.0028843723565678257, "optim/total_tokens": 6249512960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8697400093078613, "created_at": "2025-01-16T15:13:43.032234+00:00"} {"global_step": 11921, "acc_step": 0, "speed/wps": 12906.107472534037, "speed/FLOPS": 202708152706293.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04271247610449791, "optim/lr": 0.0028843393672611196, "optim/total_tokens": 6250037248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 2.80928111076355, "created_at": "2025-01-16T15:13:53.195059+00:00"} {"global_step": 11922, "acc_step": 0, "speed/wps": 12905.512861649808, "speed/FLOPS": 202698813525276.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04541432484984398, "optim/lr": 0.0028843063734377635, "optim/total_tokens": 6250561536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.900674819946289, "created_at": "2025-01-16T15:14:03.356960+00:00"} {"global_step": 11923, "acc_step": 0, "speed/wps": 12898.29283555475, "speed/FLOPS": 202585413094097.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03861609473824501, "optim/lr": 0.002884273375097865, "optim/total_tokens": 6251085824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9454236030578613, "created_at": "2025-01-16T15:14:13.523448+00:00"} {"global_step": 11924, "acc_step": 0, "speed/wps": 12901.27147977468, "speed/FLOPS": 202632196794657.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04403846338391304, "optim/lr": 0.002884240372241531, "optim/total_tokens": 6251610112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302869, "loss/out": 2.792323589324951, "created_at": "2025-01-16T15:14:23.689388+00:00"} {"global_step": 11925, "acc_step": 0, "speed/wps": 12901.676702934828, "speed/FLOPS": 202638561381222.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047813791781663895, "optim/lr": 0.00288420736486887, "optim/total_tokens": 6252134400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9388186931610107, "created_at": "2025-01-16T15:14:33.852685+00:00"} {"global_step": 11926, "acc_step": 0, "speed/wps": 12899.988592364922, "speed/FLOPS": 202612047285015.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05518544837832451, "optim/lr": 0.0028841743529799896, "optim/total_tokens": 6252658688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.7826030254364014, "created_at": "2025-01-16T15:14:44.016977+00:00"} {"global_step": 11927, "acc_step": 0, "speed/wps": 12898.389674014543, "speed/FLOPS": 202586934075178.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050043366849422455, "optim/lr": 0.002884141336574997, "optim/total_tokens": 6253182976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.9073123931884766, "created_at": "2025-01-16T15:14:54.182320+00:00"} {"global_step": 11928, "acc_step": 0, "speed/wps": 12904.497057638036, "speed/FLOPS": 202682858927410.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055149633437395096, "optim/lr": 0.002884108315654, "optim/total_tokens": 6253707264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.913484573364258, "created_at": "2025-01-16T15:15:04.344351+00:00"} {"global_step": 11929, "acc_step": 0, "speed/wps": 12906.438158054298, "speed/FLOPS": 202713346576799.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06696531921625137, "optim/lr": 0.002884075290217107, "optim/total_tokens": 6254231552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 2.8641021251678467, "created_at": "2025-01-16T15:15:14.506815+00:00"} {"global_step": 11930, "acc_step": 0, "speed/wps": 12900.81407050041, "speed/FLOPS": 202625012553459.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05569285526871681, "optim/lr": 0.002884042260264425, "optim/total_tokens": 6254755840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8225085735321045, "created_at": "2025-01-16T15:15:24.670356+00:00"} {"global_step": 11931, "acc_step": 0, "speed/wps": 12902.64790752909, "speed/FLOPS": 202653815483950.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04844094067811966, "optim/lr": 0.0028840092257960627, "optim/total_tokens": 6255280128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.908926010131836, "created_at": "2025-01-16T15:15:34.832359+00:00"} {"global_step": 11932, "acc_step": 0, "speed/wps": 12905.300122027842, "speed/FLOPS": 202695472157178.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057108718901872635, "optim/lr": 0.0028839761868121264, "optim/total_tokens": 6255804416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 2.730634927749634, "created_at": "2025-01-16T15:15:44.996204+00:00"} {"global_step": 11933, "acc_step": 0, "speed/wps": 12898.874738133207, "speed/FLOPS": 202594552673710.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06504222750663757, "optim/lr": 0.0028839431433127253, "optim/total_tokens": 6256328704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9612069129943848, "created_at": "2025-01-16T15:15:55.163736+00:00"} {"global_step": 11934, "acc_step": 0, "speed/wps": 12905.558463909912, "speed/FLOPS": 202699529771435.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053964439779520035, "optim/lr": 0.002883910095297967, "optim/total_tokens": 6256852992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.860821485519409, "created_at": "2025-01-16T15:16:05.323495+00:00"} {"global_step": 11935, "acc_step": 0, "speed/wps": 12900.843737536818, "speed/FLOPS": 202625478515032.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0665409117937088, "optim/lr": 0.0028838770427679584, "optim/total_tokens": 6257377280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.896066665649414, "created_at": "2025-01-16T15:16:15.489296+00:00"} {"global_step": 11936, "acc_step": 0, "speed/wps": 12894.103979682171, "speed/FLOPS": 202519621356527.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04251375421881676, "optim/lr": 0.0028838439857228077, "optim/total_tokens": 6257901568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.8317830562591553, "created_at": "2025-01-16T15:16:25.658052+00:00"} {"global_step": 11937, "acc_step": 0, "speed/wps": 12904.037370392389, "speed/FLOPS": 202675638907540.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0643392875790596, "optim/lr": 0.0028838109241626237, "optim/total_tokens": 6258425856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 2.923191785812378, "created_at": "2025-01-16T15:16:35.819015+00:00"} {"global_step": 11938, "acc_step": 0, "speed/wps": 12898.031084645057, "speed/FLOPS": 202581301936377.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053983982652425766, "optim/lr": 0.002883777858087513, "optim/total_tokens": 6258950144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.797574996948242, "created_at": "2025-01-16T15:16:45.985397+00:00"} {"global_step": 11939, "acc_step": 0, "speed/wps": 12903.898507419563, "speed/FLOPS": 202673457873734.8, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.078174889087677, "optim/lr": 0.0028837447874975847, "optim/total_tokens": 6259474432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.004924774169922, "created_at": "2025-01-16T15:16:56.150534+00:00"} {"global_step": 11940, "acc_step": 0, "speed/wps": 12906.181253285189, "speed/FLOPS": 202709311534376.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07029812037944794, "optim/lr": 0.0028837117123929456, "optim/total_tokens": 6259998720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305632, "loss/out": 2.807459831237793, "created_at": "2025-01-16T15:17:06.309709+00:00"} {"global_step": 11941, "acc_step": 0, "speed/wps": 12902.297552818209, "speed/FLOPS": 202648312681778.38, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.052667077630758286, "optim/lr": 0.002883678632773704, "optim/total_tokens": 6260523008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.9399185180664062, "created_at": "2025-01-16T15:17:16.479195+00:00"} {"global_step": 11942, "acc_step": 0, "speed/wps": 12902.918525153407, "speed/FLOPS": 202658065905608.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05153373256325722, "optim/lr": 0.0028836455486399674, "optim/total_tokens": 6261047296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353101, "loss/out": 2.6927549839019775, "created_at": "2025-01-16T15:17:26.642342+00:00"} {"global_step": 11943, "acc_step": 0, "speed/wps": 12904.210992356479, "speed/FLOPS": 202678365879068.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05644472315907478, "optim/lr": 0.0028836124599918456, "optim/total_tokens": 6261571584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.8174381256103516, "created_at": "2025-01-16T15:17:36.806408+00:00"} {"global_step": 11944, "acc_step": 0, "speed/wps": 12899.912347337715, "speed/FLOPS": 202610849752091.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05285944417119026, "optim/lr": 0.0028835793668294443, "optim/total_tokens": 6262095872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8921854496002197, "created_at": "2025-01-16T15:17:46.972805+00:00"} {"global_step": 11945, "acc_step": 0, "speed/wps": 12898.426117663546, "speed/FLOPS": 202587506472765.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05048810690641403, "optim/lr": 0.002883546269152872, "optim/total_tokens": 6262620160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 2.927588939666748, "created_at": "2025-01-16T15:17:57.143906+00:00"} {"global_step": 11946, "acc_step": 0, "speed/wps": 12907.470184836506, "speed/FLOPS": 202729555975567.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04302944242954254, "optim/lr": 0.0028835131669622373, "optim/total_tokens": 6263144448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.7543323040008545, "created_at": "2025-01-16T15:18:07.307009+00:00"} {"global_step": 11947, "acc_step": 0, "speed/wps": 12901.690414240444, "speed/FLOPS": 202638776736118.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044800885021686554, "optim/lr": 0.002883480060257648, "optim/total_tokens": 6263668736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.900846481323242, "created_at": "2025-01-16T15:18:17.469885+00:00"} {"global_step": 11948, "acc_step": 0, "speed/wps": 12898.043018665736, "speed/FLOPS": 202581489376571.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04444795101881027, "optim/lr": 0.0028834469490392127, "optim/total_tokens": 6264193024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 2.9818906784057617, "created_at": "2025-01-16T15:18:27.638135+00:00"} {"global_step": 11949, "acc_step": 0, "speed/wps": 12902.598276800185, "speed/FLOPS": 202653035965153.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045937683433294296, "optim/lr": 0.002883413833307038, "optim/total_tokens": 6264717312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.894529104232788, "created_at": "2025-01-16T15:18:37.802910+00:00"} {"global_step": 11950, "acc_step": 0, "speed/wps": 12898.086012529511, "speed/FLOPS": 202582164654281.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05238831043243408, "optim/lr": 0.0028833807130612334, "optim/total_tokens": 6265241600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483427, "loss/out": 2.8699021339416504, "created_at": "2025-01-16T15:18:47.970599+00:00"} {"global_step": 11951, "acc_step": 0, "speed/wps": 12899.873868671184, "speed/FLOPS": 202610245391758.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044528137892484665, "optim/lr": 0.0028833475883019055, "optim/total_tokens": 6265765888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9068243503570557, "created_at": "2025-01-16T15:18:58.140790+00:00"} {"global_step": 11952, "acc_step": 0, "speed/wps": 12901.906015431976, "speed/FLOPS": 202642163049098.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.11643268167972565, "optim/lr": 0.0028833144590291637, "optim/total_tokens": 6266290176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8696744441986084, "created_at": "2025-01-16T15:19:08.303453+00:00"} {"global_step": 11953, "acc_step": 0, "speed/wps": 12893.879016351628, "speed/FLOPS": 202516087998289.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07613058388233185, "optim/lr": 0.0028832813252431156, "optim/total_tokens": 6266814464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.796566963195801, "created_at": "2025-01-16T15:19:18.472455+00:00"} {"global_step": 11954, "acc_step": 0, "speed/wps": 12902.469474318992, "speed/FLOPS": 202651012945193.1, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06599757075309753, "optim/lr": 0.002883248186943869, "optim/total_tokens": 6267338752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.7578744888305664, "created_at": "2025-01-16T15:19:28.634611+00:00"} {"global_step": 11955, "acc_step": 0, "speed/wps": 12902.855927331571, "speed/FLOPS": 202657082720797.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0535443015396595, "optim/lr": 0.0028832150441315324, "optim/total_tokens": 6267863040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.9560890197753906, "created_at": "2025-01-16T15:19:38.797428+00:00"} {"global_step": 11956, "acc_step": 0, "speed/wps": 12893.018023959827, "speed/FLOPS": 202502564929648.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052647121250629425, "optim/lr": 0.0028831818968062138, "optim/total_tokens": 6268387328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 2.7481627464294434, "created_at": "2025-01-16T15:19:48.967078+00:00"} {"global_step": 11957, "acc_step": 0, "speed/wps": 12896.95176294903, "speed/FLOPS": 202564349705997.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04391442611813545, "optim/lr": 0.0028831487449680216, "optim/total_tokens": 6268911616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9159772396087646, "created_at": "2025-01-16T15:19:59.135397+00:00"} {"global_step": 11958, "acc_step": 0, "speed/wps": 12902.998594753297, "speed/FLOPS": 202659323508701.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05176202207803726, "optim/lr": 0.002883115588617064, "optim/total_tokens": 6269435904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8323845863342285, "created_at": "2025-01-16T15:20:09.297115+00:00"} {"global_step": 11959, "acc_step": 0, "speed/wps": 12893.636463796973, "speed/FLOPS": 202512278377116.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044246770441532135, "optim/lr": 0.002883082427753448, "optim/total_tokens": 6269960192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365596, "loss/out": 2.819056510925293, "created_at": "2025-01-16T15:20:19.467129+00:00"} {"global_step": 11960, "acc_step": 0, "speed/wps": 12904.257959105686, "speed/FLOPS": 202679103556403.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06571027636528015, "optim/lr": 0.0028830492623772835, "optim/total_tokens": 6270484480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.764437675476074, "created_at": "2025-01-16T15:20:29.638496+00:00"} {"global_step": 11961, "acc_step": 0, "speed/wps": 12901.471196510096, "speed/FLOPS": 202635333620426.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.054703906178474426, "optim/lr": 0.0028830160924886773, "optim/total_tokens": 6271008768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.966549873352051, "created_at": "2025-01-16T15:20:39.807715+00:00"} {"global_step": 11962, "acc_step": 0, "speed/wps": 12899.651747331958, "speed/FLOPS": 202606756670899.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0512295737862587, "optim/lr": 0.0028829829180877387, "optim/total_tokens": 6271533056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.887036085128784, "created_at": "2025-01-16T15:20:49.977926+00:00"} {"global_step": 11963, "acc_step": 0, "speed/wps": 12904.432163237276, "speed/FLOPS": 202681839671671.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050233013927936554, "optim/lr": 0.002882949739174575, "optim/total_tokens": 6272057344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.84260630607605, "created_at": "2025-01-16T15:21:00.140883+00:00"} {"global_step": 11964, "acc_step": 0, "speed/wps": 12903.596639130155, "speed/FLOPS": 202668716617439.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044089436531066895, "optim/lr": 0.0028829165557492952, "optim/total_tokens": 6272581632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8597586154937744, "created_at": "2025-01-16T15:21:10.304158+00:00"} {"global_step": 11965, "acc_step": 0, "speed/wps": 12903.270859647375, "speed/FLOPS": 202663599802994.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0475044809281826, "optim/lr": 0.002882883367812007, "optim/total_tokens": 6273105920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.7636215686798096, "created_at": "2025-01-16T15:21:20.468415+00:00"} {"global_step": 11966, "acc_step": 0, "speed/wps": 12899.916035238437, "speed/FLOPS": 202610907675639.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04697505757212639, "optim/lr": 0.0028828501753628195, "optim/total_tokens": 6273630208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.883662223815918, "created_at": "2025-01-16T15:21:30.635399+00:00"} {"global_step": 11967, "acc_step": 0, "speed/wps": 12901.21241745797, "speed/FLOPS": 202631269139813.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.04824208840727806, "optim/lr": 0.0028828169784018403, "optim/total_tokens": 6274154496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461122, "loss/out": 2.827998638153076, "created_at": "2025-01-16T15:21:40.804690+00:00"} {"global_step": 11968, "acc_step": 0, "speed/wps": 12898.39147556412, "speed/FLOPS": 202586962370990.3, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04390774667263031, "optim/lr": 0.002882783776929178, "optim/total_tokens": 6274678784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.8682403564453125, "created_at": "2025-01-16T15:21:50.973670+00:00"} {"global_step": 11969, "acc_step": 0, "speed/wps": 12905.137743396308, "speed/FLOPS": 202692921777638.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045334771275520325, "optim/lr": 0.0028827505709449405, "optim/total_tokens": 6275203072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.81233286857605, "created_at": "2025-01-16T15:22:01.135162+00:00"} {"global_step": 11970, "acc_step": 0, "speed/wps": 12904.74645604753, "speed/FLOPS": 202686776072140.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05078599974513054, "optim/lr": 0.0028827173604492363, "optim/total_tokens": 6275727360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8692469596862793, "created_at": "2025-01-16T15:22:11.297790+00:00"} {"global_step": 11971, "acc_step": 0, "speed/wps": 12902.126955908223, "speed/FLOPS": 202645633222882.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04686516895890236, "optim/lr": 0.0028826841454421747, "optim/total_tokens": 6276251648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8319032192230225, "created_at": "2025-01-16T15:22:21.460380+00:00"} {"global_step": 11972, "acc_step": 0, "speed/wps": 12901.64625265742, "speed/FLOPS": 202638083117773.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04504016786813736, "optim/lr": 0.0028826509259238624, "optim/total_tokens": 6276775936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464317, "loss/out": 2.7701363563537598, "created_at": "2025-01-16T15:22:31.624503+00:00"} {"global_step": 11973, "acc_step": 0, "speed/wps": 12893.828184361606, "speed/FLOPS": 202515289612035.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056134384125471115, "optim/lr": 0.0028826177018944094, "optim/total_tokens": 6277300224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9325249195098877, "created_at": "2025-01-16T15:22:41.795757+00:00"} {"global_step": 11974, "acc_step": 0, "speed/wps": 12903.898596154235, "speed/FLOPS": 202673459267434.78, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04202132672071457, "optim/lr": 0.002882584473353923, "optim/total_tokens": 6277824512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.826808452606201, "created_at": "2025-01-16T15:22:51.956998+00:00"} {"global_step": 11975, "acc_step": 0, "speed/wps": 12900.953599118133, "speed/FLOPS": 202627204042132.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04736890643835068, "optim/lr": 0.002882551240302512, "optim/total_tokens": 6278348800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.970350742340088, "created_at": "2025-01-16T15:23:02.121411+00:00"} {"global_step": 11976, "acc_step": 0, "speed/wps": 12899.521173430752, "speed/FLOPS": 202604705828353.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.045320041477680206, "optim/lr": 0.0028825180027402847, "optim/total_tokens": 6278873088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.8414759635925293, "created_at": "2025-01-16T15:23:12.291386+00:00"} {"global_step": 11977, "acc_step": 0, "speed/wps": 12900.531717198206, "speed/FLOPS": 202620577806853.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04630697891116142, "optim/lr": 0.0028824847606673497, "optim/total_tokens": 6279397376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.849694013595581, "created_at": "2025-01-16T15:23:22.455951+00:00"} {"global_step": 11978, "acc_step": 0, "speed/wps": 12894.402692698233, "speed/FLOPS": 202524313054919.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05774610489606857, "optim/lr": 0.0028824515140838157, "optim/total_tokens": 6279921664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 3.010683059692383, "created_at": "2025-01-16T15:23:32.624987+00:00"} {"global_step": 11979, "acc_step": 0, "speed/wps": 12888.80013289764, "speed/FLOPS": 202436317154527.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06046457216143608, "optim/lr": 0.0028824182629897904, "optim/total_tokens": 6280445952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.017390251159668, "created_at": "2025-01-16T15:23:42.802743+00:00"} {"global_step": 11980, "acc_step": 0, "speed/wps": 12886.98551471557, "speed/FLOPS": 202407816082431.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.048399727791547775, "optim/lr": 0.0028823850073853834, "optim/total_tokens": 6280970240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.968860149383545, "created_at": "2025-01-16T15:23:52.982774+00:00"} {"global_step": 11981, "acc_step": 0, "speed/wps": 12899.049567770358, "speed/FLOPS": 202597298613402.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06096496433019638, "optim/lr": 0.0028823517472707015, "optim/total_tokens": 6281494528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 2.77494478225708, "created_at": "2025-01-16T15:24:03.178142+00:00"} {"global_step": 11982, "acc_step": 0, "speed/wps": 12903.119327999017, "speed/FLOPS": 202661219790230.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.071834035217762, "optim/lr": 0.002882318482645855, "optim/total_tokens": 6282018816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 2.6992907524108887, "created_at": "2025-01-16T15:24:13.342479+00:00"} {"global_step": 11983, "acc_step": 0, "speed/wps": 12906.141311044885, "speed/FLOPS": 202708684186605.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.03891254961490631, "optim/lr": 0.0028822852135109516, "optim/total_tokens": 6282543104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.749119281768799, "created_at": "2025-01-16T15:24:23.503419+00:00"} {"global_step": 11984, "acc_step": 0, "speed/wps": 12907.810719941492, "speed/FLOPS": 202734904547335.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062492333352565765, "optim/lr": 0.0028822519398661, "optim/total_tokens": 6283067392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 2.872098207473755, "created_at": "2025-01-16T15:24:33.662722+00:00"} {"global_step": 11985, "acc_step": 0, "speed/wps": 12903.01242834836, "speed/FLOPS": 202659540784321.34, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04830261692404747, "optim/lr": 0.0028822186617114085, "optim/total_tokens": 6283591680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8789916038513184, "created_at": "2025-01-16T15:24:43.828973+00:00"} {"global_step": 11986, "acc_step": 0, "speed/wps": 12909.755389534832, "speed/FLOPS": 202765448255554.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05727043375372887, "optim/lr": 0.0028821853790469863, "optim/total_tokens": 6284115968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.804680585861206, "created_at": "2025-01-16T15:24:53.985546+00:00"} {"global_step": 11987, "acc_step": 0, "speed/wps": 12899.288102093828, "speed/FLOPS": 202601045122733.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06273610889911652, "optim/lr": 0.002882152091872942, "optim/total_tokens": 6284640256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306165, "loss/out": 2.920586109161377, "created_at": "2025-01-16T15:25:04.151771+00:00"} {"global_step": 11988, "acc_step": 0, "speed/wps": 12906.739645605447, "speed/FLOPS": 202718081853077.12, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06906748563051224, "optim/lr": 0.0028821188001893826, "optim/total_tokens": 6285164544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.827138900756836, "created_at": "2025-01-16T15:25:14.310631+00:00"} {"global_step": 11989, "acc_step": 0, "speed/wps": 12902.176807745764, "speed/FLOPS": 202646416214494.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08219514042139053, "optim/lr": 0.0028820855039964188, "optim/total_tokens": 6285688832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.888688564300537, "created_at": "2025-01-16T15:25:24.473382+00:00"} {"global_step": 11990, "acc_step": 0, "speed/wps": 12902.7150173229, "speed/FLOPS": 202654869535480.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05699002370238304, "optim/lr": 0.0028820522032941584, "optim/total_tokens": 6286213120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.854609489440918, "created_at": "2025-01-16T15:25:34.638799+00:00"} {"global_step": 11991, "acc_step": 0, "speed/wps": 12905.959169136542, "speed/FLOPS": 202705823397645.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07440535724163055, "optim/lr": 0.0028820188980827094, "optim/total_tokens": 6286737408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.7839388847351074, "created_at": "2025-01-16T15:25:44.798257+00:00"} {"global_step": 11992, "acc_step": 0, "speed/wps": 12896.173458844574, "speed/FLOPS": 202552125370535.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.07144778966903687, "optim/lr": 0.0028819855883621816, "optim/total_tokens": 6287261696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 2.832197666168213, "created_at": "2025-01-16T15:25:54.966183+00:00"} {"global_step": 11993, "acc_step": 0, "speed/wps": 12892.327083088432, "speed/FLOPS": 202491712753811.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0491439513862133, "optim/lr": 0.0028819522741326824, "optim/total_tokens": 6287785984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.898125648498535, "created_at": "2025-01-16T15:26:05.142929+00:00"} {"global_step": 11994, "acc_step": 0, "speed/wps": 12900.67041666253, "speed/FLOPS": 202622756272534.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06448324024677277, "optim/lr": 0.002881918955394322, "optim/total_tokens": 6288310272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.8561131954193115, "created_at": "2025-01-16T15:26:15.310964+00:00"} {"global_step": 11995, "acc_step": 0, "speed/wps": 12899.737705955784, "speed/FLOPS": 202608106769205.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04609384760260582, "optim/lr": 0.002881885632147208, "optim/total_tokens": 6288834560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.7305850982666016, "created_at": "2025-01-16T15:26:25.475869+00:00"} {"global_step": 11996, "acc_step": 0, "speed/wps": 12900.465427118648, "speed/FLOPS": 202619536630062.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055340658873319626, "optim/lr": 0.0028818523043914494, "optim/total_tokens": 6289358848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8010480403900146, "created_at": "2025-01-16T15:26:35.643375+00:00"} {"global_step": 11997, "acc_step": 0, "speed/wps": 12899.765751740255, "speed/FLOPS": 202608547266789.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0645107850432396, "optim/lr": 0.002881818972127155, "optim/total_tokens": 6289883136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8039350509643555, "created_at": "2025-01-16T15:26:45.811640+00:00"} {"global_step": 11998, "acc_step": 0, "speed/wps": 12901.6161490579, "speed/FLOPS": 202637610299375.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06908578425645828, "optim/lr": 0.002881785635354434, "optim/total_tokens": 6290407424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.830282211303711, "created_at": "2025-01-16T15:26:55.976998+00:00"} {"global_step": 11999, "acc_step": 0, "speed/wps": 12898.222561201617, "speed/FLOPS": 202584309338814.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04441377520561218, "optim/lr": 0.0028817522940733946, "optim/total_tokens": 6290931712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.9579529762268066, "created_at": "2025-01-16T15:27:06.145194+00:00"} {"global_step": 12000, "acc_step": 0, "speed/wps": 12911.123105567021, "speed/FLOPS": 202786930115278.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07810570299625397, "optim/lr": 0.0028817189482841453, "optim/total_tokens": 6291456000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450473, "loss/out": 2.8669800758361816, "created_at": "2025-01-16T15:27:16.307542+00:00"} {"global_step": 12001, "acc_step": 0, "speed/wps": 12605.25344892779, "speed/FLOPS": 197982826848806.25, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.2439, "optim/grad_norm": 0.04598310589790344, "optim/lr": 0.002881685597986795, "optim/total_tokens": 6291980288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.88397216796875, "created_at": "2025-01-16T15:27:26.710580+00:00"} {"global_step": 12002, "acc_step": 0, "speed/wps": 12913.375368376846, "speed/FLOPS": 202822304997640.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07577483355998993, "optim/lr": 0.002881652243181454, "optim/total_tokens": 6292504576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.86187744140625, "created_at": "2025-01-16T15:27:36.867194+00:00"} {"global_step": 12003, "acc_step": 0, "speed/wps": 12909.384459198018, "speed/FLOPS": 202759622284903.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08452897518873215, "optim/lr": 0.0028816188838682293, "optim/total_tokens": 6293028864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 2.9556562900543213, "created_at": "2025-01-16T15:27:47.026452+00:00"} {"global_step": 12004, "acc_step": 0, "speed/wps": 12907.578582479917, "speed/FLOPS": 202731258509510.75, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07199283689260483, "optim/lr": 0.0028815855200472304, "optim/total_tokens": 6293553152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8871965408325195, "created_at": "2025-01-16T15:27:57.191188+00:00"} {"global_step": 12005, "acc_step": 0, "speed/wps": 12907.102798459837, "speed/FLOPS": 202723785667679.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08355437964200974, "optim/lr": 0.002881552151718566, "optim/total_tokens": 6294077440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.0238137245178223, "created_at": "2025-01-16T15:28:07.349888+00:00"} {"global_step": 12006, "acc_step": 0, "speed/wps": 12914.393375325268, "speed/FLOPS": 202838294195653.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06181849166750908, "optim/lr": 0.0028815187788823454, "optim/total_tokens": 6294601728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8803422451019287, "created_at": "2025-01-16T15:28:17.502901+00:00"} {"global_step": 12007, "acc_step": 0, "speed/wps": 12908.313748802599, "speed/FLOPS": 202742805306832.84, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06897681206464767, "optim/lr": 0.002881485401538677, "optim/total_tokens": 6295126016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.7760472297668457, "created_at": "2025-01-16T15:28:27.660445+00:00"} {"global_step": 12008, "acc_step": 0, "speed/wps": 12906.026609260007, "speed/FLOPS": 202706882637457.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06210821121931076, "optim/lr": 0.00288145201968767, "optim/total_tokens": 6295650304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.9069857597351074, "created_at": "2025-01-16T15:28:37.820664+00:00"} {"global_step": 12009, "acc_step": 0, "speed/wps": 12908.956459448953, "speed/FLOPS": 202752899960710.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06678157299757004, "optim/lr": 0.002881418633329433, "optim/total_tokens": 6296174592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.899125099182129, "created_at": "2025-01-16T15:28:47.977842+00:00"} {"global_step": 12010, "acc_step": 0, "speed/wps": 12908.208166009885, "speed/FLOPS": 202741146983987.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04531975835561752, "optim/lr": 0.002881385242464075, "optim/total_tokens": 6296698880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.811983108520508, "created_at": "2025-01-16T15:28:58.141092+00:00"} {"global_step": 12011, "acc_step": 0, "speed/wps": 12913.425858144356, "speed/FLOPS": 202823098008819.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06036973372101784, "optim/lr": 0.0028813518470917053, "optim/total_tokens": 6297223168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.9093375205993652, "created_at": "2025-01-16T15:29:08.297557+00:00"} {"global_step": 12012, "acc_step": 0, "speed/wps": 12905.697854930768, "speed/FLOPS": 202701719098960.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04635929688811302, "optim/lr": 0.0028813184472124332, "optim/total_tokens": 6297747456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370389, "loss/out": 2.929659366607666, "created_at": "2025-01-16T15:29:18.459157+00:00"} {"global_step": 12013, "acc_step": 0, "speed/wps": 12907.32525105398, "speed/FLOPS": 202727279591352.56, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056348949670791626, "optim/lr": 0.002881285042826366, "optim/total_tokens": 6298271744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8259410858154297, "created_at": "2025-01-16T15:29:28.621719+00:00"} {"global_step": 12014, "acc_step": 0, "speed/wps": 12906.636468843286, "speed/FLOPS": 202716461320247.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04729150980710983, "optim/lr": 0.0028812516339336144, "optim/total_tokens": 6298796032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 2.7627243995666504, "created_at": "2025-01-16T15:29:38.782009+00:00"} {"global_step": 12015, "acc_step": 0, "speed/wps": 12909.751365613609, "speed/FLOPS": 202765385054342.28, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051243606954813004, "optim/lr": 0.002881218220534287, "optim/total_tokens": 6299320320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.7678580284118652, "created_at": "2025-01-16T15:29:48.941630+00:00"} {"global_step": 12016, "acc_step": 0, "speed/wps": 12907.557659964044, "speed/FLOPS": 202730929892648.4, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05777228996157646, "optim/lr": 0.0028811848026284925, "optim/total_tokens": 6299844608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.86370849609375, "created_at": "2025-01-16T15:29:59.099933+00:00"} {"global_step": 12017, "acc_step": 0, "speed/wps": 12907.539625979389, "speed/FLOPS": 202730646644138.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06259272247552872, "optim/lr": 0.0028811513802163395, "optim/total_tokens": 6300368896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8950400352478027, "created_at": "2025-01-16T15:30:09.260272+00:00"} {"global_step": 12018, "acc_step": 0, "speed/wps": 12910.805023508156, "speed/FLOPS": 202781934199453.8, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05108862370252609, "optim/lr": 0.0028811179532979383, "optim/total_tokens": 6300893184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.9237124919891357, "created_at": "2025-01-16T15:30:19.416086+00:00"} {"global_step": 12019, "acc_step": 0, "speed/wps": 12910.620620457672, "speed/FLOPS": 202779037896149.62, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08856523782014847, "optim/lr": 0.0028810845218733973, "optim/total_tokens": 6301417472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.9037024974823, "created_at": "2025-01-16T15:30:29.577076+00:00"} {"global_step": 12020, "acc_step": 0, "speed/wps": 12908.686998648855, "speed/FLOPS": 202748667708567.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057158391922712326, "optim/lr": 0.002881051085942825, "optim/total_tokens": 6301941760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8493857383728027, "created_at": "2025-01-16T15:30:39.736036+00:00"} {"global_step": 12021, "acc_step": 0, "speed/wps": 12911.257406295943, "speed/FLOPS": 202789039492775.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06117093935608864, "optim/lr": 0.002881017645506331, "optim/total_tokens": 6302466048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9333925247192383, "created_at": "2025-01-16T15:30:49.891394+00:00"} {"global_step": 12022, "acc_step": 0, "speed/wps": 12912.989228079972, "speed/FLOPS": 202816240133665.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050440557301044464, "optim/lr": 0.002880984200564025, "optim/total_tokens": 6302990336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9153342247009277, "created_at": "2025-01-16T15:31:00.050101+00:00"} {"global_step": 12023, "acc_step": 0, "speed/wps": 12900.432417729842, "speed/FLOPS": 202619018172251.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05461768060922623, "optim/lr": 0.002880950751116015, "optim/total_tokens": 6303514624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.8185906410217285, "created_at": "2025-01-16T15:31:10.217260+00:00"} {"global_step": 12024, "acc_step": 0, "speed/wps": 12908.269209181095, "speed/FLOPS": 202742105750873.28, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043004218488931656, "optim/lr": 0.002880917297162411, "optim/total_tokens": 6304038912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.7354929447174072, "created_at": "2025-01-16T15:31:20.377770+00:00"} {"global_step": 12025, "acc_step": 0, "speed/wps": 12911.695204495674, "speed/FLOPS": 202795915715099.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04732652008533478, "optim/lr": 0.002880883838703322, "optim/total_tokens": 6304563200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412610, "loss/out": 2.856966733932495, "created_at": "2025-01-16T15:31:30.537829+00:00"} {"global_step": 12026, "acc_step": 0, "speed/wps": 12898.006703660758, "speed/FLOPS": 202580918999516.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056410372257232666, "optim/lr": 0.002880850375738857, "optim/total_tokens": 6305087488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.900944709777832, "created_at": "2025-01-16T15:31:40.704672+00:00"} {"global_step": 12027, "acc_step": 0, "speed/wps": 12908.556840865345, "speed/FLOPS": 202746623401721.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04321815446019173, "optim/lr": 0.002880816908269125, "optim/total_tokens": 6305611776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.943960189819336, "created_at": "2025-01-16T15:31:50.862589+00:00"} {"global_step": 12028, "acc_step": 0, "speed/wps": 12911.507400119739, "speed/FLOPS": 202792965989305.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05628998205065727, "optim/lr": 0.0028807834362942347, "optim/total_tokens": 6306136064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8420238494873047, "created_at": "2025-01-16T15:32:01.018512+00:00"} {"global_step": 12029, "acc_step": 0, "speed/wps": 12911.674268632023, "speed/FLOPS": 202795586888592.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043461721390485764, "optim/lr": 0.0028807499598142968, "optim/total_tokens": 6306660352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 2.757378578186035, "created_at": "2025-01-16T15:32:11.180075+00:00"} {"global_step": 12030, "acc_step": 0, "speed/wps": 12908.254550953427, "speed/FLOPS": 202741875523265.22, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05202517285943031, "optim/lr": 0.0028807164788294194, "optim/total_tokens": 6307184640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.872157573699951, "created_at": "2025-01-16T15:32:21.340190+00:00"} {"global_step": 12031, "acc_step": 0, "speed/wps": 12907.39000646062, "speed/FLOPS": 202728296663997.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05071553960442543, "optim/lr": 0.002880682993339712, "optim/total_tokens": 6307708928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 2.913327932357788, "created_at": "2025-01-16T15:32:31.501431+00:00"} {"global_step": 12032, "acc_step": 0, "speed/wps": 12910.402321371614, "speed/FLOPS": 202775609209028.7, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05929579585790634, "optim/lr": 0.002880649503345284, "optim/total_tokens": 6308233216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299645, "loss/out": 2.76957368850708, "created_at": "2025-01-16T15:32:41.659371+00:00"} {"global_step": 12033, "acc_step": 0, "speed/wps": 12908.693509507752, "speed/FLOPS": 202748769970553.16, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052875276654958725, "optim/lr": 0.002880616008846245, "optim/total_tokens": 6308757504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.782073736190796, "created_at": "2025-01-16T15:32:51.817810+00:00"} {"global_step": 12034, "acc_step": 0, "speed/wps": 12910.324820996735, "speed/FLOPS": 202774391959144.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05000670254230499, "optim/lr": 0.0028805825098427032, "optim/total_tokens": 6309281792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 279426, "loss/out": 2.982893466949463, "created_at": "2025-01-16T15:33:01.979116+00:00"} {"global_step": 12035, "acc_step": 0, "speed/wps": 12911.046144933982, "speed/FLOPS": 202785721342781.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04843071475625038, "optim/lr": 0.0028805490063347687, "optim/total_tokens": 6309806080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.919739246368408, "created_at": "2025-01-16T15:33:12.138926+00:00"} {"global_step": 12036, "acc_step": 0, "speed/wps": 12902.018801206827, "speed/FLOPS": 202643934504676.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04307061806321144, "optim/lr": 0.002880515498322551, "optim/total_tokens": 6310330368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 2.8501229286193848, "created_at": "2025-01-16T15:33:22.301878+00:00"} {"global_step": 12037, "acc_step": 0, "speed/wps": 12897.844792197051, "speed/FLOPS": 202578375957488.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05243171751499176, "optim/lr": 0.0028804819858061588, "optim/total_tokens": 6310854656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.824462890625, "created_at": "2025-01-16T15:33:32.468761+00:00"} {"global_step": 12038, "acc_step": 0, "speed/wps": 12902.558268017825, "speed/FLOPS": 202652407572248.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06502410769462585, "optim/lr": 0.0028804484687857015, "optim/total_tokens": 6311378944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.859299421310425, "created_at": "2025-01-16T15:33:42.630979+00:00"} {"global_step": 12039, "acc_step": 0, "speed/wps": 12904.580568800218, "speed/FLOPS": 202684170584968.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053073834627866745, "optim/lr": 0.002880414947261289, "optim/total_tokens": 6311903232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 2.945307731628418, "created_at": "2025-01-16T15:33:52.799346+00:00"} {"global_step": 12040, "acc_step": 0, "speed/wps": 12904.05631863814, "speed/FLOPS": 202675936515777.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0631687194108963, "optim/lr": 0.0028803814212330302, "optim/total_tokens": 6312427520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.792975664138794, "created_at": "2025-01-16T15:34:02.961771+00:00"} {"global_step": 12041, "acc_step": 0, "speed/wps": 12908.921277670042, "speed/FLOPS": 202752347382528.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060275059193372726, "optim/lr": 0.0028803478907010343, "optim/total_tokens": 6312951808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.90360426902771, "created_at": "2025-01-16T15:34:13.121444+00:00"} {"global_step": 12042, "acc_step": 0, "speed/wps": 12906.042280257065, "speed/FLOPS": 202707128772000.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07124044746160507, "optim/lr": 0.0028803143556654118, "optim/total_tokens": 6313476096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.798222780227661, "created_at": "2025-01-16T15:34:23.286509+00:00"} {"global_step": 12043, "acc_step": 0, "speed/wps": 12909.451744476302, "speed/FLOPS": 202760679092658.28, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056297656148672104, "optim/lr": 0.0028802808161262706, "optim/total_tokens": 6314000384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388560, "loss/out": 2.8315656185150146, "created_at": "2025-01-16T15:34:33.446561+00:00"} {"global_step": 12044, "acc_step": 0, "speed/wps": 12905.212418338393, "speed/FLOPS": 202694094650217.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06425531208515167, "optim/lr": 0.0028802472720837213, "optim/total_tokens": 6314524672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.9117560386657715, "created_at": "2025-01-16T15:34:43.606640+00:00"} {"global_step": 12045, "acc_step": 0, "speed/wps": 12905.053827420028, "speed/FLOPS": 202691603761918.16, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05166188254952431, "optim/lr": 0.002880213723537873, "optim/total_tokens": 6315048960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9253034591674805, "created_at": "2025-01-16T15:34:53.768578+00:00"} {"global_step": 12046, "acc_step": 0, "speed/wps": 12906.096326019946, "speed/FLOPS": 202707977634973.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05344969034194946, "optim/lr": 0.0028801801704888344, "optim/total_tokens": 6315573248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8895716667175293, "created_at": "2025-01-16T15:35:03.928853+00:00"} {"global_step": 12047, "acc_step": 0, "speed/wps": 12910.30921420912, "speed/FLOPS": 202774146833099.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055796053260564804, "optim/lr": 0.0028801466129367162, "optim/total_tokens": 6316097536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8264248371124268, "created_at": "2025-01-16T15:35:14.089509+00:00"} {"global_step": 12048, "acc_step": 0, "speed/wps": 12904.96521236819, "speed/FLOPS": 202690211940759.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051514092832803726, "optim/lr": 0.002880113050881627, "optim/total_tokens": 6316621824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 2.8852977752685547, "created_at": "2025-01-16T15:35:24.250708+00:00"} {"global_step": 12049, "acc_step": 0, "speed/wps": 12907.83035643262, "speed/FLOPS": 202735212965411.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060122378170490265, "optim/lr": 0.0028800794843236773, "optim/total_tokens": 6317146112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8688297271728516, "created_at": "2025-01-16T15:35:34.408898+00:00"} {"global_step": 12050, "acc_step": 0, "speed/wps": 12905.53862604693, "speed/FLOPS": 202699218190536.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046108927577733994, "optim/lr": 0.0028800459132629756, "optim/total_tokens": 6317670400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.948232889175415, "created_at": "2025-01-16T15:35:44.572251+00:00"} {"global_step": 12051, "acc_step": 0, "speed/wps": 12908.616691500029, "speed/FLOPS": 202747563438182.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06099443882703781, "optim/lr": 0.0028800123376996316, "optim/total_tokens": 6318194688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.743741512298584, "created_at": "2025-01-16T15:35:54.733217+00:00"} {"global_step": 12052, "acc_step": 0, "speed/wps": 12910.967093522642, "speed/FLOPS": 202784479731738.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06543532013893127, "optim/lr": 0.002879978757633755, "optim/total_tokens": 6318718976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376441, "loss/out": 3.005347728729248, "created_at": "2025-01-16T15:36:04.892722+00:00"} {"global_step": 12053, "acc_step": 0, "speed/wps": 12896.038403486018, "speed/FLOPS": 202550004140543.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05073011666536331, "optim/lr": 0.0028799451730654557, "optim/total_tokens": 6319243264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.907217025756836, "created_at": "2025-01-16T15:36:15.060703+00:00"} {"global_step": 12054, "acc_step": 0, "speed/wps": 12896.209180939422, "speed/FLOPS": 202552686435122.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05356213450431824, "optim/lr": 0.0028799115839948427, "optim/total_tokens": 6319767552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416629, "loss/out": 2.7586324214935303, "created_at": "2025-01-16T15:36:25.228651+00:00"} {"global_step": 12055, "acc_step": 0, "speed/wps": 12902.387971754857, "speed/FLOPS": 202649732835427.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052450090646743774, "optim/lr": 0.002879877990422026, "optim/total_tokens": 6320291840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.7435688972473145, "created_at": "2025-01-16T15:36:35.391458+00:00"} {"global_step": 12056, "acc_step": 0, "speed/wps": 12899.92178406384, "speed/FLOPS": 202610997968842.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05842500180006027, "optim/lr": 0.0028798443923471148, "optim/total_tokens": 6320816128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.867281675338745, "created_at": "2025-01-16T15:36:45.557143+00:00"} {"global_step": 12057, "acc_step": 0, "speed/wps": 12910.069189694874, "speed/FLOPS": 202770376918274.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05222802981734276, "optim/lr": 0.0028798107897702193, "optim/total_tokens": 6321340416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.904841184616089, "created_at": "2025-01-16T15:36:55.713480+00:00"} {"global_step": 12058, "acc_step": 0, "speed/wps": 12900.233020490805, "speed/FLOPS": 202615886364611.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06527625769376755, "optim/lr": 0.002879777182691449, "optim/total_tokens": 6321864704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.833923816680908, "created_at": "2025-01-16T15:37:05.878110+00:00"} {"global_step": 12059, "acc_step": 0, "speed/wps": 12905.676988216685, "speed/FLOPS": 202701391358542.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05071272328495979, "optim/lr": 0.002879743571110912, "optim/total_tokens": 6322388992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 2.925569534301758, "created_at": "2025-01-16T15:37:16.040058+00:00"} {"global_step": 12060, "acc_step": 0, "speed/wps": 12904.587773638003, "speed/FLOPS": 202684283746846.44, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0624271035194397, "optim/lr": 0.002879709955028721, "optim/total_tokens": 6322913280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.934724807739258, "created_at": "2025-01-16T15:37:26.200665+00:00"} {"global_step": 12061, "acc_step": 0, "speed/wps": 12905.079134396343, "speed/FLOPS": 202692001242756.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05082535743713379, "optim/lr": 0.0028796763344449833, "optim/total_tokens": 6323437568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 2.8470911979675293, "created_at": "2025-01-16T15:37:36.361082+00:00"} {"global_step": 12062, "acc_step": 0, "speed/wps": 12902.823449856758, "speed/FLOPS": 202656572617427.06, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04765138030052185, "optim/lr": 0.0028796427093598096, "optim/total_tokens": 6323961856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.9758875370025635, "created_at": "2025-01-16T15:37:46.524446+00:00"} {"global_step": 12063, "acc_step": 0, "speed/wps": 12911.029847059974, "speed/FLOPS": 202785465362275.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04645032063126564, "optim/lr": 0.0028796090797733086, "optim/total_tokens": 6324486144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463218, "loss/out": 2.9513723850250244, "created_at": "2025-01-16T15:37:56.683044+00:00"} {"global_step": 12064, "acc_step": 0, "speed/wps": 12914.586744704755, "speed/FLOPS": 202841331327475.9, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04795326292514801, "optim/lr": 0.0028795754456855914, "optim/total_tokens": 6325010432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.917184829711914, "created_at": "2025-01-16T15:38:06.836815+00:00"} {"global_step": 12065, "acc_step": 0, "speed/wps": 12906.883570764974, "speed/FLOPS": 202720342395481.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04351973533630371, "optim/lr": 0.002879541807096767, "optim/total_tokens": 6325534720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8760151863098145, "created_at": "2025-01-16T15:38:16.995520+00:00"} {"global_step": 12066, "acc_step": 0, "speed/wps": 12911.294141825194, "speed/FLOPS": 202789616474741.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05408824607729912, "optim/lr": 0.002879508164006945, "optim/total_tokens": 6326059008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.943556308746338, "created_at": "2025-01-16T15:38:27.154455+00:00"} {"global_step": 12067, "acc_step": 0, "speed/wps": 12903.603116461338, "speed/FLOPS": 202668818352826.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04706605523824692, "optim/lr": 0.0028794745164162357, "optim/total_tokens": 6326583296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.866199016571045, "created_at": "2025-01-16T15:38:37.317267+00:00"} {"global_step": 12068, "acc_step": 0, "speed/wps": 12907.6774071845, "speed/FLOPS": 202732810687295.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05489296838641167, "optim/lr": 0.002879440864324748, "optim/total_tokens": 6327107584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.821413040161133, "created_at": "2025-01-16T15:38:47.476335+00:00"} {"global_step": 12069, "acc_step": 0, "speed/wps": 12905.124566738414, "speed/FLOPS": 202692714820119.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05350445210933685, "optim/lr": 0.002879407207732593, "optim/total_tokens": 6327631872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.746591567993164, "created_at": "2025-01-16T15:38:57.638387+00:00"} {"global_step": 12070, "acc_step": 0, "speed/wps": 12907.408472104205, "speed/FLOPS": 202728586692304.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04778876155614853, "optim/lr": 0.002879373546639879, "optim/total_tokens": 6328156160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 2.9620490074157715, "created_at": "2025-01-16T15:39:07.797141+00:00"} {"global_step": 12071, "acc_step": 0, "speed/wps": 12908.817645693962, "speed/FLOPS": 202750719699938.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05953280255198479, "optim/lr": 0.0028793398810467168, "optim/total_tokens": 6328680448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9293360710144043, "created_at": "2025-01-16T15:39:17.957677+00:00"} {"global_step": 12072, "acc_step": 0, "speed/wps": 12905.557823672496, "speed/FLOPS": 202699519715626.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0418739914894104, "optim/lr": 0.0028793062109532166, "optim/total_tokens": 6329204736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.7637507915496826, "created_at": "2025-01-16T15:39:28.119612+00:00"} {"global_step": 12073, "acc_step": 0, "speed/wps": 12908.704742205198, "speed/FLOPS": 202748946395501.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0664268210530281, "optim/lr": 0.0028792725363594868, "optim/total_tokens": 6329729024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.943208694458008, "created_at": "2025-01-16T15:39:38.278738+00:00"} {"global_step": 12074, "acc_step": 0, "speed/wps": 12908.420849973063, "speed/FLOPS": 202744487477887.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06292588263750076, "optim/lr": 0.0028792388572656392, "optim/total_tokens": 6330253312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415534, "loss/out": 2.949615955352783, "created_at": "2025-01-16T15:39:48.438567+00:00"} {"global_step": 12075, "acc_step": 0, "speed/wps": 12909.493093855137, "speed/FLOPS": 202761328541472.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04455103352665901, "optim/lr": 0.0028792051736717815, "optim/total_tokens": 6330777600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.892665147781372, "created_at": "2025-01-16T15:39:58.597188+00:00"} {"global_step": 12076, "acc_step": 0, "speed/wps": 12908.064184078243, "speed/FLOPS": 202738885549898.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05549236387014389, "optim/lr": 0.0028791714855780255, "optim/total_tokens": 6331301888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 2.811094045639038, "created_at": "2025-01-16T15:40:08.756698+00:00"} {"global_step": 12077, "acc_step": 0, "speed/wps": 12911.895045078558, "speed/FLOPS": 202799054486063.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04697585105895996, "optim/lr": 0.00287913779298448, "optim/total_tokens": 6331826176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.8605499267578125, "created_at": "2025-01-16T15:40:18.915250+00:00"} {"global_step": 12078, "acc_step": 0, "speed/wps": 12904.001297298324, "speed/FLOPS": 202675072330028.84, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054989803582429886, "optim/lr": 0.0028791040958912554, "optim/total_tokens": 6332350464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8788554668426514, "created_at": "2025-01-16T15:40:29.077506+00:00"} {"global_step": 12079, "acc_step": 0, "speed/wps": 12904.223427655968, "speed/FLOPS": 202678561192534.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04025987163186073, "optim/lr": 0.002879070394298462, "optim/total_tokens": 6332874752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 2.812568187713623, "created_at": "2025-01-16T15:40:39.238428+00:00"} {"global_step": 12080, "acc_step": 0, "speed/wps": 12905.106147862978, "speed/FLOPS": 202692425526370.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05509527027606964, "optim/lr": 0.0028790366882062088, "optim/total_tokens": 6333399040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.8792219161987305, "created_at": "2025-01-16T15:40:49.400071+00:00"} {"global_step": 12081, "acc_step": 0, "speed/wps": 12896.957661592178, "speed/FLOPS": 202564442352294.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04983679577708244, "optim/lr": 0.0028790029776146068, "optim/total_tokens": 6333923328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355742, "loss/out": 2.731447458267212, "created_at": "2025-01-16T15:40:59.566740+00:00"} {"global_step": 12082, "acc_step": 0, "speed/wps": 12909.754052569557, "speed/FLOPS": 202765427256677.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044225063174963, "optim/lr": 0.0028789692625237648, "optim/total_tokens": 6334447616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9613959789276123, "created_at": "2025-01-16T15:41:09.725267+00:00"} {"global_step": 12083, "acc_step": 0, "speed/wps": 12911.309444211704, "speed/FLOPS": 202789856819749.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049284640699625015, "optim/lr": 0.0028789355429337934, "optim/total_tokens": 6334971904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9086711406707764, "created_at": "2025-01-16T15:41:19.881671+00:00"} {"global_step": 12084, "acc_step": 0, "speed/wps": 12909.044834489028, "speed/FLOPS": 202754288012155.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0438716895878315, "optim/lr": 0.0028789018188448033, "optim/total_tokens": 6335496192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.8449692726135254, "created_at": "2025-01-16T15:41:30.040211+00:00"} {"global_step": 12085, "acc_step": 0, "speed/wps": 12899.840427191972, "speed/FLOPS": 202609720147373.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05599973350763321, "optim/lr": 0.0028788680902569032, "optim/total_tokens": 6336020480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.844747543334961, "created_at": "2025-01-16T15:41:40.206290+00:00"} {"global_step": 12086, "acc_step": 0, "speed/wps": 12905.403844276256, "speed/FLOPS": 202697101257618.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06418891251087189, "optim/lr": 0.0028788343571702043, "optim/total_tokens": 6336544768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 2.9494659900665283, "created_at": "2025-01-16T15:41:50.366751+00:00"} {"global_step": 12087, "acc_step": 0, "speed/wps": 12909.115810573869, "speed/FLOPS": 202755402789099.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06076033040881157, "optim/lr": 0.002878800619584816, "optim/total_tokens": 6337069056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8529133796691895, "created_at": "2025-01-16T15:42:00.526845+00:00"} {"global_step": 12088, "acc_step": 0, "speed/wps": 12906.900557727342, "speed/FLOPS": 202720609199067.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052762337028980255, "optim/lr": 0.0028787668775008485, "optim/total_tokens": 6337593344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.9649834632873535, "created_at": "2025-01-16T15:42:10.685620+00:00"} {"global_step": 12089, "acc_step": 0, "speed/wps": 12902.505700902815, "speed/FLOPS": 202651581933472.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07082392275333405, "optim/lr": 0.002878733130918412, "optim/total_tokens": 6338117632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.684119701385498, "created_at": "2025-01-16T15:42:20.854017+00:00"} {"global_step": 12090, "acc_step": 0, "speed/wps": 12911.469172069075, "speed/FLOPS": 202792365565238.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05927782133221626, "optim/lr": 0.002878699379837617, "optim/total_tokens": 6338641920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.939507007598877, "created_at": "2025-01-16T15:42:31.009786+00:00"} {"global_step": 12091, "acc_step": 0, "speed/wps": 12908.954306834348, "speed/FLOPS": 202752866150940.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05089747533202171, "optim/lr": 0.0028786656242585727, "optim/total_tokens": 6339166208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.7444019317626953, "created_at": "2025-01-16T15:42:41.168388+00:00"} {"global_step": 12092, "acc_step": 0, "speed/wps": 12904.66079983824, "speed/FLOPS": 202685430723669.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04745393246412277, "optim/lr": 0.0028786318641813896, "optim/total_tokens": 6339690496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8536789417266846, "created_at": "2025-01-16T15:42:51.333369+00:00"} {"global_step": 12093, "acc_step": 0, "speed/wps": 12906.006771141288, "speed/FLOPS": 202706571052542.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05040682852268219, "optim/lr": 0.0028785980996061783, "optim/total_tokens": 6340214784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.812101364135742, "created_at": "2025-01-16T15:43:01.502766+00:00"} {"global_step": 12094, "acc_step": 0, "speed/wps": 12907.37290338457, "speed/FLOPS": 202728028036685.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04841487109661102, "optim/lr": 0.0028785643305330485, "optim/total_tokens": 6340739072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 2.9418814182281494, "created_at": "2025-01-16T15:43:11.664071+00:00"} {"global_step": 12095, "acc_step": 0, "speed/wps": 12907.84650535154, "speed/FLOPS": 202735466606374.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044435128569602966, "optim/lr": 0.002878530556962111, "optim/total_tokens": 6341263360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.791306972503662, "created_at": "2025-01-16T15:43:21.823047+00:00"} {"global_step": 12096, "acc_step": 0, "speed/wps": 12903.738290086598, "speed/FLOPS": 202670941440358.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04875177890062332, "optim/lr": 0.002878496778893475, "optim/total_tokens": 6341787648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8373565673828125, "created_at": "2025-01-16T15:43:31.988329+00:00"} {"global_step": 12097, "acc_step": 0, "speed/wps": 12912.250151642309, "speed/FLOPS": 202804631922614.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044962212443351746, "optim/lr": 0.002878462996327251, "optim/total_tokens": 6342311936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.85844087600708, "created_at": "2025-01-16T15:43:42.145835+00:00"} {"global_step": 12098, "acc_step": 0, "speed/wps": 12912.965111591217, "speed/FLOPS": 202815861351070.34, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04801688715815544, "optim/lr": 0.0028784292092635497, "optim/total_tokens": 6342836224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.897383213043213, "created_at": "2025-01-16T15:43:52.302777+00:00"} {"global_step": 12099, "acc_step": 0, "speed/wps": 12904.29916163201, "speed/FLOPS": 202679750698697.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04072355106472969, "optim/lr": 0.002878395417702481, "optim/total_tokens": 6343360512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425680, "loss/out": 2.799515724182129, "created_at": "2025-01-16T15:44:02.466505+00:00"} {"global_step": 12100, "acc_step": 0, "speed/wps": 12905.949478589699, "speed/FLOPS": 202705671194291.53, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04489327594637871, "optim/lr": 0.0028783616216441553, "optim/total_tokens": 6343884800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.7585716247558594, "created_at": "2025-01-16T15:44:12.628218+00:00"} {"global_step": 12101, "acc_step": 0, "speed/wps": 12908.320371759275, "speed/FLOPS": 202742909329468.34, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04120909050107002, "optim/lr": 0.0028783278210886827, "optim/total_tokens": 6344409088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 2.688563585281372, "created_at": "2025-01-16T15:44:22.786964+00:00"} {"global_step": 12102, "acc_step": 0, "speed/wps": 12911.139197580938, "speed/FLOPS": 202787182862470.16, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04352964460849762, "optim/lr": 0.0028782940160361737, "optim/total_tokens": 6344933376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.806577682495117, "created_at": "2025-01-16T15:44:32.942790+00:00"} {"global_step": 12103, "acc_step": 0, "speed/wps": 12908.7530938214, "speed/FLOPS": 202749705824076.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047104161232709885, "optim/lr": 0.0028782602064867384, "optim/total_tokens": 6345457664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462696, "loss/out": 2.8442671298980713, "created_at": "2025-01-16T15:44:43.100568+00:00"} {"global_step": 12104, "acc_step": 0, "speed/wps": 12906.618694279878, "speed/FLOPS": 202716182146304.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04831537976861, "optim/lr": 0.0028782263924404874, "optim/total_tokens": 6345981952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8778951168060303, "created_at": "2025-01-16T15:44:53.264691+00:00"} {"global_step": 12105, "acc_step": 0, "speed/wps": 12908.37991474186, "speed/FLOPS": 202743844533830.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.03837287053465843, "optim/lr": 0.0028781925738975305, "optim/total_tokens": 6346506240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.743727207183838, "created_at": "2025-01-16T15:45:03.422526+00:00"} {"global_step": 12106, "acc_step": 0, "speed/wps": 12909.102522775938, "speed/FLOPS": 202755194085973.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05071329325437546, "optim/lr": 0.0028781587508579782, "optim/total_tokens": 6347030528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.855767250061035, "created_at": "2025-01-16T15:45:13.580995+00:00"} {"global_step": 12107, "acc_step": 0, "speed/wps": 12912.191464991183, "speed/FLOPS": 202803710168115.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045395053923130035, "optim/lr": 0.002878124923321941, "optim/total_tokens": 6347554816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8922688961029053, "created_at": "2025-01-16T15:45:23.736444+00:00"} {"global_step": 12108, "acc_step": 0, "speed/wps": 12908.71935172678, "speed/FLOPS": 202749175858113.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05333811417222023, "optim/lr": 0.00287809109128953, "optim/total_tokens": 6348079104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.918736219406128, "created_at": "2025-01-16T15:45:33.893783+00:00"} {"global_step": 12109, "acc_step": 0, "speed/wps": 12910.757549546941, "speed/FLOPS": 202781188555656.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05438091233372688, "optim/lr": 0.0028780572547608537, "optim/total_tokens": 6348603392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.876803159713745, "created_at": "2025-01-16T15:45:44.049523+00:00"} {"global_step": 12110, "acc_step": 0, "speed/wps": 12904.31043756093, "speed/FLOPS": 202679927802656.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04699556156992912, "optim/lr": 0.0028780234137360246, "optim/total_tokens": 6349127680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 2.881643056869507, "created_at": "2025-01-16T15:45:54.212749+00:00"} {"global_step": 12111, "acc_step": 0, "speed/wps": 12911.138777098688, "speed/FLOPS": 202787176258218.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05505197122693062, "optim/lr": 0.0028779895682151516, "optim/total_tokens": 6349651968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478609, "loss/out": 2.7408571243286133, "created_at": "2025-01-16T15:46:04.371962+00:00"} {"global_step": 12112, "acc_step": 0, "speed/wps": 12910.131768705878, "speed/FLOPS": 202771359807635.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04708898067474365, "optim/lr": 0.002877955718198346, "optim/total_tokens": 6350176256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370300, "loss/out": 2.8658947944641113, "created_at": "2025-01-16T15:46:14.529802+00:00"} {"global_step": 12113, "acc_step": 0, "speed/wps": 12903.280403027215, "speed/FLOPS": 202663749694888.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04808998480439186, "optim/lr": 0.002877921863685718, "optim/total_tokens": 6350700544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7135109901428223, "created_at": "2025-01-16T15:46:24.698073+00:00"} {"global_step": 12114, "acc_step": 0, "speed/wps": 12903.853184274974, "speed/FLOPS": 202672746011468.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04834586754441261, "optim/lr": 0.0028778880046773773, "optim/total_tokens": 6351224832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381609, "loss/out": 2.743800163269043, "created_at": "2025-01-16T15:46:34.863228+00:00"} {"global_step": 12115, "acc_step": 0, "speed/wps": 12901.818590869883, "speed/FLOPS": 202640789926217.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05279822647571564, "optim/lr": 0.002877854141173436, "optim/total_tokens": 6351749120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.826664924621582, "created_at": "2025-01-16T15:46:45.027659+00:00"} {"global_step": 12116, "acc_step": 0, "speed/wps": 12899.014248448868, "speed/FLOPS": 202596743874924.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052283842116594315, "optim/lr": 0.002877820273174002, "optim/total_tokens": 6352273408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.811264991760254, "created_at": "2025-01-16T15:46:55.194094+00:00"} {"global_step": 12117, "acc_step": 0, "speed/wps": 12911.648450117156, "speed/FLOPS": 202795181373337.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0517672598361969, "optim/lr": 0.0028777864006791893, "optim/total_tokens": 6352797696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.864762783050537, "created_at": "2025-01-16T15:47:05.349108+00:00"} {"global_step": 12118, "acc_step": 0, "speed/wps": 12905.577774021005, "speed/FLOPS": 202699833063264.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05189308896660805, "optim/lr": 0.0028777525236891054, "optim/total_tokens": 6353321984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.988041639328003, "created_at": "2025-01-16T15:47:15.513827+00:00"} {"global_step": 12119, "acc_step": 0, "speed/wps": 12898.301404612259, "speed/FLOPS": 202585547682920.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04971783608198166, "optim/lr": 0.0028777186422038626, "optim/total_tokens": 6353846272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.9440865516662598, "created_at": "2025-01-16T15:47:25.679624+00:00"} {"global_step": 12120, "acc_step": 0, "speed/wps": 12905.281135811036, "speed/FLOPS": 202695173952553.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0463968925178051, "optim/lr": 0.002877684756223571, "optim/total_tokens": 6354370560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.831058979034424, "created_at": "2025-01-16T15:47:35.842837+00:00"} {"global_step": 12121, "acc_step": 0, "speed/wps": 12904.539950183876, "speed/FLOPS": 202683532613783.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051808375865221024, "optim/lr": 0.00287765086574834, "optim/total_tokens": 6354894848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 2.9138121604919434, "created_at": "2025-01-16T15:47:46.004276+00:00"} {"global_step": 12122, "acc_step": 0, "speed/wps": 12904.642584624986, "speed/FLOPS": 202685144628715.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05170265585184097, "optim/lr": 0.0028776169707782815, "optim/total_tokens": 6355419136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.912627935409546, "created_at": "2025-01-16T15:47:56.170415+00:00"} {"global_step": 12123, "acc_step": 0, "speed/wps": 12904.941763651224, "speed/FLOPS": 202689843646437.34, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0522306002676487, "optim/lr": 0.0028775830713135063, "optim/total_tokens": 6355943424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9178619384765625, "created_at": "2025-01-16T15:48:06.330814+00:00"} {"global_step": 12124, "acc_step": 0, "speed/wps": 12904.746532961019, "speed/FLOPS": 202686777280172.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056266311556100845, "optim/lr": 0.002877549167354124, "optim/total_tokens": 6356467712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 2.9140560626983643, "created_at": "2025-01-16T15:48:16.492166+00:00"} {"global_step": 12125, "acc_step": 0, "speed/wps": 12901.166411881974, "speed/FLOPS": 202630546559024.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04737289622426033, "optim/lr": 0.0028775152589002456, "optim/total_tokens": 6356992000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.80465030670166, "created_at": "2025-01-16T15:48:26.657646+00:00"} {"global_step": 12126, "acc_step": 0, "speed/wps": 12911.081897354781, "speed/FLOPS": 202786282883679.6, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05497162416577339, "optim/lr": 0.0028774813459519818, "optim/total_tokens": 6357516288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 2.8466298580169678, "created_at": "2025-01-16T15:48:36.813744+00:00"} {"global_step": 12127, "acc_step": 0, "speed/wps": 12913.0199321556, "speed/FLOPS": 202816722383365.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04436583071947098, "optim/lr": 0.002877447428509443, "optim/total_tokens": 6358040576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.9303770065307617, "created_at": "2025-01-16T15:48:46.968324+00:00"} {"global_step": 12128, "acc_step": 0, "speed/wps": 12908.855373773431, "speed/FLOPS": 202751312271270.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04755312576889992, "optim/lr": 0.002877413506572741, "optim/total_tokens": 6358564864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9468038082122803, "created_at": "2025-01-16T15:48:57.130338+00:00"} {"global_step": 12129, "acc_step": 0, "speed/wps": 12908.999907369347, "speed/FLOPS": 202753582370004.3, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04733600467443466, "optim/lr": 0.002877379580141985, "optim/total_tokens": 6359089152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.7343344688415527, "created_at": "2025-01-16T15:49:07.288045+00:00"} {"global_step": 12130, "acc_step": 0, "speed/wps": 12911.435748621629, "speed/FLOPS": 202791840604068.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.038310132920742035, "optim/lr": 0.0028773456492172866, "optim/total_tokens": 6359613440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 2.9151129722595215, "created_at": "2025-01-16T15:49:17.443363+00:00"} {"global_step": 12131, "acc_step": 0, "speed/wps": 12908.886798220867, "speed/FLOPS": 202751805835399.16, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05084380879998207, "optim/lr": 0.0028773117137987554, "optim/total_tokens": 6360137728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.871544361114502, "created_at": "2025-01-16T15:49:27.600919+00:00"} {"global_step": 12132, "acc_step": 0, "speed/wps": 12900.254148665019, "speed/FLOPS": 202616218211621.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0575411431491375, "optim/lr": 0.002877277773886504, "optim/total_tokens": 6360662016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.7540111541748047, "created_at": "2025-01-16T15:49:37.766038+00:00"} {"global_step": 12133, "acc_step": 0, "speed/wps": 12910.618746797354, "speed/FLOPS": 202779008467739.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05598266422748566, "optim/lr": 0.002877243829480641, "optim/total_tokens": 6361186304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.964454412460327, "created_at": "2025-01-16T15:49:47.922502+00:00"} {"global_step": 12134, "acc_step": 0, "speed/wps": 12912.001619301425, "speed/FLOPS": 202800728380682.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06344926357269287, "optim/lr": 0.0028772098805812793, "optim/total_tokens": 6361710592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.665299415588379, "created_at": "2025-01-16T15:49:58.077766+00:00"} {"global_step": 12135, "acc_step": 0, "speed/wps": 12910.064284488624, "speed/FLOPS": 202770299875270.0, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06211194023489952, "optim/lr": 0.002877175927188527, "optim/total_tokens": 6362234880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.0545883178710938, "created_at": "2025-01-16T15:50:08.237469+00:00"} {"global_step": 12136, "acc_step": 0, "speed/wps": 12912.31073785592, "speed/FLOPS": 202805583512353.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05472169071435928, "optim/lr": 0.002877141969302498, "optim/total_tokens": 6362759168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8380327224731445, "created_at": "2025-01-16T15:50:18.392833+00:00"} {"global_step": 12137, "acc_step": 0, "speed/wps": 12915.09948464795, "speed/FLOPS": 202849384612862.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07171332836151123, "optim/lr": 0.0028771080069233, "optim/total_tokens": 6363283456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8432486057281494, "created_at": "2025-01-16T15:50:28.545266+00:00"} {"global_step": 12138, "acc_step": 0, "speed/wps": 12895.928900930403, "speed/FLOPS": 202548284252435.62, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0021, "optim/grad_norm": 0.05294375494122505, "optim/lr": 0.002877074040051047, "optim/total_tokens": 6363807744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9286417961120605, "created_at": "2025-01-16T15:50:38.719267+00:00"} {"global_step": 12139, "acc_step": 0, "speed/wps": 12896.60170450884, "speed/FLOPS": 202558851557163.97, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.047997426241636276, "optim/lr": 0.0028770400686858465, "optim/total_tokens": 6364332032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409644, "loss/out": 2.9525563716888428, "created_at": "2025-01-16T15:50:48.891898+00:00"} {"global_step": 12140, "acc_step": 0, "speed/wps": 12906.393476325298, "speed/FLOPS": 202712644788845.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05572177469730377, "optim/lr": 0.0028770060928278124, "optim/total_tokens": 6364856320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.856740951538086, "created_at": "2025-01-16T15:50:59.051788+00:00"} {"global_step": 12141, "acc_step": 0, "speed/wps": 12899.894190458075, "speed/FLOPS": 202610564573346.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048377446830272675, "optim/lr": 0.002876972112477053, "optim/total_tokens": 6365380608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440624, "loss/out": 2.883091926574707, "created_at": "2025-01-16T15:51:09.222694+00:00"} {"global_step": 12142, "acc_step": 0, "speed/wps": 12903.07450244654, "speed/FLOPS": 202660515743332.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06057801842689514, "optim/lr": 0.002876938127633681, "optim/total_tokens": 6365904896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.996281862258911, "created_at": "2025-01-16T15:51:19.386047+00:00"} {"global_step": 12143, "acc_step": 0, "speed/wps": 12909.546465129448, "speed/FLOPS": 202762166810675.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05421380326151848, "optim/lr": 0.002876904138297806, "optim/total_tokens": 6366429184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8257079124450684, "created_at": "2025-01-16T15:51:29.545407+00:00"} {"global_step": 12144, "acc_step": 0, "speed/wps": 12911.783906075276, "speed/FLOPS": 202797308895296.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07315468788146973, "optim/lr": 0.00287687014446954, "optim/total_tokens": 6366953472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 2.889202356338501, "created_at": "2025-01-16T15:51:39.703378+00:00"} {"global_step": 12145, "acc_step": 0, "speed/wps": 12905.611124542616, "speed/FLOPS": 202700356879034.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04645666480064392, "optim/lr": 0.0028768361461489938, "optim/total_tokens": 6367477760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.912738800048828, "created_at": "2025-01-16T15:51:49.865926+00:00"} {"global_step": 12146, "acc_step": 0, "speed/wps": 12912.131806284651, "speed/FLOPS": 202802773146149.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0707826167345047, "optim/lr": 0.002876802143336277, "optim/total_tokens": 6368002048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429128, "loss/out": 2.7975292205810547, "created_at": "2025-01-16T15:52:00.025444+00:00"} {"global_step": 12147, "acc_step": 0, "speed/wps": 12903.310560531541, "speed/FLOPS": 202664223359934.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06288010627031326, "optim/lr": 0.002876768136031502, "optim/total_tokens": 6368526336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.7011301517486572, "created_at": "2025-01-16T15:52:10.192746+00:00"} {"global_step": 12148, "acc_step": 0, "speed/wps": 12905.422492399302, "speed/FLOPS": 202697394152016.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04618304222822189, "optim/lr": 0.0028767341242347793, "optim/total_tokens": 6369050624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421131, "loss/out": 2.8343167304992676, "created_at": "2025-01-16T15:52:20.356102+00:00"} {"global_step": 12149, "acc_step": 0, "speed/wps": 12908.756222001224, "speed/FLOPS": 202749754956439.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06294053792953491, "optim/lr": 0.002876700107946219, "optim/total_tokens": 6369574912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.646199941635132, "created_at": "2025-01-16T15:52:30.516181+00:00"} {"global_step": 12150, "acc_step": 0, "speed/wps": 12908.471118673757, "speed/FLOPS": 202745277016908.25, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045126110315322876, "optim/lr": 0.002876666087165934, "optim/total_tokens": 6370099200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8192410469055176, "created_at": "2025-01-16T15:52:40.674944+00:00"} {"global_step": 12151, "acc_step": 0, "speed/wps": 12902.767544741258, "speed/FLOPS": 202655694550766.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07235197722911835, "optim/lr": 0.002876632061894033, "optim/total_tokens": 6370623488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.9568238258361816, "created_at": "2025-01-16T15:52:50.836778+00:00"} {"global_step": 12152, "acc_step": 0, "speed/wps": 12914.042645534515, "speed/FLOPS": 202832785502336.66, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05303356051445007, "optim/lr": 0.002876598032130629, "optim/total_tokens": 6371147776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.875924587249756, "created_at": "2025-01-16T15:53:00.990824+00:00"} {"global_step": 12153, "acc_step": 0, "speed/wps": 12908.258189164811, "speed/FLOPS": 202741932666374.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0733785405755043, "optim/lr": 0.0028765639978758323, "optim/total_tokens": 6371672064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323796, "loss/out": 2.949739933013916, "created_at": "2025-01-16T15:53:11.149273+00:00"} {"global_step": 12154, "acc_step": 0, "speed/wps": 12910.238259345177, "speed/FLOPS": 202773032389458.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053355179727077484, "optim/lr": 0.0028765299591297535, "optim/total_tokens": 6372196352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.8567392826080322, "created_at": "2025-01-16T15:53:21.311163+00:00"} {"global_step": 12155, "acc_step": 0, "speed/wps": 12906.287102364515, "speed/FLOPS": 202710974039618.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054769039154052734, "optim/lr": 0.0028764959158925035, "optim/total_tokens": 6372720640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 2.9598588943481445, "created_at": "2025-01-16T15:53:31.472726+00:00"} {"global_step": 12156, "acc_step": 0, "speed/wps": 12915.096968503181, "speed/FLOPS": 202849345093351.16, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055283788591623306, "optim/lr": 0.0028764618681641945, "optim/total_tokens": 6373244928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.881214141845703, "created_at": "2025-01-16T15:53:41.627646+00:00"} {"global_step": 12157, "acc_step": 0, "speed/wps": 12907.237838263563, "speed/FLOPS": 202725906653361.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04589207470417023, "optim/lr": 0.002876427815944937, "optim/total_tokens": 6373769216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 2.8164329528808594, "created_at": "2025-01-16T15:53:51.786896+00:00"} {"global_step": 12158, "acc_step": 0, "speed/wps": 12906.789467993627, "speed/FLOPS": 202718864382146.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05059867724776268, "optim/lr": 0.002876393759234842, "optim/total_tokens": 6374293504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506128, "loss/out": 2.870022773742676, "created_at": "2025-01-16T15:54:01.945724+00:00"} {"global_step": 12159, "acc_step": 0, "speed/wps": 12907.392269827902, "speed/FLOPS": 202728332213290.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04544981196522713, "optim/lr": 0.002876359698034021, "optim/total_tokens": 6374817792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8365116119384766, "created_at": "2025-01-16T15:54:12.104115+00:00"} {"global_step": 12160, "acc_step": 0, "speed/wps": 12909.56650478269, "speed/FLOPS": 202762481560967.28, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04306851327419281, "optim/lr": 0.002876325632342584, "optim/total_tokens": 6375342080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8986904621124268, "created_at": "2025-01-16T15:54:22.262209+00:00"} {"global_step": 12161, "acc_step": 0, "speed/wps": 12909.434261746357, "speed/FLOPS": 202760404502360.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0438980832695961, "optim/lr": 0.002876291562160644, "optim/total_tokens": 6375866368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 2.8372344970703125, "created_at": "2025-01-16T15:54:32.422027+00:00"} {"global_step": 12162, "acc_step": 0, "speed/wps": 12915.040257728297, "speed/FLOPS": 202848454372702.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.045500047504901886, "optim/lr": 0.0028762574874883103, "optim/total_tokens": 6376390656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.871025800704956, "created_at": "2025-01-16T15:54:42.574410+00:00"} {"global_step": 12163, "acc_step": 0, "speed/wps": 12911.628657492776, "speed/FLOPS": 202794870502973.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047746773809194565, "optim/lr": 0.002876223408325695, "optim/total_tokens": 6376914944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453393, "loss/out": 2.8906004428863525, "created_at": "2025-01-16T15:54:52.732670+00:00"} {"global_step": 12164, "acc_step": 0, "speed/wps": 12912.848908225262, "speed/FLOPS": 202814036217527.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06847869604825974, "optim/lr": 0.0028761893246729096, "optim/total_tokens": 6377439232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.888547658920288, "created_at": "2025-01-16T15:55:02.887165+00:00"} {"global_step": 12165, "acc_step": 0, "speed/wps": 12910.010888800396, "speed/FLOPS": 202769461222613.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05548427999019623, "optim/lr": 0.0028761552365300647, "optim/total_tokens": 6377963520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9056618213653564, "created_at": "2025-01-16T15:55:13.043466+00:00"} {"global_step": 12166, "acc_step": 0, "speed/wps": 12913.503479881978, "speed/FLOPS": 202824317164872.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057152748107910156, "optim/lr": 0.002876121143897272, "optim/total_tokens": 6378487808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9392268657684326, "created_at": "2025-01-16T15:55:23.200630+00:00"} {"global_step": 12167, "acc_step": 0, "speed/wps": 12912.492495169332, "speed/FLOPS": 202808438260721.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0491044744849205, "optim/lr": 0.0028760870467746413, "optim/total_tokens": 6379012096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320126, "loss/out": 2.8010432720184326, "created_at": "2025-01-16T15:55:33.357533+00:00"} {"global_step": 12168, "acc_step": 0, "speed/wps": 12909.1901715947, "speed/FLOPS": 202756570731113.94, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053595032542943954, "optim/lr": 0.002876052945162286, "optim/total_tokens": 6379536384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.653165340423584, "created_at": "2025-01-16T15:55:43.516490+00:00"} {"global_step": 12169, "acc_step": 0, "speed/wps": 12911.811839607726, "speed/FLOPS": 202797747629807.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05226529389619827, "optim/lr": 0.002876018839060316, "optim/total_tokens": 6380060672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384291, "loss/out": 2.882211685180664, "created_at": "2025-01-16T15:55:53.672547+00:00"} {"global_step": 12170, "acc_step": 0, "speed/wps": 12912.233211013221, "speed/FLOPS": 202804365846756.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04897885024547577, "optim/lr": 0.0028759847284688432, "optim/total_tokens": 6380584960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9005351066589355, "created_at": "2025-01-16T15:56:03.833081+00:00"} {"global_step": 12171, "acc_step": 0, "speed/wps": 12911.711716086134, "speed/FLOPS": 202796175052317.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04361068829894066, "optim/lr": 0.0028759506133879784, "optim/total_tokens": 6381109248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.7949185371398926, "created_at": "2025-01-16T15:56:13.990186+00:00"} {"global_step": 12172, "acc_step": 0, "speed/wps": 12911.748926843666, "speed/FLOPS": 202796759498397.1, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0580904558300972, "optim/lr": 0.0028759164938178335, "optim/total_tokens": 6381633536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.88139271736145, "created_at": "2025-01-16T15:56:24.145009+00:00"} {"global_step": 12173, "acc_step": 0, "speed/wps": 12909.479433456576, "speed/FLOPS": 202761113986142.12, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059083253145217896, "optim/lr": 0.0028758823697585196, "optim/total_tokens": 6382157824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8254013061523438, "created_at": "2025-01-16T15:56:34.305898+00:00"} {"global_step": 12174, "acc_step": 0, "speed/wps": 12909.463780162667, "speed/FLOPS": 202760868129651.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05129487067461014, "optim/lr": 0.002875848241210147, "optim/total_tokens": 6382682112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.9130358695983887, "created_at": "2025-01-16T15:56:44.462619+00:00"} {"global_step": 12175, "acc_step": 0, "speed/wps": 12912.38031638958, "speed/FLOPS": 202806676338835.8, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06741868704557419, "optim/lr": 0.0028758141081728283, "optim/total_tokens": 6383206400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.872282028198242, "created_at": "2025-01-16T15:56:54.618086+00:00"} {"global_step": 12176, "acc_step": 0, "speed/wps": 12911.41368831719, "speed/FLOPS": 202791494116672.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05119584873318672, "optim/lr": 0.0028757799706466743, "optim/total_tokens": 6383730688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.9045064449310303, "created_at": "2025-01-16T15:57:04.774922+00:00"} {"global_step": 12177, "acc_step": 0, "speed/wps": 12913.895857768353, "speed/FLOPS": 202830479998761.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0516597256064415, "optim/lr": 0.002875745828631797, "optim/total_tokens": 6384254976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8842947483062744, "created_at": "2025-01-16T15:57:14.930455+00:00"} {"global_step": 12178, "acc_step": 0, "speed/wps": 12912.453495970058, "speed/FLOPS": 202807825724706.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0677572563290596, "optim/lr": 0.002875711682128308, "optim/total_tokens": 6384779264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8774313926696777, "created_at": "2025-01-16T15:57:25.087531+00:00"} {"global_step": 12179, "acc_step": 0, "speed/wps": 12913.269301102333, "speed/FLOPS": 202820639065342.62, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050889257341623306, "optim/lr": 0.0028756775311363173, "optim/total_tokens": 6385303552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.920328140258789, "created_at": "2025-01-16T15:57:35.241481+00:00"} {"global_step": 12180, "acc_step": 0, "speed/wps": 12908.414744302581, "speed/FLOPS": 202744391579942.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05122358351945877, "optim/lr": 0.0028756433756559367, "optim/total_tokens": 6385827840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8677854537963867, "created_at": "2025-01-16T15:57:45.405127+00:00"} {"global_step": 12181, "acc_step": 0, "speed/wps": 12913.76189672911, "speed/FLOPS": 202828375956558.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05691052973270416, "optim/lr": 0.0028756092156872783, "optim/total_tokens": 6386352128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9418838024139404, "created_at": "2025-01-16T15:57:55.563630+00:00"} {"global_step": 12182, "acc_step": 0, "speed/wps": 12911.008144656884, "speed/FLOPS": 202785124496212.28, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046469613909721375, "optim/lr": 0.002875575051230454, "optim/total_tokens": 6386876416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.855098009109497, "created_at": "2025-01-16T15:58:05.719271+00:00"} {"global_step": 12183, "acc_step": 0, "speed/wps": 12897.077348129875, "speed/FLOPS": 202566322193835.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055165089666843414, "optim/lr": 0.002875540882285574, "optim/total_tokens": 6387400704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.9538216590881348, "created_at": "2025-01-16T15:58:15.886092+00:00"} {"global_step": 12184, "acc_step": 0, "speed/wps": 12902.707277534746, "speed/FLOPS": 202654747971472.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05561311915516853, "optim/lr": 0.0028755067088527503, "optim/total_tokens": 6387924992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352528, "loss/out": 2.8427681922912598, "created_at": "2025-01-16T15:58:26.053971+00:00"} {"global_step": 12185, "acc_step": 0, "speed/wps": 12910.966682527003, "speed/FLOPS": 202784473276487.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04680779203772545, "optim/lr": 0.0028754725309320947, "optim/total_tokens": 6388449280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.8059701919555664, "created_at": "2025-01-16T15:58:36.212298+00:00"} {"global_step": 12186, "acc_step": 0, "speed/wps": 12910.625941798064, "speed/FLOPS": 202779121475112.75, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059194330126047134, "optim/lr": 0.002875438348523718, "optim/total_tokens": 6388973568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419297, "loss/out": 2.8344078063964844, "created_at": "2025-01-16T15:58:46.368174+00:00"} {"global_step": 12187, "acc_step": 0, "speed/wps": 12909.927012465278, "speed/FLOPS": 202768143829511.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05175649747252464, "optim/lr": 0.0028754041616277327, "optim/total_tokens": 6389497856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.662289619445801, "created_at": "2025-01-16T15:58:56.525167+00:00"} {"global_step": 12188, "acc_step": 0, "speed/wps": 12914.142359646683, "speed/FLOPS": 202834351649489.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07401135563850403, "optim/lr": 0.00287536997024425, "optim/total_tokens": 6390022144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 2.985187530517578, "created_at": "2025-01-16T15:59:06.678463+00:00"} {"global_step": 12189, "acc_step": 0, "speed/wps": 12912.003476778526, "speed/FLOPS": 202800757554912.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06302022933959961, "optim/lr": 0.0028753357743733807, "optim/total_tokens": 6390546432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.852334499359131, "created_at": "2025-01-16T15:59:16.835741+00:00"} {"global_step": 12190, "acc_step": 0, "speed/wps": 12915.052802796205, "speed/FLOPS": 202848651410233.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04581553488969803, "optim/lr": 0.0028753015740152375, "optim/total_tokens": 6391070720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 2.9148285388946533, "created_at": "2025-01-16T15:59:26.988470+00:00"} {"global_step": 12191, "acc_step": 0, "speed/wps": 12905.898854358622, "speed/FLOPS": 202704876071177.47, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04481009021401405, "optim/lr": 0.0028752673691699307, "optim/total_tokens": 6391595008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.811271905899048, "created_at": "2025-01-16T15:59:37.151105+00:00"} {"global_step": 12192, "acc_step": 0, "speed/wps": 12895.301947840333, "speed/FLOPS": 202538437092632.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05038100853562355, "optim/lr": 0.0028752331598375733, "optim/total_tokens": 6392119296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.821793556213379, "created_at": "2025-01-16T15:59:47.322915+00:00"} {"global_step": 12193, "acc_step": 0, "speed/wps": 12901.274664644552, "speed/FLOPS": 202632246817415.47, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05095934867858887, "optim/lr": 0.002875198946018276, "optim/total_tokens": 6392643584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 2.9385526180267334, "created_at": "2025-01-16T15:59:57.486760+00:00"} {"global_step": 12194, "acc_step": 0, "speed/wps": 12910.817722612128, "speed/FLOPS": 202782133656332.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06595738977193832, "optim/lr": 0.0028751647277121505, "optim/total_tokens": 6393167872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.91705060005188, "created_at": "2025-01-16T16:00:07.643658+00:00"} {"global_step": 12195, "acc_step": 0, "speed/wps": 12912.49501740762, "speed/FLOPS": 202808477875939.88, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05052408576011658, "optim/lr": 0.002875130504919309, "optim/total_tokens": 6393692160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.9420104026794434, "created_at": "2025-01-16T16:00:17.801680+00:00"} {"global_step": 12196, "acc_step": 0, "speed/wps": 12910.495896506638, "speed/FLOPS": 202777078935109.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05378645285964012, "optim/lr": 0.0028750962776398622, "optim/total_tokens": 6394216448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8358359336853027, "created_at": "2025-01-16T16:00:27.960787+00:00"} {"global_step": 12197, "acc_step": 0, "speed/wps": 12908.16152093932, "speed/FLOPS": 202740414359057.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05305841937661171, "optim/lr": 0.002875062045873923, "optim/total_tokens": 6394740736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.7877798080444336, "created_at": "2025-01-16T16:00:38.119521+00:00"} {"global_step": 12198, "acc_step": 0, "speed/wps": 12906.743678302262, "speed/FLOPS": 202718145192122.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05573968589305878, "optim/lr": 0.0028750278096216025, "optim/total_tokens": 6395265024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.844613790512085, "created_at": "2025-01-16T16:00:48.285143+00:00"} {"global_step": 12199, "acc_step": 0, "speed/wps": 12910.817876583884, "speed/FLOPS": 202782136074670.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05175430700182915, "optim/lr": 0.002874993568883012, "optim/total_tokens": 6395789312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380714, "loss/out": 2.949882984161377, "created_at": "2025-01-16T16:00:58.444911+00:00"} {"global_step": 12200, "acc_step": 0, "speed/wps": 12908.902778096526, "speed/FLOPS": 202752056821305.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04618113115429878, "optim/lr": 0.0028749593236582632, "optim/total_tokens": 6396313600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9624085426330566, "created_at": "2025-01-16T16:01:08.604185+00:00"} {"global_step": 12201, "acc_step": 0, "speed/wps": 12906.200500164729, "speed/FLOPS": 202709613833067.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05886421725153923, "optim/lr": 0.0028749250739474688, "optim/total_tokens": 6396837888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 2.827078104019165, "created_at": "2025-01-16T16:01:18.767499+00:00"} {"global_step": 12202, "acc_step": 0, "speed/wps": 12906.936171191188, "speed/FLOPS": 202721168557455.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05242898687720299, "optim/lr": 0.0028748908197507392, "optim/total_tokens": 6397362176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.921945571899414, "created_at": "2025-01-16T16:01:28.927991+00:00"} {"global_step": 12203, "acc_step": 0, "speed/wps": 12913.777398111126, "speed/FLOPS": 202828619427064.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07703753560781479, "optim/lr": 0.002874856561068188, "optim/total_tokens": 6397886464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8315176963806152, "created_at": "2025-01-16T16:01:39.082933+00:00"} {"global_step": 12204, "acc_step": 0, "speed/wps": 12910.793361957905, "speed/FLOPS": 202781751038782.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053747568279504776, "optim/lr": 0.002874822297899925, "optim/total_tokens": 6398410752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9185595512390137, "created_at": "2025-01-16T16:01:49.239535+00:00"} {"global_step": 12205, "acc_step": 0, "speed/wps": 12907.577598738968, "speed/FLOPS": 202731243058507.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07424183189868927, "optim/lr": 0.0028747880302460635, "optim/total_tokens": 6398935040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.763964891433716, "created_at": "2025-01-16T16:01:59.400210+00:00"} {"global_step": 12206, "acc_step": 0, "speed/wps": 12916.511316161434, "speed/FLOPS": 202871559366841.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051916901022195816, "optim/lr": 0.0028747537581067144, "optim/total_tokens": 6399459328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.9609813690185547, "created_at": "2025-01-16T16:02:09.560705+00:00"} {"global_step": 12207, "acc_step": 0, "speed/wps": 12893.228818588432, "speed/FLOPS": 202505875748953.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05748531222343445, "optim/lr": 0.0028747194814819897, "optim/total_tokens": 6399983616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.889982223510742, "created_at": "2025-01-16T16:02:19.734220+00:00"} {"global_step": 12208, "acc_step": 0, "speed/wps": 12915.332434359925, "speed/FLOPS": 202853043408195.3, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06120702996850014, "optim/lr": 0.0028746852003720016, "optim/total_tokens": 6400507904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8123061656951904, "created_at": "2025-01-16T16:02:29.887463+00:00"} {"global_step": 12209, "acc_step": 0, "speed/wps": 12914.484561152085, "speed/FLOPS": 202839726394366.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06456691026687622, "optim/lr": 0.002874650914776861, "optim/total_tokens": 6401032192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9186978340148926, "created_at": "2025-01-16T16:02:40.040305+00:00"} {"global_step": 12210, "acc_step": 0, "speed/wps": 12909.03401680856, "speed/FLOPS": 202754118105618.44, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07387710362672806, "optim/lr": 0.0028746166246966806, "optim/total_tokens": 6401556480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 2.90376615524292, "created_at": "2025-01-16T16:02:50.199650+00:00"} {"global_step": 12211, "acc_step": 0, "speed/wps": 12909.89433359181, "speed/FLOPS": 202767630562898.53, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05711238458752632, "optim/lr": 0.0028745823301315723, "optim/total_tokens": 6402080768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.844925880432129, "created_at": "2025-01-16T16:03:00.357544+00:00"} {"global_step": 12212, "acc_step": 0, "speed/wps": 12911.675334731848, "speed/FLOPS": 202795603633154.84, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06874759495258331, "optim/lr": 0.0028745480310816483, "optim/total_tokens": 6402605056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410747, "loss/out": 2.82383394241333, "created_at": "2025-01-16T16:03:10.515719+00:00"} {"global_step": 12213, "acc_step": 0, "speed/wps": 12909.744747109438, "speed/FLOPS": 202765281101639.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07935426384210587, "optim/lr": 0.002874513727547019, "optim/total_tokens": 6403129344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.8392653465270996, "created_at": "2025-01-16T16:03:20.679718+00:00"} {"global_step": 12214, "acc_step": 0, "speed/wps": 12911.199617019647, "speed/FLOPS": 202788131832780.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05570138245820999, "optim/lr": 0.0028744794195277984, "optim/total_tokens": 6403653632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302164, "loss/out": 2.8120133876800537, "created_at": "2025-01-16T16:03:30.835647+00:00"} {"global_step": 12215, "acc_step": 0, "speed/wps": 12906.927830944685, "speed/FLOPS": 202721037562423.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0874970555305481, "optim/lr": 0.0028744451070240964, "optim/total_tokens": 6404177920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9137070178985596, "created_at": "2025-01-16T16:03:40.995213+00:00"} {"global_step": 12216, "acc_step": 0, "speed/wps": 12912.862802047404, "speed/FLOPS": 202814254439096.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049520429223775864, "optim/lr": 0.0028744107900360266, "optim/total_tokens": 6404702208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459275, "loss/out": 2.8961243629455566, "created_at": "2025-01-16T16:03:51.149697+00:00"} {"global_step": 12217, "acc_step": 0, "speed/wps": 12910.667758235006, "speed/FLOPS": 202779778259716.75, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06133035570383072, "optim/lr": 0.0028743764685637, "optim/total_tokens": 6405226496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9570107460021973, "created_at": "2025-01-16T16:04:01.308917+00:00"} {"global_step": 12218, "acc_step": 0, "speed/wps": 12912.274208596655, "speed/FLOPS": 202805009770140.56, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05023766681551933, "optim/lr": 0.0028743421426072288, "optim/total_tokens": 6405750784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9045708179473877, "created_at": "2025-01-16T16:04:11.466441+00:00"} {"global_step": 12219, "acc_step": 0, "speed/wps": 12914.004416689619, "speed/FLOPS": 202832185065795.62, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06280071288347244, "optim/lr": 0.002874307812166725, "optim/total_tokens": 6406275072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.808901786804199, "created_at": "2025-01-16T16:04:21.625030+00:00"} {"global_step": 12220, "acc_step": 0, "speed/wps": 12913.565440495286, "speed/FLOPS": 202825290341447.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04336443170905113, "optim/lr": 0.0028742734772423005, "optim/total_tokens": 6406799360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374860, "loss/out": 2.9555888175964355, "created_at": "2025-01-16T16:04:31.778641+00:00"} {"global_step": 12221, "acc_step": 0, "speed/wps": 12912.792778767738, "speed/FLOPS": 202813154627267.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04911475256085396, "optim/lr": 0.0028742391378340677, "optim/total_tokens": 6407323648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8292958736419678, "created_at": "2025-01-16T16:04:41.935270+00:00"} {"global_step": 12222, "acc_step": 0, "speed/wps": 12903.323402238759, "speed/FLOPS": 202664425056592.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047756627202034, "optim/lr": 0.002874204793942139, "optim/total_tokens": 6407847936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.868457794189453, "created_at": "2025-01-16T16:04:52.099475+00:00"} {"global_step": 12223, "acc_step": 0, "speed/wps": 12899.264618436142, "speed/FLOPS": 202600676279619.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049452636390924454, "optim/lr": 0.0028741704455666254, "optim/total_tokens": 6408372224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.947514772415161, "created_at": "2025-01-16T16:05:02.269857+00:00"} {"global_step": 12224, "acc_step": 0, "speed/wps": 12912.720870129519, "speed/FLOPS": 202812025203291.44, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05801402032375336, "optim/lr": 0.002874136092707639, "optim/total_tokens": 6408896512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.8166537284851074, "created_at": "2025-01-16T16:05:12.428866+00:00"} {"global_step": 12225, "acc_step": 0, "speed/wps": 12906.4781577613, "speed/FLOPS": 202713974827162.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056767769157886505, "optim/lr": 0.0028741017353652925, "optim/total_tokens": 6409420800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 2.8693947792053223, "created_at": "2025-01-16T16:05:22.588253+00:00"} {"global_step": 12226, "acc_step": 0, "speed/wps": 12914.418778197765, "speed/FLOPS": 202838693182673.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06299291551113129, "optim/lr": 0.002874067373539699, "optim/total_tokens": 6409945088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 2.9364662170410156, "created_at": "2025-01-16T16:05:32.746689+00:00"} {"global_step": 12227, "acc_step": 0, "speed/wps": 12907.771910019635, "speed/FLOPS": 202734294984181.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04989631474018097, "optim/lr": 0.0028740330072309682, "optim/total_tokens": 6410469376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303224, "loss/out": 2.9219894409179688, "created_at": "2025-01-16T16:05:42.905607+00:00"} {"global_step": 12228, "acc_step": 0, "speed/wps": 12908.15943252099, "speed/FLOPS": 202740381557577.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057230472564697266, "optim/lr": 0.002873998636439214, "optim/total_tokens": 6410993664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.830997943878174, "created_at": "2025-01-16T16:05:53.068236+00:00"} {"global_step": 12229, "acc_step": 0, "speed/wps": 12911.506861161048, "speed/FLOPS": 202792957524218.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0634080171585083, "optim/lr": 0.002873964261164547, "optim/total_tokens": 6411517952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485334, "loss/out": 2.892073631286621, "created_at": "2025-01-16T16:06:03.223532+00:00"} {"global_step": 12230, "acc_step": 0, "speed/wps": 12905.025409552914, "speed/FLOPS": 202691157420264.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06535033881664276, "optim/lr": 0.0028739298814070818, "optim/total_tokens": 6412042240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8143014907836914, "created_at": "2025-01-16T16:06:13.383960+00:00"} {"global_step": 12231, "acc_step": 0, "speed/wps": 12906.626418639635, "speed/FLOPS": 202716303467988.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052392710000276566, "optim/lr": 0.0028738954971669287, "optim/total_tokens": 6412566528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9329686164855957, "created_at": "2025-01-16T16:06:23.542879+00:00"} {"global_step": 12232, "acc_step": 0, "speed/wps": 12911.841836809535, "speed/FLOPS": 202798218777082.75, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0631856769323349, "optim/lr": 0.0028738611084442, "optim/total_tokens": 6413090816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.8601131439208984, "created_at": "2025-01-16T16:06:33.700766+00:00"} {"global_step": 12233, "acc_step": 0, "speed/wps": 12908.07224284262, "speed/FLOPS": 202739012123867.4, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05457257851958275, "optim/lr": 0.002873826715239008, "optim/total_tokens": 6413615104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.920783281326294, "created_at": "2025-01-16T16:06:43.859646+00:00"} {"global_step": 12234, "acc_step": 0, "speed/wps": 12909.553574859257, "speed/FLOPS": 202762278478752.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06289993971586227, "optim/lr": 0.002873792317551465, "optim/total_tokens": 6414139392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.974201202392578, "created_at": "2025-01-16T16:06:54.018669+00:00"} {"global_step": 12235, "acc_step": 0, "speed/wps": 12904.387213790544, "speed/FLOPS": 202681133678844.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05120448023080826, "optim/lr": 0.0028737579153816843, "optim/total_tokens": 6414663680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.9625513553619385, "created_at": "2025-01-16T16:07:04.179486+00:00"} {"global_step": 12236, "acc_step": 0, "speed/wps": 12905.844896644065, "speed/FLOPS": 202704028591124.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06354159861803055, "optim/lr": 0.0028737235087297765, "optim/total_tokens": 6415187968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 2.894784927368164, "created_at": "2025-01-16T16:07:14.341798+00:00"} {"global_step": 12237, "acc_step": 0, "speed/wps": 12910.159168086851, "speed/FLOPS": 202771790152563.88, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04416793957352638, "optim/lr": 0.002873689097595855, "optim/total_tokens": 6415712256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.626438617706299, "created_at": "2025-01-16T16:07:24.502102+00:00"} {"global_step": 12238, "acc_step": 0, "speed/wps": 12911.980218076395, "speed/FLOPS": 202800392245034.56, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06450087577104568, "optim/lr": 0.0028736546819800313, "optim/total_tokens": 6416236544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 2.8571226596832275, "created_at": "2025-01-16T16:07:34.658071+00:00"} {"global_step": 12239, "acc_step": 0, "speed/wps": 12911.168591134909, "speed/FLOPS": 202787644528626.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056470245122909546, "optim/lr": 0.002873620261882418, "optim/total_tokens": 6416760832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.898020029067993, "created_at": "2025-01-16T16:07:44.818632+00:00"} {"global_step": 12240, "acc_step": 0, "speed/wps": 12911.673206086041, "speed/FLOPS": 202795570199847.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047402411699295044, "optim/lr": 0.0028735858373031276, "optim/total_tokens": 6417285120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 2.8891775608062744, "created_at": "2025-01-16T16:07:54.977409+00:00"} {"global_step": 12241, "acc_step": 0, "speed/wps": 12913.962348750494, "speed/FLOPS": 202831524331004.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043855201452970505, "optim/lr": 0.0028735514082422722, "optim/total_tokens": 6417809408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.9010767936706543, "created_at": "2025-01-16T16:08:05.130936+00:00"} {"global_step": 12242, "acc_step": 0, "speed/wps": 12904.936732175422, "speed/FLOPS": 202689764620195.9, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05702918395400047, "optim/lr": 0.0028735169746999644, "optim/total_tokens": 6418333696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 2.970778465270996, "created_at": "2025-01-16T16:08:15.293204+00:00"} {"global_step": 12243, "acc_step": 0, "speed/wps": 12908.984841348893, "speed/FLOPS": 202753345737450.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04902389273047447, "optim/lr": 0.0028734825366763155, "optim/total_tokens": 6418857984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.831526041030884, "created_at": "2025-01-16T16:08:25.450397+00:00"} {"global_step": 12244, "acc_step": 0, "speed/wps": 12907.846650965463, "speed/FLOPS": 202735468893440.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05189415067434311, "optim/lr": 0.0028734480941714395, "optim/total_tokens": 6419382272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.8571739196777344, "created_at": "2025-01-16T16:08:35.609013+00:00"} {"global_step": 12245, "acc_step": 0, "speed/wps": 12905.951598247639, "speed/FLOPS": 202705704486432.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04908082261681557, "optim/lr": 0.0028734136471854476, "optim/total_tokens": 6419906560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.854267120361328, "created_at": "2025-01-16T16:08:45.768632+00:00"} {"global_step": 12246, "acc_step": 0, "speed/wps": 12899.59349792384, "speed/FLOPS": 202605841783901.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.04234858974814415, "optim/lr": 0.0028733791957184524, "optim/total_tokens": 6420430848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.7492835521698, "created_at": "2025-01-16T16:08:55.936275+00:00"} {"global_step": 12247, "acc_step": 0, "speed/wps": 12909.919416788463, "speed/FLOPS": 202768024528969.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062117546796798706, "optim/lr": 0.002873344739770567, "optim/total_tokens": 6420955136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.7862212657928467, "created_at": "2025-01-16T16:09:06.092884+00:00"} {"global_step": 12248, "acc_step": 0, "speed/wps": 12907.829204546611, "speed/FLOPS": 202735194873459.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04900549724698067, "optim/lr": 0.0028733102793419025, "optim/total_tokens": 6421479424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8291738033294678, "created_at": "2025-01-16T16:09:16.253357+00:00"} {"global_step": 12249, "acc_step": 0, "speed/wps": 12909.8590294626, "speed/FLOPS": 202767076063036.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05694714933633804, "optim/lr": 0.0028732758144325724, "optim/total_tokens": 6422003712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 2.8921141624450684, "created_at": "2025-01-16T16:09:26.410455+00:00"} {"global_step": 12250, "acc_step": 0, "speed/wps": 12911.279263519782, "speed/FLOPS": 202789382790510.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052434422075748444, "optim/lr": 0.002873241345042689, "optim/total_tokens": 6422528000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8704681396484375, "created_at": "2025-01-16T16:09:36.567741+00:00"} {"global_step": 12251, "acc_step": 0, "speed/wps": 12909.093727410214, "speed/FLOPS": 202755055942668.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08110316842794418, "optim/lr": 0.002873206871172365, "optim/total_tokens": 6423052288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 2.8184831142425537, "created_at": "2025-01-16T16:09:46.734832+00:00"} {"global_step": 12252, "acc_step": 0, "speed/wps": 12911.645678269522, "speed/FLOPS": 202795137837661.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06739243865013123, "optim/lr": 0.0028731723928217116, "optim/total_tokens": 6423576576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9615325927734375, "created_at": "2025-01-16T16:09:56.895998+00:00"} {"global_step": 12253, "acc_step": 0, "speed/wps": 12907.809094520504, "speed/FLOPS": 202734879017864.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05262426659464836, "optim/lr": 0.0028731379099908427, "optim/total_tokens": 6424100864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 2.839921474456787, "created_at": "2025-01-16T16:10:07.055928+00:00"} {"global_step": 12254, "acc_step": 0, "speed/wps": 12907.2344728877, "speed/FLOPS": 202725853795508.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06966391205787659, "optim/lr": 0.0028731034226798704, "optim/total_tokens": 6424625152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8892900943756104, "created_at": "2025-01-16T16:10:17.216683+00:00"} {"global_step": 12255, "acc_step": 0, "speed/wps": 12911.36407896679, "speed/FLOPS": 202790714933654.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08462151885032654, "optim/lr": 0.0028730689308889066, "optim/total_tokens": 6425149440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 2.7946434020996094, "created_at": "2025-01-16T16:10:27.372024+00:00"} {"global_step": 12256, "acc_step": 0, "speed/wps": 12909.338019952698, "speed/FLOPS": 202758892892741.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05661286786198616, "optim/lr": 0.0028730344346180646, "optim/total_tokens": 6425673728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.8391852378845215, "created_at": "2025-01-16T16:10:37.529554+00:00"} {"global_step": 12257, "acc_step": 0, "speed/wps": 12901.151114606895, "speed/FLOPS": 202630306294298.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05272073298692703, "optim/lr": 0.0028729999338674567, "optim/total_tokens": 6426198016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9087395668029785, "created_at": "2025-01-16T16:10:47.695417+00:00"} {"global_step": 12258, "acc_step": 0, "speed/wps": 12910.62518735893, "speed/FLOPS": 202779109625609.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04789670929312706, "optim/lr": 0.0028729654286371955, "optim/total_tokens": 6426722304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.8310513496398926, "created_at": "2025-01-16T16:10:57.852585+00:00"} {"global_step": 12259, "acc_step": 0, "speed/wps": 12910.081095079384, "speed/FLOPS": 202770563908698.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05603279545903206, "optim/lr": 0.0028729309189273935, "optim/total_tokens": 6427246592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.938413619995117, "created_at": "2025-01-16T16:11:08.010537+00:00"} {"global_step": 12260, "acc_step": 0, "speed/wps": 12910.615988419786, "speed/FLOPS": 202778965143629.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06121096760034561, "optim/lr": 0.0028728964047381633, "optim/total_tokens": 6427770880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.8573172092437744, "created_at": "2025-01-16T16:11:18.169940+00:00"} {"global_step": 12261, "acc_step": 0, "speed/wps": 12907.300589882912, "speed/FLOPS": 202726892253772.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05771905556321144, "optim/lr": 0.0028728618860696174, "optim/total_tokens": 6428295168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457122, "loss/out": 2.7169241905212402, "created_at": "2025-01-16T16:11:28.330888+00:00"} {"global_step": 12262, "acc_step": 0, "speed/wps": 12903.125797753084, "speed/FLOPS": 202661321406608.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04083783179521561, "optim/lr": 0.002872827362921869, "optim/total_tokens": 6428819456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.943171262741089, "created_at": "2025-01-16T16:11:38.494338+00:00"} {"global_step": 12263, "acc_step": 0, "speed/wps": 12908.648170433413, "speed/FLOPS": 202748057858088.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05117648094892502, "optim/lr": 0.0028727928352950296, "optim/total_tokens": 6429343744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9829089641571045, "created_at": "2025-01-16T16:11:48.653300+00:00"} {"global_step": 12264, "acc_step": 0, "speed/wps": 12911.010496941331, "speed/FLOPS": 202785161442071.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0469285249710083, "optim/lr": 0.002872758303189213, "optim/total_tokens": 6429868032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.9965767860412598, "created_at": "2025-01-16T16:11:58.811210+00:00"} {"global_step": 12265, "acc_step": 0, "speed/wps": 12905.93970639568, "speed/FLOPS": 202705517708556.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04489802569150925, "optim/lr": 0.0028727237666045314, "optim/total_tokens": 6430392320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.9134578704833984, "created_at": "2025-01-16T16:12:08.971458+00:00"} {"global_step": 12266, "acc_step": 0, "speed/wps": 12910.615045668426, "speed/FLOPS": 202778950336423.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0556185320019722, "optim/lr": 0.002872689225541098, "optim/total_tokens": 6430916608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.854530096054077, "created_at": "2025-01-16T16:12:19.132322+00:00"} {"global_step": 12267, "acc_step": 0, "speed/wps": 12907.489266278972, "speed/FLOPS": 202729855675841.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051911868155002594, "optim/lr": 0.0028726546799990238, "optim/total_tokens": 6431440896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.879931926727295, "created_at": "2025-01-16T16:12:29.295583+00:00"} {"global_step": 12268, "acc_step": 0, "speed/wps": 12907.890447757736, "speed/FLOPS": 202736156782246.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07749336212873459, "optim/lr": 0.002872620129978423, "optim/total_tokens": 6431965184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 2.8612868785858154, "created_at": "2025-01-16T16:12:39.456313+00:00"} {"global_step": 12269, "acc_step": 0, "speed/wps": 12907.814136523284, "speed/FLOPS": 202734958209447.0, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046123042702674866, "optim/lr": 0.0028725855754794085, "optim/total_tokens": 6432489472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8934640884399414, "created_at": "2025-01-16T16:12:49.615275+00:00"} {"global_step": 12270, "acc_step": 0, "speed/wps": 12913.784926046883, "speed/FLOPS": 202828737663639.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08099588751792908, "optim/lr": 0.0028725510165020928, "optim/total_tokens": 6433013760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375384, "loss/out": 2.7014870643615723, "created_at": "2025-01-16T16:12:59.770152+00:00"} {"global_step": 12271, "acc_step": 0, "speed/wps": 12909.517125143087, "speed/FLOPS": 202761705985872.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05229005217552185, "optim/lr": 0.002872516453046588, "optim/total_tokens": 6433538048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.957298517227173, "created_at": "2025-01-16T16:13:09.927137+00:00"} {"global_step": 12272, "acc_step": 0, "speed/wps": 12909.029516156781, "speed/FLOPS": 202754047416697.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0754975825548172, "optim/lr": 0.002872481885113007, "optim/total_tokens": 6434062336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 2.912470817565918, "created_at": "2025-01-16T16:13:20.086157+00:00"} {"global_step": 12273, "acc_step": 0, "speed/wps": 12904.71872465282, "speed/FLOPS": 202686340512479.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050927937030792236, "optim/lr": 0.0028724473127014634, "optim/total_tokens": 6434586624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.977083444595337, "created_at": "2025-01-16T16:13:30.253989+00:00"} {"global_step": 12274, "acc_step": 0, "speed/wps": 12910.315460235694, "speed/FLOPS": 202774244935529.34, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10242770612239838, "optim/lr": 0.0028724127358120692, "optim/total_tokens": 6435110912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 2.801029920578003, "created_at": "2025-01-16T16:13:40.410370+00:00"} {"global_step": 12275, "acc_step": 0, "speed/wps": 12907.007514710995, "speed/FLOPS": 202722289105470.47, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04901628568768501, "optim/lr": 0.002872378154444938, "optim/total_tokens": 6435635200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8092150688171387, "created_at": "2025-01-16T16:13:50.569084+00:00"} {"global_step": 12276, "acc_step": 0, "speed/wps": 12906.367199574705, "speed/FLOPS": 202712232076369.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07181236147880554, "optim/lr": 0.0028723435686001816, "optim/total_tokens": 6436159488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.825161933898926, "created_at": "2025-01-16T16:14:00.731119+00:00"} {"global_step": 12277, "acc_step": 0, "speed/wps": 12912.692153103419, "speed/FLOPS": 202811574162935.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043479666113853455, "optim/lr": 0.0028723089782779127, "optim/total_tokens": 6436683776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.9642608165740967, "created_at": "2025-01-16T16:14:10.886754+00:00"} {"global_step": 12278, "acc_step": 0, "speed/wps": 12906.68456823862, "speed/FLOPS": 202717216787346.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05641566216945648, "optim/lr": 0.002872274383478246, "optim/total_tokens": 6437208064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.812452793121338, "created_at": "2025-01-16T16:14:21.048674+00:00"} {"global_step": 12279, "acc_step": 0, "speed/wps": 12908.259280747021, "speed/FLOPS": 202741949811173.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04557003825902939, "optim/lr": 0.002872239784201293, "optim/total_tokens": 6437732352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 2.8892500400543213, "created_at": "2025-01-16T16:14:31.207539+00:00"} {"global_step": 12280, "acc_step": 0, "speed/wps": 12904.207866379158, "speed/FLOPS": 202678316781299.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04665330424904823, "optim/lr": 0.0028722051804471664, "optim/total_tokens": 6438256640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9218807220458984, "created_at": "2025-01-16T16:14:41.370827+00:00"} {"global_step": 12281, "acc_step": 0, "speed/wps": 12907.711109501239, "speed/FLOPS": 202733340028491.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047010187059640884, "optim/lr": 0.0028721705722159797, "optim/total_tokens": 6438780928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 2.7369446754455566, "created_at": "2025-01-16T16:14:51.532831+00:00"} {"global_step": 12282, "acc_step": 0, "speed/wps": 12906.11870428198, "speed/FLOPS": 202708329116329.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04431089386343956, "optim/lr": 0.002872135959507845, "optim/total_tokens": 6439305216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.8671765327453613, "created_at": "2025-01-16T16:15:01.692128+00:00"} {"global_step": 12283, "acc_step": 0, "speed/wps": 12904.825904621, "speed/FLOPS": 202688023921160.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0473535917699337, "optim/lr": 0.002872101342322877, "optim/total_tokens": 6439829504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445406, "loss/out": 3.0512380599975586, "created_at": "2025-01-16T16:15:11.854142+00:00"} {"global_step": 12284, "acc_step": 0, "speed/wps": 12912.262506482979, "speed/FLOPS": 202804825972364.9, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05940201133489609, "optim/lr": 0.0028720667206611865, "optim/total_tokens": 6440353792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 3.001594066619873, "created_at": "2025-01-16T16:15:22.008696+00:00"} {"global_step": 12285, "acc_step": 0, "speed/wps": 12914.78100634875, "speed/FLOPS": 202844382473538.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05577319860458374, "optim/lr": 0.002872032094522888, "optim/total_tokens": 6440878080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.839052677154541, "created_at": "2025-01-16T16:15:32.161274+00:00"} {"global_step": 12286, "acc_step": 0, "speed/wps": 12903.901898268023, "speed/FLOPS": 202673511131669.38, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05939073488116264, "optim/lr": 0.002871997463908093, "optim/total_tokens": 6441402368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.7803382873535156, "created_at": "2025-01-16T16:15:42.326437+00:00"} {"global_step": 12287, "acc_step": 0, "speed/wps": 12902.688083657253, "speed/FLOPS": 202654446505251.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07655907422304153, "optim/lr": 0.002871962828816916, "optim/total_tokens": 6441926656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485528, "loss/out": 2.8729443550109863, "created_at": "2025-01-16T16:15:52.491602+00:00"} {"global_step": 12288, "acc_step": 0, "speed/wps": 12903.145229531036, "speed/FLOPS": 202661626609383.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04952193796634674, "optim/lr": 0.00287192818924947, "optim/total_tokens": 6442450944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.914292812347412, "created_at": "2025-01-16T16:16:02.656149+00:00"} {"global_step": 12289, "acc_step": 0, "speed/wps": 12903.792809911301, "speed/FLOPS": 202671797749123.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0777120292186737, "optim/lr": 0.0028718935452058665, "optim/total_tokens": 6442975232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.88790225982666, "created_at": "2025-01-16T16:16:12.820137+00:00"} {"global_step": 12290, "acc_step": 0, "speed/wps": 12904.896604758003, "speed/FLOPS": 202689134363964.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056341469287872314, "optim/lr": 0.0028718588966862198, "optim/total_tokens": 6443499520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 2.9430127143859863, "created_at": "2025-01-16T16:16:22.983211+00:00"} {"global_step": 12291, "acc_step": 0, "speed/wps": 12909.180141095552, "speed/FLOPS": 202756413188341.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057986196130514145, "optim/lr": 0.0028718242436906426, "optim/total_tokens": 6444023808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.985746145248413, "created_at": "2025-01-16T16:16:33.146943+00:00"} {"global_step": 12292, "acc_step": 0, "speed/wps": 12909.04361489411, "speed/FLOPS": 202754268856741.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05414910987019539, "optim/lr": 0.0028717895862192476, "optim/total_tokens": 6444548096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442200, "loss/out": 2.8840889930725098, "created_at": "2025-01-16T16:16:43.307342+00:00"} {"global_step": 12293, "acc_step": 0, "speed/wps": 12909.340745812782, "speed/FLOPS": 202758935706119.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0534638948738575, "optim/lr": 0.002871754924272149, "optim/total_tokens": 6445072384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.8441169261932373, "created_at": "2025-01-16T16:16:53.465965+00:00"} {"global_step": 12294, "acc_step": 0, "speed/wps": 12909.419201816869, "speed/FLOPS": 202760167965473.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05528838187456131, "optim/lr": 0.0028717202578494583, "optim/total_tokens": 6445596672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 2.8798153400421143, "created_at": "2025-01-16T16:17:03.623098+00:00"} {"global_step": 12295, "acc_step": 0, "speed/wps": 12901.150208716055, "speed/FLOPS": 202630292066038.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054829832166433334, "optim/lr": 0.00287168558695129, "optim/total_tokens": 6446120960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.838387966156006, "created_at": "2025-01-16T16:17:13.787592+00:00"} {"global_step": 12296, "acc_step": 0, "speed/wps": 12912.252222429948, "speed/FLOPS": 202804664447180.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.043152615427970886, "optim/lr": 0.0028716509115777563, "optim/total_tokens": 6446645248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.860351800918579, "created_at": "2025-01-16T16:17:23.942133+00:00"} {"global_step": 12297, "acc_step": 0, "speed/wps": 12905.378237809524, "speed/FLOPS": 202696699072871.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06233399733901024, "optim/lr": 0.0028716162317289707, "optim/total_tokens": 6447169536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8966879844665527, "created_at": "2025-01-16T16:17:34.104834+00:00"} {"global_step": 12298, "acc_step": 0, "speed/wps": 12905.276189274071, "speed/FLOPS": 202695096260393.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04773251712322235, "optim/lr": 0.0028715815474050464, "optim/total_tokens": 6447693824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.8980722427368164, "created_at": "2025-01-16T16:17:44.269278+00:00"} {"global_step": 12299, "acc_step": 0, "speed/wps": 12906.967414660041, "speed/FLOPS": 202721659279067.0, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05115307867527008, "optim/lr": 0.0028715468586060964, "optim/total_tokens": 6448218112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.899672269821167, "created_at": "2025-01-16T16:17:54.427991+00:00"} {"global_step": 12300, "acc_step": 0, "speed/wps": 12906.317441992294, "speed/FLOPS": 202711450565164.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04511338844895363, "optim/lr": 0.002871512165332234, "optim/total_tokens": 6448742400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9100890159606934, "created_at": "2025-01-16T16:18:04.592003+00:00"} {"global_step": 12301, "acc_step": 0, "speed/wps": 12907.15827239348, "speed/FLOPS": 202724656962036.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06334786117076874, "optim/lr": 0.0028714774675835724, "optim/total_tokens": 6449266688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 2.760786533355713, "created_at": "2025-01-16T16:18:14.752349+00:00"} {"global_step": 12302, "acc_step": 0, "speed/wps": 12911.213170903166, "speed/FLOPS": 202788344715145.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050310637801885605, "optim/lr": 0.002871442765360225, "optim/total_tokens": 6449790976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8347702026367188, "created_at": "2025-01-16T16:18:24.910192+00:00"} {"global_step": 12303, "acc_step": 0, "speed/wps": 12909.50076836878, "speed/FLOPS": 202761449080255.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047386154532432556, "optim/lr": 0.002871408058662305, "optim/total_tokens": 6450315264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372023, "loss/out": 2.7458741664886475, "created_at": "2025-01-16T16:18:35.074069+00:00"} {"global_step": 12304, "acc_step": 0, "speed/wps": 12907.214925806096, "speed/FLOPS": 202725546781731.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05148349702358246, "optim/lr": 0.002871373347489925, "optim/total_tokens": 6450839552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 2.9007930755615234, "created_at": "2025-01-16T16:18:45.243252+00:00"} {"global_step": 12305, "acc_step": 0, "speed/wps": 12912.016649718193, "speed/FLOPS": 202800964454031.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04901738092303276, "optim/lr": 0.002871338631843198, "optim/total_tokens": 6451363840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.9219701290130615, "created_at": "2025-01-16T16:18:55.398762+00:00"} {"global_step": 12306, "acc_step": 0, "speed/wps": 12910.253543833105, "speed/FLOPS": 202773272453344.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05535384640097618, "optim/lr": 0.002871303911722239, "optim/total_tokens": 6451888128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.8863677978515625, "created_at": "2025-01-16T16:19:05.569755+00:00"} {"global_step": 12307, "acc_step": 0, "speed/wps": 12907.153344540755, "speed/FLOPS": 202724579563338.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055349286645650864, "optim/lr": 0.00287126918712716, "optim/total_tokens": 6452412416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.9195117950439453, "created_at": "2025-01-16T16:19:15.729500+00:00"} {"global_step": 12308, "acc_step": 0, "speed/wps": 12904.949332172075, "speed/FLOPS": 202689962520457.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059416577219963074, "optim/lr": 0.0028712344580580744, "optim/total_tokens": 6452936704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.873668909072876, "created_at": "2025-01-16T16:19:25.892954+00:00"} {"global_step": 12309, "acc_step": 0, "speed/wps": 12913.199628864459, "speed/FLOPS": 202819544767107.12, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05076175555586815, "optim/lr": 0.0028711997245150955, "optim/total_tokens": 6453460992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.935042381286621, "created_at": "2025-01-16T16:19:36.047147+00:00"} {"global_step": 12310, "acc_step": 0, "speed/wps": 12910.729714142195, "speed/FLOPS": 202780751362375.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04704851657152176, "optim/lr": 0.0028711649864983364, "optim/total_tokens": 6453985280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.9396986961364746, "created_at": "2025-01-16T16:19:46.204425+00:00"} {"global_step": 12311, "acc_step": 0, "speed/wps": 12906.582217340903, "speed/FLOPS": 202715609225853.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05360914394259453, "optim/lr": 0.0028711302440079115, "optim/total_tokens": 6454509568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.935664415359497, "created_at": "2025-01-16T16:19:56.363412+00:00"} {"global_step": 12312, "acc_step": 0, "speed/wps": 12904.040160271097, "speed/FLOPS": 202675682726419.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05156954750418663, "optim/lr": 0.0028710954970439335, "optim/total_tokens": 6455033856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476786, "loss/out": 2.852956771850586, "created_at": "2025-01-16T16:20:06.524423+00:00"} {"global_step": 12313, "acc_step": 0, "speed/wps": 12901.296644192285, "speed/FLOPS": 202632592036415.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06293212622404099, "optim/lr": 0.0028710607456065153, "optim/total_tokens": 6455558144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.9341416358947754, "created_at": "2025-01-16T16:20:16.692719+00:00"} {"global_step": 12314, "acc_step": 0, "speed/wps": 12906.47591365222, "speed/FLOPS": 202713939580346.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04852821305394173, "optim/lr": 0.0028710259896957712, "optim/total_tokens": 6456082432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.8416614532470703, "created_at": "2025-01-16T16:20:26.856492+00:00"} {"global_step": 12315, "acc_step": 0, "speed/wps": 12905.03289060494, "speed/FLOPS": 202691274920467.0, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056584253907203674, "optim/lr": 0.0028709912293118133, "optim/total_tokens": 6456606720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.916205883026123, "created_at": "2025-01-16T16:20:37.016728+00:00"} {"global_step": 12316, "acc_step": 0, "speed/wps": 12900.243838797913, "speed/FLOPS": 202616056280990.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05262523889541626, "optim/lr": 0.0028709564644547564, "optim/total_tokens": 6457131008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.952486991882324, "created_at": "2025-01-16T16:20:47.181126+00:00"} {"global_step": 12317, "acc_step": 0, "speed/wps": 12909.978423850478, "speed/FLOPS": 202768951315963.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05584512650966644, "optim/lr": 0.002870921695124713, "optim/total_tokens": 6457655296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.773392677307129, "created_at": "2025-01-16T16:20:57.346691+00:00"} {"global_step": 12318, "acc_step": 0, "speed/wps": 12904.39203304157, "speed/FLOPS": 202681209371803.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0686279684305191, "optim/lr": 0.0028708869213217975, "optim/total_tokens": 6458179584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.8931925296783447, "created_at": "2025-01-16T16:21:07.507298+00:00"} {"global_step": 12319, "acc_step": 0, "speed/wps": 12901.461674729191, "speed/FLOPS": 202635184067773.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05118643864989281, "optim/lr": 0.0028708521430461224, "optim/total_tokens": 6458703872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340384, "loss/out": 2.9100797176361084, "created_at": "2025-01-16T16:21:17.671599+00:00"} {"global_step": 12320, "acc_step": 0, "speed/wps": 12910.801829193546, "speed/FLOPS": 202781884028353.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057683903723955154, "optim/lr": 0.0028708173602978015, "optim/total_tokens": 6459228160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.85662841796875, "created_at": "2025-01-16T16:21:27.828247+00:00"} {"global_step": 12321, "acc_step": 0, "speed/wps": 12900.700634103072, "speed/FLOPS": 202623230878960.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04318162798881531, "optim/lr": 0.0028707825730769475, "optim/total_tokens": 6459752448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.705265522003174, "created_at": "2025-01-16T16:21:37.994026+00:00"} {"global_step": 12322, "acc_step": 0, "speed/wps": 12904.192922761344, "speed/FLOPS": 202678082071246.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05264043062925339, "optim/lr": 0.0028707477813836755, "optim/total_tokens": 6460276736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.9450933933258057, "created_at": "2025-01-16T16:21:48.155490+00:00"} {"global_step": 12323, "acc_step": 0, "speed/wps": 12903.80625004998, "speed/FLOPS": 202672008844970.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07312562316656113, "optim/lr": 0.002870712985218098, "optim/total_tokens": 6460801024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387212, "loss/out": 2.94269061088562, "created_at": "2025-01-16T16:21:58.318560+00:00"} {"global_step": 12324, "acc_step": 0, "speed/wps": 12905.258584130092, "speed/FLOPS": 202694819747414.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06111394986510277, "optim/lr": 0.002870678184580329, "optim/total_tokens": 6461325312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.833407402038574, "created_at": "2025-01-16T16:22:08.481824+00:00"} {"global_step": 12325, "acc_step": 0, "speed/wps": 12906.616724723492, "speed/FLOPS": 202716151211715.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04453153908252716, "optim/lr": 0.0028706433794704815, "optim/total_tokens": 6461849600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9810986518859863, "created_at": "2025-01-16T16:22:18.643327+00:00"} {"global_step": 12326, "acc_step": 0, "speed/wps": 12913.092281321604, "speed/FLOPS": 202817858726435.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05125187337398529, "optim/lr": 0.00287060856988867, "optim/total_tokens": 6462373888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.944688320159912, "created_at": "2025-01-16T16:22:28.799517+00:00"} {"global_step": 12327, "acc_step": 0, "speed/wps": 12911.57658502136, "speed/FLOPS": 202794052633255.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05141789838671684, "optim/lr": 0.0028705737558350065, "optim/total_tokens": 6462898176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 2.9254672527313232, "created_at": "2025-01-16T16:22:38.958129+00:00"} {"global_step": 12328, "acc_step": 0, "speed/wps": 12909.228508114258, "speed/FLOPS": 202757172858833.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05472497642040253, "optim/lr": 0.0028705389373096055, "optim/total_tokens": 6463422464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9563283920288086, "created_at": "2025-01-16T16:22:49.115091+00:00"} {"global_step": 12329, "acc_step": 0, "speed/wps": 12911.911693409002, "speed/FLOPS": 202799315970970.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04534820839762688, "optim/lr": 0.002870504114312581, "optim/total_tokens": 6463946752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.7463510036468506, "created_at": "2025-01-16T16:22:59.270678+00:00"} {"global_step": 12330, "acc_step": 0, "speed/wps": 12905.613184917247, "speed/FLOPS": 202700389240049.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054151587188243866, "optim/lr": 0.002870469286844046, "optim/total_tokens": 6464471040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8768482208251953, "created_at": "2025-01-16T16:23:09.433644+00:00"} {"global_step": 12331, "acc_step": 0, "speed/wps": 12905.137022732419, "speed/FLOPS": 202692910458621.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04925073683261871, "optim/lr": 0.0028704344549041146, "optim/total_tokens": 6464995328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.896829605102539, "created_at": "2025-01-16T16:23:19.597424+00:00"} {"global_step": 12332, "acc_step": 0, "speed/wps": 12910.646935849265, "speed/FLOPS": 202779451215535.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050788480788469315, "optim/lr": 0.0028703996184928997, "optim/total_tokens": 6465519616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 2.9283981323242188, "created_at": "2025-01-16T16:23:29.755423+00:00"} {"global_step": 12333, "acc_step": 0, "speed/wps": 12902.376117223304, "speed/FLOPS": 202649546643719.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050736796110868454, "optim/lr": 0.002870364777610516, "optim/total_tokens": 6466043904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.9439961910247803, "created_at": "2025-01-16T16:23:39.924957+00:00"} {"global_step": 12334, "acc_step": 0, "speed/wps": 12901.841117519514, "speed/FLOPS": 202641143738204.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056553978472948074, "optim/lr": 0.0028703299322570767, "optim/total_tokens": 6466568192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312903, "loss/out": 2.865481376647949, "created_at": "2025-01-16T16:23:50.090451+00:00"} {"global_step": 12335, "acc_step": 0, "speed/wps": 12898.791667284007, "speed/FLOPS": 202593247931872.2, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04265493527054787, "optim/lr": 0.002870295082432695, "optim/total_tokens": 6467092480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.871760368347168, "created_at": "2025-01-16T16:24:00.255964+00:00"} {"global_step": 12336, "acc_step": 0, "speed/wps": 12910.478960480334, "speed/FLOPS": 202776812931544.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.039223309606313705, "optim/lr": 0.002870260228137485, "optim/total_tokens": 6467616768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7555878162384033, "created_at": "2025-01-16T16:24:10.411939+00:00"} {"global_step": 12337, "acc_step": 0, "speed/wps": 12901.617936133649, "speed/FLOPS": 202637638367856.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05072096362709999, "optim/lr": 0.0028702253693715604, "optim/total_tokens": 6468141056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 2.8359427452087402, "created_at": "2025-01-16T16:24:20.576029+00:00"} {"global_step": 12338, "acc_step": 0, "speed/wps": 12909.839513510002, "speed/FLOPS": 202766769538183.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05693046748638153, "optim/lr": 0.0028701905061350353, "optim/total_tokens": 6468665344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.830665111541748, "created_at": "2025-01-16T16:24:30.737462+00:00"} {"global_step": 12339, "acc_step": 0, "speed/wps": 12904.931292458701, "speed/FLOPS": 202689679181969.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044235341250896454, "optim/lr": 0.002870155638428023, "optim/total_tokens": 6469189632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 2.900928020477295, "created_at": "2025-01-16T16:24:40.897844+00:00"} {"global_step": 12340, "acc_step": 0, "speed/wps": 12906.4484352153, "speed/FLOPS": 202713507993734.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04139318689703941, "optim/lr": 0.002870120766250637, "optim/total_tokens": 6469713920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.842817544937134, "created_at": "2025-01-16T16:24:51.056941+00:00"} {"global_step": 12341, "acc_step": 0, "speed/wps": 12907.389654881152, "speed/FLOPS": 202728291141958.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.043135277926921844, "optim/lr": 0.002870085889602992, "optim/total_tokens": 6470238208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 2.872979164123535, "created_at": "2025-01-16T16:25:01.219394+00:00"} {"global_step": 12342, "acc_step": 0, "speed/wps": 12902.185034217182, "speed/FLOPS": 202646545422532.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04479010030627251, "optim/lr": 0.0028700510084852006, "optim/total_tokens": 6470762496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8194313049316406, "created_at": "2025-01-16T16:25:11.382599+00:00"} {"global_step": 12343, "acc_step": 0, "speed/wps": 12904.273035392263, "speed/FLOPS": 202679340350201.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04651375859975815, "optim/lr": 0.002870016122897378, "optim/total_tokens": 6471286784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.7980895042419434, "created_at": "2025-01-16T16:25:21.546103+00:00"} {"global_step": 12344, "acc_step": 0, "speed/wps": 12905.433473255438, "speed/FLOPS": 202697566621451.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05019281804561615, "optim/lr": 0.002869981232839637, "optim/total_tokens": 6471811072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9490067958831787, "created_at": "2025-01-16T16:25:31.707563+00:00"} {"global_step": 12345, "acc_step": 0, "speed/wps": 12905.078316704683, "speed/FLOPS": 202691988399785.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05131883546710014, "optim/lr": 0.0028699463383120914, "optim/total_tokens": 6472335360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.855929374694824, "created_at": "2025-01-16T16:25:41.875777+00:00"} {"global_step": 12346, "acc_step": 0, "speed/wps": 12906.586837015815, "speed/FLOPS": 202715681784195.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0466560460627079, "optim/lr": 0.0028699114393148555, "optim/total_tokens": 6472859648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 2.925015449523926, "created_at": "2025-01-16T16:25:52.038991+00:00"} {"global_step": 12347, "acc_step": 0, "speed/wps": 12905.292032406322, "speed/FLOPS": 202695345098555.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04806336387991905, "optim/lr": 0.0028698765358480434, "optim/total_tokens": 6473383936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8402202129364014, "created_at": "2025-01-16T16:26:02.199969+00:00"} {"global_step": 12348, "acc_step": 0, "speed/wps": 12910.315457867082, "speed/FLOPS": 202774244898327.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041680317372083664, "optim/lr": 0.002869841627911768, "optim/total_tokens": 6473908224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 2.9179039001464844, "created_at": "2025-01-16T16:26:12.362300+00:00"} {"global_step": 12349, "acc_step": 0, "speed/wps": 12907.37174447761, "speed/FLOPS": 202728009834459.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060652099549770355, "optim/lr": 0.0028698067155061443, "optim/total_tokens": 6474432512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.8737733364105225, "created_at": "2025-01-16T16:26:22.523544+00:00"} {"global_step": 12350, "acc_step": 0, "speed/wps": 12911.530846643127, "speed/FLOPS": 202793334249174.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05820757523179054, "optim/lr": 0.002869771798631285, "optim/total_tokens": 6474956800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.926011562347412, "created_at": "2025-01-16T16:26:32.678629+00:00"} {"global_step": 12351, "acc_step": 0, "speed/wps": 12904.249304041055, "speed/FLOPS": 202678967616720.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05916424095630646, "optim/lr": 0.002869736877287305, "optim/total_tokens": 6475481088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8621597290039062, "created_at": "2025-01-16T16:26:42.842597+00:00"} {"global_step": 12352, "acc_step": 0, "speed/wps": 12908.478879586055, "speed/FLOPS": 202745398912700.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055710289627313614, "optim/lr": 0.0028697019514743187, "optim/total_tokens": 6476005376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8685102462768555, "created_at": "2025-01-16T16:26:53.003138+00:00"} {"global_step": 12353, "acc_step": 0, "speed/wps": 12905.856538605325, "speed/FLOPS": 202704211444124.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.049515463411808014, "optim/lr": 0.0028696670211924387, "optim/total_tokens": 6476529664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.849440336227417, "created_at": "2025-01-16T16:27:03.163303+00:00"} {"global_step": 12354, "acc_step": 0, "speed/wps": 12890.812634802613, "speed/FLOPS": 202467926262412.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046321578323841095, "optim/lr": 0.0028696320864417795, "optim/total_tokens": 6477053952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 2.7351388931274414, "created_at": "2025-01-16T16:27:13.335218+00:00"} {"global_step": 12355, "acc_step": 0, "speed/wps": 12892.589875712873, "speed/FLOPS": 202495840273092.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0024, "optim/grad_norm": 0.059059105813503265, "optim/lr": 0.0028695971472224554, "optim/total_tokens": 6477578240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9045019149780273, "created_at": "2025-01-16T16:27:23.514767+00:00"} {"global_step": 12356, "acc_step": 0, "speed/wps": 12895.36909112505, "speed/FLOPS": 202539491670183.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04599815979599953, "optim/lr": 0.00286956220353458, "optim/total_tokens": 6478102528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 2.9371447563171387, "created_at": "2025-01-16T16:27:33.683521+00:00"} {"global_step": 12357, "acc_step": 0, "speed/wps": 12891.927743513068, "speed/FLOPS": 202485440577029.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06762079149484634, "optim/lr": 0.0028695272553782673, "optim/total_tokens": 6478626816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9287002086639404, "created_at": "2025-01-16T16:27:43.855624+00:00"} {"global_step": 12358, "acc_step": 0, "speed/wps": 12904.006527996862, "speed/FLOPS": 202675154485337.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06168780475854874, "optim/lr": 0.002869492302753632, "optim/total_tokens": 6479151104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483184, "loss/out": 2.839567184448242, "created_at": "2025-01-16T16:27:54.020226+00:00"} {"global_step": 12359, "acc_step": 0, "speed/wps": 12903.820576369473, "speed/FLOPS": 202672233859504.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05531727150082588, "optim/lr": 0.0028694573456607875, "optim/total_tokens": 6479675392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.94541335105896, "created_at": "2025-01-16T16:28:04.181604+00:00"} {"global_step": 12360, "acc_step": 0, "speed/wps": 12902.934596721952, "speed/FLOPS": 202658318331676.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06349611282348633, "optim/lr": 0.0028694223840998474, "optim/total_tokens": 6480199680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378926, "loss/out": 2.815217971801758, "created_at": "2025-01-16T16:28:14.343704+00:00"} {"global_step": 12361, "acc_step": 0, "speed/wps": 12908.04630376771, "speed/FLOPS": 202738604715051.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049406345933675766, "optim/lr": 0.002869387418070927, "optim/total_tokens": 6480723968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.94472599029541, "created_at": "2025-01-16T16:28:24.502518+00:00"} {"global_step": 12362, "acc_step": 0, "speed/wps": 12906.200477677257, "speed/FLOPS": 202709613479870.53, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05550282076001167, "optim/lr": 0.002869352447574139, "optim/total_tokens": 6481248256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.886234760284424, "created_at": "2025-01-16T16:28:34.662156+00:00"} {"global_step": 12363, "acc_step": 0, "speed/wps": 12907.362334729003, "speed/FLOPS": 202727862041427.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06506196409463882, "optim/lr": 0.002869317472609599, "optim/total_tokens": 6481772544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.9100379943847656, "created_at": "2025-01-16T16:28:44.825116+00:00"} {"global_step": 12364, "acc_step": 0, "speed/wps": 12906.524266327124, "speed/FLOPS": 202714699025548.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057262931019067764, "optim/lr": 0.00286928249317742, "optim/total_tokens": 6482296832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.0314812660217285, "created_at": "2025-01-16T16:28:54.985973+00:00"} {"global_step": 12365, "acc_step": 0, "speed/wps": 12899.687439039995, "speed/FLOPS": 202607317258219.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06550506502389908, "optim/lr": 0.0028692475092777165, "optim/total_tokens": 6482821120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 2.820990562438965, "created_at": "2025-01-16T16:29:05.152786+00:00"} {"global_step": 12366, "acc_step": 0, "speed/wps": 12906.170938676223, "speed/FLOPS": 202709149529268.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05658363923430443, "optim/lr": 0.0028692125209106023, "optim/total_tokens": 6483345408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 2.6655914783477783, "created_at": "2025-01-16T16:29:15.313884+00:00"} {"global_step": 12367, "acc_step": 0, "speed/wps": 12900.646844011637, "speed/FLOPS": 202622386031659.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05265723168849945, "optim/lr": 0.0028691775280761922, "optim/total_tokens": 6483869696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 2.8424503803253174, "created_at": "2025-01-16T16:29:25.480382+00:00"} {"global_step": 12368, "acc_step": 0, "speed/wps": 12906.922256976699, "speed/FLOPS": 202720950015596.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04865559935569763, "optim/lr": 0.0028691425307746004, "optim/total_tokens": 6484393984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.893557071685791, "created_at": "2025-01-16T16:29:35.641265+00:00"} {"global_step": 12369, "acc_step": 0, "speed/wps": 12910.86110962925, "speed/FLOPS": 202782815109056.0, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044443678110837936, "optim/lr": 0.0028691075290059403, "optim/total_tokens": 6484918272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.8236799240112305, "created_at": "2025-01-16T16:29:45.801402+00:00"} {"global_step": 12370, "acc_step": 0, "speed/wps": 12903.838343226762, "speed/FLOPS": 202672512912412.62, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04811399430036545, "optim/lr": 0.0028690725227703264, "optim/total_tokens": 6485442560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8312740325927734, "created_at": "2025-01-16T16:29:55.972543+00:00"} {"global_step": 12371, "acc_step": 0, "speed/wps": 12909.021218199296, "speed/FLOPS": 202753917085872.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04139291122555733, "optim/lr": 0.002869037512067873, "optim/total_tokens": 6485966848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.0322515964508057, "created_at": "2025-01-16T16:30:06.130668+00:00"} {"global_step": 12372, "acc_step": 0, "speed/wps": 12912.88241955436, "speed/FLOPS": 202814562559000.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053491055965423584, "optim/lr": 0.0028690024968986948, "optim/total_tokens": 6486491136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.7555065155029297, "created_at": "2025-01-16T16:30:16.288053+00:00"} {"global_step": 12373, "acc_step": 0, "speed/wps": 12904.00251712382, "speed/FLOPS": 202675091489064.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057617105543613434, "optim/lr": 0.0028689674772629055, "optim/total_tokens": 6487015424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.7429392337799072, "created_at": "2025-01-16T16:30:26.450452+00:00"} {"global_step": 12374, "acc_step": 0, "speed/wps": 12903.393224663034, "speed/FLOPS": 202665521713712.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06069448962807655, "optim/lr": 0.0028689324531606195, "optim/total_tokens": 6487539712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.818047523498535, "created_at": "2025-01-16T16:30:36.614659+00:00"} {"global_step": 12375, "acc_step": 0, "speed/wps": 12911.204437806906, "speed/FLOPS": 202788207549868.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049081820994615555, "optim/lr": 0.00286889742459195, "optim/total_tokens": 6488064000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.892467498779297, "created_at": "2025-01-16T16:30:46.771876+00:00"} {"global_step": 12376, "acc_step": 0, "speed/wps": 12908.48423941606, "speed/FLOPS": 202745483096195.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04577725753188133, "optim/lr": 0.0028688623915570136, "optim/total_tokens": 6488588288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.8106529712677, "created_at": "2025-01-16T16:30:56.929271+00:00"} {"global_step": 12377, "acc_step": 0, "speed/wps": 12900.663801553548, "speed/FLOPS": 202622652373157.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05554305389523506, "optim/lr": 0.0028688273540559226, "optim/total_tokens": 6489112576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.5893797874450684, "created_at": "2025-01-16T16:31:07.095399+00:00"} {"global_step": 12378, "acc_step": 0, "speed/wps": 12903.052284977764, "speed/FLOPS": 202660166787454.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06237388774752617, "optim/lr": 0.002868792312088793, "optim/total_tokens": 6489636864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.860386371612549, "created_at": "2025-01-16T16:31:17.263277+00:00"} {"global_step": 12379, "acc_step": 0, "speed/wps": 12905.583071086767, "speed/FLOPS": 202699916260961.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049929406493902206, "optim/lr": 0.002868757265655737, "optim/total_tokens": 6490161152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.7966041564941406, "created_at": "2025-01-16T16:31:27.423232+00:00"} {"global_step": 12380, "acc_step": 0, "speed/wps": 12905.472311719577, "speed/FLOPS": 202698176632900.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053739238530397415, "optim/lr": 0.0028687222147568705, "optim/total_tokens": 6490685440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9658122062683105, "created_at": "2025-01-16T16:31:37.586791+00:00"} {"global_step": 12381, "acc_step": 0, "speed/wps": 12896.016843623805, "speed/FLOPS": 202549665513281.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05032505840063095, "optim/lr": 0.002868687159392308, "optim/total_tokens": 6491209728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.868889808654785, "created_at": "2025-01-16T16:31:47.755655+00:00"} {"global_step": 12382, "acc_step": 0, "speed/wps": 12908.021913318707, "speed/FLOPS": 202738221629534.34, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04441116750240326, "optim/lr": 0.0028686520995621625, "optim/total_tokens": 6491734016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 2.9229230880737305, "created_at": "2025-01-16T16:31:57.917059+00:00"} {"global_step": 12383, "acc_step": 0, "speed/wps": 12903.960224453242, "speed/FLOPS": 202674427224556.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044813305139541626, "optim/lr": 0.0028686170352665496, "optim/total_tokens": 6492258304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434742, "loss/out": 2.807239055633545, "created_at": "2025-01-16T16:32:08.082720+00:00"} {"global_step": 12384, "acc_step": 0, "speed/wps": 12900.16954533513, "speed/FLOPS": 202614889400067.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05154493823647499, "optim/lr": 0.0028685819665055837, "optim/total_tokens": 6492782592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9363794326782227, "created_at": "2025-01-16T16:32:18.246715+00:00"} {"global_step": 12385, "acc_step": 0, "speed/wps": 12909.839776406545, "speed/FLOPS": 202766773667334.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05404650792479515, "optim/lr": 0.002868546893279379, "optim/total_tokens": 6493306880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9968700408935547, "created_at": "2025-01-16T16:32:28.404825+00:00"} {"global_step": 12386, "acc_step": 0, "speed/wps": 12907.486624065818, "speed/FLOPS": 202729814176253.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05921068787574768, "optim/lr": 0.0028685118155880487, "optim/total_tokens": 6493831168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.7834975719451904, "created_at": "2025-01-16T16:32:38.566279+00:00"} {"global_step": 12387, "acc_step": 0, "speed/wps": 12904.9351962271, "speed/FLOPS": 202689740496016.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07237278670072556, "optim/lr": 0.002868476733431709, "optim/total_tokens": 6494355456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 2.869783401489258, "created_at": "2025-01-16T16:32:48.726689+00:00"} {"global_step": 12388, "acc_step": 0, "speed/wps": 12910.843183668365, "speed/FLOPS": 202782533557207.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05017613619565964, "optim/lr": 0.0028684416468104743, "optim/total_tokens": 6494879744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8435261249542236, "created_at": "2025-01-16T16:32:58.882681+00:00"} {"global_step": 12389, "acc_step": 0, "speed/wps": 12907.528801351124, "speed/FLOPS": 202730476628476.6, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05966929346323013, "optim/lr": 0.0028684065557244578, "optim/total_tokens": 6495404032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8832156658172607, "created_at": "2025-01-16T16:33:09.043510+00:00"} {"global_step": 12390, "acc_step": 0, "speed/wps": 12906.329251690142, "speed/FLOPS": 202711636052696.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.054809972643852234, "optim/lr": 0.0028683714601737743, "optim/total_tokens": 6495928320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329371, "loss/out": 2.865471839904785, "created_at": "2025-01-16T16:33:19.203620+00:00"} {"global_step": 12391, "acc_step": 0, "speed/wps": 12897.359409589772, "speed/FLOPS": 202570752356655.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06842394173145294, "optim/lr": 0.0028683363601585397, "optim/total_tokens": 6496452608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.982771396636963, "created_at": "2025-01-16T16:33:29.371968+00:00"} {"global_step": 12392, "acc_step": 0, "speed/wps": 12904.71449324721, "speed/FLOPS": 202686274052440.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0724605917930603, "optim/lr": 0.0028683012556788667, "optim/total_tokens": 6496976896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 2.8656675815582275, "created_at": "2025-01-16T16:33:39.534928+00:00"} {"global_step": 12393, "acc_step": 0, "speed/wps": 12891.978063342296, "speed/FLOPS": 202486230919094.62, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05463740974664688, "optim/lr": 0.0028682661467348707, "optim/total_tokens": 6497501184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8189749717712402, "created_at": "2025-01-16T16:33:49.705483+00:00"} {"global_step": 12394, "acc_step": 0, "speed/wps": 12891.085235826542, "speed/FLOPS": 202472207836086.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048300981521606445, "optim/lr": 0.0028682310333266662, "optim/total_tokens": 6498025472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8533318042755127, "created_at": "2025-01-16T16:33:59.877256+00:00"} {"global_step": 12395, "acc_step": 0, "speed/wps": 12893.479781437021, "speed/FLOPS": 202509817465349.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05103352293372154, "optim/lr": 0.002868195915454368, "optim/total_tokens": 6498549760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 2.865882635116577, "created_at": "2025-01-16T16:34:10.050151+00:00"} {"global_step": 12396, "acc_step": 0, "speed/wps": 12898.983570745633, "speed/FLOPS": 202596262039440.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05253119394183159, "optim/lr": 0.0028681607931180905, "optim/total_tokens": 6499074048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.827132225036621, "created_at": "2025-01-16T16:34:20.217552+00:00"} {"global_step": 12397, "acc_step": 0, "speed/wps": 12909.431616368558, "speed/FLOPS": 202760362953067.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056934673339128494, "optim/lr": 0.002868125666317948, "optim/total_tokens": 6499598336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.7127482891082764, "created_at": "2025-01-16T16:34:30.378063+00:00"} {"global_step": 12398, "acc_step": 0, "speed/wps": 12904.558941319383, "speed/FLOPS": 202683830895662.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04376015067100525, "optim/lr": 0.002868090535054055, "optim/total_tokens": 6500122624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9265711307525635, "created_at": "2025-01-16T16:34:40.540446+00:00"} {"global_step": 12399, "acc_step": 0, "speed/wps": 12913.204901383078, "speed/FLOPS": 202819627579257.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06046279892325401, "optim/lr": 0.0028680553993265267, "optim/total_tokens": 6500646912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 2.818755865097046, "created_at": "2025-01-16T16:34:50.697500+00:00"} {"global_step": 12400, "acc_step": 0, "speed/wps": 12900.609290428445, "speed/FLOPS": 202621796201031.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.060019634664058685, "optim/lr": 0.0028680202591354776, "optim/total_tokens": 6501171200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.7678213119506836, "created_at": "2025-01-16T16:35:00.865104+00:00"} {"global_step": 12401, "acc_step": 0, "speed/wps": 12899.702420701276, "speed/FLOPS": 202607552565796.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04957786202430725, "optim/lr": 0.0028679851144810223, "optim/total_tokens": 6501695488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.856130838394165, "created_at": "2025-01-16T16:35:11.039009+00:00"} {"global_step": 12402, "acc_step": 0, "speed/wps": 12905.428340806398, "speed/FLOPS": 202697486009286.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.064211905002594, "optim/lr": 0.0028679499653632753, "optim/total_tokens": 6502219776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 2.848615884780884, "created_at": "2025-01-16T16:35:21.199741+00:00"} {"global_step": 12403, "acc_step": 0, "speed/wps": 12907.821556881097, "speed/FLOPS": 202735074756363.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0414864718914032, "optim/lr": 0.0028679148117823515, "optim/total_tokens": 6502744064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8991479873657227, "created_at": "2025-01-16T16:35:31.357690+00:00"} {"global_step": 12404, "acc_step": 0, "speed/wps": 12884.86763425887, "speed/FLOPS": 202374551859585.97, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05417503044009209, "optim/lr": 0.0028678796537383654, "optim/total_tokens": 6503268352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7989368438720703, "created_at": "2025-01-16T16:35:41.535546+00:00"} {"global_step": 12405, "acc_step": 0, "speed/wps": 12891.285786244982, "speed/FLOPS": 202475357755992.9, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05880540609359741, "optim/lr": 0.0028678444912314314, "optim/total_tokens": 6503792640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 2.848583221435547, "created_at": "2025-01-16T16:35:51.707813+00:00"} {"global_step": 12406, "acc_step": 0, "speed/wps": 12892.252390671121, "speed/FLOPS": 202490539606759.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04242146387696266, "optim/lr": 0.002867809324261665, "optim/total_tokens": 6504316928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.783949851989746, "created_at": "2025-01-16T16:36:01.878845+00:00"} {"global_step": 12407, "acc_step": 0, "speed/wps": 12892.172250077012, "speed/FLOPS": 202489280888603.88, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05523254722356796, "optim/lr": 0.002867774152829181, "optim/total_tokens": 6504841216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415004, "loss/out": 2.8903112411499023, "created_at": "2025-01-16T16:36:12.050658+00:00"} {"global_step": 12408, "acc_step": 0, "speed/wps": 12902.278285641289, "speed/FLOPS": 202648010064288.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04603772982954979, "optim/lr": 0.002867738976934093, "optim/total_tokens": 6505365504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.8307714462280273, "created_at": "2025-01-16T16:36:22.217269+00:00"} {"global_step": 12409, "acc_step": 0, "speed/wps": 12904.052133810048, "speed/FLOPS": 202675870787302.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052643243223428726, "optim/lr": 0.0028677037965765166, "optim/total_tokens": 6505889792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.884787082672119, "created_at": "2025-01-16T16:36:32.378650+00:00"} {"global_step": 12410, "acc_step": 0, "speed/wps": 12909.128979990322, "speed/FLOPS": 202755609632881.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05121547356247902, "optim/lr": 0.0028676686117565667, "optim/total_tokens": 6506414080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.924757480621338, "created_at": "2025-01-16T16:36:42.536397+00:00"} {"global_step": 12411, "acc_step": 0, "speed/wps": 12897.426099160997, "speed/FLOPS": 202571799808013.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.06389457732439041, "optim/lr": 0.0028676334224743572, "optim/total_tokens": 6506938368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9158554077148438, "created_at": "2025-01-16T16:36:52.705615+00:00"} {"global_step": 12412, "acc_step": 0, "speed/wps": 12906.939816945243, "speed/FLOPS": 202721225819032.88, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04405229538679123, "optim/lr": 0.0028675982287300044, "optim/total_tokens": 6507462656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309674, "loss/out": 2.8614206314086914, "created_at": "2025-01-16T16:37:02.868601+00:00"} {"global_step": 12413, "acc_step": 0, "speed/wps": 12902.487542636405, "speed/FLOPS": 202651296732946.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05125994235277176, "optim/lr": 0.002867563030523622, "optim/total_tokens": 6507986944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.91546368598938, "created_at": "2025-01-16T16:37:13.032166+00:00"} {"global_step": 12414, "acc_step": 0, "speed/wps": 12898.730744454939, "speed/FLOPS": 202592291055124.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045660436153411865, "optim/lr": 0.0028675278278553246, "optim/total_tokens": 6508511232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 2.8760690689086914, "created_at": "2025-01-16T16:37:23.198163+00:00"} {"global_step": 12415, "acc_step": 0, "speed/wps": 12908.456372374118, "speed/FLOPS": 202745045406008.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.058746758848428726, "optim/lr": 0.002867492620725228, "optim/total_tokens": 6509035520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.796576499938965, "created_at": "2025-01-16T16:37:33.362139+00:00"} {"global_step": 12416, "acc_step": 0, "speed/wps": 12901.390915245085, "speed/FLOPS": 202634072692841.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.050667595118284225, "optim/lr": 0.0028674574091334466, "optim/total_tokens": 6509559808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 2.9832539558410645, "created_at": "2025-01-16T16:37:43.529842+00:00"} {"global_step": 12417, "acc_step": 0, "speed/wps": 12896.253904444498, "speed/FLOPS": 202553388879226.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07275108247995377, "optim/lr": 0.0028674221930800954, "optim/total_tokens": 6510084096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8161425590515137, "created_at": "2025-01-16T16:37:53.697684+00:00"} {"global_step": 12418, "acc_step": 0, "speed/wps": 12909.462343789235, "speed/FLOPS": 202760845569432.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058318883180618286, "optim/lr": 0.0028673869725652896, "optim/total_tokens": 6510608384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8834469318389893, "created_at": "2025-01-16T16:38:03.858229+00:00"} {"global_step": 12419, "acc_step": 0, "speed/wps": 12908.691763095601, "speed/FLOPS": 202748742540750.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.064633809030056, "optim/lr": 0.0028673517475891427, "optim/total_tokens": 6511132672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 3.0144331455230713, "created_at": "2025-01-16T16:38:14.018713+00:00"} {"global_step": 12420, "acc_step": 0, "speed/wps": 12899.873881677486, "speed/FLOPS": 202610245596040.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05810829624533653, "optim/lr": 0.0028673165181517715, "optim/total_tokens": 6511656960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8647220134735107, "created_at": "2025-01-16T16:38:24.183513+00:00"} {"global_step": 12421, "acc_step": 0, "speed/wps": 12904.642725433569, "speed/FLOPS": 202685146840307.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046877216547727585, "optim/lr": 0.00286728128425329, "optim/total_tokens": 6512181248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8313701152801514, "created_at": "2025-01-16T16:38:34.345543+00:00"} {"global_step": 12422, "acc_step": 0, "speed/wps": 12904.586398695435, "speed/FLOPS": 202684262151484.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05939968675374985, "optim/lr": 0.002867246045893813, "optim/total_tokens": 6512705536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 2.964916706085205, "created_at": "2025-01-16T16:38:44.508359+00:00"} {"global_step": 12423, "acc_step": 0, "speed/wps": 12902.564872058083, "speed/FLOPS": 202652511297775.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04597535356879234, "optim/lr": 0.002867210803073456, "optim/total_tokens": 6513229824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.857442617416382, "created_at": "2025-01-16T16:38:54.674611+00:00"} {"global_step": 12424, "acc_step": 0, "speed/wps": 12911.655333543926, "speed/FLOPS": 202795289487014.0, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06464099884033203, "optim/lr": 0.0028671755557923336, "optim/total_tokens": 6513754112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 2.8217856884002686, "created_at": "2025-01-16T16:39:04.833296+00:00"} {"global_step": 12425, "acc_step": 0, "speed/wps": 12910.810896929468, "speed/FLOPS": 202782026449606.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05446622893214226, "optim/lr": 0.0028671403040505606, "optim/total_tokens": 6514278400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.8531689643859863, "created_at": "2025-01-16T16:39:14.990056+00:00"} {"global_step": 12426, "acc_step": 0, "speed/wps": 12911.958074173466, "speed/FLOPS": 202800044444610.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057117484509944916, "optim/lr": 0.0028671050478482527, "optim/total_tokens": 6514802688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 2.822144031524658, "created_at": "2025-01-16T16:39:25.148921+00:00"} {"global_step": 12427, "acc_step": 0, "speed/wps": 12905.780454975318, "speed/FLOPS": 202703016446164.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07263102382421494, "optim/lr": 0.002867069787185525, "optim/total_tokens": 6515326976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 3.049520492553711, "created_at": "2025-01-16T16:39:35.308522+00:00"} {"global_step": 12428, "acc_step": 0, "speed/wps": 12908.77744677689, "speed/FLOPS": 202750088320706.88, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04962144419550896, "optim/lr": 0.002867034522062492, "optim/total_tokens": 6515851264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425394, "loss/out": 2.8992788791656494, "created_at": "2025-01-16T16:39:45.468490+00:00"} {"global_step": 12429, "acc_step": 0, "speed/wps": 12901.679944775438, "speed/FLOPS": 202638612298784.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053991928696632385, "optim/lr": 0.0028669992524792682, "optim/total_tokens": 6516375552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.8916268348693848, "created_at": "2025-01-16T16:39:55.636296+00:00"} {"global_step": 12430, "acc_step": 0, "speed/wps": 12905.093990132309, "speed/FLOPS": 202692234572503.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055088456720113754, "optim/lr": 0.00286696397843597, "optim/total_tokens": 6516899840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.8604140281677246, "created_at": "2025-01-16T16:40:05.805843+00:00"} {"global_step": 12431, "acc_step": 0, "speed/wps": 12910.198220932534, "speed/FLOPS": 202772403531169.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05888836085796356, "optim/lr": 0.002866928699932711, "optim/total_tokens": 6517424128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 2.797330617904663, "created_at": "2025-01-16T16:40:15.963015+00:00"} {"global_step": 12432, "acc_step": 0, "speed/wps": 12907.55614470017, "speed/FLOPS": 202730906093347.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04863974079489708, "optim/lr": 0.0028668934169696074, "optim/total_tokens": 6517948416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8403143882751465, "created_at": "2025-01-16T16:40:26.125004+00:00"} {"global_step": 12433, "acc_step": 0, "speed/wps": 12903.279450699722, "speed/FLOPS": 202663734737276.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060950882732868195, "optim/lr": 0.002866858129546775, "optim/total_tokens": 6518472704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 2.82716703414917, "created_at": "2025-01-16T16:40:36.289375+00:00"} {"global_step": 12434, "acc_step": 0, "speed/wps": 12910.547259127216, "speed/FLOPS": 202777885655645.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05044763907790184, "optim/lr": 0.0028668228376643274, "optim/total_tokens": 6518996992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 2.893411636352539, "created_at": "2025-01-16T16:40:46.448759+00:00"} {"global_step": 12435, "acc_step": 0, "speed/wps": 12910.8234468107, "speed/FLOPS": 202782223562736.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053067199885845184, "optim/lr": 0.00286678754132238, "optim/total_tokens": 6519521280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 2.831801652908325, "created_at": "2025-01-16T16:40:56.609628+00:00"} {"global_step": 12436, "acc_step": 0, "speed/wps": 12915.067009507635, "speed/FLOPS": 202848874546158.12, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051571574062108994, "optim/lr": 0.0028667522405210486, "optim/total_tokens": 6520045568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.7840981483459473, "created_at": "2025-01-16T16:41:06.764065+00:00"} {"global_step": 12437, "acc_step": 0, "speed/wps": 12901.14451792691, "speed/FLOPS": 202630202684374.56, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04449588805437088, "optim/lr": 0.0028667169352604486, "optim/total_tokens": 6520569856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.8264317512512207, "created_at": "2025-01-16T16:41:16.929679+00:00"} {"global_step": 12438, "acc_step": 0, "speed/wps": 12910.641273394229, "speed/FLOPS": 202779362278898.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0681523010134697, "optim/lr": 0.0028666816255406937, "optim/total_tokens": 6521094144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8053536415100098, "created_at": "2025-01-16T16:41:27.088563+00:00"} {"global_step": 12439, "acc_step": 0, "speed/wps": 12906.16481501371, "speed/FLOPS": 202709053348733.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046404823660850525, "optim/lr": 0.0028666463113619003, "optim/total_tokens": 6521618432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9480645656585693, "created_at": "2025-01-16T16:41:37.249909+00:00"} {"global_step": 12440, "acc_step": 0, "speed/wps": 12909.621630432366, "speed/FLOPS": 202763347385046.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06888353824615479, "optim/lr": 0.002866610992724183, "optim/total_tokens": 6522142720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8120601177215576, "created_at": "2025-01-16T16:41:47.406712+00:00"} {"global_step": 12441, "acc_step": 0, "speed/wps": 12913.37754853325, "speed/FLOPS": 202822339239992.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05736514925956726, "optim/lr": 0.002866575669627658, "optim/total_tokens": 6522667008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.818986415863037, "created_at": "2025-01-16T16:41:57.560738+00:00"} {"global_step": 12442, "acc_step": 0, "speed/wps": 12911.97202886298, "speed/FLOPS": 202800263622184.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06443358212709427, "optim/lr": 0.0028665403420724406, "optim/total_tokens": 6523191296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.901453971862793, "created_at": "2025-01-16T16:42:07.716350+00:00"} {"global_step": 12443, "acc_step": 0, "speed/wps": 12914.03180051637, "speed/FLOPS": 202832615166424.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07419566810131073, "optim/lr": 0.0028665050100586446, "optim/total_tokens": 6523715584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.892580986022949, "created_at": "2025-01-16T16:42:17.871598+00:00"} {"global_step": 12444, "acc_step": 0, "speed/wps": 12907.5722645217, "speed/FLOPS": 202731159277295.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0730394572019577, "optim/lr": 0.0028664696735863863, "optim/total_tokens": 6524239872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.886563777923584, "created_at": "2025-01-16T16:42:28.032665+00:00"} {"global_step": 12445, "acc_step": 0, "speed/wps": 12908.839514786721, "speed/FLOPS": 202751063184091.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058317434042692184, "optim/lr": 0.0028664343326557803, "optim/total_tokens": 6524764160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.84456467628479, "created_at": "2025-01-16T16:42:38.189949+00:00"} {"global_step": 12446, "acc_step": 0, "speed/wps": 12909.192976727047, "speed/FLOPS": 202756614789572.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.11351147294044495, "optim/lr": 0.0028663989872669436, "optim/total_tokens": 6525288448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8097333908081055, "created_at": "2025-01-16T16:42:48.348330+00:00"} {"global_step": 12447, "acc_step": 0, "speed/wps": 12911.57978682658, "speed/FLOPS": 202794102922006.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07713336497545242, "optim/lr": 0.002866363637419989, "optim/total_tokens": 6525812736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.835226535797119, "created_at": "2025-01-16T16:42:58.503936+00:00"} {"global_step": 12448, "acc_step": 0, "speed/wps": 12911.458385860546, "speed/FLOPS": 202792196153012.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05928843840956688, "optim/lr": 0.0028663282831150336, "optim/total_tokens": 6526337024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 2.864229917526245, "created_at": "2025-01-16T16:43:08.660463+00:00"} {"global_step": 12449, "acc_step": 0, "speed/wps": 12915.292932064087, "speed/FLOPS": 202852422970357.3, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06149284914135933, "optim/lr": 0.0028662929243521923, "optim/total_tokens": 6526861312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.845158576965332, "created_at": "2025-01-16T16:43:18.813919+00:00"} {"global_step": 12450, "acc_step": 0, "speed/wps": 12907.299648799435, "speed/FLOPS": 202726877472763.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052757903933525085, "optim/lr": 0.002866257561131581, "optim/total_tokens": 6527385600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289738, "loss/out": 2.876882553100586, "created_at": "2025-01-16T16:43:28.976785+00:00"} {"global_step": 12451, "acc_step": 0, "speed/wps": 12914.023076641522, "speed/FLOPS": 202832478145981.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05507253110408783, "optim/lr": 0.0028662221934533144, "optim/total_tokens": 6527909888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.847842216491699, "created_at": "2025-01-16T16:43:39.164698+00:00"} {"global_step": 12452, "acc_step": 0, "speed/wps": 12897.632471964136, "speed/FLOPS": 202575041176472.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04887774959206581, "optim/lr": 0.0028661868213175077, "optim/total_tokens": 6528434176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457635, "loss/out": 2.7759222984313965, "created_at": "2025-01-16T16:43:49.331853+00:00"} {"global_step": 12453, "acc_step": 0, "speed/wps": 12898.191653211623, "speed/FLOPS": 202583823886359.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07475810497999191, "optim/lr": 0.0028661514447242764, "optim/total_tokens": 6528958464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8655850887298584, "created_at": "2025-01-16T16:43:59.498030+00:00"} {"global_step": 12454, "acc_step": 0, "speed/wps": 12890.788433414205, "speed/FLOPS": 202467546146351.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06734368950128555, "optim/lr": 0.0028661160636737366, "optim/total_tokens": 6529482752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.9408631324768066, "created_at": "2025-01-16T16:44:09.674117+00:00"} {"global_step": 12455, "acc_step": 0, "speed/wps": 12893.987269708672, "speed/FLOPS": 202517788266017.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046908244490623474, "optim/lr": 0.0028660806781660034, "optim/total_tokens": 6530007040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8330068588256836, "created_at": "2025-01-16T16:44:19.846490+00:00"} {"global_step": 12456, "acc_step": 0, "speed/wps": 12892.209315228905, "speed/FLOPS": 202489863047747.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0526486411690712, "optim/lr": 0.002866045288201192, "optim/total_tokens": 6530531328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 2.7451412677764893, "created_at": "2025-01-16T16:44:30.024782+00:00"} {"global_step": 12457, "acc_step": 0, "speed/wps": 12894.841158606572, "speed/FLOPS": 202531199764525.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07839898765087128, "optim/lr": 0.0028660098937794178, "optim/total_tokens": 6531055616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.7979419231414795, "created_at": "2025-01-16T16:44:40.193170+00:00"} {"global_step": 12458, "acc_step": 0, "speed/wps": 12892.426274350279, "speed/FLOPS": 202493270688881.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062008656561374664, "optim/lr": 0.002865974494900797, "optim/total_tokens": 6531579904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.813547134399414, "created_at": "2025-01-16T16:44:50.371145+00:00"} {"global_step": 12459, "acc_step": 0, "speed/wps": 12891.341316775413, "speed/FLOPS": 202476229939280.28, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0511014424264431, "optim/lr": 0.002865939091565444, "optim/total_tokens": 6532104192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.843904733657837, "created_at": "2025-01-16T16:45:00.544768+00:00"} {"global_step": 12460, "acc_step": 0, "speed/wps": 12896.590560226765, "speed/FLOPS": 202558676520900.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054483477026224136, "optim/lr": 0.0028659036837734753, "optim/total_tokens": 6532628480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.8928565979003906, "created_at": "2025-01-16T16:45:10.717696+00:00"} {"global_step": 12461, "acc_step": 0, "speed/wps": 12896.91805061102, "speed/FLOPS": 202563820207403.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05387875437736511, "optim/lr": 0.0028658682715250058, "optim/total_tokens": 6533152768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.888291358947754, "created_at": "2025-01-16T16:45:20.885415+00:00"} {"global_step": 12462, "acc_step": 0, "speed/wps": 12894.577436571164, "speed/FLOPS": 202527057647567.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0526980385184288, "optim/lr": 0.0028658328548201517, "optim/total_tokens": 6533677056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 2.7531259059906006, "created_at": "2025-01-16T16:45:31.053765+00:00"} {"global_step": 12463, "acc_step": 0, "speed/wps": 12897.142150850712, "speed/FLOPS": 202567340009614.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04473228380084038, "optim/lr": 0.002865797433659028, "optim/total_tokens": 6534201344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.7460849285125732, "created_at": "2025-01-16T16:45:41.222741+00:00"} {"global_step": 12464, "acc_step": 0, "speed/wps": 12889.591574013628, "speed/FLOPS": 202448747824807.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052001260221004486, "optim/lr": 0.00286576200804175, "optim/total_tokens": 6534725632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.17606258392334, "created_at": "2025-01-16T16:45:51.396486+00:00"} {"global_step": 12465, "acc_step": 0, "speed/wps": 12875.59330163191, "speed/FLOPS": 202228885721411.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05768642574548721, "optim/lr": 0.0028657265779684343, "optim/total_tokens": 6535249920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.8703410625457764, "created_at": "2025-01-16T16:46:01.597434+00:00"} {"global_step": 12466, "acc_step": 0, "speed/wps": 12902.267678013899, "speed/FLOPS": 202647843456924.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059731949120759964, "optim/lr": 0.0028656911434391955, "optim/total_tokens": 6535774208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.9206361770629883, "created_at": "2025-01-16T16:46:11.762048+00:00"} {"global_step": 12467, "acc_step": 0, "speed/wps": 12882.313700734661, "speed/FLOPS": 202334438824115.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05002239719033241, "optim/lr": 0.002865655704454149, "optim/total_tokens": 6536298496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 2.9327340126037598, "created_at": "2025-01-16T16:46:21.941085+00:00"} {"global_step": 12468, "acc_step": 0, "speed/wps": 12885.354224176075, "speed/FLOPS": 202382194422880.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05802099034190178, "optim/lr": 0.002865620261013412, "optim/total_tokens": 6536822784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.7085375785827637, "created_at": "2025-01-16T16:46:32.117415+00:00"} {"global_step": 12469, "acc_step": 0, "speed/wps": 12884.140837827246, "speed/FLOPS": 202363136522904.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055224284529685974, "optim/lr": 0.0028655848131170984, "optim/total_tokens": 6537347072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358724, "loss/out": 2.845682144165039, "created_at": "2025-01-16T16:46:42.294963+00:00"} {"global_step": 12470, "acc_step": 0, "speed/wps": 12884.786257188223, "speed/FLOPS": 202373273720867.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04377853870391846, "optim/lr": 0.002865549360765326, "optim/total_tokens": 6537871360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.940368175506592, "created_at": "2025-01-16T16:46:52.478513+00:00"} {"global_step": 12471, "acc_step": 0, "speed/wps": 12891.361209024308, "speed/FLOPS": 202476542374383.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042829181998968124, "optim/lr": 0.0028655139039582074, "optim/total_tokens": 6538395648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.897279739379883, "created_at": "2025-01-16T16:47:02.650631+00:00"} {"global_step": 12472, "acc_step": 0, "speed/wps": 12888.444024011269, "speed/FLOPS": 202430723975121.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046866051852703094, "optim/lr": 0.0028654784426958607, "optim/total_tokens": 6538919936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338751, "loss/out": 2.775730609893799, "created_at": "2025-01-16T16:47:12.824108+00:00"} {"global_step": 12473, "acc_step": 0, "speed/wps": 12891.67100394801, "speed/FLOPS": 202481408129363.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04787636920809746, "optim/lr": 0.002865442976978401, "optim/total_tokens": 6539444224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.897609233856201, "created_at": "2025-01-16T16:47:22.995585+00:00"} {"global_step": 12474, "acc_step": 0, "speed/wps": 12893.147492964294, "speed/FLOPS": 202504598418272.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05476262420415878, "optim/lr": 0.002865407506805944, "optim/total_tokens": 6539968512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.8561573028564453, "created_at": "2025-01-16T16:47:33.165680+00:00"} {"global_step": 12475, "acc_step": 0, "speed/wps": 12887.977152148991, "speed/FLOPS": 202423391110975.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04718852415680885, "optim/lr": 0.0028653720321786045, "optim/total_tokens": 6540492800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467785, "loss/out": 2.8747377395629883, "created_at": "2025-01-16T16:47:43.341532+00:00"} {"global_step": 12476, "acc_step": 0, "speed/wps": 12892.273167937346, "speed/FLOPS": 202490865942276.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05853629112243652, "optim/lr": 0.0028653365530965, "optim/total_tokens": 6541017088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8750104904174805, "created_at": "2025-01-16T16:47:53.511877+00:00"} {"global_step": 12477, "acc_step": 0, "speed/wps": 12894.101325226218, "speed/FLOPS": 202519579664649.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053590282797813416, "optim/lr": 0.0028653010695597445, "optim/total_tokens": 6541541376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8831770420074463, "created_at": "2025-01-16T16:48:03.682430+00:00"} {"global_step": 12478, "acc_step": 0, "speed/wps": 12894.284124934102, "speed/FLOPS": 202522450785254.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05408424511551857, "optim/lr": 0.002865265581568455, "optim/total_tokens": 6542065664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 2.869276285171509, "created_at": "2025-01-16T16:48:13.855168+00:00"} {"global_step": 12479, "acc_step": 0, "speed/wps": 12895.266762467278, "speed/FLOPS": 202537884457998.62, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04737576097249985, "optim/lr": 0.002865230089122747, "optim/total_tokens": 6542589952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.833120346069336, "created_at": "2025-01-16T16:48:24.025986+00:00"} {"global_step": 12480, "acc_step": 0, "speed/wps": 12892.636939790504, "speed/FLOPS": 202496579479104.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0476091243326664, "optim/lr": 0.0028651945922227354, "optim/total_tokens": 6543114240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8795278072357178, "created_at": "2025-01-16T16:48:34.196642+00:00"} {"global_step": 12481, "acc_step": 0, "speed/wps": 12895.354429096913, "speed/FLOPS": 202539261382883.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0560813769698143, "optim/lr": 0.0028651590908685375, "optim/total_tokens": 6543638528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.790300130844116, "created_at": "2025-01-16T16:48:44.364503+00:00"} {"global_step": 12482, "acc_step": 0, "speed/wps": 12889.7516515518, "speed/FLOPS": 202451262062514.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.050885822623968124, "optim/lr": 0.0028651235850602683, "optim/total_tokens": 6544162816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.7576744556427, "created_at": "2025-01-16T16:48:54.536825+00:00"} {"global_step": 12483, "acc_step": 0, "speed/wps": 12891.777727565328, "speed/FLOPS": 202483084370427.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058064401149749756, "optim/lr": 0.0028650880747980436, "optim/total_tokens": 6544687104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.863290309906006, "created_at": "2025-01-16T16:49:04.708256+00:00"} {"global_step": 12484, "acc_step": 0, "speed/wps": 12890.905531808628, "speed/FLOPS": 202469385337545.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05447029322385788, "optim/lr": 0.0028650525600819794, "optim/total_tokens": 6545211392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 2.952033042907715, "created_at": "2025-01-16T16:49:14.879850+00:00"} {"global_step": 12485, "acc_step": 0, "speed/wps": 12889.31793242959, "speed/FLOPS": 202444449907707.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05675363168120384, "optim/lr": 0.0028650170409121917, "optim/total_tokens": 6545735680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.818840980529785, "created_at": "2025-01-16T16:49:25.052652+00:00"} {"global_step": 12486, "acc_step": 0, "speed/wps": 12896.622611612143, "speed/FLOPS": 202559179931950.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04899370297789574, "optim/lr": 0.002864981517288796, "optim/total_tokens": 6546259968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.8893537521362305, "created_at": "2025-01-16T16:49:35.219913+00:00"} {"global_step": 12487, "acc_step": 0, "speed/wps": 12892.184730681704, "speed/FLOPS": 202489476913650.72, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06069605052471161, "optim/lr": 0.002864945989211909, "optim/total_tokens": 6546784256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 2.8002309799194336, "created_at": "2025-01-16T16:49:45.390153+00:00"} {"global_step": 12488, "acc_step": 0, "speed/wps": 12895.90376335667, "speed/FLOPS": 202547889432297.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04152792692184448, "optim/lr": 0.0028649104566816457, "optim/total_tokens": 6547308544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.9417104721069336, "created_at": "2025-01-16T16:49:55.557499+00:00"} {"global_step": 12489, "acc_step": 0, "speed/wps": 12891.497804282264, "speed/FLOPS": 202478687790611.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05302155017852783, "optim/lr": 0.0028648749196981227, "optim/total_tokens": 6547832832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.8517158031463623, "created_at": "2025-01-16T16:50:05.728362+00:00"} {"global_step": 12490, "acc_step": 0, "speed/wps": 12893.333023711433, "speed/FLOPS": 202507512433602.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05022862181067467, "optim/lr": 0.0028648393782614557, "optim/total_tokens": 6548357120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339571, "loss/out": 2.839069128036499, "created_at": "2025-01-16T16:50:15.900946+00:00"} {"global_step": 12491, "acc_step": 0, "speed/wps": 12892.604376736377, "speed/FLOPS": 202496068031593.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050239261239767075, "optim/lr": 0.00286480383237176, "optim/total_tokens": 6548881408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.71256685256958, "created_at": "2025-01-16T16:50:26.070928+00:00"} {"global_step": 12492, "acc_step": 0, "speed/wps": 12893.252815347809, "speed/FLOPS": 202506252651034.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04195300117135048, "optim/lr": 0.0028647682820291524, "optim/total_tokens": 6549405696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429961, "loss/out": 2.8391714096069336, "created_at": "2025-01-16T16:50:36.245171+00:00"} {"global_step": 12493, "acc_step": 0, "speed/wps": 12889.518068992445, "speed/FLOPS": 202447593327443.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0540425069630146, "optim/lr": 0.002864732727233749, "optim/total_tokens": 6549929984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 2.9764533042907715, "created_at": "2025-01-16T16:50:46.421090+00:00"} {"global_step": 12494, "acc_step": 0, "speed/wps": 12894.84085496914, "speed/FLOPS": 202531194995482.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053616102784872055, "optim/lr": 0.0028646971679856656, "optim/total_tokens": 6550454272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8152542114257812, "created_at": "2025-01-16T16:50:56.591939+00:00"} {"global_step": 12495, "acc_step": 0, "speed/wps": 12891.704628634578, "speed/FLOPS": 202481936251271.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.039980221539735794, "optim/lr": 0.002864661604285018, "optim/total_tokens": 6550978560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.677863597869873, "created_at": "2025-01-16T16:51:06.765100+00:00"} {"global_step": 12496, "acc_step": 0, "speed/wps": 12893.071138619429, "speed/FLOPS": 202503399168364.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05363459885120392, "optim/lr": 0.002864626036131922, "optim/total_tokens": 6551502848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 2.8998661041259766, "created_at": "2025-01-16T16:51:16.943998+00:00"} {"global_step": 12497, "acc_step": 0, "speed/wps": 12889.794006968237, "speed/FLOPS": 202451927312532.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05284690484404564, "optim/lr": 0.0028645904635264943, "optim/total_tokens": 6552027136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8721795082092285, "created_at": "2025-01-16T16:51:27.116610+00:00"} {"global_step": 12498, "acc_step": 0, "speed/wps": 12890.520810868533, "speed/FLOPS": 202463342766519.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04884679615497589, "optim/lr": 0.002864554886468851, "optim/total_tokens": 6552551424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9535365104675293, "created_at": "2025-01-16T16:51:37.291689+00:00"} {"global_step": 12499, "acc_step": 0, "speed/wps": 12889.236264227922, "speed/FLOPS": 202443167196376.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06772570312023163, "optim/lr": 0.002864519304959107, "optim/total_tokens": 6553075712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351882, "loss/out": 2.885498285293579, "created_at": "2025-01-16T16:51:47.464908+00:00"} {"global_step": 12500, "acc_step": 0, "speed/wps": 12883.94836494229, "speed/FLOPS": 202360113471760.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05388658121228218, "optim/lr": 0.0028644837189973796, "optim/total_tokens": 6553600000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8699028491973877, "created_at": "2025-01-16T16:51:57.649274+00:00"} {"global_step": 12501, "acc_step": 0, "speed/wps": 5351.633207464991, "speed/FLOPS": 84054753437899.92, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061394695192575455, "optim/lr": 0.0028644481285837846, "optim/total_tokens": 6554124288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.8901584148406982, "created_at": "2025-01-16T16:52:22.149891+00:00"} {"global_step": 12502, "acc_step": 0, "speed/wps": 12946.007218673249, "speed/FLOPS": 203334833047406.88, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049841541796922684, "optim/lr": 0.002864412533718438, "optim/total_tokens": 6554648576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485284, "loss/out": 2.6964032649993896, "created_at": "2025-01-16T16:52:32.284267+00:00"} {"global_step": 12503, "acc_step": 0, "speed/wps": 12928.514228819735, "speed/FLOPS": 203060081603870.38, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06524696201086044, "optim/lr": 0.002864376934401456, "optim/total_tokens": 6555172864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.940803050994873, "created_at": "2025-01-16T16:52:42.431594+00:00"} {"global_step": 12504, "acc_step": 0, "speed/wps": 12913.613622356917, "speed/FLOPS": 202826047103792.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05579818785190582, "optim/lr": 0.002864341330632955, "optim/total_tokens": 6555697152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 2.7904181480407715, "created_at": "2025-01-16T16:52:52.585075+00:00"} {"global_step": 12505, "acc_step": 0, "speed/wps": 12910.254831161457, "speed/FLOPS": 202773292672605.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04991166293621063, "optim/lr": 0.0028643057224130505, "optim/total_tokens": 6556221440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9640073776245117, "created_at": "2025-01-16T16:53:02.743305+00:00"} {"global_step": 12506, "acc_step": 0, "speed/wps": 12909.376579967528, "speed/FLOPS": 202759498530761.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048059288412332535, "optim/lr": 0.002864270109741859, "optim/total_tokens": 6556745728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.803617477416992, "created_at": "2025-01-16T16:53:12.900826+00:00"} {"global_step": 12507, "acc_step": 0, "speed/wps": 12912.684225999315, "speed/FLOPS": 202811449656872.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04803672432899475, "optim/lr": 0.002864234492619497, "optim/total_tokens": 6557270016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.844241142272949, "created_at": "2025-01-16T16:53:23.057697+00:00"} {"global_step": 12508, "acc_step": 0, "speed/wps": 12906.484391796259, "speed/FLOPS": 202714072741248.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046590279787778854, "optim/lr": 0.002864198871046081, "optim/total_tokens": 6557794304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.916889190673828, "created_at": "2025-01-16T16:53:33.218452+00:00"} {"global_step": 12509, "acc_step": 0, "speed/wps": 12915.640599438062, "speed/FLOPS": 202857883564210.62, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04354189708828926, "optim/lr": 0.0028641632450217254, "optim/total_tokens": 6558318592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 2.984327793121338, "created_at": "2025-01-16T16:53:43.372150+00:00"} {"global_step": 12510, "acc_step": 0, "speed/wps": 12913.581485277144, "speed/FLOPS": 202825542346794.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04300042241811752, "optim/lr": 0.0028641276145465484, "optim/total_tokens": 6558842880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.8425118923187256, "created_at": "2025-01-16T16:53:53.527182+00:00"} {"global_step": 12511, "acc_step": 0, "speed/wps": 12906.280404562343, "speed/FLOPS": 202710868841431.84, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045931100845336914, "optim/lr": 0.002864091979620666, "optim/total_tokens": 6559367168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316449, "loss/out": 2.8078837394714355, "created_at": "2025-01-16T16:54:03.688677+00:00"} {"global_step": 12512, "acc_step": 0, "speed/wps": 12908.898916921107, "speed/FLOPS": 202751996176239.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050829775631427765, "optim/lr": 0.0028640563402441933, "optim/total_tokens": 6559891456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.8528804779052734, "created_at": "2025-01-16T16:54:13.849539+00:00"} {"global_step": 12513, "acc_step": 0, "speed/wps": 12899.513711773223, "speed/FLOPS": 202604588632768.4, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04892893508076668, "optim/lr": 0.002864020696417248, "optim/total_tokens": 6560415744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.789912223815918, "created_at": "2025-01-16T16:54:24.021322+00:00"} {"global_step": 12514, "acc_step": 0, "speed/wps": 12912.602836019058, "speed/FLOPS": 202810171315390.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05103747174143791, "optim/lr": 0.0028639850481399455, "optim/total_tokens": 6560940032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8053503036499023, "created_at": "2025-01-16T16:54:34.178845+00:00"} {"global_step": 12515, "acc_step": 0, "speed/wps": 12895.858213939879, "speed/FLOPS": 202547174016115.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04439575970172882, "optim/lr": 0.0028639493954124022, "optim/total_tokens": 6561464320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.836338520050049, "created_at": "2025-01-16T16:54:44.347813+00:00"} {"global_step": 12516, "acc_step": 0, "speed/wps": 12901.769356824494, "speed/FLOPS": 202640016637879.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05702662095427513, "optim/lr": 0.0028639137382347347, "optim/total_tokens": 6561988608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 2.813340663909912, "created_at": "2025-01-16T16:54:54.515582+00:00"} {"global_step": 12517, "acc_step": 0, "speed/wps": 12904.295105615389, "speed/FLOPS": 202679686993383.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04686124250292778, "optim/lr": 0.002863878076607059, "optim/total_tokens": 6562512896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.876525402069092, "created_at": "2025-01-16T16:55:04.677175+00:00"} {"global_step": 12518, "acc_step": 0, "speed/wps": 12909.303346449784, "speed/FLOPS": 202758348297731.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06642143428325653, "optim/lr": 0.002863842410529492, "optim/total_tokens": 6563037184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.9307470321655273, "created_at": "2025-01-16T16:55:14.837946+00:00"} {"global_step": 12519, "acc_step": 0, "speed/wps": 12895.287417116973, "speed/FLOPS": 202538208867654.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.06894102692604065, "optim/lr": 0.002863806740002149, "optim/total_tokens": 6563561472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.7649660110473633, "created_at": "2025-01-16T16:55:25.008798+00:00"} {"global_step": 12520, "acc_step": 0, "speed/wps": 12892.287080110067, "speed/FLOPS": 202491084452066.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0557723194360733, "optim/lr": 0.002863771065025148, "optim/total_tokens": 6564085760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.756096839904785, "created_at": "2025-01-16T16:55:35.179100+00:00"} {"global_step": 12521, "acc_step": 0, "speed/wps": 12894.70406008068, "speed/FLOPS": 202529046443783.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.07193227112293243, "optim/lr": 0.002863735385598604, "optim/total_tokens": 6564610048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8739852905273438, "created_at": "2025-01-16T16:55:45.350381+00:00"} {"global_step": 12522, "acc_step": 0, "speed/wps": 12905.30562359163, "speed/FLOPS": 202695558566797.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04930207505822182, "optim/lr": 0.0028636997017226335, "optim/total_tokens": 6565134336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 2.8876490592956543, "created_at": "2025-01-16T16:55:55.510362+00:00"} {"global_step": 12523, "acc_step": 0, "speed/wps": 12902.468537488465, "speed/FLOPS": 202650998230982.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04839431121945381, "optim/lr": 0.002863664013397354, "optim/total_tokens": 6565658624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.7823305130004883, "created_at": "2025-01-16T16:56:05.675042+00:00"} {"global_step": 12524, "acc_step": 0, "speed/wps": 12899.810185403927, "speed/FLOPS": 202609245158536.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04474710673093796, "optim/lr": 0.002863628320622881, "optim/total_tokens": 6566182912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340917, "loss/out": 2.8249363899230957, "created_at": "2025-01-16T16:56:15.839580+00:00"} {"global_step": 12525, "acc_step": 0, "speed/wps": 12892.982514427704, "speed/FLOPS": 202502007203651.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052039001137018204, "optim/lr": 0.002863592623399331, "optim/total_tokens": 6566707200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.879167079925537, "created_at": "2025-01-16T16:56:26.011066+00:00"} {"global_step": 12526, "acc_step": 0, "speed/wps": 12896.890512400425, "speed/FLOPS": 202563387681964.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06570222228765488, "optim/lr": 0.0028635569217268213, "optim/total_tokens": 6567231488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.9310007095336914, "created_at": "2025-01-16T16:56:36.178038+00:00"} {"global_step": 12527, "acc_step": 0, "speed/wps": 12896.068765849954, "speed/FLOPS": 202550481023191.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.041768886148929596, "optim/lr": 0.002863521215605467, "optim/total_tokens": 6567755776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.9180431365966797, "created_at": "2025-01-16T16:56:46.346026+00:00"} {"global_step": 12528, "acc_step": 0, "speed/wps": 12906.945998160236, "speed/FLOPS": 202721322903508.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0791093036532402, "optim/lr": 0.002863485505035386, "optim/total_tokens": 6568280064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.7849507331848145, "created_at": "2025-01-16T16:56:56.505255+00:00"} {"global_step": 12529, "acc_step": 0, "speed/wps": 12899.143687760958, "speed/FLOPS": 202598776897189.5, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059133052825927734, "optim/lr": 0.002863449790016694, "optim/total_tokens": 6568804352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.73317813873291, "created_at": "2025-01-16T16:57:06.670200+00:00"} {"global_step": 12530, "acc_step": 0, "speed/wps": 12903.73453138036, "speed/FLOPS": 202670882404712.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05238001048564911, "optim/lr": 0.0028634140705495075, "optim/total_tokens": 6569328640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 276728, "loss/out": 2.824976921081543, "created_at": "2025-01-16T16:57:16.835523+00:00"} {"global_step": 12531, "acc_step": 0, "speed/wps": 12900.363933985895, "speed/FLOPS": 202617942540946.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04822830855846405, "optim/lr": 0.0028633783466339435, "optim/total_tokens": 6569852928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.8147382736206055, "created_at": "2025-01-16T16:57:27.004373+00:00"} {"global_step": 12532, "acc_step": 0, "speed/wps": 12902.658101766163, "speed/FLOPS": 202653975598452.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0445062480866909, "optim/lr": 0.0028633426182701183, "optim/total_tokens": 6570377216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.8121490478515625, "created_at": "2025-01-16T16:57:37.166761+00:00"} {"global_step": 12533, "acc_step": 0, "speed/wps": 12900.456338433969, "speed/FLOPS": 202619393879780.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04801354557275772, "optim/lr": 0.0028633068854581486, "optim/total_tokens": 6570901504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.862274408340454, "created_at": "2025-01-16T16:57:47.332615+00:00"} {"global_step": 12534, "acc_step": 0, "speed/wps": 12903.407216448459, "speed/FLOPS": 202665741473929.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05320518836379051, "optim/lr": 0.0028632711481981505, "optim/total_tokens": 6571425792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.882333517074585, "created_at": "2025-01-16T16:57:57.503439+00:00"} {"global_step": 12535, "acc_step": 0, "speed/wps": 12897.85156044201, "speed/FLOPS": 202578482262076.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044401612132787704, "optim/lr": 0.0028632354064902408, "optim/total_tokens": 6571950080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.8443081378936768, "created_at": "2025-01-16T16:58:07.671120+00:00"} {"global_step": 12536, "acc_step": 0, "speed/wps": 12903.790824653754, "speed/FLOPS": 202671766567926.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054997872561216354, "optim/lr": 0.002863199660334537, "optim/total_tokens": 6572474368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 2.9428932666778564, "created_at": "2025-01-16T16:58:17.837221+00:00"} {"global_step": 12537, "acc_step": 0, "speed/wps": 12901.011108676927, "speed/FLOPS": 202628107308780.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049626339226961136, "optim/lr": 0.0028631639097311543, "optim/total_tokens": 6572998656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.751715660095215, "created_at": "2025-01-16T16:58:28.001544+00:00"} {"global_step": 12538, "acc_step": 0, "speed/wps": 12902.093113605728, "speed/FLOPS": 202645101683017.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054984986782073975, "optim/lr": 0.00286312815468021, "optim/total_tokens": 6573522944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 2.9046332836151123, "created_at": "2025-01-16T16:58:38.164409+00:00"} {"global_step": 12539, "acc_step": 0, "speed/wps": 12892.147368065449, "speed/FLOPS": 202488890082420.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.057559169828891754, "optim/lr": 0.002863092395181821, "optim/total_tokens": 6574047232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8665099143981934, "created_at": "2025-01-16T16:58:48.335962+00:00"} {"global_step": 12540, "acc_step": 0, "speed/wps": 12900.989490793736, "speed/FLOPS": 202627767770218.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04174511507153511, "optim/lr": 0.0028630566312361043, "optim/total_tokens": 6574571520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.793935537338257, "created_at": "2025-01-16T16:58:58.500661+00:00"} {"global_step": 12541, "acc_step": 0, "speed/wps": 12909.1358275765, "speed/FLOPS": 202755717183631.56, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06778649985790253, "optim/lr": 0.002863020862843176, "optim/total_tokens": 6575095808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370920, "loss/out": 2.7904422283172607, "created_at": "2025-01-16T16:59:08.659289+00:00"} {"global_step": 12542, "acc_step": 0, "speed/wps": 12904.337619374892, "speed/FLOPS": 202680354730396.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046561360359191895, "optim/lr": 0.002862985090003152, "optim/total_tokens": 6575620096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.9071269035339355, "created_at": "2025-01-16T16:59:18.822702+00:00"} {"global_step": 12543, "acc_step": 0, "speed/wps": 12904.553448674946, "speed/FLOPS": 202683744626134.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06389415264129639, "optim/lr": 0.0028629493127161503, "optim/total_tokens": 6576144384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 2.921882152557373, "created_at": "2025-01-16T16:59:28.984929+00:00"} {"global_step": 12544, "acc_step": 0, "speed/wps": 12899.394561705065, "speed/FLOPS": 202602717217218.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05026787891983986, "optim/lr": 0.002862913530982287, "optim/total_tokens": 6576668672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8423404693603516, "created_at": "2025-01-16T16:59:39.150169+00:00"} {"global_step": 12545, "acc_step": 0, "speed/wps": 12904.813393612623, "speed/FLOPS": 202687827418582.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048572998493909836, "optim/lr": 0.0028628777448016797, "optim/total_tokens": 6577192960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.944162130355835, "created_at": "2025-01-16T16:59:49.310921+00:00"} {"global_step": 12546, "acc_step": 0, "speed/wps": 12914.29631537288, "speed/FLOPS": 202836769735727.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049841903150081635, "optim/lr": 0.002862841954174444, "optim/total_tokens": 6577717248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.8638687133789062, "created_at": "2025-01-16T16:59:59.464867+00:00"} {"global_step": 12547, "acc_step": 0, "speed/wps": 12906.753583099264, "speed/FLOPS": 202718300760569.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04594563692808151, "optim/lr": 0.0028628061591006973, "optim/total_tokens": 6578241536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.79293155670166, "created_at": "2025-01-16T17:00:09.626542+00:00"} {"global_step": 12548, "acc_step": 0, "speed/wps": 12900.765719211618, "speed/FLOPS": 202624253130026.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05345585197210312, "optim/lr": 0.0028627703595805565, "optim/total_tokens": 6578765824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8832478523254395, "created_at": "2025-01-16T17:00:19.792649+00:00"} {"global_step": 12549, "acc_step": 0, "speed/wps": 12912.376658083906, "speed/FLOPS": 202806618880118.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051112111657857895, "optim/lr": 0.0028627345556141376, "optim/total_tokens": 6579290112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 2.8936564922332764, "created_at": "2025-01-16T17:00:29.954926+00:00"} {"global_step": 12550, "acc_step": 0, "speed/wps": 12906.327019468588, "speed/FLOPS": 202711600992590.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04174289107322693, "optim/lr": 0.0028626987472015583, "optim/total_tokens": 6579814400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.8054540157318115, "created_at": "2025-01-16T17:00:40.114101+00:00"} {"global_step": 12551, "acc_step": 0, "speed/wps": 12901.561511614276, "speed/FLOPS": 202636752143244.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051338277757167816, "optim/lr": 0.0028626629343429352, "optim/total_tokens": 6580338688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 2.8601279258728027, "created_at": "2025-01-16T17:00:50.277115+00:00"} {"global_step": 12552, "acc_step": 0, "speed/wps": 12902.640570006852, "speed/FLOPS": 202653700238081.03, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05708729848265648, "optim/lr": 0.002862627117038385, "optim/total_tokens": 6580862976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8309006690979004, "created_at": "2025-01-16T17:01:00.442060+00:00"} {"global_step": 12553, "acc_step": 0, "speed/wps": 12906.87988716507, "speed/FLOPS": 202720284539483.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061414800584316254, "optim/lr": 0.0028625912952880243, "optim/total_tokens": 6581387264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474921, "loss/out": 2.8067874908447266, "created_at": "2025-01-16T17:01:10.601333+00:00"} {"global_step": 12554, "acc_step": 0, "speed/wps": 12904.895423812754, "speed/FLOPS": 202689115815596.56, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054487861692905426, "optim/lr": 0.0028625554690919707, "optim/total_tokens": 6581911552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.814094066619873, "created_at": "2025-01-16T17:01:20.763072+00:00"} {"global_step": 12555, "acc_step": 0, "speed/wps": 12895.85871377997, "speed/FLOPS": 202547181866790.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.061540719121694565, "optim/lr": 0.0028625196384503408, "optim/total_tokens": 6582435840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.747535228729248, "created_at": "2025-01-16T17:01:30.933438+00:00"} {"global_step": 12556, "acc_step": 0, "speed/wps": 12895.962774713416, "speed/FLOPS": 202548816286745.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49779409170150757, "optim/lr": 0.0028624838033632512, "optim/total_tokens": 6582960128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 2.898573637008667, "created_at": "2025-01-16T17:01:41.105451+00:00"} {"global_step": 12557, "acc_step": 0, "speed/wps": 12902.220003047552, "speed/FLOPS": 202647094656065.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08591517060995102, "optim/lr": 0.002862447963830819, "optim/total_tokens": 6583484416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9192516803741455, "created_at": "2025-01-16T17:01:51.267833+00:00"} {"global_step": 12558, "acc_step": 0, "speed/wps": 12904.71841936675, "speed/FLOPS": 202686335717542.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04666653275489807, "optim/lr": 0.0028624121198531608, "optim/total_tokens": 6584008704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.839479923248291, "created_at": "2025-01-16T17:02:01.428522+00:00"} {"global_step": 12559, "acc_step": 0, "speed/wps": 12902.93567202559, "speed/FLOPS": 202658335220797.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04837789386510849, "optim/lr": 0.002862376271430395, "optim/total_tokens": 6584532992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.7440590858459473, "created_at": "2025-01-16T17:02:11.590387+00:00"} {"global_step": 12560, "acc_step": 0, "speed/wps": 12897.727090172557, "speed/FLOPS": 202576527285454.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04737699031829834, "optim/lr": 0.002862340418562636, "optim/total_tokens": 6585057280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.828300714492798, "created_at": "2025-01-16T17:02:21.757973+00:00"} {"global_step": 12561, "acc_step": 0, "speed/wps": 12889.198741634247, "speed/FLOPS": 202442577852481.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047763749957084656, "optim/lr": 0.0028623045612500035, "optim/total_tokens": 6585581568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8794660568237305, "created_at": "2025-01-16T17:02:31.932349+00:00"} {"global_step": 12562, "acc_step": 0, "speed/wps": 12890.207224761498, "speed/FLOPS": 202458417465796.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04609108343720436, "optim/lr": 0.0028622686994926125, "optim/total_tokens": 6586105856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.901109218597412, "created_at": "2025-01-16T17:02:42.105334+00:00"} {"global_step": 12563, "acc_step": 0, "speed/wps": 12895.697568759979, "speed/FLOPS": 202544650862815.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04586448147892952, "optim/lr": 0.0028622328332905808, "optim/total_tokens": 6586630144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.7495837211608887, "created_at": "2025-01-16T17:02:52.275695+00:00"} {"global_step": 12564, "acc_step": 0, "speed/wps": 12903.808008154698, "speed/FLOPS": 202672036458421.1, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050574664026498795, "optim/lr": 0.0028621969626440257, "optim/total_tokens": 6587154432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 2.925464630126953, "created_at": "2025-01-16T17:03:02.439908+00:00"} {"global_step": 12565, "acc_step": 0, "speed/wps": 12895.763376187391, "speed/FLOPS": 202545684458891.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07024059444665909, "optim/lr": 0.002862161087553063, "optim/total_tokens": 6587678720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 2.812755584716797, "created_at": "2025-01-16T17:03:12.610024+00:00"} {"global_step": 12566, "acc_step": 0, "speed/wps": 12890.9980428628, "speed/FLOPS": 202470838350774.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057355500757694244, "optim/lr": 0.0028621252080178114, "optim/total_tokens": 6588203008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9360008239746094, "created_at": "2025-01-16T17:03:22.786409+00:00"} {"global_step": 12567, "acc_step": 0, "speed/wps": 12903.91297237031, "speed/FLOPS": 202673685065663.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054234523326158524, "optim/lr": 0.002862089324038387, "optim/total_tokens": 6588727296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.5911107063293457, "created_at": "2025-01-16T17:03:32.948179+00:00"} {"global_step": 12568, "acc_step": 0, "speed/wps": 12901.62074980322, "speed/FLOPS": 202637682560403.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0595843605697155, "optim/lr": 0.0028620534356149072, "optim/total_tokens": 6589251584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 2.905515432357788, "created_at": "2025-01-16T17:03:43.112368+00:00"} {"global_step": 12569, "acc_step": 0, "speed/wps": 12897.62592616116, "speed/FLOPS": 202574938365641.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04326795041561127, "optim/lr": 0.002862017542747489, "optim/total_tokens": 6589775872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.7541663646698, "created_at": "2025-01-16T17:03:53.283228+00:00"} {"global_step": 12570, "acc_step": 0, "speed/wps": 12901.061832935418, "speed/FLOPS": 202628904002962.1, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04977209493517876, "optim/lr": 0.002861981645436249, "optim/total_tokens": 6590300160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 2.9731171131134033, "created_at": "2025-01-16T17:04:03.450685+00:00"} {"global_step": 12571, "acc_step": 0, "speed/wps": 12900.299238416226, "speed/FLOPS": 202616926408123.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045700326561927795, "optim/lr": 0.002861945743681305, "optim/total_tokens": 6590824448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.8281302452087402, "created_at": "2025-01-16T17:04:13.615054+00:00"} {"global_step": 12572, "acc_step": 0, "speed/wps": 12902.358838176617, "speed/FLOPS": 202649275252547.1, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.043723996728658676, "optim/lr": 0.0028619098374827743, "optim/total_tokens": 6591348736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.814068555831909, "created_at": "2025-01-16T17:04:23.779491+00:00"} {"global_step": 12573, "acc_step": 0, "speed/wps": 12908.837841745228, "speed/FLOPS": 202751036906676.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05273555964231491, "optim/lr": 0.0028618739268407734, "optim/total_tokens": 6591873024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 2.8458595275878906, "created_at": "2025-01-16T17:04:33.936836+00:00"} {"global_step": 12574, "acc_step": 0, "speed/wps": 12892.932923775264, "speed/FLOPS": 202501228314310.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06490080058574677, "optim/lr": 0.00286183801175542, "optim/total_tokens": 6592397312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.79257869720459, "created_at": "2025-01-16T17:04:44.106836+00:00"} {"global_step": 12575, "acc_step": 0, "speed/wps": 12906.529327447597, "speed/FLOPS": 202714778517400.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04958263412117958, "optim/lr": 0.002861802092226831, "optim/total_tokens": 6592921600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8008928298950195, "created_at": "2025-01-16T17:04:54.271375+00:00"} {"global_step": 12576, "acc_step": 0, "speed/wps": 12904.714474314753, "speed/FLOPS": 202686273755079.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04889804124832153, "optim/lr": 0.002861766168255123, "optim/total_tokens": 6593445888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.7644567489624023, "created_at": "2025-01-16T17:05:04.432416+00:00"} {"global_step": 12577, "acc_step": 0, "speed/wps": 12903.187827988122, "speed/FLOPS": 202662295676688.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06937454640865326, "optim/lr": 0.002861730239840414, "optim/total_tokens": 6593970176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.926544666290283, "created_at": "2025-01-16T17:05:14.595204+00:00"} {"global_step": 12578, "acc_step": 0, "speed/wps": 12895.87304018241, "speed/FLOPS": 202547406882627.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046812091022729874, "optim/lr": 0.0028616943069828217, "optim/total_tokens": 6594494464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 2.85697865486145, "created_at": "2025-01-16T17:05:24.762970+00:00"} {"global_step": 12579, "acc_step": 0, "speed/wps": 12899.320145607602, "speed/FLOPS": 202601548410149.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05771131068468094, "optim/lr": 0.002861658369682462, "optim/total_tokens": 6595018752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.865499496459961, "created_at": "2025-01-16T17:05:34.928193+00:00"} {"global_step": 12580, "acc_step": 0, "speed/wps": 12901.304357452453, "speed/FLOPS": 202632713183765.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05266352742910385, "optim/lr": 0.0028616224279394524, "optim/total_tokens": 6595543040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.915808916091919, "created_at": "2025-01-16T17:05:45.092031+00:00"} {"global_step": 12581, "acc_step": 0, "speed/wps": 12902.776805840856, "speed/FLOPS": 202655840009061.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04599491134285927, "optim/lr": 0.0028615864817539117, "optim/total_tokens": 6596067328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 2.778740882873535, "created_at": "2025-01-16T17:05:55.253950+00:00"} {"global_step": 12582, "acc_step": 0, "speed/wps": 12902.705315086498, "speed/FLOPS": 202654717148525.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06170345097780228, "optim/lr": 0.0028615505311259557, "optim/total_tokens": 6596591616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.943085193634033, "created_at": "2025-01-16T17:06:05.417207+00:00"} {"global_step": 12583, "acc_step": 0, "speed/wps": 12898.057992961434, "speed/FLOPS": 202581724568462.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0543326772749424, "optim/lr": 0.002861514576055702, "optim/total_tokens": 6597115904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 2.7730484008789062, "created_at": "2025-01-16T17:06:15.584688+00:00"} {"global_step": 12584, "acc_step": 0, "speed/wps": 12894.124725114107, "speed/FLOPS": 202519947192043.3, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05829022824764252, "optim/lr": 0.002861478616543268, "optim/total_tokens": 6597640192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8931777477264404, "created_at": "2025-01-16T17:06:25.754084+00:00"} {"global_step": 12585, "acc_step": 0, "speed/wps": 12892.74412821353, "speed/FLOPS": 202498263020580.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0565185509622097, "optim/lr": 0.0028614426525887713, "optim/total_tokens": 6598164480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.7863359451293945, "created_at": "2025-01-16T17:06:35.924887+00:00"} {"global_step": 12586, "acc_step": 0, "speed/wps": 12903.574335865304, "speed/FLOPS": 202668366314016.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04251401871442795, "optim/lr": 0.0028614066841923286, "optim/total_tokens": 6598688768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390425, "loss/out": 2.994436264038086, "created_at": "2025-01-16T17:06:46.088017+00:00"} {"global_step": 12587, "acc_step": 0, "speed/wps": 12903.66460943234, "speed/FLOPS": 202669784184435.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03865981101989746, "optim/lr": 0.0028613707113540575, "optim/total_tokens": 6599213056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.82399320602417, "created_at": "2025-01-16T17:06:56.250131+00:00"} {"global_step": 12588, "acc_step": 0, "speed/wps": 12903.555562945714, "speed/FLOPS": 202668071459517.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04341275244951248, "optim/lr": 0.0028613347340740755, "optim/total_tokens": 6599737344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.8217339515686035, "created_at": "2025-01-16T17:07:06.418236+00:00"} {"global_step": 12589, "acc_step": 0, "speed/wps": 12899.406468691384, "speed/FLOPS": 202602904232800.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06281092017889023, "optim/lr": 0.0028612987523525006, "optim/total_tokens": 6600261632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 2.907731294631958, "created_at": "2025-01-16T17:07:16.584086+00:00"} {"global_step": 12590, "acc_step": 0, "speed/wps": 12901.154269851695, "speed/FLOPS": 202630355851753.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050891779363155365, "optim/lr": 0.002861262766189449, "optim/total_tokens": 6600785920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 2.950758695602417, "created_at": "2025-01-16T17:07:26.752674+00:00"} {"global_step": 12591, "acc_step": 0, "speed/wps": 12901.11184092624, "speed/FLOPS": 202629689447176.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05803167074918747, "optim/lr": 0.0028612267755850386, "optim/total_tokens": 6601310208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460308, "loss/out": 2.8954195976257324, "created_at": "2025-01-16T17:07:36.918145+00:00"} {"global_step": 12592, "acc_step": 0, "speed/wps": 12902.926143351082, "speed/FLOPS": 202658185559871.06, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10124891251325607, "optim/lr": 0.0028611907805393875, "optim/total_tokens": 6601834496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.790700912475586, "created_at": "2025-01-16T17:07:47.086960+00:00"} {"global_step": 12593, "acc_step": 0, "speed/wps": 12902.98611804476, "speed/FLOPS": 202659127544849.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04950680956244469, "optim/lr": 0.0028611547810526113, "optim/total_tokens": 6602358784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.949660062789917, "created_at": "2025-01-16T17:07:57.257703+00:00"} {"global_step": 12594, "acc_step": 0, "speed/wps": 12902.129317968467, "speed/FLOPS": 202645670322284.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08175777643918991, "optim/lr": 0.0028611187771248295, "optim/total_tokens": 6602883072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 3.0544207096099854, "created_at": "2025-01-16T17:08:07.420489+00:00"} {"global_step": 12595, "acc_step": 0, "speed/wps": 12902.603856497231, "speed/FLOPS": 202653123601962.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045973025262355804, "optim/lr": 0.002861082768756159, "optim/total_tokens": 6603407360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 2.8818469047546387, "created_at": "2025-01-16T17:08:17.584803+00:00"} {"global_step": 12596, "acc_step": 0, "speed/wps": 12905.695876187743, "speed/FLOPS": 202701688020082.6, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06989352405071259, "optim/lr": 0.0028610467559467166, "optim/total_tokens": 6603931648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.898777484893799, "created_at": "2025-01-16T17:08:27.749937+00:00"} {"global_step": 12597, "acc_step": 0, "speed/wps": 12907.016048003687, "speed/FLOPS": 202722423132558.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051179178059101105, "optim/lr": 0.00286101073869662, "optim/total_tokens": 6604455936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 2.836827278137207, "created_at": "2025-01-16T17:08:37.911142+00:00"} {"global_step": 12598, "acc_step": 0, "speed/wps": 12901.56606737798, "speed/FLOPS": 202636823697773.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.078132763504982, "optim/lr": 0.002860974717005988, "optim/total_tokens": 6604980224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.7923855781555176, "created_at": "2025-01-16T17:08:48.076889+00:00"} {"global_step": 12599, "acc_step": 0, "speed/wps": 12904.353558409386, "speed/FLOPS": 202680605074835.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08082127571105957, "optim/lr": 0.002860938690874936, "optim/total_tokens": 6605504512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.8534531593322754, "created_at": "2025-01-16T17:08:58.238965+00:00"} {"global_step": 12600, "acc_step": 0, "speed/wps": 12901.129768257384, "speed/FLOPS": 202629971020546.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05218598246574402, "optim/lr": 0.0028609026603035833, "optim/total_tokens": 6606028800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 2.853780746459961, "created_at": "2025-01-16T17:09:08.403442+00:00"} {"global_step": 12601, "acc_step": 0, "speed/wps": 12896.724219610387, "speed/FLOPS": 202560775825187.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0747295692563057, "optim/lr": 0.002860866625292046, "optim/total_tokens": 6606553088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8544812202453613, "created_at": "2025-01-16T17:09:18.570457+00:00"} {"global_step": 12602, "acc_step": 0, "speed/wps": 12900.866117691045, "speed/FLOPS": 202625830026108.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06122535839676857, "optim/lr": 0.002860830585840443, "optim/total_tokens": 6607077376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.923102617263794, "created_at": "2025-01-16T17:09:28.736740+00:00"} {"global_step": 12603, "acc_step": 0, "speed/wps": 12904.290093575033, "speed/FLOPS": 202679608272402.25, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06295217573642731, "optim/lr": 0.0028607945419488914, "optim/total_tokens": 6607601664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9421889781951904, "created_at": "2025-01-16T17:09:38.897548+00:00"} {"global_step": 12604, "acc_step": 0, "speed/wps": 12886.71106338674, "speed/FLOPS": 202403505447176.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06106113642454147, "optim/lr": 0.0028607584936175086, "optim/total_tokens": 6608125952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9028775691986084, "created_at": "2025-01-16T17:09:49.076466+00:00"} {"global_step": 12605, "acc_step": 0, "speed/wps": 12892.159743514983, "speed/FLOPS": 202489084455861.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04778147116303444, "optim/lr": 0.0028607224408464123, "optim/total_tokens": 6608650240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 2.8088958263397217, "created_at": "2025-01-16T17:09:59.250528+00:00"} {"global_step": 12606, "acc_step": 0, "speed/wps": 12886.649104151935, "speed/FLOPS": 202402532292252.72, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05276450142264366, "optim/lr": 0.00286068638363572, "optim/total_tokens": 6609174528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288679, "loss/out": 2.814586639404297, "created_at": "2025-01-16T17:10:09.428547+00:00"} {"global_step": 12607, "acc_step": 0, "speed/wps": 12885.31520115797, "speed/FLOPS": 202381581512757.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053396619856357574, "optim/lr": 0.0028606503219855495, "optim/total_tokens": 6609698816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.7611441612243652, "created_at": "2025-01-16T17:10:19.606245+00:00"} {"global_step": 12608, "acc_step": 0, "speed/wps": 12904.322271983074, "speed/FLOPS": 202680113678518.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061156924813985825, "optim/lr": 0.0028606142558960193, "optim/total_tokens": 6610223104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.9074342250823975, "created_at": "2025-01-16T17:10:29.770756+00:00"} {"global_step": 12609, "acc_step": 0, "speed/wps": 12902.63685334151, "speed/FLOPS": 202653641862744.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05304458737373352, "optim/lr": 0.0028605781853672453, "optim/total_tokens": 6610747392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343732, "loss/out": 2.728163003921509, "created_at": "2025-01-16T17:10:39.941647+00:00"} {"global_step": 12610, "acc_step": 0, "speed/wps": 12905.751600758354, "speed/FLOPS": 202702563251037.62, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0544426366686821, "optim/lr": 0.0028605421103993466, "optim/total_tokens": 6611271680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.787241220474243, "created_at": "2025-01-16T17:10:50.110025+00:00"} {"global_step": 12611, "acc_step": 0, "speed/wps": 12905.0732851224, "speed/FLOPS": 202691909371871.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04474177956581116, "optim/lr": 0.00286050603099244, "optim/total_tokens": 6611795968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8417086601257324, "created_at": "2025-01-16T17:11:00.270961+00:00"} {"global_step": 12612, "acc_step": 0, "speed/wps": 12901.053995791463, "speed/FLOPS": 202628780909846.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04552828520536423, "optim/lr": 0.0028604699471466437, "optim/total_tokens": 6612320256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 2.8271875381469727, "created_at": "2025-01-16T17:11:10.440655+00:00"} {"global_step": 12613, "acc_step": 0, "speed/wps": 12905.144764252494, "speed/FLOPS": 202693032049832.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0486709401011467, "optim/lr": 0.002860433858862076, "optim/total_tokens": 6612844544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.921231269836426, "created_at": "2025-01-16T17:11:20.602625+00:00"} {"global_step": 12614, "acc_step": 0, "speed/wps": 12905.471032445319, "speed/FLOPS": 202698156540140.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04987821355462074, "optim/lr": 0.0028603977661388535, "optim/total_tokens": 6613368832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.7562406063079834, "created_at": "2025-01-16T17:11:30.771210+00:00"} {"global_step": 12615, "acc_step": 0, "speed/wps": 12903.451113254696, "speed/FLOPS": 202666430933591.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04137399047613144, "optim/lr": 0.002860361668977095, "optim/total_tokens": 6613893120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.8077735900878906, "created_at": "2025-01-16T17:11:40.934900+00:00"} {"global_step": 12616, "acc_step": 0, "speed/wps": 12904.601986916885, "speed/FLOPS": 202684506985922.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04277310520410538, "optim/lr": 0.002860325567376917, "optim/total_tokens": 6614417408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.791409492492676, "created_at": "2025-01-16T17:11:51.101466+00:00"} {"global_step": 12617, "acc_step": 0, "speed/wps": 12906.479631346008, "speed/FLOPS": 202713997971835.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04789687693119049, "optim/lr": 0.002860289461338438, "optim/total_tokens": 6614941696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 2.8459653854370117, "created_at": "2025-01-16T17:12:01.262788+00:00"} {"global_step": 12618, "acc_step": 0, "speed/wps": 12904.640160114639, "speed/FLOPS": 202685106548448.6, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05927126854658127, "optim/lr": 0.0028602533508617764, "optim/total_tokens": 6615465984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8094589710235596, "created_at": "2025-01-16T17:12:11.424468+00:00"} {"global_step": 12619, "acc_step": 0, "speed/wps": 12901.928163901232, "speed/FLOPS": 202642510921243.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045500922948122025, "optim/lr": 0.0028602172359470496, "optim/total_tokens": 6615990272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.860037326812744, "created_at": "2025-01-16T17:12:21.588189+00:00"} {"global_step": 12620, "acc_step": 0, "speed/wps": 12905.112824329452, "speed/FLOPS": 202692530389450.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05825364962220192, "optim/lr": 0.0028601811165943755, "optim/total_tokens": 6616514560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8593668937683105, "created_at": "2025-01-16T17:12:31.751320+00:00"} {"global_step": 12621, "acc_step": 0, "speed/wps": 12898.973063113132, "speed/FLOPS": 202596097002632.56, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.03847265616059303, "optim/lr": 0.0028601449928038706, "optim/total_tokens": 6617038848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8419201374053955, "created_at": "2025-01-16T17:12:41.917208+00:00"} {"global_step": 12622, "acc_step": 0, "speed/wps": 12901.310515531091, "speed/FLOPS": 202632809904852.84, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06061723083257675, "optim/lr": 0.0028601088645756542, "optim/total_tokens": 6617563136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439286, "loss/out": 2.7897589206695557, "created_at": "2025-01-16T17:12:52.080529+00:00"} {"global_step": 12623, "acc_step": 0, "speed/wps": 12901.58973209121, "speed/FLOPS": 202637195384613.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0571226142346859, "optim/lr": 0.0028600727319098448, "optim/total_tokens": 6618087424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.874556541442871, "created_at": "2025-01-16T17:13:02.247278+00:00"} {"global_step": 12624, "acc_step": 0, "speed/wps": 12905.7006112072, "speed/FLOPS": 202701762390069.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05495765432715416, "optim/lr": 0.0028600365948065584, "optim/total_tokens": 6618611712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 2.727318525314331, "created_at": "2025-01-16T17:13:12.408234+00:00"} {"global_step": 12625, "acc_step": 0, "speed/wps": 12900.447268682035, "speed/FLOPS": 202619251426862.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052050817757844925, "optim/lr": 0.0028600004532659145, "optim/total_tokens": 6619136000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.7826828956604004, "created_at": "2025-01-16T17:13:22.574737+00:00"} {"global_step": 12626, "acc_step": 0, "speed/wps": 12904.278434325304, "speed/FLOPS": 202679425147863.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06330455839633942, "optim/lr": 0.0028599643072880297, "optim/total_tokens": 6619660288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 2.9500010013580322, "created_at": "2025-01-16T17:13:32.736303+00:00"} {"global_step": 12627, "acc_step": 0, "speed/wps": 12898.792545653218, "speed/FLOPS": 202593261727867.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0473342202603817, "optim/lr": 0.002859928156873023, "optim/total_tokens": 6620184576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 2.816248893737793, "created_at": "2025-01-16T17:13:42.904969+00:00"} {"global_step": 12628, "acc_step": 0, "speed/wps": 12898.756225245284, "speed/FLOPS": 202592691265951.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04993937164545059, "optim/lr": 0.0028598920020210124, "optim/total_tokens": 6620708864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8744688034057617, "created_at": "2025-01-16T17:13:53.072240+00:00"} {"global_step": 12629, "acc_step": 0, "speed/wps": 12898.423885810687, "speed/FLOPS": 202587471418449.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05169886350631714, "optim/lr": 0.002859855842732115, "optim/total_tokens": 6621233152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8350284099578857, "created_at": "2025-01-16T17:14:03.242314+00:00"} {"global_step": 12630, "acc_step": 0, "speed/wps": 12896.71459374477, "speed/FLOPS": 202560624637740.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04710399731993675, "optim/lr": 0.002859819679006449, "optim/total_tokens": 6621757440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 2.7872233390808105, "created_at": "2025-01-16T17:14:13.412248+00:00"} {"global_step": 12631, "acc_step": 0, "speed/wps": 12900.295244037101, "speed/FLOPS": 202616863670910.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047703973948955536, "optim/lr": 0.0028597835108441333, "optim/total_tokens": 6622281728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.9404375553131104, "created_at": "2025-01-16T17:14:23.577165+00:00"} {"global_step": 12632, "acc_step": 0, "speed/wps": 12900.015245170578, "speed/FLOPS": 202612465903953.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054752450436353683, "optim/lr": 0.0028597473382452844, "optim/total_tokens": 6622806016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.825845241546631, "created_at": "2025-01-16T17:14:33.741536+00:00"} {"global_step": 12633, "acc_step": 0, "speed/wps": 12904.915727107978, "speed/FLOPS": 202689434706747.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052043404430150986, "optim/lr": 0.002859711161210022, "optim/total_tokens": 6623330304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.9210071563720703, "created_at": "2025-01-16T17:14:43.903185+00:00"} {"global_step": 12634, "acc_step": 0, "speed/wps": 12902.84073961138, "speed/FLOPS": 202656844176782.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0659298449754715, "optim/lr": 0.002859674979738462, "optim/total_tokens": 6623854592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.825960636138916, "created_at": "2025-01-16T17:14:54.071588+00:00"} {"global_step": 12635, "acc_step": 0, "speed/wps": 12904.60598278023, "speed/FLOPS": 202684569746447.1, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06583468616008759, "optim/lr": 0.0028596387938307245, "optim/total_tokens": 6624378880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.924821138381958, "created_at": "2025-01-16T17:15:04.235627+00:00"} {"global_step": 12636, "acc_step": 0, "speed/wps": 12903.249917970072, "speed/FLOPS": 202663270885175.25, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054714493453502655, "optim/lr": 0.002859602603486927, "optim/total_tokens": 6624903168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8016934394836426, "created_at": "2025-01-16T17:15:14.398641+00:00"} {"global_step": 12637, "acc_step": 0, "speed/wps": 12903.37630608317, "speed/FLOPS": 202665255984166.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05010881647467613, "optim/lr": 0.0028595664087071866, "optim/total_tokens": 6625427456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 2.8233280181884766, "created_at": "2025-01-16T17:15:24.560519+00:00"} {"global_step": 12638, "acc_step": 0, "speed/wps": 12899.697123732376, "speed/FLOPS": 202607469369621.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04436928778886795, "optim/lr": 0.0028595302094916227, "optim/total_tokens": 6625951744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.8201310634613037, "created_at": "2025-01-16T17:15:34.724677+00:00"} {"global_step": 12639, "acc_step": 0, "speed/wps": 12903.736875099668, "speed/FLOPS": 202670919216044.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05462351068854332, "optim/lr": 0.0028594940058403527, "optim/total_tokens": 6626476032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 2.8742480278015137, "created_at": "2025-01-16T17:15:44.885826+00:00"} {"global_step": 12640, "acc_step": 0, "speed/wps": 12900.913874893793, "speed/FLOPS": 202626580118602.72, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05911625549197197, "optim/lr": 0.002859457797753495, "optim/total_tokens": 6627000320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.900397777557373, "created_at": "2025-01-16T17:15:55.058490+00:00"} {"global_step": 12641, "acc_step": 0, "speed/wps": 12902.671923984999, "speed/FLOPS": 202654192695392.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04449122026562691, "optim/lr": 0.0028594215852311674, "optim/total_tokens": 6627524608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.8800859451293945, "created_at": "2025-01-16T17:16:05.230916+00:00"} {"global_step": 12642, "acc_step": 0, "speed/wps": 12899.822184200935, "speed/FLOPS": 202609433616131.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.044619329273700714, "optim/lr": 0.002859385368273488, "optim/total_tokens": 6628048896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.805738687515259, "created_at": "2025-01-16T17:16:15.396896+00:00"} {"global_step": 12643, "acc_step": 0, "speed/wps": 12901.162706709505, "speed/FLOPS": 202630488364199.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040690820664167404, "optim/lr": 0.0028593491468805756, "optim/total_tokens": 6628573184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 2.937985420227051, "created_at": "2025-01-16T17:16:25.563964+00:00"} {"global_step": 12644, "acc_step": 0, "speed/wps": 12890.875924631659, "speed/FLOPS": 202468920316146.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05169132351875305, "optim/lr": 0.0028593129210525477, "optim/total_tokens": 6629097472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 2.802058219909668, "created_at": "2025-01-16T17:16:35.735399+00:00"} {"global_step": 12645, "acc_step": 0, "speed/wps": 12893.890688711746, "speed/FLOPS": 202516271328744.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07526251673698425, "optim/lr": 0.002859276690789523, "optim/total_tokens": 6629621760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.9827990531921387, "created_at": "2025-01-16T17:16:45.907086+00:00"} {"global_step": 12646, "acc_step": 0, "speed/wps": 12902.054830099458, "speed/FLOPS": 202644500387944.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05608060583472252, "optim/lr": 0.002859240456091619, "optim/total_tokens": 6630146048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.7733590602874756, "created_at": "2025-01-16T17:16:56.070025+00:00"} {"global_step": 12647, "acc_step": 0, "speed/wps": 12894.170422949017, "speed/FLOPS": 202520664939335.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043088238686323166, "optim/lr": 0.0028592042169589553, "optim/total_tokens": 6630670336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316449, "loss/out": 2.8693995475769043, "created_at": "2025-01-16T17:17:06.239731+00:00"} {"global_step": 12648, "acc_step": 0, "speed/wps": 12890.08769993784, "speed/FLOPS": 202456540164196.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056360967457294464, "optim/lr": 0.002859167973391649, "optim/total_tokens": 6631194624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8927974700927734, "created_at": "2025-01-16T17:17:16.415462+00:00"} {"global_step": 12649, "acc_step": 0, "speed/wps": 12896.915040441681, "speed/FLOPS": 202563772928557.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.054673440754413605, "optim/lr": 0.002859131725389818, "optim/total_tokens": 6631718912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8876471519470215, "created_at": "2025-01-16T17:17:26.586559+00:00"} {"global_step": 12650, "acc_step": 0, "speed/wps": 12902.219953369064, "speed/FLOPS": 202647093875796.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062010787427425385, "optim/lr": 0.002859095472953582, "optim/total_tokens": 6632243200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.9048798084259033, "created_at": "2025-01-16T17:17:36.749233+00:00"} {"global_step": 12651, "acc_step": 0, "speed/wps": 12900.423521837836, "speed/FLOPS": 202618878450043.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047270555049180984, "optim/lr": 0.002859059216083058, "optim/total_tokens": 6632767488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8318238258361816, "created_at": "2025-01-16T17:17:46.917336+00:00"} {"global_step": 12652, "acc_step": 0, "speed/wps": 12898.908076481943, "speed/FLOPS": 202595076298288.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05911331996321678, "optim/lr": 0.0028590229547783646, "optim/total_tokens": 6633291776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419023, "loss/out": 2.809556484222412, "created_at": "2025-01-16T17:17:57.087526+00:00"} {"global_step": 12653, "acc_step": 0, "speed/wps": 12902.394801545748, "speed/FLOPS": 202649840106678.16, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0472896471619606, "optim/lr": 0.0028589866890396205, "optim/total_tokens": 6633816064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.80098819732666, "created_at": "2025-01-16T17:18:07.250180+00:00"} {"global_step": 12654, "acc_step": 0, "speed/wps": 12907.951316262506, "speed/FLOPS": 202737112805756.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052902381867170334, "optim/lr": 0.0028589504188669436, "optim/total_tokens": 6634340352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.7431299686431885, "created_at": "2025-01-16T17:18:17.409597+00:00"} {"global_step": 12655, "acc_step": 0, "speed/wps": 12904.125326690992, "speed/FLOPS": 202677020382074.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05442536994814873, "optim/lr": 0.0028589141442604527, "optim/total_tokens": 6634864640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 2.8261680603027344, "created_at": "2025-01-16T17:18:27.571350+00:00"} {"global_step": 12656, "acc_step": 0, "speed/wps": 12909.094751641065, "speed/FLOPS": 202755072029621.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05129789188504219, "optim/lr": 0.0028588778652202653, "optim/total_tokens": 6635388928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8516149520874023, "created_at": "2025-01-16T17:18:37.729740+00:00"} {"global_step": 12657, "acc_step": 0, "speed/wps": 12898.310974923632, "speed/FLOPS": 202585697997811.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055803921073675156, "optim/lr": 0.002858841581746501, "optim/total_tokens": 6635913216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 2.8694918155670166, "created_at": "2025-01-16T17:18:47.896479+00:00"} {"global_step": 12658, "acc_step": 0, "speed/wps": 12898.007156385458, "speed/FLOPS": 202580926110179.53, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04913131892681122, "optim/lr": 0.002858805293839277, "optim/total_tokens": 6636437504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.833207607269287, "created_at": "2025-01-16T17:18:58.064548+00:00"} {"global_step": 12659, "acc_step": 0, "speed/wps": 12901.145368233616, "speed/FLOPS": 202630216039609.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06505835801362991, "optim/lr": 0.002858769001498712, "optim/total_tokens": 6636961792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8752312660217285, "created_at": "2025-01-16T17:19:08.231700+00:00"} {"global_step": 12660, "acc_step": 0, "speed/wps": 12901.590915979637, "speed/FLOPS": 202637213979207.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0541975162923336, "optim/lr": 0.0028587327047249252, "optim/total_tokens": 6637486080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 2.8788414001464844, "created_at": "2025-01-16T17:19:18.397715+00:00"} {"global_step": 12661, "acc_step": 0, "speed/wps": 12905.681656948747, "speed/FLOPS": 202701464687394.88, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05234922468662262, "optim/lr": 0.0028586964035180345, "optim/total_tokens": 6638010368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.869486093521118, "created_at": "2025-01-16T17:19:28.557600+00:00"} {"global_step": 12662, "acc_step": 0, "speed/wps": 12900.160395580291, "speed/FLOPS": 202614745690595.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04795806482434273, "optim/lr": 0.0028586600978781577, "optim/total_tokens": 6638534656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8209941387176514, "created_at": "2025-01-16T17:19:38.731842+00:00"} {"global_step": 12663, "acc_step": 0, "speed/wps": 12906.276637273593, "speed/FLOPS": 202710809670985.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0624762587249279, "optim/lr": 0.0028586237878054145, "optim/total_tokens": 6639058944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 2.7317113876342773, "created_at": "2025-01-16T17:19:48.891203+00:00"} {"global_step": 12664, "acc_step": 0, "speed/wps": 12900.886502915746, "speed/FLOPS": 202626150204074.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0534973107278347, "optim/lr": 0.0028585874732999224, "optim/total_tokens": 6639583232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.88838267326355, "created_at": "2025-01-16T17:19:59.056429+00:00"} {"global_step": 12665, "acc_step": 0, "speed/wps": 12903.48671241746, "speed/FLOPS": 202666990067361.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052810780704021454, "optim/lr": 0.0028585511543618002, "optim/total_tokens": 6640107520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459779, "loss/out": 2.847552537918091, "created_at": "2025-01-16T17:20:09.218248+00:00"} {"global_step": 12666, "acc_step": 0, "speed/wps": 12898.467430723436, "speed/FLOPS": 202588155351140.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04372364655137062, "optim/lr": 0.002858514830991166, "optim/total_tokens": 6640631808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8520078659057617, "created_at": "2025-01-16T17:20:19.386584+00:00"} {"global_step": 12667, "acc_step": 0, "speed/wps": 12899.208822185365, "speed/FLOPS": 202599799922828.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04747200384736061, "optim/lr": 0.002858478503188139, "optim/total_tokens": 6641156096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.923219919204712, "created_at": "2025-01-16T17:20:29.553702+00:00"} {"global_step": 12668, "acc_step": 0, "speed/wps": 12905.113896446122, "speed/FLOPS": 202692547228516.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05105987563729286, "optim/lr": 0.0028584421709528375, "optim/total_tokens": 6641680384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371737, "loss/out": 2.843235731124878, "created_at": "2025-01-16T17:20:39.714565+00:00"} {"global_step": 12669, "acc_step": 0, "speed/wps": 12902.182848375873, "speed/FLOPS": 202646511090891.1, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04885576292872429, "optim/lr": 0.00285840583428538, "optim/total_tokens": 6642204672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.7200703620910645, "created_at": "2025-01-16T17:20:49.877410+00:00"} {"global_step": 12670, "acc_step": 0, "speed/wps": 12901.623143611338, "speed/FLOPS": 202637720158449.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05187666416168213, "optim/lr": 0.002858369493185885, "optim/total_tokens": 6642728960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419583, "loss/out": 2.8846487998962402, "created_at": "2025-01-16T17:21:00.041957+00:00"} {"global_step": 12671, "acc_step": 0, "speed/wps": 12901.588306931102, "speed/FLOPS": 202637173000515.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052396245300769806, "optim/lr": 0.002858333147654471, "optim/total_tokens": 6643253248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.7057313919067383, "created_at": "2025-01-16T17:21:10.206448+00:00"} {"global_step": 12672, "acc_step": 0, "speed/wps": 12900.94508330046, "speed/FLOPS": 202627070289514.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04883045703172684, "optim/lr": 0.0028582967976912566, "optim/total_tokens": 6643777536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.942305088043213, "created_at": "2025-01-16T17:21:20.369969+00:00"} {"global_step": 12673, "acc_step": 0, "speed/wps": 12901.301305025025, "speed/FLOPS": 202632665241198.5, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04032348096370697, "optim/lr": 0.00285826044329636, "optim/total_tokens": 6644301824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372699, "loss/out": 2.851729393005371, "created_at": "2025-01-16T17:21:30.534492+00:00"} {"global_step": 12674, "acc_step": 0, "speed/wps": 12900.550286205264, "speed/FLOPS": 202620869458626.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044549617916345596, "optim/lr": 0.002858224084469901, "optim/total_tokens": 6644826112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.870814561843872, "created_at": "2025-01-16T17:21:40.700135+00:00"} {"global_step": 12675, "acc_step": 0, "speed/wps": 12908.829104757799, "speed/FLOPS": 202750899680282.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04056728258728981, "optim/lr": 0.002858187721211997, "optim/total_tokens": 6645350400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8015103340148926, "created_at": "2025-01-16T17:21:50.857766+00:00"} {"global_step": 12676, "acc_step": 0, "speed/wps": 12906.633269489277, "speed/FLOPS": 202716411069996.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045647792518138885, "optim/lr": 0.002858151353522768, "optim/total_tokens": 6645874688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9074363708496094, "created_at": "2025-01-16T17:22:01.019062+00:00"} {"global_step": 12677, "acc_step": 0, "speed/wps": 12898.358612731025, "speed/FLOPS": 202586446215038.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06029350683093071, "optim/lr": 0.0028581149814023315, "optim/total_tokens": 6646398976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.762594699859619, "created_at": "2025-01-16T17:22:11.185596+00:00"} {"global_step": 12678, "acc_step": 0, "speed/wps": 12899.597865487978, "speed/FLOPS": 202605910382497.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05388252064585686, "optim/lr": 0.002858078604850806, "optim/total_tokens": 6646923264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 2.9472572803497314, "created_at": "2025-01-16T17:22:21.350991+00:00"} {"global_step": 12679, "acc_step": 0, "speed/wps": 12907.858042009584, "speed/FLOPS": 202735647805441.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06272365897893906, "optim/lr": 0.002858042223868311, "optim/total_tokens": 6647447552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8368983268737793, "created_at": "2025-01-16T17:22:31.510585+00:00"} {"global_step": 12680, "acc_step": 0, "speed/wps": 12909.510492669193, "speed/FLOPS": 202761601813756.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06565713882446289, "optim/lr": 0.002858005838454965, "optim/total_tokens": 6647971840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.721742868423462, "created_at": "2025-01-16T17:22:41.667617+00:00"} {"global_step": 12681, "acc_step": 0, "speed/wps": 12898.381731362007, "speed/FLOPS": 202586809324906.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08245272189378738, "optim/lr": 0.002857969448610886, "optim/total_tokens": 6648496128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 2.9614768028259277, "created_at": "2025-01-16T17:22:51.835694+00:00"} {"global_step": 12682, "acc_step": 0, "speed/wps": 12903.816027295215, "speed/FLOPS": 202672162410042.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07641682773828506, "optim/lr": 0.0028579330543361935, "optim/total_tokens": 6649020416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9210901260375977, "created_at": "2025-01-16T17:23:02.000272+00:00"} {"global_step": 12683, "acc_step": 0, "speed/wps": 12907.006410319796, "speed/FLOPS": 202722271759489.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04960501939058304, "optim/lr": 0.0028578966556310065, "optim/total_tokens": 6649544704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.7170395851135254, "created_at": "2025-01-16T17:23:12.161443+00:00"} {"global_step": 12684, "acc_step": 0, "speed/wps": 12906.397405833706, "speed/FLOPS": 202712706507175.2, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055322643369436264, "optim/lr": 0.0028578602524954432, "optim/total_tokens": 6650068992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.806067943572998, "created_at": "2025-01-16T17:23:22.321056+00:00"} {"global_step": 12685, "acc_step": 0, "speed/wps": 12905.050398099254, "speed/FLOPS": 202691549899723.03, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05948150157928467, "optim/lr": 0.0028578238449296217, "optim/total_tokens": 6650593280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447274, "loss/out": 2.871249198913574, "created_at": "2025-01-16T17:23:32.483220+00:00"} {"global_step": 12686, "acc_step": 0, "speed/wps": 12906.180319466212, "speed/FLOPS": 202709296867465.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059094928205013275, "optim/lr": 0.0028577874329336625, "optim/total_tokens": 6651117568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.7858686447143555, "created_at": "2025-01-16T17:23:42.642595+00:00"} {"global_step": 12687, "acc_step": 0, "speed/wps": 12894.429454742725, "speed/FLOPS": 202524733389603.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051398348063230515, "optim/lr": 0.002857751016507683, "optim/total_tokens": 6651641856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 2.844099998474121, "created_at": "2025-01-16T17:23:52.811647+00:00"} {"global_step": 12688, "acc_step": 0, "speed/wps": 12905.393364050025, "speed/FLOPS": 202696936651264.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05586123839020729, "optim/lr": 0.0028577145956518025, "optim/total_tokens": 6652166144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394694, "loss/out": 2.815317153930664, "created_at": "2025-01-16T17:24:02.977871+00:00"} {"global_step": 12689, "acc_step": 0, "speed/wps": 12904.82456749325, "speed/FLOPS": 202688002919731.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06060120463371277, "optim/lr": 0.00285767817036614, "optim/total_tokens": 6652690432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.831317186355591, "created_at": "2025-01-16T17:24:13.138769+00:00"} {"global_step": 12690, "acc_step": 0, "speed/wps": 12899.366850942657, "speed/FLOPS": 202602281981616.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07345954328775406, "optim/lr": 0.002857641740650814, "optim/total_tokens": 6653214720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.8357858657836914, "created_at": "2025-01-16T17:24:23.307322+00:00"} {"global_step": 12691, "acc_step": 0, "speed/wps": 12901.021962584548, "speed/FLOPS": 202628277784314.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04856273904442787, "optim/lr": 0.0028576053065059435, "optim/total_tokens": 6653739008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331536, "loss/out": 2.8682661056518555, "created_at": "2025-01-16T17:24:33.471357+00:00"} {"global_step": 12692, "acc_step": 0, "speed/wps": 12905.666251920939, "speed/FLOPS": 202701222730264.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06942014396190643, "optim/lr": 0.002857568867931647, "optim/total_tokens": 6654263296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.93383526802063, "created_at": "2025-01-16T17:24:43.635970+00:00"} {"global_step": 12693, "acc_step": 0, "speed/wps": 12901.50457784903, "speed/FLOPS": 202635857920225.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0566977821290493, "optim/lr": 0.0028575324249280444, "optim/total_tokens": 6654787584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9436206817626953, "created_at": "2025-01-16T17:24:53.802393+00:00"} {"global_step": 12694, "acc_step": 0, "speed/wps": 12892.09023865008, "speed/FLOPS": 202487992786448.4, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06932664662599564, "optim/lr": 0.002857495977495254, "optim/total_tokens": 6655311872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 3.028796672821045, "created_at": "2025-01-16T17:25:03.974091+00:00"} {"global_step": 12695, "acc_step": 0, "speed/wps": 12896.949089592603, "speed/FLOPS": 202564307717261.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0631793811917305, "optim/lr": 0.0028574595256333946, "optim/total_tokens": 6655836160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9376227855682373, "created_at": "2025-01-16T17:25:14.141190+00:00"} {"global_step": 12696, "acc_step": 0, "speed/wps": 12910.327213298671, "speed/FLOPS": 202774429533533.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054336633533239365, "optim/lr": 0.002857423069342584, "optim/total_tokens": 6656360448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.864006280899048, "created_at": "2025-01-16T17:25:24.299417+00:00"} {"global_step": 12697, "acc_step": 0, "speed/wps": 12896.191023847394, "speed/FLOPS": 202552401253041.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06534206122159958, "optim/lr": 0.002857386608622944, "optim/total_tokens": 6656884736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.797290325164795, "created_at": "2025-01-16T17:25:34.467594+00:00"} {"global_step": 12698, "acc_step": 0, "speed/wps": 12898.759139334848, "speed/FLOPS": 202592737035732.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04099712148308754, "optim/lr": 0.0028573501434745914, "optim/total_tokens": 6657409024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8774399757385254, "created_at": "2025-01-16T17:25:44.632755+00:00"} {"global_step": 12699, "acc_step": 0, "speed/wps": 12909.656695241965, "speed/FLOPS": 202763898126064.9, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05726942420005798, "optim/lr": 0.0028573136738976454, "optim/total_tokens": 6657933312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.7191548347473145, "created_at": "2025-01-16T17:25:54.789481+00:00"} {"global_step": 12700, "acc_step": 0, "speed/wps": 12905.587064007399, "speed/FLOPS": 202699978975266.7, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05368127301335335, "optim/lr": 0.0028572771998922257, "optim/total_tokens": 6658457600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.7941720485687256, "created_at": "2025-01-16T17:26:04.949668+00:00"} {"global_step": 12701, "acc_step": 0, "speed/wps": 12901.340938205198, "speed/FLOPS": 202633287734754.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04501508176326752, "optim/lr": 0.0028572407214584507, "optim/total_tokens": 6658981888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9052774906158447, "created_at": "2025-01-16T17:26:15.114942+00:00"} {"global_step": 12702, "acc_step": 0, "speed/wps": 12903.456498486476, "speed/FLOPS": 202666515516056.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061092883348464966, "optim/lr": 0.0028572042385964394, "optim/total_tokens": 6659506176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.824383497238159, "created_at": "2025-01-16T17:26:25.280692+00:00"} {"global_step": 12703, "acc_step": 0, "speed/wps": 12903.3845045084, "speed/FLOPS": 202665384751700.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04791777953505516, "optim/lr": 0.0028571677513063113, "optim/total_tokens": 6660030464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 2.9061038494110107, "created_at": "2025-01-16T17:26:35.442273+00:00"} {"global_step": 12704, "acc_step": 0, "speed/wps": 12905.705310726322, "speed/FLOPS": 202701836202475.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05420432984828949, "optim/lr": 0.002857131259588185, "optim/total_tokens": 6660554752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.8352043628692627, "created_at": "2025-01-16T17:26:45.606006+00:00"} {"global_step": 12705, "acc_step": 0, "speed/wps": 12898.708129362532, "speed/FLOPS": 202591935854022.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04111771285533905, "optim/lr": 0.0028570947634421803, "optim/total_tokens": 6661079040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.8836536407470703, "created_at": "2025-01-16T17:26:55.772452+00:00"} {"global_step": 12706, "acc_step": 0, "speed/wps": 12904.74935391179, "speed/FLOPS": 202686821587080.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0511552169919014, "optim/lr": 0.002857058262868415, "optim/total_tokens": 6661603328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.76185941696167, "created_at": "2025-01-16T17:27:05.936740+00:00"} {"global_step": 12707, "acc_step": 0, "speed/wps": 12908.777582939627, "speed/FLOPS": 202750090459329.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055843133479356766, "optim/lr": 0.002857021757867009, "optim/total_tokens": 6662127616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.798069477081299, "created_at": "2025-01-16T17:27:16.096521+00:00"} {"global_step": 12708, "acc_step": 0, "speed/wps": 12905.00630107787, "speed/FLOPS": 202690857295406.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05653214827179909, "optim/lr": 0.002856985248438082, "optim/total_tokens": 6662651904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.860574245452881, "created_at": "2025-01-16T17:27:26.258450+00:00"} {"global_step": 12709, "acc_step": 0, "speed/wps": 12903.027491077999, "speed/FLOPS": 202659777365188.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05148785188794136, "optim/lr": 0.0028569487345817515, "optim/total_tokens": 6663176192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282658, "loss/out": 2.8720688819885254, "created_at": "2025-01-16T17:27:36.423735+00:00"} {"global_step": 12710, "acc_step": 0, "speed/wps": 12903.647710253215, "speed/FLOPS": 202669518759605.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07758774608373642, "optim/lr": 0.0028569122162981383, "optim/total_tokens": 6663700480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8953168392181396, "created_at": "2025-01-16T17:27:46.588225+00:00"} {"global_step": 12711, "acc_step": 0, "speed/wps": 12897.212336607636, "speed/FLOPS": 202568442373372.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048738665878772736, "optim/lr": 0.0028568756935873608, "optim/total_tokens": 6664224768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482092, "loss/out": 2.8200366497039795, "created_at": "2025-01-16T17:27:56.754822+00:00"} {"global_step": 12712, "acc_step": 0, "speed/wps": 12900.28627738628, "speed/FLOPS": 202616722837337.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08401304483413696, "optim/lr": 0.0028568391664495372, "optim/total_tokens": 6664749056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.87445068359375, "created_at": "2025-01-16T17:28:06.919980+00:00"} {"global_step": 12713, "acc_step": 0, "speed/wps": 12896.250352172516, "speed/FLOPS": 202553333085913.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08683903515338898, "optim/lr": 0.0028568026348847887, "optim/total_tokens": 6665273344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.921552896499634, "created_at": "2025-01-16T17:28:17.087625+00:00"} {"global_step": 12714, "acc_step": 0, "speed/wps": 12906.00046655685, "speed/FLOPS": 202706472030380.44, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05016467347741127, "optim/lr": 0.0028567660988932327, "optim/total_tokens": 6665797632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9081764221191406, "created_at": "2025-01-16T17:28:27.248034+00:00"} {"global_step": 12715, "acc_step": 0, "speed/wps": 12906.392808782542, "speed/FLOPS": 202712634304169.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06779685616493225, "optim/lr": 0.0028567295584749895, "optim/total_tokens": 6666321920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.7888741493225098, "created_at": "2025-01-16T17:28:37.408363+00:00"} {"global_step": 12716, "acc_step": 0, "speed/wps": 12906.070007872324, "speed/FLOPS": 202707564272299.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05825603008270264, "optim/lr": 0.002856693013630178, "optim/total_tokens": 6666846208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 2.8869552612304688, "created_at": "2025-01-16T17:28:47.570181+00:00"} {"global_step": 12717, "acc_step": 0, "speed/wps": 12904.064080158552, "speed/FLOPS": 202676058421121.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06425795704126358, "optim/lr": 0.002856656464358917, "optim/total_tokens": 6667370496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.8200416564941406, "created_at": "2025-01-16T17:28:57.734081+00:00"} {"global_step": 12718, "acc_step": 0, "speed/wps": 12901.269995552531, "speed/FLOPS": 202632173482909.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058437660336494446, "optim/lr": 0.002856619910661327, "optim/total_tokens": 6667894784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.7735061645507812, "created_at": "2025-01-16T17:29:07.897908+00:00"} {"global_step": 12719, "acc_step": 0, "speed/wps": 12899.026391180223, "speed/FLOPS": 202596934593206.7, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.060918815433979034, "optim/lr": 0.002856583352537526, "optim/total_tokens": 6668419072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 2.8315296173095703, "created_at": "2025-01-16T17:29:18.063098+00:00"} {"global_step": 12720, "acc_step": 0, "speed/wps": 12906.15272748433, "speed/FLOPS": 202708863497474.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06067601218819618, "optim/lr": 0.002856546789987633, "optim/total_tokens": 6668943360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.841646671295166, "created_at": "2025-01-16T17:29:28.222393+00:00"} {"global_step": 12721, "acc_step": 0, "speed/wps": 12898.59218599651, "speed/FLOPS": 202590114804136.0, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057625990360975266, "optim/lr": 0.002856510223011769, "optim/total_tokens": 6669467648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.999659299850464, "created_at": "2025-01-16T17:29:38.390752+00:00"} {"global_step": 12722, "acc_step": 0, "speed/wps": 12882.99024025634, "speed/FLOPS": 202345064806966.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0765749141573906, "optim/lr": 0.002856473651610052, "optim/total_tokens": 6669991936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.822484016418457, "created_at": "2025-01-16T17:29:48.571501+00:00"} {"global_step": 12723, "acc_step": 0, "speed/wps": 12902.717100431157, "speed/FLOPS": 202654902253558.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04458821192383766, "optim/lr": 0.0028564370757826014, "optim/total_tokens": 6670516224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8672537803649902, "created_at": "2025-01-16T17:29:58.738114+00:00"} {"global_step": 12724, "acc_step": 0, "speed/wps": 12897.827747504056, "speed/FLOPS": 202578108247163.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056535281240940094, "optim/lr": 0.002856400495529537, "optim/total_tokens": 6671040512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.897667169570923, "created_at": "2025-01-16T17:30:08.903889+00:00"} {"global_step": 12725, "acc_step": 0, "speed/wps": 12906.281821290726, "speed/FLOPS": 202710891093097.88, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047297846525907516, "optim/lr": 0.0028563639108509778, "optim/total_tokens": 6671564800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 2.955932140350342, "created_at": "2025-01-16T17:30:19.068184+00:00"} {"global_step": 12726, "acc_step": 0, "speed/wps": 12908.056016405913, "speed/FLOPS": 202738757265381.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046241648495197296, "optim/lr": 0.002856327321747043, "optim/total_tokens": 6672089088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.830679416656494, "created_at": "2025-01-16T17:30:29.227662+00:00"} {"global_step": 12727, "acc_step": 0, "speed/wps": 12905.80692106038, "speed/FLOPS": 202703432132399.0, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04775131121277809, "optim/lr": 0.0028562907282178524, "optim/total_tokens": 6672613376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.849045753479004, "created_at": "2025-01-16T17:30:39.391771+00:00"} {"global_step": 12728, "acc_step": 0, "speed/wps": 12904.23911793817, "speed/FLOPS": 202678807629977.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05010060966014862, "optim/lr": 0.002856254130263525, "optim/total_tokens": 6673137664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481801, "loss/out": 2.7655773162841797, "created_at": "2025-01-16T17:30:49.555432+00:00"} {"global_step": 12729, "acc_step": 0, "speed/wps": 12904.161030144067, "speed/FLOPS": 202677581153867.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04779753461480141, "optim/lr": 0.0028562175278841807, "optim/total_tokens": 6673661952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.769503355026245, "created_at": "2025-01-16T17:30:59.718823+00:00"} {"global_step": 12730, "acc_step": 0, "speed/wps": 12905.946789666725, "speed/FLOPS": 202705628961061.25, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050842780619859695, "optim/lr": 0.002856180921079939, "optim/total_tokens": 6674186240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 2.8993582725524902, "created_at": "2025-01-16T17:31:09.883145+00:00"} {"global_step": 12731, "acc_step": 0, "speed/wps": 12901.552974888147, "speed/FLOPS": 202636618062229.47, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.03805350512266159, "optim/lr": 0.0028561443098509185, "optim/total_tokens": 6674710528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9541029930114746, "created_at": "2025-01-16T17:31:20.051223+00:00"} {"global_step": 12732, "acc_step": 0, "speed/wps": 12906.706230068772, "speed/FLOPS": 202717557016155.88, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04560038447380066, "optim/lr": 0.002856107694197239, "optim/total_tokens": 6675234816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.8591885566711426, "created_at": "2025-01-16T17:31:30.211276+00:00"} {"global_step": 12733, "acc_step": 0, "speed/wps": 12907.431759322339, "speed/FLOPS": 202728952450064.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057203926146030426, "optim/lr": 0.0028560710741190203, "optim/total_tokens": 6675759104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 2.8998656272888184, "created_at": "2025-01-16T17:31:40.369440+00:00"} {"global_step": 12734, "acc_step": 0, "speed/wps": 12904.075804094819, "speed/FLOPS": 202676242561650.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05433419719338417, "optim/lr": 0.002856034449616382, "optim/total_tokens": 6676283392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.929201126098633, "created_at": "2025-01-16T17:31:50.530517+00:00"} {"global_step": 12735, "acc_step": 0, "speed/wps": 12907.747714756508, "speed/FLOPS": 202733914964326.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04778534919023514, "optim/lr": 0.0028559978206894434, "optim/total_tokens": 6676807680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393176, "loss/out": 2.7925305366516113, "created_at": "2025-01-16T17:32:00.691320+00:00"} {"global_step": 12736, "acc_step": 0, "speed/wps": 12906.926544279742, "speed/FLOPS": 202721017353582.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05216824263334274, "optim/lr": 0.0028559611873383234, "optim/total_tokens": 6677331968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.9180145263671875, "created_at": "2025-01-16T17:32:10.854550+00:00"} {"global_step": 12737, "acc_step": 0, "speed/wps": 12896.455654488353, "speed/FLOPS": 202556557640898.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050032854080200195, "optim/lr": 0.002855924549563142, "optim/total_tokens": 6677856256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.9053590297698975, "created_at": "2025-01-16T17:32:21.021472+00:00"} {"global_step": 12738, "acc_step": 0, "speed/wps": 12907.254098155636, "speed/FLOPS": 202726162037309.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048670440912246704, "optim/lr": 0.0028558879073640187, "optim/total_tokens": 6678380544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.7646572589874268, "created_at": "2025-01-16T17:32:31.181967+00:00"} {"global_step": 12739, "acc_step": 0, "speed/wps": 12910.153346182375, "speed/FLOPS": 202771698711553.94, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06177496537566185, "optim/lr": 0.0028558512607410737, "optim/total_tokens": 6678904832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.9329042434692383, "created_at": "2025-01-16T17:32:41.340525+00:00"} {"global_step": 12740, "acc_step": 0, "speed/wps": 12909.09825178842, "speed/FLOPS": 202755127004245.25, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06502796709537506, "optim/lr": 0.002855814609694425, "optim/total_tokens": 6679429120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381895, "loss/out": 2.861905813217163, "created_at": "2025-01-16T17:32:51.498124+00:00"} {"global_step": 12741, "acc_step": 0, "speed/wps": 12908.3326456764, "speed/FLOPS": 202743102108202.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04700436815619469, "optim/lr": 0.002855777954224194, "optim/total_tokens": 6679953408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.7427420616149902, "created_at": "2025-01-16T17:33:01.657457+00:00"} {"global_step": 12742, "acc_step": 0, "speed/wps": 12908.009760748559, "speed/FLOPS": 202738030756720.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056590184569358826, "optim/lr": 0.0028557412943304994, "optim/total_tokens": 6680477696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377626, "loss/out": 2.8486721515655518, "created_at": "2025-01-16T17:33:11.816863+00:00"} {"global_step": 12743, "acc_step": 0, "speed/wps": 12900.960401373572, "speed/FLOPS": 202627310880901.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05691490322351456, "optim/lr": 0.0028557046300134604, "optim/total_tokens": 6681001984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.8072776794433594, "created_at": "2025-01-16T17:33:21.982732+00:00"} {"global_step": 12744, "acc_step": 0, "speed/wps": 12900.85694802692, "speed/FLOPS": 202625686003932.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048358190804719925, "optim/lr": 0.0028556679612731974, "optim/total_tokens": 6681526272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.928133964538574, "created_at": "2025-01-16T17:33:32.149624+00:00"} {"global_step": 12745, "acc_step": 0, "speed/wps": 12909.709525507797, "speed/FLOPS": 202764727897986.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05080718174576759, "optim/lr": 0.00285563128810983, "optim/total_tokens": 6682050560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.977904796600342, "created_at": "2025-01-16T17:33:42.306477+00:00"} {"global_step": 12746, "acc_step": 0, "speed/wps": 12901.726647094196, "speed/FLOPS": 202639345822876.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048915039747953415, "optim/lr": 0.0028555946105234768, "optim/total_tokens": 6682574848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.8823819160461426, "created_at": "2025-01-16T17:33:52.472094+00:00"} {"global_step": 12747, "acc_step": 0, "speed/wps": 12904.76468102032, "speed/FLOPS": 202687062320381.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06472475826740265, "optim/lr": 0.0028555579285142584, "optim/total_tokens": 6683099136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9417591094970703, "created_at": "2025-01-16T17:34:02.636626+00:00"} {"global_step": 12748, "acc_step": 0, "speed/wps": 12904.38321806268, "speed/FLOPS": 202681070920447.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060616374015808105, "optim/lr": 0.002855521242082295, "optim/total_tokens": 6683623424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 2.911465644836426, "created_at": "2025-01-16T17:34:12.802132+00:00"} {"global_step": 12749, "acc_step": 0, "speed/wps": 12905.738274913594, "speed/FLOPS": 202702353950334.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.061004459857940674, "optim/lr": 0.0028554845512277047, "optim/total_tokens": 6684147712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 2.9208872318267822, "created_at": "2025-01-16T17:34:22.963351+00:00"} {"global_step": 12750, "acc_step": 0, "speed/wps": 12904.890718966886, "speed/FLOPS": 202689041919527.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0642746165394783, "optim/lr": 0.0028554478559506087, "optim/total_tokens": 6684672000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.9806737899780273, "created_at": "2025-01-16T17:34:33.123664+00:00"} {"global_step": 12751, "acc_step": 0, "speed/wps": 12902.285460703333, "speed/FLOPS": 202648122758497.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045072901993989944, "optim/lr": 0.0028554111562511257, "optim/total_tokens": 6685196288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8355612754821777, "created_at": "2025-01-16T17:34:43.286289+00:00"} {"global_step": 12752, "acc_step": 0, "speed/wps": 12907.38298908008, "speed/FLOPS": 202728186446392.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06639391928911209, "optim/lr": 0.002855374452129377, "optim/total_tokens": 6685720576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 2.918829917907715, "created_at": "2025-01-16T17:34:53.452960+00:00"} {"global_step": 12753, "acc_step": 0, "speed/wps": 12899.015876384556, "speed/FLOPS": 202596769443891.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04965544864535332, "optim/lr": 0.00285533774358548, "optim/total_tokens": 6686244864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 2.8442063331604004, "created_at": "2025-01-16T17:35:03.617743+00:00"} {"global_step": 12754, "acc_step": 0, "speed/wps": 12907.477836820803, "speed/FLOPS": 202729676160495.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07529056072235107, "optim/lr": 0.002855301030619556, "optim/total_tokens": 6686769152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.91355562210083, "created_at": "2025-01-16T17:35:13.778662+00:00"} {"global_step": 12755, "acc_step": 0, "speed/wps": 12899.9473003026, "speed/FLOPS": 202611398736435.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06000976637005806, "optim/lr": 0.002855264313231725, "optim/total_tokens": 6687293440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 2.806820869445801, "created_at": "2025-01-16T17:35:23.947515+00:00"} {"global_step": 12756, "acc_step": 0, "speed/wps": 12898.997842615932, "speed/FLOPS": 202596486198774.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05307132750749588, "optim/lr": 0.002855227591422106, "optim/total_tokens": 6687817728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8203392028808594, "created_at": "2025-01-16T17:35:34.114435+00:00"} {"global_step": 12757, "acc_step": 0, "speed/wps": 12899.489906885161, "speed/FLOPS": 202604214744290.06, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06897509843111038, "optim/lr": 0.002855190865190819, "optim/total_tokens": 6688342016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.875316619873047, "created_at": "2025-01-16T17:35:44.285660+00:00"} {"global_step": 12758, "acc_step": 0, "speed/wps": 12907.936188760606, "speed/FLOPS": 202736875207551.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04659763351082802, "optim/lr": 0.0028551541345379836, "optim/total_tokens": 6688866304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318033, "loss/out": 2.771117687225342, "created_at": "2025-01-16T17:35:54.444518+00:00"} {"global_step": 12759, "acc_step": 0, "speed/wps": 12903.598153464534, "speed/FLOPS": 202668740402141.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06396392732858658, "optim/lr": 0.0028551173994637207, "optim/total_tokens": 6689390592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.7630221843719482, "created_at": "2025-01-16T17:36:04.605763+00:00"} {"global_step": 12760, "acc_step": 0, "speed/wps": 12905.941074525623, "speed/FLOPS": 202705539196916.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04455012083053589, "optim/lr": 0.0028550806599681486, "optim/total_tokens": 6689914880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.792689323425293, "created_at": "2025-01-16T17:36:14.765488+00:00"} {"global_step": 12761, "acc_step": 0, "speed/wps": 12905.07934976522, "speed/FLOPS": 202692004625420.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0749964565038681, "optim/lr": 0.0028550439160513886, "optim/total_tokens": 6690439168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8685100078582764, "created_at": "2025-01-16T17:36:24.928229+00:00"} {"global_step": 12762, "acc_step": 0, "speed/wps": 12905.898298115624, "speed/FLOPS": 202704867334616.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.037672266364097595, "optim/lr": 0.0028550071677135597, "optim/total_tokens": 6690963456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.944680690765381, "created_at": "2025-01-16T17:36:35.087791+00:00"} {"global_step": 12763, "acc_step": 0, "speed/wps": 12902.229511753967, "speed/FLOPS": 202647244003366.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06380993872880936, "optim/lr": 0.0028549704149547824, "optim/total_tokens": 6691487744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.8097519874572754, "created_at": "2025-01-16T17:36:45.251514+00:00"} {"global_step": 12764, "acc_step": 0, "speed/wps": 12903.161093457442, "speed/FLOPS": 202661875774146.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053249698132276535, "optim/lr": 0.002854933657775176, "optim/total_tokens": 6692012032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.798755407333374, "created_at": "2025-01-16T17:36:55.417952+00:00"} {"global_step": 12765, "acc_step": 0, "speed/wps": 12900.963839164297, "speed/FLOPS": 202627364876129.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06597673147916794, "optim/lr": 0.002854896896174861, "optim/total_tokens": 6692536320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 2.8622140884399414, "created_at": "2025-01-16T17:37:05.581516+00:00"} {"global_step": 12766, "acc_step": 0, "speed/wps": 12902.18797942661, "speed/FLOPS": 202646591681093.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042302101850509644, "optim/lr": 0.002854860130153957, "optim/total_tokens": 6693060608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.948582172393799, "created_at": "2025-01-16T17:37:15.744193+00:00"} {"global_step": 12767, "acc_step": 0, "speed/wps": 12907.415564090215, "speed/FLOPS": 202728698081689.72, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06479462236166, "optim/lr": 0.002854823359712584, "optim/total_tokens": 6693584896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8593831062316895, "created_at": "2025-01-16T17:37:25.904952+00:00"} {"global_step": 12768, "acc_step": 0, "speed/wps": 12908.500887239828, "speed/FLOPS": 202745744573144.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04903336241841316, "optim/lr": 0.0028547865848508615, "optim/total_tokens": 6694109184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.906912326812744, "created_at": "2025-01-16T17:37:36.065392+00:00"} {"global_step": 12769, "acc_step": 0, "speed/wps": 12909.415948980919, "speed/FLOPS": 202760116875215.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06404469162225723, "optim/lr": 0.0028547498055689104, "optim/total_tokens": 6694633472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.7958626747131348, "created_at": "2025-01-16T17:37:46.230261+00:00"} {"global_step": 12770, "acc_step": 0, "speed/wps": 12904.080679905042, "speed/FLOPS": 202676319142950.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049775131046772, "optim/lr": 0.00285471302186685, "optim/total_tokens": 6695157760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.7942252159118652, "created_at": "2025-01-16T17:37:56.393847+00:00"} {"global_step": 12771, "acc_step": 0, "speed/wps": 12903.435591697917, "speed/FLOPS": 202666187146213.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046657148748636246, "optim/lr": 0.002854676233744801, "optim/total_tokens": 6695682048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.7921440601348877, "created_at": "2025-01-16T17:38:06.555513+00:00"} {"global_step": 12772, "acc_step": 0, "speed/wps": 12905.99209201419, "speed/FLOPS": 202706340496680.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06360315531492233, "optim/lr": 0.0028546394412028823, "optim/total_tokens": 6696206336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.8553576469421387, "created_at": "2025-01-16T17:38:16.714970+00:00"} {"global_step": 12773, "acc_step": 0, "speed/wps": 12902.335417855886, "speed/FLOPS": 202648907404227.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043810222297906876, "optim/lr": 0.002854602644241215, "optim/total_tokens": 6696730624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.924104690551758, "created_at": "2025-01-16T17:38:26.879289+00:00"} {"global_step": 12774, "acc_step": 0, "speed/wps": 12908.715530917512, "speed/FLOPS": 202749115847054.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06739915162324905, "optim/lr": 0.0028545658428599193, "optim/total_tokens": 6697254912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8474502563476562, "created_at": "2025-01-16T17:38:37.038973+00:00"} {"global_step": 12775, "acc_step": 0, "speed/wps": 12905.805038145676, "speed/FLOPS": 202703402558636.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04821343347430229, "optim/lr": 0.0028545290370591146, "optim/total_tokens": 6697779200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9395055770874023, "created_at": "2025-01-16T17:38:47.201149+00:00"} {"global_step": 12776, "acc_step": 0, "speed/wps": 12905.759370235608, "speed/FLOPS": 202702685281354.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048346806317567825, "optim/lr": 0.002854492226838921, "optim/total_tokens": 6698303488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 2.8377156257629395, "created_at": "2025-01-16T17:38:57.364797+00:00"} {"global_step": 12777, "acc_step": 0, "speed/wps": 12902.247172507101, "speed/FLOPS": 202647521389762.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04531970992684364, "optim/lr": 0.0028544554121994583, "optim/total_tokens": 6698827776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.7965993881225586, "created_at": "2025-01-16T17:39:07.527579+00:00"} {"global_step": 12778, "acc_step": 0, "speed/wps": 12910.925721672182, "speed/FLOPS": 202783829929976.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05920805782079697, "optim/lr": 0.0028544185931408473, "optim/total_tokens": 6699352064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8137047290802, "created_at": "2025-01-16T17:39:17.683374+00:00"} {"global_step": 12779, "acc_step": 0, "speed/wps": 12908.34962579778, "speed/FLOPS": 202743368804342.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06487728655338287, "optim/lr": 0.0028543817696632086, "optim/total_tokens": 6699876352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 2.792768955230713, "created_at": "2025-01-16T17:39:27.841629+00:00"} {"global_step": 12780, "acc_step": 0, "speed/wps": 12901.575015702037, "speed/FLOPS": 202636964243498.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.058319807052612305, "optim/lr": 0.0028543449417666607, "optim/total_tokens": 6700400640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.9644763469696045, "created_at": "2025-01-16T17:39:38.004638+00:00"} {"global_step": 12781, "acc_step": 0, "speed/wps": 12905.538078118989, "speed/FLOPS": 202699209584575.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05212878808379173, "optim/lr": 0.0028543081094513255, "optim/total_tokens": 6700924928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 3.069077968597412, "created_at": "2025-01-16T17:39:48.165459+00:00"} {"global_step": 12782, "acc_step": 0, "speed/wps": 12910.18684352191, "speed/FLOPS": 202772224833301.5, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06455392390489578, "optim/lr": 0.0028542712727173216, "optim/total_tokens": 6701449216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.820274829864502, "created_at": "2025-01-16T17:39:58.322719+00:00"} {"global_step": 12783, "acc_step": 0, "speed/wps": 12905.785024370349, "speed/FLOPS": 202703088214792.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06338757276535034, "optim/lr": 0.0028542344315647707, "optim/total_tokens": 6701973504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.900738000869751, "created_at": "2025-01-16T17:40:08.482929+00:00"} {"global_step": 12784, "acc_step": 0, "speed/wps": 12902.863955935234, "speed/FLOPS": 202657208821050.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04304148629307747, "optim/lr": 0.0028541975859937914, "optim/total_tokens": 6702497792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 2.886058807373047, "created_at": "2025-01-16T17:40:18.648722+00:00"} {"global_step": 12785, "acc_step": 0, "speed/wps": 12908.75515519938, "speed/FLOPS": 202749738200850.25, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0622384138405323, "optim/lr": 0.0028541607360045057, "optim/total_tokens": 6703022080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.7817115783691406, "created_at": "2025-01-16T17:40:28.806635+00:00"} {"global_step": 12786, "acc_step": 0, "speed/wps": 12900.288379812322, "speed/FLOPS": 202616755858827.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04517773911356926, "optim/lr": 0.002854123881597033, "optim/total_tokens": 6703546368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.7886202335357666, "created_at": "2025-01-16T17:40:38.974121+00:00"} {"global_step": 12787, "acc_step": 0, "speed/wps": 12908.18067782857, "speed/FLOPS": 202740715244327.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.041570622473955154, "optim/lr": 0.0028540870227714925, "optim/total_tokens": 6704070656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.883997917175293, "created_at": "2025-01-16T17:40:49.132546+00:00"} {"global_step": 12788, "acc_step": 0, "speed/wps": 12906.717589509013, "speed/FLOPS": 202717735431774.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040667545050382614, "optim/lr": 0.002854050159528006, "optim/total_tokens": 6704594944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.6845459938049316, "created_at": "2025-01-16T17:40:59.298784+00:00"} {"global_step": 12789, "acc_step": 0, "speed/wps": 12903.372911961835, "speed/FLOPS": 202665202674827.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0481812059879303, "optim/lr": 0.002854013291866693, "optim/total_tokens": 6705119232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283142, "loss/out": 2.81166410446167, "created_at": "2025-01-16T17:41:09.462924+00:00"} {"global_step": 12790, "acc_step": 0, "speed/wps": 12897.044599543153, "speed/FLOPS": 202565807832279.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04031507670879364, "optim/lr": 0.0028539764197876736, "optim/total_tokens": 6705643520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.825916051864624, "created_at": "2025-01-16T17:41:19.630125+00:00"} {"global_step": 12791, "acc_step": 0, "speed/wps": 12904.940160252465, "speed/FLOPS": 202689818462856.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0538569875061512, "optim/lr": 0.002853939543291069, "optim/total_tokens": 6706167808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8481619358062744, "created_at": "2025-01-16T17:41:29.790404+00:00"} {"global_step": 12792, "acc_step": 0, "speed/wps": 12903.049510904852, "speed/FLOPS": 202660123216827.8, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046246420592069626, "optim/lr": 0.0028539026623769993, "optim/total_tokens": 6706692096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 2.9427037239074707, "created_at": "2025-01-16T17:41:39.957008+00:00"} {"global_step": 12793, "acc_step": 0, "speed/wps": 12906.847662857614, "speed/FLOPS": 202719778412453.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051488496363162994, "optim/lr": 0.002853865777045584, "optim/total_tokens": 6707216384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 2.8231711387634277, "created_at": "2025-01-16T17:41:50.118227+00:00"} {"global_step": 12794, "acc_step": 0, "speed/wps": 12905.905388442197, "speed/FLOPS": 202704978697938.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053479306399822235, "optim/lr": 0.002853828887296944, "optim/total_tokens": 6707740672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9721553325653076, "created_at": "2025-01-16T17:42:00.278551+00:00"} {"global_step": 12795, "acc_step": 0, "speed/wps": 12907.433518415124, "speed/FLOPS": 202728980079034.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1244778111577034, "optim/lr": 0.0028537919931311997, "optim/total_tokens": 6708264960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.8876712322235107, "created_at": "2025-01-16T17:42:10.438607+00:00"} {"global_step": 12796, "acc_step": 0, "speed/wps": 12899.061063881778, "speed/FLOPS": 202597479175629.88, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06888101249933243, "optim/lr": 0.0028537550945484714, "optim/total_tokens": 6708789248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.887526512145996, "created_at": "2025-01-16T17:42:20.605193+00:00"} {"global_step": 12797, "acc_step": 0, "speed/wps": 12903.947227670506, "speed/FLOPS": 202674223092224.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05099503695964813, "optim/lr": 0.00285371819154888, "optim/total_tokens": 6709313536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 2.7075910568237305, "created_at": "2025-01-16T17:42:30.766495+00:00"} {"global_step": 12798, "acc_step": 0, "speed/wps": 12905.873240010033, "speed/FLOPS": 202704473762635.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06034066528081894, "optim/lr": 0.002853681284132544, "optim/total_tokens": 6709837824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.916857957839966, "created_at": "2025-01-16T17:42:40.926456+00:00"} {"global_step": 12799, "acc_step": 0, "speed/wps": 12903.916617599803, "speed/FLOPS": 202673742319001.5, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0482451505959034, "optim/lr": 0.002853644372299587, "optim/total_tokens": 6710362112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.822080135345459, "created_at": "2025-01-16T17:42:51.087736+00:00"} {"global_step": 12800, "acc_step": 0, "speed/wps": 12906.718686748909, "speed/FLOPS": 202717752665434.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04946131259202957, "optim/lr": 0.0028536074560501267, "optim/total_tokens": 6710886400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.825608968734741, "created_at": "2025-01-16T17:43:01.246596+00:00"} {"global_step": 12801, "acc_step": 0, "speed/wps": 12909.17479844908, "speed/FLOPS": 202756329274737.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07518832385540009, "optim/lr": 0.0028535705353842844, "optim/total_tokens": 6711410688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8479371070861816, "created_at": "2025-01-16T17:43:11.406071+00:00"} {"global_step": 12802, "acc_step": 0, "speed/wps": 12908.5451478566, "speed/FLOPS": 202746439746951.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05462510138750076, "optim/lr": 0.0028535336103021804, "optim/total_tokens": 6711934976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294366, "loss/out": 2.8033151626586914, "created_at": "2025-01-16T17:43:21.565863+00:00"} {"global_step": 12803, "acc_step": 0, "speed/wps": 12905.546160948392, "speed/FLOPS": 202699336536518.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08509206026792526, "optim/lr": 0.002853496680803936, "optim/total_tokens": 6712459264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9443840980529785, "created_at": "2025-01-16T17:43:31.736286+00:00"} {"global_step": 12804, "acc_step": 0, "speed/wps": 12910.95595756361, "speed/FLOPS": 202784304826199.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07519233226776123, "optim/lr": 0.002853459746889671, "optim/total_tokens": 6712983552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.567068099975586, "created_at": "2025-01-16T17:43:41.894703+00:00"} {"global_step": 12805, "acc_step": 0, "speed/wps": 12901.955649114761, "speed/FLOPS": 202642942614290.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0583922378718853, "optim/lr": 0.002853422808559506, "optim/total_tokens": 6713507840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309920, "loss/out": 2.770589828491211, "created_at": "2025-01-16T17:43:52.059021+00:00"} {"global_step": 12806, "acc_step": 0, "speed/wps": 12902.49933704622, "speed/FLOPS": 202651481980359.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06029188260436058, "optim/lr": 0.002853385865813561, "optim/total_tokens": 6714032128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.7669999599456787, "created_at": "2025-01-16T17:44:02.221205+00:00"} {"global_step": 12807, "acc_step": 0, "speed/wps": 12901.034013305294, "speed/FLOPS": 202628467057442.8, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06464815139770508, "optim/lr": 0.002853348918651958, "optim/total_tokens": 6714556416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.926414966583252, "created_at": "2025-01-16T17:44:12.384591+00:00"} {"global_step": 12808, "acc_step": 0, "speed/wps": 12899.16769353793, "speed/FLOPS": 202599153940904.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052610982209444046, "optim/lr": 0.0028533119670748155, "optim/total_tokens": 6715080704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 2.9682259559631348, "created_at": "2025-01-16T17:44:22.553210+00:00"} {"global_step": 12809, "acc_step": 0, "speed/wps": 12904.42812606866, "speed/FLOPS": 202681776262391.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046814121305942535, "optim/lr": 0.0028532750110822564, "optim/total_tokens": 6715604992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.8141748905181885, "created_at": "2025-01-16T17:44:32.716710+00:00"} {"global_step": 12810, "acc_step": 0, "speed/wps": 12901.545458841174, "speed/FLOPS": 202636500012383.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05337992310523987, "optim/lr": 0.0028532380506743993, "optim/total_tokens": 6716129280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.765723943710327, "created_at": "2025-01-16T17:44:42.879671+00:00"} {"global_step": 12811, "acc_step": 0, "speed/wps": 12905.320625405988, "speed/FLOPS": 202695794190906.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05870996415615082, "optim/lr": 0.002853201085851365, "optim/total_tokens": 6716653568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.778419256210327, "created_at": "2025-01-16T17:44:53.039798+00:00"} {"global_step": 12812, "acc_step": 0, "speed/wps": 12908.565701802649, "speed/FLOPS": 202746762574918.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055598802864551544, "optim/lr": 0.0028531641166132755, "optim/total_tokens": 6717177856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.7812447547912598, "created_at": "2025-01-16T17:45:03.197670+00:00"} {"global_step": 12813, "acc_step": 0, "speed/wps": 12901.479108670117, "speed/FLOPS": 202635457891772.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04936310648918152, "optim/lr": 0.0028531271429602504, "optim/total_tokens": 6717702144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409693, "loss/out": 3.009923219680786, "created_at": "2025-01-16T17:45:13.360868+00:00"} {"global_step": 12814, "acc_step": 0, "speed/wps": 12902.980865697193, "speed/FLOPS": 202659045049513.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052294667810201645, "optim/lr": 0.0028530901648924105, "optim/total_tokens": 6718226432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8189547061920166, "created_at": "2025-01-16T17:45:23.523570+00:00"} {"global_step": 12815, "acc_step": 0, "speed/wps": 12905.468504666398, "speed/FLOPS": 202698116837898.66, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05906684696674347, "optim/lr": 0.002853053182409876, "optim/total_tokens": 6718750720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.8353686332702637, "created_at": "2025-01-16T17:45:33.683762+00:00"} {"global_step": 12816, "acc_step": 0, "speed/wps": 12904.00661554987, "speed/FLOPS": 202675155860477.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0636797845363617, "optim/lr": 0.0028530161955127686, "optim/total_tokens": 6719275008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.9963583946228027, "created_at": "2025-01-16T17:45:43.845140+00:00"} {"global_step": 12817, "acc_step": 0, "speed/wps": 12902.557070940144, "speed/FLOPS": 202652388770498.66, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061070654541254044, "optim/lr": 0.002852979204201208, "optim/total_tokens": 6719799296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9181690216064453, "created_at": "2025-01-16T17:45:54.011155+00:00"} {"global_step": 12818, "acc_step": 0, "speed/wps": 12903.81818174585, "speed/FLOPS": 202672196248650.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05781121179461479, "optim/lr": 0.0028529422084753154, "optim/total_tokens": 6720323584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 2.7707629203796387, "created_at": "2025-01-16T17:46:04.172212+00:00"} {"global_step": 12819, "acc_step": 0, "speed/wps": 12904.956238040933, "speed/FLOPS": 202690070986617.62, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06443105638027191, "optim/lr": 0.0028529052083352117, "optim/total_tokens": 6720847872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8774001598358154, "created_at": "2025-01-16T17:46:14.333887+00:00"} {"global_step": 12820, "acc_step": 0, "speed/wps": 12901.289220690709, "speed/FLOPS": 202632475440122.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04839614778757095, "optim/lr": 0.002852868203781017, "optim/total_tokens": 6721372160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481042, "loss/out": 2.7966067790985107, "created_at": "2025-01-16T17:46:24.498045+00:00"} {"global_step": 12821, "acc_step": 0, "speed/wps": 12911.317766432025, "speed/FLOPS": 202789987531655.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06420277059078217, "optim/lr": 0.002852831194812852, "optim/total_tokens": 6721896448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 2.8556911945343018, "created_at": "2025-01-16T17:46:34.662084+00:00"} {"global_step": 12822, "acc_step": 0, "speed/wps": 12913.119118632405, "speed/FLOPS": 202818280243280.06, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04987107217311859, "optim/lr": 0.0028527941814308383, "optim/total_tokens": 6722420736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.8114919662475586, "created_at": "2025-01-16T17:46:44.819944+00:00"} {"global_step": 12823, "acc_step": 0, "speed/wps": 12904.894728025472, "speed/FLOPS": 202689104887301.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053384825587272644, "optim/lr": 0.0028527571636350955, "optim/total_tokens": 6722945024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421973, "loss/out": 2.855245590209961, "created_at": "2025-01-16T17:46:54.983878+00:00"} {"global_step": 12824, "acc_step": 0, "speed/wps": 12911.907661035855, "speed/FLOPS": 202799252637009.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06499508023262024, "optim/lr": 0.0028527201414257457, "optim/total_tokens": 6723469312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.075833797454834, "created_at": "2025-01-16T17:47:05.138741+00:00"} {"global_step": 12825, "acc_step": 0, "speed/wps": 12908.944460230166, "speed/FLOPS": 202752711496491.12, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05023334175348282, "optim/lr": 0.002852683114802909, "optim/total_tokens": 6723993600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.7973928451538086, "created_at": "2025-01-16T17:47:15.297091+00:00"} {"global_step": 12826, "acc_step": 0, "speed/wps": 12909.659883069027, "speed/FLOPS": 202763948195269.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05842285975813866, "optim/lr": 0.0028526460837667057, "optim/total_tokens": 6724517888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 2.8520212173461914, "created_at": "2025-01-16T17:47:25.454831+00:00"} {"global_step": 12827, "acc_step": 0, "speed/wps": 12911.069091131138, "speed/FLOPS": 202786081744339.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06412608921527863, "optim/lr": 0.0028526090483172576, "optim/total_tokens": 6725042176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.7692835330963135, "created_at": "2025-01-16T17:47:35.614108+00:00"} {"global_step": 12828, "acc_step": 0, "speed/wps": 12910.962747866912, "speed/FLOPS": 202784411477244.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04838552698493004, "optim/lr": 0.0028525720084546847, "optim/total_tokens": 6725566464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 2.925337314605713, "created_at": "2025-01-16T17:47:45.769534+00:00"} {"global_step": 12829, "acc_step": 0, "speed/wps": 12909.014606364095, "speed/FLOPS": 202753813237915.38, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0576992928981781, "optim/lr": 0.0028525349641791084, "optim/total_tokens": 6726090752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 2.8505892753601074, "created_at": "2025-01-16T17:47:55.929940+00:00"} {"global_step": 12830, "acc_step": 0, "speed/wps": 12904.341341752966, "speed/FLOPS": 202680413195459.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045078083872795105, "optim/lr": 0.0028524979154906497, "optim/total_tokens": 6726615040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.832343339920044, "created_at": "2025-01-16T17:48:06.091439+00:00"} {"global_step": 12831, "acc_step": 0, "speed/wps": 12907.931859351384, "speed/FLOPS": 202736807208230.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07334180176258087, "optim/lr": 0.0028524608623894283, "optim/total_tokens": 6727139328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.771477699279785, "created_at": "2025-01-16T17:48:16.249783+00:00"} {"global_step": 12832, "acc_step": 0, "speed/wps": 12906.59004344854, "speed/FLOPS": 202715732145627.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050824567675590515, "optim/lr": 0.0028524238048755666, "optim/total_tokens": 6727663616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8544843196868896, "created_at": "2025-01-16T17:48:26.411070+00:00"} {"global_step": 12833, "acc_step": 0, "speed/wps": 12908.336571629932, "speed/FLOPS": 202743163770697.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07254102826118469, "optim/lr": 0.002852386742949185, "optim/total_tokens": 6728187904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 2.9102225303649902, "created_at": "2025-01-16T17:48:36.570604+00:00"} {"global_step": 12834, "acc_step": 0, "speed/wps": 12908.081326870999, "speed/FLOPS": 202739154801015.88, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055085983127355576, "optim/lr": 0.002852349676610404, "optim/total_tokens": 6728712192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.8573427200317383, "created_at": "2025-01-16T17:48:46.729185+00:00"} {"global_step": 12835, "acc_step": 0, "speed/wps": 12906.12820808397, "speed/FLOPS": 202708478386599.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0556953065097332, "optim/lr": 0.0028523126058593445, "optim/total_tokens": 6729236480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.8358588218688965, "created_at": "2025-01-16T17:48:56.888472+00:00"} {"global_step": 12836, "acc_step": 0, "speed/wps": 12908.086687554765, "speed/FLOPS": 202739238997920.78, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07141498476266861, "optim/lr": 0.0028522755306961277, "optim/total_tokens": 6729760768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.7550370693206787, "created_at": "2025-01-16T17:49:07.046412+00:00"} {"global_step": 12837, "acc_step": 0, "speed/wps": 12905.691241765666, "speed/FLOPS": 202701615230115.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.044958364218473434, "optim/lr": 0.0028522384511208755, "optim/total_tokens": 6730285056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.7744932174682617, "created_at": "2025-01-16T17:49:17.208663+00:00"} {"global_step": 12838, "acc_step": 0, "speed/wps": 12912.919955766843, "speed/FLOPS": 202815152116797.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054834917187690735, "optim/lr": 0.002852201367133707, "optim/total_tokens": 6730809344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381526, "loss/out": 2.9242827892303467, "created_at": "2025-01-16T17:49:27.363820+00:00"} {"global_step": 12839, "acc_step": 0, "speed/wps": 12897.95352350525, "speed/FLOPS": 202580083732096.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.045982055366039276, "optim/lr": 0.002852164278734745, "optim/total_tokens": 6731333632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.6362786293029785, "created_at": "2025-01-16T17:49:37.529664+00:00"} {"global_step": 12840, "acc_step": 0, "speed/wps": 12893.477622168719, "speed/FLOPS": 202509783551073.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059277281165122986, "optim/lr": 0.0028521271859241097, "optim/total_tokens": 6731857920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.8688178062438965, "created_at": "2025-01-16T17:49:47.699814+00:00"} {"global_step": 12841, "acc_step": 0, "speed/wps": 12904.377148158042, "speed/FLOPS": 202680975584253.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05430992692708969, "optim/lr": 0.002852090088701922, "optim/total_tokens": 6732382208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417722, "loss/out": 2.777001142501831, "created_at": "2025-01-16T17:49:57.865612+00:00"} {"global_step": 12842, "acc_step": 0, "speed/wps": 12904.068371478801, "speed/FLOPS": 202676125822202.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06174826622009277, "optim/lr": 0.0028520529870683024, "optim/total_tokens": 6732906496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.7391467094421387, "created_at": "2025-01-16T17:50:08.027587+00:00"} {"global_step": 12843, "acc_step": 0, "speed/wps": 12904.102458411273, "speed/FLOPS": 202676661204317.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05122523382306099, "optim/lr": 0.0028520158810233735, "optim/total_tokens": 6733430784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 3.0109615325927734, "created_at": "2025-01-16T17:50:18.188500+00:00"} {"global_step": 12844, "acc_step": 0, "speed/wps": 12910.105498047968, "speed/FLOPS": 202770947190852.84, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062032926827669144, "optim/lr": 0.0028519787705672548, "optim/total_tokens": 6733955072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 2.8370518684387207, "created_at": "2025-01-16T17:50:28.345471+00:00"} {"global_step": 12845, "acc_step": 0, "speed/wps": 12913.014259325228, "speed/FLOPS": 202816633283769.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055913399904966354, "optim/lr": 0.0028519416557000687, "optim/total_tokens": 6734479360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.7233190536499023, "created_at": "2025-01-16T17:50:38.500638+00:00"} {"global_step": 12846, "acc_step": 0, "speed/wps": 12902.794843080024, "speed/FLOPS": 202656123308688.62, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04542100429534912, "optim/lr": 0.0028519045364219352, "optim/total_tokens": 6735003648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401173, "loss/out": 2.8307065963745117, "created_at": "2025-01-16T17:50:48.662631+00:00"} {"global_step": 12847, "acc_step": 0, "speed/wps": 12901.649927357006, "speed/FLOPS": 202638140833979.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05908774584531784, "optim/lr": 0.002851867412732976, "optim/total_tokens": 6735527936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.764176845550537, "created_at": "2025-01-16T17:50:58.825409+00:00"} {"global_step": 12848, "acc_step": 0, "speed/wps": 12908.227395256881, "speed/FLOPS": 202741449005735.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045198697596788406, "optim/lr": 0.002851830284633312, "optim/total_tokens": 6736052224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.831672191619873, "created_at": "2025-01-16T17:51:08.984408+00:00"} {"global_step": 12849, "acc_step": 0, "speed/wps": 12905.847995010885, "speed/FLOPS": 202704077255233.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05814557150006294, "optim/lr": 0.002851793152123064, "optim/total_tokens": 6736576512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.849503993988037, "created_at": "2025-01-16T17:51:19.144044+00:00"} {"global_step": 12850, "acc_step": 0, "speed/wps": 12902.905874706517, "speed/FLOPS": 202657867212956.16, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04555182158946991, "optim/lr": 0.002851756015202354, "optim/total_tokens": 6737100800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.9260191917419434, "created_at": "2025-01-16T17:51:29.305987+00:00"} {"global_step": 12851, "acc_step": 0, "speed/wps": 12894.955033103804, "speed/FLOPS": 202532988319984.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051440950483083725, "optim/lr": 0.002851718873871303, "optim/total_tokens": 6737625088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447797, "loss/out": 2.8805055618286133, "created_at": "2025-01-16T17:51:39.476210+00:00"} {"global_step": 12852, "acc_step": 0, "speed/wps": 12892.070802364104, "speed/FLOPS": 202487687512868.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046886466443538666, "optim/lr": 0.002851681728130032, "optim/total_tokens": 6738149376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.762784481048584, "created_at": "2025-01-16T17:51:49.665911+00:00"} {"global_step": 12853, "acc_step": 0, "speed/wps": 12900.9381249598, "speed/FLOPS": 202626960999212.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0501415990293026, "optim/lr": 0.0028516445779786623, "optim/total_tokens": 6738673664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8580918312072754, "created_at": "2025-01-16T17:51:59.834371+00:00"} {"global_step": 12854, "acc_step": 0, "speed/wps": 12906.815151080744, "speed/FLOPS": 202719267770322.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04340926185250282, "optim/lr": 0.002851607423417314, "optim/total_tokens": 6739197952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.841221332550049, "created_at": "2025-01-16T17:52:09.994410+00:00"} {"global_step": 12855, "acc_step": 0, "speed/wps": 12906.330292052413, "speed/FLOPS": 202711652393016.0, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0438169501721859, "optim/lr": 0.00285157026444611, "optim/total_tokens": 6739722240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8633646965026855, "created_at": "2025-01-16T17:52:20.159521+00:00"} {"global_step": 12856, "acc_step": 0, "speed/wps": 12902.302025120374, "speed/FLOPS": 202648382925429.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05218885466456413, "optim/lr": 0.0028515331010651704, "optim/total_tokens": 6740246528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.7956113815307617, "created_at": "2025-01-16T17:52:30.324662+00:00"} {"global_step": 12857, "acc_step": 0, "speed/wps": 12900.745632371436, "speed/FLOPS": 202623937638599.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060193438082933426, "optim/lr": 0.0028514959332746168, "optim/total_tokens": 6740770816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.8572840690612793, "created_at": "2025-01-16T17:52:40.497455+00:00"} {"global_step": 12858, "acc_step": 0, "speed/wps": 12902.243821561477, "speed/FLOPS": 202647468758556.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05100686475634575, "optim/lr": 0.002851458761074571, "optim/total_tokens": 6741295104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 2.811415672302246, "created_at": "2025-01-16T17:52:50.662077+00:00"} {"global_step": 12859, "acc_step": 0, "speed/wps": 12909.189406666463, "speed/FLOPS": 202756558716864.84, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04416161775588989, "optim/lr": 0.002851421584465153, "optim/total_tokens": 6741819392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8419952392578125, "created_at": "2025-01-16T17:53:00.823728+00:00"} {"global_step": 12860, "acc_step": 0, "speed/wps": 12902.377201896657, "speed/FLOPS": 202649563680004.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07656322419643402, "optim/lr": 0.0028513844034464856, "optim/total_tokens": 6742343680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.893801689147949, "created_at": "2025-01-16T17:53:10.986914+00:00"} {"global_step": 12861, "acc_step": 0, "speed/wps": 12894.811496723209, "speed/FLOPS": 202530733883887.34, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0691523477435112, "optim/lr": 0.0028513472180186888, "optim/total_tokens": 6742867968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.9879605770111084, "created_at": "2025-01-16T17:53:21.155880+00:00"} {"global_step": 12862, "acc_step": 0, "speed/wps": 12902.405703913111, "speed/FLOPS": 202650011343339.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.062479812651872635, "optim/lr": 0.002851310028181885, "optim/total_tokens": 6743392256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.970424175262451, "created_at": "2025-01-16T17:53:31.321852+00:00"} {"global_step": 12863, "acc_step": 0, "speed/wps": 12897.465346858638, "speed/FLOPS": 202572416247036.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045257773250341415, "optim/lr": 0.002851272833936195, "optim/total_tokens": 6743916544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 2.8176321983337402, "created_at": "2025-01-16T17:53:41.491139+00:00"} {"global_step": 12864, "acc_step": 0, "speed/wps": 12908.027755794821, "speed/FLOPS": 202738313393650.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05835772678256035, "optim/lr": 0.00285123563528174, "optim/total_tokens": 6744440832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.7975220680236816, "created_at": "2025-01-16T17:53:51.651071+00:00"} {"global_step": 12865, "acc_step": 0, "speed/wps": 12906.536306024338, "speed/FLOPS": 202714888125537.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043471742421388626, "optim/lr": 0.0028511984322186416, "optim/total_tokens": 6744965120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.6591548919677734, "created_at": "2025-01-16T17:54:01.813013+00:00"} {"global_step": 12866, "acc_step": 0, "speed/wps": 12907.796361083734, "speed/FLOPS": 202734679021742.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058251988142728806, "optim/lr": 0.0028511612247470214, "optim/total_tokens": 6745489408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.792668342590332, "created_at": "2025-01-16T17:54:11.977756+00:00"} {"global_step": 12867, "acc_step": 0, "speed/wps": 12903.020199259778, "speed/FLOPS": 202659662837163.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05127720162272453, "optim/lr": 0.0028511240128670005, "optim/total_tokens": 6746013696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.848818778991699, "created_at": "2025-01-16T17:54:22.140117+00:00"} {"global_step": 12868, "acc_step": 0, "speed/wps": 12907.532758778183, "speed/FLOPS": 202730538785306.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051052071154117584, "optim/lr": 0.0028510867965787, "optim/total_tokens": 6746537984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.837534189224243, "created_at": "2025-01-16T17:54:32.298841+00:00"} {"global_step": 12869, "acc_step": 0, "speed/wps": 12905.667664964132, "speed/FLOPS": 202701244924049.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05111176520586014, "optim/lr": 0.002851049575882242, "optim/total_tokens": 6747062272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.8782010078430176, "created_at": "2025-01-16T17:54:42.462180+00:00"} {"global_step": 12870, "acc_step": 0, "speed/wps": 12904.286351112123, "speed/FLOPS": 202679549491879.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041681740432977676, "optim/lr": 0.0028510123507777476, "optim/total_tokens": 6747586560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.845423936843872, "created_at": "2025-01-16T17:54:52.625928+00:00"} {"global_step": 12871, "acc_step": 0, "speed/wps": 12907.750501504474, "speed/FLOPS": 202733958734032.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05157863721251488, "optim/lr": 0.002850975121265338, "optim/total_tokens": 6748110848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.749696731567383, "created_at": "2025-01-16T17:55:02.787713+00:00"} {"global_step": 12872, "acc_step": 0, "speed/wps": 12906.120503257123, "speed/FLOPS": 202708357371705.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04343102499842644, "optim/lr": 0.0028509378873451353, "optim/total_tokens": 6748635136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 2.8447391986846924, "created_at": "2025-01-16T17:55:12.947712+00:00"} {"global_step": 12873, "acc_step": 0, "speed/wps": 12905.412460653864, "speed/FLOPS": 202697236589669.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05008392781019211, "optim/lr": 0.0028509006490172605, "optim/total_tokens": 6749159424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.8149333000183105, "created_at": "2025-01-16T17:55:23.110452+00:00"} {"global_step": 12874, "acc_step": 0, "speed/wps": 12903.539922831329, "speed/FLOPS": 202667825810029.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04957348853349686, "optim/lr": 0.0028508634062818354, "optim/total_tokens": 6749683712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.9371542930603027, "created_at": "2025-01-16T17:55:33.274838+00:00"} {"global_step": 12875, "acc_step": 0, "speed/wps": 12908.077017490636, "speed/FLOPS": 202739087116275.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06716732680797577, "optim/lr": 0.002850826159138981, "optim/total_tokens": 6750208000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.973264694213867, "created_at": "2025-01-16T17:55:43.436156+00:00"} {"global_step": 12876, "acc_step": 0, "speed/wps": 12893.600955583166, "speed/FLOPS": 202511720671825.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046739622950553894, "optim/lr": 0.002850788907588819, "optim/total_tokens": 6750732288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.87357759475708, "created_at": "2025-01-16T17:55:53.608903+00:00"} {"global_step": 12877, "acc_step": 0, "speed/wps": 12906.013068630786, "speed/FLOPS": 202706669963268.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058340057730674744, "optim/lr": 0.0028507516516314718, "optim/total_tokens": 6751256576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420398, "loss/out": 2.8872077465057373, "created_at": "2025-01-16T17:56:03.769395+00:00"} {"global_step": 12878, "acc_step": 0, "speed/wps": 12906.379095765235, "speed/FLOPS": 202712418922389.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04560840129852295, "optim/lr": 0.0028507143912670594, "optim/total_tokens": 6751780864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.8621602058410645, "created_at": "2025-01-16T17:56:13.928644+00:00"} {"global_step": 12879, "acc_step": 0, "speed/wps": 12899.38567315588, "speed/FLOPS": 202602577610339.56, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06632915884256363, "optim/lr": 0.002850677126495705, "optim/total_tokens": 6752305152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.877647638320923, "created_at": "2025-01-16T17:56:24.093744+00:00"} {"global_step": 12880, "acc_step": 0, "speed/wps": 12906.358225664571, "speed/FLOPS": 202712091128779.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05496988445520401, "optim/lr": 0.0028506398573175293, "optim/total_tokens": 6752829440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381895, "loss/out": 2.8744239807128906, "created_at": "2025-01-16T17:56:34.253197+00:00"} {"global_step": 12881, "acc_step": 0, "speed/wps": 12905.64650485219, "speed/FLOPS": 202700912575414.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05556255951523781, "optim/lr": 0.0028506025837326536, "optim/total_tokens": 6753353728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8681588172912598, "created_at": "2025-01-16T17:56:44.414135+00:00"} {"global_step": 12882, "acc_step": 0, "speed/wps": 12907.147578614562, "speed/FLOPS": 202724489001543.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07054334133863449, "optim/lr": 0.0028505653057412004, "optim/total_tokens": 6753878016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441962, "loss/out": 2.8392767906188965, "created_at": "2025-01-16T17:56:54.573517+00:00"} {"global_step": 12883, "acc_step": 0, "speed/wps": 12900.837025279541, "speed/FLOPS": 202625373089808.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040022145956754684, "optim/lr": 0.0028505280233432906, "optim/total_tokens": 6754402304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.861605405807495, "created_at": "2025-01-16T17:57:04.739883+00:00"} {"global_step": 12884, "acc_step": 0, "speed/wps": 12901.07927525663, "speed/FLOPS": 202629177958585.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05995049700140953, "optim/lr": 0.0028504907365390457, "optim/total_tokens": 6754926592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472483, "loss/out": 2.8857388496398926, "created_at": "2025-01-16T17:57:14.904049+00:00"} {"global_step": 12885, "acc_step": 0, "speed/wps": 12903.921373793282, "speed/FLOPS": 202673817021555.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04708244279026985, "optim/lr": 0.0028504534453285884, "optim/total_tokens": 6755450880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.8777668476104736, "created_at": "2025-01-16T17:57:25.067956+00:00"} {"global_step": 12886, "acc_step": 0, "speed/wps": 12901.241150856848, "speed/FLOPS": 202631720437326.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06197643652558327, "optim/lr": 0.0028504161497120394, "optim/total_tokens": 6755975168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 3.0031838417053223, "created_at": "2025-01-16T17:57:35.231222+00:00"} {"global_step": 12887, "acc_step": 0, "speed/wps": 12899.43938533899, "speed/FLOPS": 202603421233983.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0525585301220417, "optim/lr": 0.0028503788496895208, "optim/total_tokens": 6756499456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.9228157997131348, "created_at": "2025-01-16T17:57:45.400003+00:00"} {"global_step": 12888, "acc_step": 0, "speed/wps": 12891.963690088383, "speed/FLOPS": 202486005167391.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03963568061590195, "optim/lr": 0.002850341545261154, "optim/total_tokens": 6757023744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 2.825550079345703, "created_at": "2025-01-16T17:57:55.571212+00:00"} {"global_step": 12889, "acc_step": 0, "speed/wps": 12904.769885122118, "speed/FLOPS": 202687144057951.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04843422770500183, "optim/lr": 0.002850304236427061, "optim/total_tokens": 6757548032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.9427716732025146, "created_at": "2025-01-16T17:58:05.734972+00:00"} {"global_step": 12890, "acc_step": 0, "speed/wps": 12900.143681762078, "speed/FLOPS": 202614483177112.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04515056312084198, "optim/lr": 0.0028502669231873635, "optim/total_tokens": 6758072320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.8386428356170654, "created_at": "2025-01-16T17:58:15.901134+00:00"} {"global_step": 12891, "acc_step": 0, "speed/wps": 12893.165971031198, "speed/FLOPS": 202504888641705.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0519060455262661, "optim/lr": 0.002850229605542183, "optim/total_tokens": 6758596608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.745429039001465, "created_at": "2025-01-16T17:58:26.072562+00:00"} {"global_step": 12892, "acc_step": 0, "speed/wps": 12901.102327954077, "speed/FLOPS": 202629540032876.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06294234842061996, "optim/lr": 0.0028501922834916417, "optim/total_tokens": 6759120896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 2.847254753112793, "created_at": "2025-01-16T17:58:36.237114+00:00"} {"global_step": 12893, "acc_step": 0, "speed/wps": 12903.346233449514, "speed/FLOPS": 202664783652131.25, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04432426765561104, "optim/lr": 0.002850154957035861, "optim/total_tokens": 6759645184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.8463022708892822, "created_at": "2025-01-16T17:58:46.400205+00:00"} {"global_step": 12894, "acc_step": 0, "speed/wps": 12893.643931650462, "speed/FLOPS": 202512395670017.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04906267672777176, "optim/lr": 0.0028501176261749622, "optim/total_tokens": 6760169472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.962282657623291, "created_at": "2025-01-16T17:58:56.569955+00:00"} {"global_step": 12895, "acc_step": 0, "speed/wps": 12901.989772197501, "speed/FLOPS": 202643478564195.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04786059260368347, "optim/lr": 0.002850080290909068, "optim/total_tokens": 6760693760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.786975383758545, "created_at": "2025-01-16T17:59:06.742659+00:00"} {"global_step": 12896, "acc_step": 0, "speed/wps": 12903.59314669785, "speed/FLOPS": 202668661763990.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047634292393922806, "optim/lr": 0.0028500429512383005, "optim/total_tokens": 6761218048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8052988052368164, "created_at": "2025-01-16T17:59:16.904186+00:00"} {"global_step": 12897, "acc_step": 0, "speed/wps": 12903.832751813912, "speed/FLOPS": 202672425091590.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04221456125378609, "optim/lr": 0.00285000560716278, "optim/total_tokens": 6761742336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.9895126819610596, "created_at": "2025-01-16T17:59:27.070430+00:00"} {"global_step": 12898, "acc_step": 0, "speed/wps": 12901.687788593534, "speed/FLOPS": 202638735496726.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.04678163677453995, "optim/lr": 0.00284996825868263, "optim/total_tokens": 6762266624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 2.8329551219940186, "created_at": "2025-01-16T17:59:37.233901+00:00"} {"global_step": 12899, "acc_step": 0, "speed/wps": 12905.153931729505, "speed/FLOPS": 202693176037656.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045647092163562775, "optim/lr": 0.0028499309057979714, "optim/total_tokens": 6762790912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.830183267593384, "created_at": "2025-01-16T17:59:47.395296+00:00"} {"global_step": 12900, "acc_step": 0, "speed/wps": 12894.095341777003, "speed/FLOPS": 202519485686357.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06439739465713501, "optim/lr": 0.0028498935485089263, "optim/total_tokens": 6763315200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.7000575065612793, "created_at": "2025-01-16T17:59:57.565176+00:00"} {"global_step": 12901, "acc_step": 0, "speed/wps": 12895.685187699077, "speed/FLOPS": 202544456401240.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09481477737426758, "optim/lr": 0.002849856186815617, "optim/total_tokens": 6763839488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 2.8648157119750977, "created_at": "2025-01-16T18:00:07.733921+00:00"} {"global_step": 12902, "acc_step": 0, "speed/wps": 12890.20399812884, "speed/FLOPS": 202458366787096.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09445984661579132, "optim/lr": 0.002849818820718164, "optim/total_tokens": 6764363776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.846595048904419, "created_at": "2025-01-16T18:00:17.907087+00:00"} {"global_step": 12903, "acc_step": 0, "speed/wps": 12896.756664024248, "speed/FLOPS": 202561285409290.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06950834393501282, "optim/lr": 0.002849781450216691, "optim/total_tokens": 6764888064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.757187604904175, "created_at": "2025-01-16T18:00:28.076906+00:00"} {"global_step": 12904, "acc_step": 0, "speed/wps": 12890.86625790307, "speed/FLOPS": 202468768486890.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0625014528632164, "optim/lr": 0.002849744075311319, "optim/total_tokens": 6765412352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432114, "loss/out": 2.8860955238342285, "created_at": "2025-01-16T18:00:38.248799+00:00"} {"global_step": 12905, "acc_step": 0, "speed/wps": 12900.445680589886, "speed/FLOPS": 202619226483693.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07039079070091248, "optim/lr": 0.0028497066960021696, "optim/total_tokens": 6765936640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.838820457458496, "created_at": "2025-01-16T18:00:48.415708+00:00"} {"global_step": 12906, "acc_step": 0, "speed/wps": 12897.089946966424, "speed/FLOPS": 202566520075875.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053341709077358246, "optim/lr": 0.002849669312289366, "optim/total_tokens": 6766460928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.897700548171997, "created_at": "2025-01-16T18:00:58.582856+00:00"} {"global_step": 12907, "acc_step": 0, "speed/wps": 12905.248528971943, "speed/FLOPS": 202694661817338.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07154759019613266, "optim/lr": 0.002849631924173028, "optim/total_tokens": 6766985216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8650474548339844, "created_at": "2025-01-16T18:01:08.743348+00:00"} {"global_step": 12908, "acc_step": 0, "speed/wps": 12903.392406001969, "speed/FLOPS": 202665508855515.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04815002158284187, "optim/lr": 0.00284959453165328, "optim/total_tokens": 6767509504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8428211212158203, "created_at": "2025-01-16T18:01:18.907463+00:00"} {"global_step": 12909, "acc_step": 0, "speed/wps": 12906.271633162376, "speed/FLOPS": 202710731074542.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.068950355052948, "optim/lr": 0.002849557134730243, "optim/total_tokens": 6768033792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.900334596633911, "created_at": "2025-01-16T18:01:29.067494+00:00"} {"global_step": 12910, "acc_step": 0, "speed/wps": 12902.427045022314, "speed/FLOPS": 202650346534784.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048429496586322784, "optim/lr": 0.0028495197334040385, "optim/total_tokens": 6768558080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 2.920868158340454, "created_at": "2025-01-16T18:01:39.229893+00:00"} {"global_step": 12911, "acc_step": 0, "speed/wps": 12905.266311595453, "speed/FLOPS": 202694941117877.06, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05084647238254547, "optim/lr": 0.0028494823276747896, "optim/total_tokens": 6769082368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9465079307556152, "created_at": "2025-01-16T18:01:49.392429+00:00"} {"global_step": 12912, "acc_step": 0, "speed/wps": 12905.628286856141, "speed/FLOPS": 202700626436752.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05843575298786163, "optim/lr": 0.0028494449175426174, "optim/total_tokens": 6769606656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.89666485786438, "created_at": "2025-01-16T18:01:59.554522+00:00"} {"global_step": 12913, "acc_step": 0, "speed/wps": 12907.143619056242, "speed/FLOPS": 202724426811239.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04670136049389839, "optim/lr": 0.002849407503007644, "optim/total_tokens": 6770130944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.6221463680267334, "created_at": "2025-01-16T18:02:09.715726+00:00"} {"global_step": 12914, "acc_step": 0, "speed/wps": 12902.97194144798, "speed/FLOPS": 202658904881916.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.040535058826208115, "optim/lr": 0.0028493700840699922, "optim/total_tokens": 6770655232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.870547294616699, "created_at": "2025-01-16T18:02:19.880477+00:00"} {"global_step": 12915, "acc_step": 0, "speed/wps": 12903.32868444081, "speed/FLOPS": 202664508020834.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05590621381998062, "optim/lr": 0.0028493326607297834, "optim/total_tokens": 6771179520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373603, "loss/out": 2.726106643676758, "created_at": "2025-01-16T18:02:30.044069+00:00"} {"global_step": 12916, "acc_step": 0, "speed/wps": 12899.04695502062, "speed/FLOPS": 202597257576577.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04760853573679924, "optim/lr": 0.0028492952329871404, "optim/total_tokens": 6771703808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8715901374816895, "created_at": "2025-01-16T18:02:40.208968+00:00"} {"global_step": 12917, "acc_step": 0, "speed/wps": 12902.527189019625, "speed/FLOPS": 202651919433874.84, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.038120243698358536, "optim/lr": 0.0028492578008421847, "optim/total_tokens": 6772228096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8930187225341797, "created_at": "2025-01-16T18:02:50.371757+00:00"} {"global_step": 12918, "acc_step": 0, "speed/wps": 12902.60918897456, "speed/FLOPS": 202653207355846.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05740373581647873, "optim/lr": 0.002849220364295038, "optim/total_tokens": 6772752384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 2.67374849319458, "created_at": "2025-01-16T18:03:00.536680+00:00"} {"global_step": 12919, "acc_step": 0, "speed/wps": 12899.257304860166, "speed/FLOPS": 202600561409858.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0435279980301857, "optim/lr": 0.002849182923345824, "optim/total_tokens": 6773276672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8979172706604004, "created_at": "2025-01-16T18:03:10.702618+00:00"} {"global_step": 12920, "acc_step": 0, "speed/wps": 12899.280057840788, "speed/FLOPS": 202600918776685.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048755813390016556, "optim/lr": 0.0028491454779946634, "optim/total_tokens": 6773800960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.754279375076294, "created_at": "2025-01-16T18:03:20.867475+00:00"} {"global_step": 12921, "acc_step": 0, "speed/wps": 12905.970242036168, "speed/FLOPS": 202705997312750.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056865423917770386, "optim/lr": 0.0028491080282416793, "optim/total_tokens": 6774325248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.9549269676208496, "created_at": "2025-01-16T18:03:31.031118+00:00"} {"global_step": 12922, "acc_step": 0, "speed/wps": 12903.398692659102, "speed/FLOPS": 202665607596104.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0456097275018692, "optim/lr": 0.002849070574086993, "optim/total_tokens": 6774849536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.820308208465576, "created_at": "2025-01-16T18:03:41.193686+00:00"} {"global_step": 12923, "acc_step": 0, "speed/wps": 12901.660943189296, "speed/FLOPS": 202638313852763.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04272861406207085, "optim/lr": 0.0028490331155307277, "optim/total_tokens": 6775373824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.844606399536133, "created_at": "2025-01-16T18:03:51.360059+00:00"} {"global_step": 12924, "acc_step": 0, "speed/wps": 12905.534956232394, "speed/FLOPS": 202699160551056.62, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04799734801054001, "optim/lr": 0.0028489956525730047, "optim/total_tokens": 6775898112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.735696315765381, "created_at": "2025-01-16T18:04:01.524436+00:00"} {"global_step": 12925, "acc_step": 0, "speed/wps": 12896.470958356871, "speed/FLOPS": 202556798009183.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04769456014037132, "optim/lr": 0.0028489581852139466, "optim/total_tokens": 6776422400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.856368064880371, "created_at": "2025-01-16T18:04:11.692511+00:00"} {"global_step": 12926, "acc_step": 0, "speed/wps": 12896.351111720529, "speed/FLOPS": 202554915653072.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05673322454094887, "optim/lr": 0.0028489207134536755, "optim/total_tokens": 6776946688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506365, "loss/out": 2.784102439880371, "created_at": "2025-01-16T18:04:21.886374+00:00"} {"global_step": 12927, "acc_step": 0, "speed/wps": 12899.55591736762, "speed/FLOPS": 202605251529624.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056973423808813095, "optim/lr": 0.0028488832372923147, "optim/total_tokens": 6777470976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 2.677595376968384, "created_at": "2025-01-16T18:04:32.054600+00:00"} {"global_step": 12928, "acc_step": 0, "speed/wps": 12907.062541567198, "speed/FLOPS": 202723153377861.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.050551384687423706, "optim/lr": 0.0028488457567299843, "optim/total_tokens": 6777995264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.741865396499634, "created_at": "2025-01-16T18:04:42.213849+00:00"} {"global_step": 12929, "acc_step": 0, "speed/wps": 12906.279858938546, "speed/FLOPS": 202710860271660.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050176844000816345, "optim/lr": 0.0028488082717668087, "optim/total_tokens": 6778519552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.9473648071289062, "created_at": "2025-01-16T18:04:52.374015+00:00"} {"global_step": 12930, "acc_step": 0, "speed/wps": 12894.745534788639, "speed/FLOPS": 202529697861063.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04730083420872688, "optim/lr": 0.0028487707824029094, "optim/total_tokens": 6779043840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 2.858524799346924, "created_at": "2025-01-16T18:05:02.542281+00:00"} {"global_step": 12931, "acc_step": 0, "speed/wps": 12903.646371003444, "speed/FLOPS": 202669497724847.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05050106719136238, "optim/lr": 0.0028487332886384083, "optim/total_tokens": 6779568128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.6747171878814697, "created_at": "2025-01-16T18:05:12.706440+00:00"} {"global_step": 12932, "acc_step": 0, "speed/wps": 12899.364242793574, "speed/FLOPS": 202602241017051.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04651430621743202, "optim/lr": 0.0028486957904734285, "optim/total_tokens": 6780092416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.872699737548828, "created_at": "2025-01-16T18:05:22.871331+00:00"} {"global_step": 12933, "acc_step": 0, "speed/wps": 12898.868420417324, "speed/FLOPS": 202594453445300.78, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06028687208890915, "optim/lr": 0.0028486582879080915, "optim/total_tokens": 6780616704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.8466501235961914, "created_at": "2025-01-16T18:05:33.036515+00:00"} {"global_step": 12934, "acc_step": 0, "speed/wps": 12908.151790383708, "speed/FLOPS": 202740261527310.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05636608228087425, "optim/lr": 0.0028486207809425203, "optim/total_tokens": 6781140992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.804871082305908, "created_at": "2025-01-16T18:05:43.194350+00:00"} {"global_step": 12935, "acc_step": 0, "speed/wps": 12904.780883800364, "speed/FLOPS": 202687316807307.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06221218779683113, "optim/lr": 0.0028485832695768374, "optim/total_tokens": 6781665280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.882014274597168, "created_at": "2025-01-16T18:05:53.355205+00:00"} {"global_step": 12936, "acc_step": 0, "speed/wps": 12902.368026535769, "speed/FLOPS": 202649419568353.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05261910334229469, "optim/lr": 0.002848545753811165, "optim/total_tokens": 6782189568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 2.8347957134246826, "created_at": "2025-01-16T18:06:03.520842+00:00"} {"global_step": 12937, "acc_step": 0, "speed/wps": 12906.068186424247, "speed/FLOPS": 202707535663954.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04382726177573204, "optim/lr": 0.002848508233645625, "optim/total_tokens": 6782713856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9054317474365234, "created_at": "2025-01-16T18:06:13.684586+00:00"} {"global_step": 12938, "acc_step": 0, "speed/wps": 12907.575025150236, "speed/FLOPS": 202731202636759.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06322222948074341, "optim/lr": 0.0028484707090803405, "optim/total_tokens": 6783238144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.8043813705444336, "created_at": "2025-01-16T18:06:23.842821+00:00"} {"global_step": 12939, "acc_step": 0, "speed/wps": 12893.977714081615, "speed/FLOPS": 202517638181763.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05543658882379532, "optim/lr": 0.002848433180115433, "optim/total_tokens": 6783762432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 2.858819007873535, "created_at": "2025-01-16T18:06:34.013016+00:00"} {"global_step": 12940, "acc_step": 0, "speed/wps": 12896.999231165219, "speed/FLOPS": 202565095259561.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05287233740091324, "optim/lr": 0.0028483956467510265, "optim/total_tokens": 6784286720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.875262975692749, "created_at": "2025-01-16T18:06:44.182002+00:00"} {"global_step": 12941, "acc_step": 0, "speed/wps": 12898.18412807765, "speed/FLOPS": 202583705693789.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06096433848142624, "optim/lr": 0.0028483581089872416, "optim/total_tokens": 6784811008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468571, "loss/out": 2.8329672813415527, "created_at": "2025-01-16T18:06:54.347626+00:00"} {"global_step": 12942, "acc_step": 0, "speed/wps": 12894.49299983303, "speed/FLOPS": 202525731452566.5, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06095292419195175, "optim/lr": 0.0028483205668242027, "optim/total_tokens": 6785335296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 2.9311420917510986, "created_at": "2025-01-16T18:07:04.516528+00:00"} {"global_step": 12943, "acc_step": 0, "speed/wps": 12895.33627913792, "speed/FLOPS": 202538976312837.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07731320708990097, "optim/lr": 0.002848283020262031, "optim/total_tokens": 6785859584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8800950050354004, "created_at": "2025-01-16T18:07:14.686354+00:00"} {"global_step": 12944, "acc_step": 0, "speed/wps": 12898.385698549639, "speed/FLOPS": 202586871635039.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05102573335170746, "optim/lr": 0.0028482454693008487, "optim/total_tokens": 6786383872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.844050884246826, "created_at": "2025-01-16T18:07:24.853161+00:00"} {"global_step": 12945, "acc_step": 0, "speed/wps": 12898.010219074798, "speed/FLOPS": 202580974213924.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0445697121322155, "optim/lr": 0.0028482079139407796, "optim/total_tokens": 6786908160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 2.8861093521118164, "created_at": "2025-01-16T18:07:35.021710+00:00"} {"global_step": 12946, "acc_step": 0, "speed/wps": 12898.036338873904, "speed/FLOPS": 202581384461261.5, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044536277651786804, "optim/lr": 0.002848170354181945, "optim/total_tokens": 6787432448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8766520023345947, "created_at": "2025-01-16T18:07:45.193003+00:00"} {"global_step": 12947, "acc_step": 0, "speed/wps": 12904.234522424373, "speed/FLOPS": 202678735451118.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0499354749917984, "optim/lr": 0.0028481327900244683, "optim/total_tokens": 6787956736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.736345052719116, "created_at": "2025-01-16T18:07:55.354157+00:00"} {"global_step": 12948, "acc_step": 0, "speed/wps": 12906.569285109432, "speed/FLOPS": 202715406107387.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06046079099178314, "optim/lr": 0.002848095221468472, "optim/total_tokens": 6788481024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.918156147003174, "created_at": "2025-01-16T18:08:05.514356+00:00"} {"global_step": 12949, "acc_step": 0, "speed/wps": 12900.643810815849, "speed/FLOPS": 202622338391151.7, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04170987010002136, "optim/lr": 0.002848057648514078, "optim/total_tokens": 6789005312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.808638334274292, "created_at": "2025-01-16T18:08:15.678264+00:00"} {"global_step": 12950, "acc_step": 0, "speed/wps": 12892.65946982115, "speed/FLOPS": 202496933344195.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044209059327840805, "optim/lr": 0.0028480200711614103, "optim/total_tokens": 6789529600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9745736122131348, "created_at": "2025-01-16T18:08:25.848290+00:00"} {"global_step": 12951, "acc_step": 0, "speed/wps": 12897.181669838556, "speed/FLOPS": 202567960709623.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04272887110710144, "optim/lr": 0.002847982489410589, "optim/total_tokens": 6790053888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285883, "loss/out": 2.862030267715454, "created_at": "2025-01-16T18:08:36.018216+00:00"} {"global_step": 12952, "acc_step": 0, "speed/wps": 12897.800230230687, "speed/FLOPS": 202577676050572.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046223144978284836, "optim/lr": 0.0028479449032617395, "optim/total_tokens": 6790578176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8671634197235107, "created_at": "2025-01-16T18:08:46.187732+00:00"} {"global_step": 12953, "acc_step": 0, "speed/wps": 12903.429024623663, "speed/FLOPS": 202666084001288.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04197464510798454, "optim/lr": 0.0028479073127149826, "optim/total_tokens": 6791102464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.7866616249084473, "created_at": "2025-01-16T18:08:56.349201+00:00"} {"global_step": 12954, "acc_step": 0, "speed/wps": 12901.356770494049, "speed/FLOPS": 202633536402607.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04730575159192085, "optim/lr": 0.0028478697177704414, "optim/total_tokens": 6791626752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333121, "loss/out": 2.846088409423828, "created_at": "2025-01-16T18:09:06.520251+00:00"} {"global_step": 12955, "acc_step": 0, "speed/wps": 12896.058746277684, "speed/FLOPS": 202550323652040.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04278537258505821, "optim/lr": 0.0028478321184282392, "optim/total_tokens": 6792151040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.7745327949523926, "created_at": "2025-01-16T18:09:16.687806+00:00"} {"global_step": 12956, "acc_step": 0, "speed/wps": 12901.057808518824, "speed/FLOPS": 202628840793969.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05434607341885567, "optim/lr": 0.002847794514688498, "optim/total_tokens": 6792675328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.76015305519104, "created_at": "2025-01-16T18:09:26.851159+00:00"} {"global_step": 12957, "acc_step": 0, "speed/wps": 12900.575061125597, "speed/FLOPS": 202621258582794.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04195186495780945, "optim/lr": 0.0028477569065513404, "optim/total_tokens": 6793199616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315105, "loss/out": 2.9154701232910156, "created_at": "2025-01-16T18:09:37.016145+00:00"} {"global_step": 12958, "acc_step": 0, "speed/wps": 12903.07227016772, "speed/FLOPS": 202660480682325.66, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05116496607661247, "optim/lr": 0.0028477192940168893, "optim/total_tokens": 6793723904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.901479959487915, "created_at": "2025-01-16T18:09:47.184225+00:00"} {"global_step": 12959, "acc_step": 0, "speed/wps": 12900.229996949261, "speed/FLOPS": 202615838875737.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049244269728660583, "optim/lr": 0.0028476816770852675, "optim/total_tokens": 6794248192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.7800745964050293, "created_at": "2025-01-16T18:09:57.348985+00:00"} {"global_step": 12960, "acc_step": 0, "speed/wps": 12897.530977305798, "speed/FLOPS": 202573447063394.84, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0644591748714447, "optim/lr": 0.002847644055756598, "optim/total_tokens": 6794772480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 2.8463029861450195, "created_at": "2025-01-16T18:10:07.517619+00:00"} {"global_step": 12961, "acc_step": 0, "speed/wps": 12904.727670250993, "speed/FLOPS": 202686481015391.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09262308478355408, "optim/lr": 0.0028476064300310034, "optim/total_tokens": 6795296768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.778297185897827, "created_at": "2025-01-16T18:10:17.683598+00:00"} {"global_step": 12962, "acc_step": 0, "speed/wps": 12897.211569552514, "speed/FLOPS": 202568430325717.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07938190549612045, "optim/lr": 0.002847568799908606, "optim/total_tokens": 6795821056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.7692222595214844, "created_at": "2025-01-16T18:10:27.853952+00:00"} {"global_step": 12963, "acc_step": 0, "speed/wps": 12902.429533758414, "speed/FLOPS": 202650385623805.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0557270273566246, "optim/lr": 0.0028475311653895284, "optim/total_tokens": 6796345344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.9095029830932617, "created_at": "2025-01-16T18:10:38.019834+00:00"} {"global_step": 12964, "acc_step": 0, "speed/wps": 12901.742852890831, "speed/FLOPS": 202639600357181.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0564417839050293, "optim/lr": 0.0028474935264738944, "optim/total_tokens": 6796869632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.931384563446045, "created_at": "2025-01-16T18:10:48.182606+00:00"} {"global_step": 12965, "acc_step": 0, "speed/wps": 12903.360241695455, "speed/FLOPS": 202665003670883.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07029969990253448, "optim/lr": 0.0028474558831618263, "optim/total_tokens": 6797393920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.8555984497070312, "created_at": "2025-01-16T18:10:58.347849+00:00"} {"global_step": 12966, "acc_step": 0, "speed/wps": 12904.29723064595, "speed/FLOPS": 202679720369908.38, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0730309933423996, "optim/lr": 0.0028474182354534467, "optim/total_tokens": 6797918208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349477, "loss/out": 2.81705379486084, "created_at": "2025-01-16T18:11:08.510917+00:00"} {"global_step": 12967, "acc_step": 0, "speed/wps": 12902.341199591807, "speed/FLOPS": 202648998214334.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058894041925668716, "optim/lr": 0.002847380583348879, "optim/total_tokens": 6798442496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8458445072174072, "created_at": "2025-01-16T18:11:18.676716+00:00"} {"global_step": 12968, "acc_step": 0, "speed/wps": 12892.947468090852, "speed/FLOPS": 202501456752772.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07203780859708786, "optim/lr": 0.0028473429268482454, "optim/total_tokens": 6798966784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.7741546630859375, "created_at": "2025-01-16T18:11:28.850474+00:00"} {"global_step": 12969, "acc_step": 0, "speed/wps": 12901.249112405721, "speed/FLOPS": 202631845484392.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06810420006513596, "optim/lr": 0.0028473052659516687, "optim/total_tokens": 6799491072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312703, "loss/out": 2.807199478149414, "created_at": "2025-01-16T18:11:39.017470+00:00"} {"global_step": 12970, "acc_step": 0, "speed/wps": 12904.554402373751, "speed/FLOPS": 202683759605284.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07761400192975998, "optim/lr": 0.0028472676006592724, "optim/total_tokens": 6800015360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.9341866970062256, "created_at": "2025-01-16T18:11:49.178212+00:00"} {"global_step": 12971, "acc_step": 0, "speed/wps": 12905.088569216568, "speed/FLOPS": 202692149429573.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0650143027305603, "optim/lr": 0.0028472299309711792, "optim/total_tokens": 6800539648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.904843807220459, "created_at": "2025-01-16T18:11:59.338378+00:00"} {"global_step": 12972, "acc_step": 0, "speed/wps": 12896.908575732537, "speed/FLOPS": 202563671391417.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07639087736606598, "optim/lr": 0.0028471922568875123, "optim/total_tokens": 6801063936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283915, "loss/out": 2.8244829177856445, "created_at": "2025-01-16T18:12:09.510973+00:00"} {"global_step": 12973, "acc_step": 0, "speed/wps": 12898.754071302445, "speed/FLOPS": 202592657435319.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053801387548446655, "optim/lr": 0.0028471545784083934, "optim/total_tokens": 6801588224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.936342716217041, "created_at": "2025-01-16T18:12:19.679247+00:00"} {"global_step": 12974, "acc_step": 0, "speed/wps": 12901.76010545021, "speed/FLOPS": 202639871332333.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0664382204413414, "optim/lr": 0.0028471168955339463, "optim/total_tokens": 6802112512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.9618842601776123, "created_at": "2025-01-16T18:12:29.846662+00:00"} {"global_step": 12975, "acc_step": 0, "speed/wps": 12901.537614013445, "speed/FLOPS": 202636376798584.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.045680560171604156, "optim/lr": 0.0028470792082642945, "optim/total_tokens": 6802636800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 2.833880662918091, "created_at": "2025-01-16T18:12:40.014018+00:00"} {"global_step": 12976, "acc_step": 0, "speed/wps": 12901.581482724825, "speed/FLOPS": 202637065816978.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07321479171514511, "optim/lr": 0.00284704151659956, "optim/total_tokens": 6803161088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.809791088104248, "created_at": "2025-01-16T18:12:50.178134+00:00"} {"global_step": 12977, "acc_step": 0, "speed/wps": 12894.738387013362, "speed/FLOPS": 202529585595430.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04743357375264168, "optim/lr": 0.0028470038205398664, "optim/total_tokens": 6803685376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.7833640575408936, "created_at": "2025-01-16T18:13:00.346581+00:00"} {"global_step": 12978, "acc_step": 0, "speed/wps": 12889.224870613918, "speed/FLOPS": 202442988244012.2, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05672759190201759, "optim/lr": 0.002846966120085336, "optim/total_tokens": 6804209664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356847, "loss/out": 2.742081642150879, "created_at": "2025-01-16T18:13:10.519553+00:00"} {"global_step": 12979, "acc_step": 0, "speed/wps": 12894.908646811253, "speed/FLOPS": 202532259759518.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054242637008428574, "optim/lr": 0.0028469284152360923, "optim/total_tokens": 6804733952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.9353322982788086, "created_at": "2025-01-16T18:13:20.689423+00:00"} {"global_step": 12980, "acc_step": 0, "speed/wps": 12886.922592306875, "speed/FLOPS": 202406827799538.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.05338435247540474, "optim/lr": 0.0028468907059922585, "optim/total_tokens": 6805258240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.7570455074310303, "created_at": "2025-01-16T18:13:30.865573+00:00"} {"global_step": 12981, "acc_step": 0, "speed/wps": 12885.076243120127, "speed/FLOPS": 202377828348410.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06618527323007584, "optim/lr": 0.0028468529923539572, "optim/total_tokens": 6805782528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400071, "loss/out": 2.8250882625579834, "created_at": "2025-01-16T18:13:41.043228+00:00"} {"global_step": 12982, "acc_step": 0, "speed/wps": 12902.865639258, "speed/FLOPS": 202657235259947.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06040750816464424, "optim/lr": 0.0028468152743213117, "optim/total_tokens": 6806306816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.906907558441162, "created_at": "2025-01-16T18:13:51.205454+00:00"} {"global_step": 12983, "acc_step": 0, "speed/wps": 12908.571670273554, "speed/FLOPS": 202746856317955.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05582038685679436, "optim/lr": 0.002846777551894445, "optim/total_tokens": 6806831104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7311909198760986, "created_at": "2025-01-16T18:14:01.367971+00:00"} {"global_step": 12984, "acc_step": 0, "speed/wps": 12898.272184279329, "speed/FLOPS": 202585088737439.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05775020644068718, "optim/lr": 0.0028467398250734805, "optim/total_tokens": 6807355392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 2.916503429412842, "created_at": "2025-01-16T18:14:11.534279+00:00"} {"global_step": 12985, "acc_step": 0, "speed/wps": 12896.841483780012, "speed/FLOPS": 202562617620108.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.049343883991241455, "optim/lr": 0.0028467020938585403, "optim/total_tokens": 6807879680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8676974773406982, "created_at": "2025-01-16T18:14:21.703372+00:00"} {"global_step": 12986, "acc_step": 0, "speed/wps": 12892.28955548792, "speed/FLOPS": 202491123331276.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04891171306371689, "optim/lr": 0.002846664358249749, "optim/total_tokens": 6808403968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.822411060333252, "created_at": "2025-01-16T18:14:31.874795+00:00"} {"global_step": 12987, "acc_step": 0, "speed/wps": 12892.705511593847, "speed/FLOPS": 202497656493503.9, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061498504132032394, "optim/lr": 0.002846626618247228, "optim/total_tokens": 6808928256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.848662853240967, "created_at": "2025-01-16T18:14:42.046683+00:00"} {"global_step": 12988, "acc_step": 0, "speed/wps": 12898.0111954477, "speed/FLOPS": 202580989549202.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05086135119199753, "optim/lr": 0.002846588873851102, "optim/total_tokens": 6809452544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.951770067214966, "created_at": "2025-01-16T18:14:52.215305+00:00"} {"global_step": 12989, "acc_step": 0, "speed/wps": 12884.743972423736, "speed/FLOPS": 202372609580536.0, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05442905053496361, "optim/lr": 0.002846551125061493, "optim/total_tokens": 6809976832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.876542329788208, "created_at": "2025-01-16T18:15:02.391928+00:00"} {"global_step": 12990, "acc_step": 0, "speed/wps": 12889.096552194407, "speed/FLOPS": 202440972826905.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05605083703994751, "optim/lr": 0.0028465133718785247, "optim/total_tokens": 6810501120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8919479846954346, "created_at": "2025-01-16T18:15:12.564681+00:00"} {"global_step": 12991, "acc_step": 0, "speed/wps": 12907.319830631062, "speed/FLOPS": 202727194456162.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046597886830568314, "optim/lr": 0.00284647561430232, "optim/total_tokens": 6811025408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 2.792886972427368, "created_at": "2025-01-16T18:15:22.723382+00:00"} {"global_step": 12992, "acc_step": 0, "speed/wps": 12902.286293417943, "speed/FLOPS": 202648135837425.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05437823012471199, "optim/lr": 0.0028464378523330026, "optim/total_tokens": 6811549696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.770846128463745, "created_at": "2025-01-16T18:15:32.886303+00:00"} {"global_step": 12993, "acc_step": 0, "speed/wps": 12893.07308160211, "speed/FLOPS": 202503429685577.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05009171739220619, "optim/lr": 0.002846400085970696, "optim/total_tokens": 6812073984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.7845425605773926, "created_at": "2025-01-16T18:15:43.055986+00:00"} {"global_step": 12994, "acc_step": 0, "speed/wps": 12896.138298258114, "speed/FLOPS": 202551573125209.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0440247543156147, "optim/lr": 0.002846362315215522, "optim/total_tokens": 6812598272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 2.7942304611206055, "created_at": "2025-01-16T18:15:53.224127+00:00"} {"global_step": 12995, "acc_step": 0, "speed/wps": 12896.597043525742, "speed/FLOPS": 202558778350019.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0469636432826519, "optim/lr": 0.0028463245400676047, "optim/total_tokens": 6813122560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9711153507232666, "created_at": "2025-01-16T18:16:03.394112+00:00"} {"global_step": 12996, "acc_step": 0, "speed/wps": 12890.555711777508, "speed/FLOPS": 202463890933253.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.039098385721445084, "optim/lr": 0.0028462867605270675, "optim/total_tokens": 6813646848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.9365346431732178, "created_at": "2025-01-16T18:16:13.566522+00:00"} {"global_step": 12997, "acc_step": 0, "speed/wps": 12900.647343040477, "speed/FLOPS": 202622393869593.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048288632184267044, "optim/lr": 0.0028462489765940335, "optim/total_tokens": 6814171136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9064271450042725, "created_at": "2025-01-16T18:16:23.730282+00:00"} {"global_step": 12998, "acc_step": 0, "speed/wps": 12902.324821962959, "speed/FLOPS": 202648740981168.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0490332692861557, "optim/lr": 0.002846211188268626, "optim/total_tokens": 6814695424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 2.849318742752075, "created_at": "2025-01-16T18:16:33.892943+00:00"} {"global_step": 12999, "acc_step": 0, "speed/wps": 12901.118022473478, "speed/FLOPS": 202629786536869.84, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055569518357515335, "optim/lr": 0.0028461733955509683, "optim/total_tokens": 6815219712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.890770435333252, "created_at": "2025-01-16T18:16:44.056124+00:00"} {"global_step": 13000, "acc_step": 0, "speed/wps": 12903.459241975132, "speed/FLOPS": 202666558606316.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05863424390554428, "optim/lr": 0.0028461355984411834, "optim/total_tokens": 6815744000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.7923712730407715, "created_at": "2025-01-16T18:16:54.217687+00:00"} {"global_step": 13001, "acc_step": 0, "speed/wps": 12602.532097934463, "speed/FLOPS": 197940084291927.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.245, "optim/grad_norm": 0.046546854078769684, "optim/lr": 0.0028460977969393944, "optim/total_tokens": 6816268288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 2.8119282722473145, "created_at": "2025-01-16T18:17:04.624684+00:00"} {"global_step": 13002, "acc_step": 0, "speed/wps": 12904.040999127994, "speed/FLOPS": 202675695901819.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05502935126423836, "optim/lr": 0.0028460599910457257, "optim/total_tokens": 6816792576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.876922130584717, "created_at": "2025-01-16T18:17:14.786097+00:00"} {"global_step": 13003, "acc_step": 0, "speed/wps": 12901.157995123967, "speed/FLOPS": 202630414362273.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05187620595097542, "optim/lr": 0.0028460221807603, "optim/total_tokens": 6817316864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8176331520080566, "created_at": "2025-01-16T18:17:24.952208+00:00"} {"global_step": 13004, "acc_step": 0, "speed/wps": 12902.684035750777, "speed/FLOPS": 202654382927317.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05072484537959099, "optim/lr": 0.0028459843660832407, "optim/total_tokens": 6817841152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416874, "loss/out": 2.817826271057129, "created_at": "2025-01-16T18:17:35.120192+00:00"} {"global_step": 13005, "acc_step": 0, "speed/wps": 12899.031907494316, "speed/FLOPS": 202597021234499.72, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049353428184986115, "optim/lr": 0.002845946547014671, "optim/total_tokens": 6818365440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.973539352416992, "created_at": "2025-01-16T18:17:45.286770+00:00"} {"global_step": 13006, "acc_step": 0, "speed/wps": 12899.092366180419, "speed/FLOPS": 202597970821243.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055199023336172104, "optim/lr": 0.0028459087235547147, "optim/total_tokens": 6818889728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.8768482208251953, "created_at": "2025-01-16T18:17:55.452812+00:00"} {"global_step": 13007, "acc_step": 0, "speed/wps": 12899.15753673658, "speed/FLOPS": 202598994414382.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05773826315999031, "optim/lr": 0.0028458708957034944, "optim/total_tokens": 6819414016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 2.9145071506500244, "created_at": "2025-01-16T18:18:05.627030+00:00"} {"global_step": 13008, "acc_step": 0, "speed/wps": 12902.644559924003, "speed/FLOPS": 202653762905212.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05404021590948105, "optim/lr": 0.002845833063461135, "optim/total_tokens": 6819938304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.7065680027008057, "created_at": "2025-01-16T18:18:15.789922+00:00"} {"global_step": 13009, "acc_step": 0, "speed/wps": 12887.958156678289, "speed/FLOPS": 202423092761005.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05940083786845207, "optim/lr": 0.0028457952268277588, "optim/total_tokens": 6820462592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456312, "loss/out": 2.88549542427063, "created_at": "2025-01-16T18:18:25.967572+00:00"} {"global_step": 13010, "acc_step": 0, "speed/wps": 12897.275254348844, "speed/FLOPS": 202569430582955.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07133321464061737, "optim/lr": 0.0028457573858034894, "optim/total_tokens": 6820986880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.902407169342041, "created_at": "2025-01-16T18:18:36.133917+00:00"} {"global_step": 13011, "acc_step": 0, "speed/wps": 12900.499192297468, "speed/FLOPS": 202620066958593.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04603198915719986, "optim/lr": 0.00284571954038845, "optim/total_tokens": 6821511168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 2.9285433292388916, "created_at": "2025-01-16T18:18:46.302032+00:00"} {"global_step": 13012, "acc_step": 0, "speed/wps": 12904.420316781325, "speed/FLOPS": 202681653606802.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0606902651488781, "optim/lr": 0.0028456816905827642, "optim/total_tokens": 6822035456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8799149990081787, "created_at": "2025-01-16T18:18:56.465687+00:00"} {"global_step": 13013, "acc_step": 0, "speed/wps": 12901.510768051468, "speed/FLOPS": 202635955145861.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04809364676475525, "optim/lr": 0.002845643836386557, "optim/total_tokens": 6822559744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.653407096862793, "created_at": "2025-01-16T18:19:06.629972+00:00"} {"global_step": 13014, "acc_step": 0, "speed/wps": 12906.813235912918, "speed/FLOPS": 202719237689979.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054970066994428635, "optim/lr": 0.0028456059777999493, "optim/total_tokens": 6823084032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.861006736755371, "created_at": "2025-01-16T18:19:16.795499+00:00"} {"global_step": 13015, "acc_step": 0, "speed/wps": 12899.50812765456, "speed/FLOPS": 202604500926511.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052412353456020355, "optim/lr": 0.002845568114823066, "optim/total_tokens": 6823608320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.8846988677978516, "created_at": "2025-01-16T18:19:26.962414+00:00"} {"global_step": 13016, "acc_step": 0, "speed/wps": 12895.597892323312, "speed/FLOPS": 202543085307407.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06149369850754738, "optim/lr": 0.002845530247456031, "optim/total_tokens": 6824132608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8628017902374268, "created_at": "2025-01-16T18:19:37.134723+00:00"} {"global_step": 13017, "acc_step": 0, "speed/wps": 12895.935414287545, "speed/FLOPS": 202548386553659.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.045972347259521484, "optim/lr": 0.002845492375698968, "optim/total_tokens": 6824656896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277752, "loss/out": 2.744196891784668, "created_at": "2025-01-16T18:19:47.302224+00:00"} {"global_step": 13018, "acc_step": 0, "speed/wps": 12891.575573569646, "speed/FLOPS": 202479909264136.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04794906824827194, "optim/lr": 0.002845454499551999, "optim/total_tokens": 6825181184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8343515396118164, "created_at": "2025-01-16T18:19:57.475418+00:00"} {"global_step": 13019, "acc_step": 0, "speed/wps": 12897.49172566286, "speed/FLOPS": 202572830562405.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051091358065605164, "optim/lr": 0.0028454166190152496, "optim/total_tokens": 6825705472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.72072696685791, "created_at": "2025-01-16T18:20:07.643482+00:00"} {"global_step": 13020, "acc_step": 0, "speed/wps": 12902.755665853816, "speed/FLOPS": 202655507976515.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07266180217266083, "optim/lr": 0.0028453787340888413, "optim/total_tokens": 6826229760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385871, "loss/out": 2.742766857147217, "created_at": "2025-01-16T18:20:17.809602+00:00"} {"global_step": 13021, "acc_step": 0, "speed/wps": 12899.815387875253, "speed/FLOPS": 202609326870497.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04368466138839722, "optim/lr": 0.0028453408447728992, "optim/total_tokens": 6826754048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9456889629364014, "created_at": "2025-01-16T18:20:27.974395+00:00"} {"global_step": 13022, "acc_step": 0, "speed/wps": 12902.052216132124, "speed/FLOPS": 202644459331995.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05648212134838104, "optim/lr": 0.002845302951067547, "optim/total_tokens": 6827278336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.824258804321289, "created_at": "2025-01-16T18:20:38.143529+00:00"} {"global_step": 13023, "acc_step": 0, "speed/wps": 12883.218278069777, "speed/FLOPS": 202348646454183.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.043930534273386, "optim/lr": 0.0028452650529729065, "optim/total_tokens": 6827802624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482614, "loss/out": 2.7144956588745117, "created_at": "2025-01-16T18:20:48.324428+00:00"} {"global_step": 13024, "acc_step": 0, "speed/wps": 12885.276286910148, "speed/FLOPS": 202380970311021.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06062612682580948, "optim/lr": 0.002845227150489104, "optim/total_tokens": 6828326912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.852139472961426, "created_at": "2025-01-16T18:20:58.503014+00:00"} {"global_step": 13025, "acc_step": 0, "speed/wps": 12885.36615008604, "speed/FLOPS": 202382381735684.2, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.04373912513256073, "optim/lr": 0.002845189243616261, "optim/total_tokens": 6828851200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8701560497283936, "created_at": "2025-01-16T18:21:08.682644+00:00"} {"global_step": 13026, "acc_step": 0, "speed/wps": 12897.358409676948, "speed/FLOPS": 202570736651650.97, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060178451240062714, "optim/lr": 0.002845151332354502, "optim/total_tokens": 6829375488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.904231071472168, "created_at": "2025-01-16T18:21:18.851499+00:00"} {"global_step": 13027, "acc_step": 0, "speed/wps": 12902.70453554875, "speed/FLOPS": 202654704904814.4, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05691540241241455, "optim/lr": 0.002845113416703951, "optim/total_tokens": 6829899776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 2.877577304840088, "created_at": "2025-01-16T18:21:29.022736+00:00"} {"global_step": 13028, "acc_step": 0, "speed/wps": 12896.236427899712, "speed/FLOPS": 202553114386074.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057252436876297, "optim/lr": 0.0028450754966647318, "optim/total_tokens": 6830424064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8803653717041016, "created_at": "2025-01-16T18:21:39.192423+00:00"} {"global_step": 13029, "acc_step": 0, "speed/wps": 12899.40206460541, "speed/FLOPS": 202602835060578.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08061960339546204, "optim/lr": 0.0028450375722369674, "optim/total_tokens": 6830948352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.7554380893707275, "created_at": "2025-01-16T18:21:49.357316+00:00"} {"global_step": 13030, "acc_step": 0, "speed/wps": 12899.435055716107, "speed/FLOPS": 202603353231306.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050825148820877075, "optim/lr": 0.0028449996434207815, "optim/total_tokens": 6831472640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.801269054412842, "created_at": "2025-01-16T18:21:59.524862+00:00"} {"global_step": 13031, "acc_step": 0, "speed/wps": 12905.132387528345, "speed/FLOPS": 202692837656372.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06628566980361938, "optim/lr": 0.002844961710216299, "optim/total_tokens": 6831996928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.843228816986084, "created_at": "2025-01-16T18:22:09.685448+00:00"} {"global_step": 13032, "acc_step": 0, "speed/wps": 12904.950591226436, "speed/FLOPS": 202689982295636.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05145900323987007, "optim/lr": 0.0028449237726236422, "optim/total_tokens": 6832521216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.9039382934570312, "created_at": "2025-01-16T18:22:19.846347+00:00"} {"global_step": 13033, "acc_step": 0, "speed/wps": 12901.481886798561, "speed/FLOPS": 202635501526097.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05457894504070282, "optim/lr": 0.0028448858306429356, "optim/total_tokens": 6833045504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.6646933555603027, "created_at": "2025-01-16T18:22:30.010586+00:00"} {"global_step": 13034, "acc_step": 0, "speed/wps": 12901.703256271217, "speed/FLOPS": 202638978437858.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.08690499514341354, "optim/lr": 0.0028448478842743033, "optim/total_tokens": 6833569792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8433384895324707, "created_at": "2025-01-16T18:22:40.174236+00:00"} {"global_step": 13035, "acc_step": 0, "speed/wps": 12897.258549187845, "speed/FLOPS": 202569168205446.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08050048351287842, "optim/lr": 0.0028448099335178683, "optim/total_tokens": 6834094080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.906576633453369, "created_at": "2025-01-16T18:22:50.340872+00:00"} {"global_step": 13036, "acc_step": 0, "speed/wps": 12897.332422635283, "speed/FLOPS": 202570328489451.7, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050093408674001694, "optim/lr": 0.002844771978373755, "optim/total_tokens": 6834618368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389370, "loss/out": 2.9193270206451416, "created_at": "2025-01-16T18:23:00.513074+00:00"} {"global_step": 13037, "acc_step": 0, "speed/wps": 12905.287696486967, "speed/FLOPS": 202695276996984.22, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06586500257253647, "optim/lr": 0.0028447340188420874, "optim/total_tokens": 6835142656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.753115653991699, "created_at": "2025-01-16T18:23:10.676267+00:00"} {"global_step": 13038, "acc_step": 0, "speed/wps": 12894.902850468054, "speed/FLOPS": 202532168719983.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046874698251485825, "optim/lr": 0.0028446960549229888, "optim/total_tokens": 6835666944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.7972030639648438, "created_at": "2025-01-16T18:23:20.847696+00:00"} {"global_step": 13039, "acc_step": 0, "speed/wps": 12897.79517831791, "speed/FLOPS": 202577596703340.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08541601896286011, "optim/lr": 0.0028446580866165836, "optim/total_tokens": 6836191232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8197860717773438, "created_at": "2025-01-16T18:23:31.015018+00:00"} {"global_step": 13040, "acc_step": 0, "speed/wps": 12901.276619561744, "speed/FLOPS": 202632277522076.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047910477966070175, "optim/lr": 0.0028446201139229956, "optim/total_tokens": 6836715520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.9096736907958984, "created_at": "2025-01-16T18:23:41.183163+00:00"} {"global_step": 13041, "acc_step": 0, "speed/wps": 12899.301255013013, "speed/FLOPS": 202601251707403.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054966386407613754, "optim/lr": 0.0028445821368423474, "optim/total_tokens": 6837239808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8101115226745605, "created_at": "2025-01-16T18:23:51.351615+00:00"} {"global_step": 13042, "acc_step": 0, "speed/wps": 12901.732456670092, "speed/FLOPS": 202639437070248.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0441221222281456, "optim/lr": 0.002844544155374765, "optim/total_tokens": 6837764096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8698465824127197, "created_at": "2025-01-16T18:24:01.515853+00:00"} {"global_step": 13043, "acc_step": 0, "speed/wps": 12903.600297194747, "speed/FLOPS": 202668774072370.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05588109791278839, "optim/lr": 0.002844506169520371, "optim/total_tokens": 6838288384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.892688751220703, "created_at": "2025-01-16T18:24:11.677330+00:00"} {"global_step": 13044, "acc_step": 0, "speed/wps": 12904.114184783613, "speed/FLOPS": 202676845383108.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046682003885507584, "optim/lr": 0.00284446817927929, "optim/total_tokens": 6838812672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.779029369354248, "created_at": "2025-01-16T18:24:21.840658+00:00"} {"global_step": 13045, "acc_step": 0, "speed/wps": 12900.570645598962, "speed/FLOPS": 202621189230880.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06161440908908844, "optim/lr": 0.0028444301846516453, "optim/total_tokens": 6839336960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.8395321369171143, "created_at": "2025-01-16T18:24:32.006795+00:00"} {"global_step": 13046, "acc_step": 0, "speed/wps": 12894.066007008209, "speed/FLOPS": 202519024943503.06, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05133171007037163, "optim/lr": 0.002844392185637561, "optim/total_tokens": 6839861248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311645, "loss/out": 2.838949680328369, "created_at": "2025-01-16T18:24:42.176073+00:00"} {"global_step": 13047, "acc_step": 0, "speed/wps": 12896.619875757056, "speed/FLOPS": 202559136961587.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04976844787597656, "optim/lr": 0.0028443541822371615, "optim/total_tokens": 6840385536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8097290992736816, "created_at": "2025-01-16T18:24:52.345832+00:00"} {"global_step": 13048, "acc_step": 0, "speed/wps": 12903.772763333945, "speed/FLOPS": 202671482890079.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05088043957948685, "optim/lr": 0.0028443161744505707, "optim/total_tokens": 6840909824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368524, "loss/out": 2.829460620880127, "created_at": "2025-01-16T18:25:02.507851+00:00"} {"global_step": 13049, "acc_step": 0, "speed/wps": 12906.395133346625, "speed/FLOPS": 202712670814642.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04777999967336655, "optim/lr": 0.002844278162277912, "optim/total_tokens": 6841434112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8479208946228027, "created_at": "2025-01-16T18:25:12.672160+00:00"} {"global_step": 13050, "acc_step": 0, "speed/wps": 12895.550721550446, "speed/FLOPS": 202542344425600.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04662228375673294, "optim/lr": 0.00284424014571931, "optim/total_tokens": 6841958400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8017218112945557, "created_at": "2025-01-16T18:25:22.842794+00:00"} {"global_step": 13051, "acc_step": 0, "speed/wps": 12905.372621394092, "speed/FLOPS": 202696610859349.66, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06666281074285507, "optim/lr": 0.0028442021247748893, "optim/total_tokens": 6842482688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 2.8972666263580322, "created_at": "2025-01-16T18:25:33.005003+00:00"} {"global_step": 13052, "acc_step": 0, "speed/wps": 12906.508480533224, "speed/FLOPS": 202714451087963.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051611073315143585, "optim/lr": 0.0028441640994447724, "optim/total_tokens": 6843006976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9074320793151855, "created_at": "2025-01-16T18:25:43.164890+00:00"} {"global_step": 13053, "acc_step": 0, "speed/wps": 12903.92217122507, "speed/FLOPS": 202673829546317.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06851320713758469, "optim/lr": 0.0028441260697290847, "optim/total_tokens": 6843531264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.8849616050720215, "created_at": "2025-01-16T18:25:53.326122+00:00"} {"global_step": 13054, "acc_step": 0, "speed/wps": 12905.812897631751, "speed/FLOPS": 202703526002665.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04972146451473236, "optim/lr": 0.0028440880356279496, "optim/total_tokens": 6844055552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8722825050354004, "created_at": "2025-01-16T18:26:03.502559+00:00"} {"global_step": 13055, "acc_step": 0, "speed/wps": 12905.49712921703, "speed/FLOPS": 202698566425800.6, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05080481246113777, "optim/lr": 0.0028440499971414913, "optim/total_tokens": 6844579840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.8000786304473877, "created_at": "2025-01-16T18:26:13.665730+00:00"} {"global_step": 13056, "acc_step": 0, "speed/wps": 12898.745016947958, "speed/FLOPS": 202592515224240.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05702554062008858, "optim/lr": 0.002844011954269834, "optim/total_tokens": 6845104128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8091650009155273, "created_at": "2025-01-16T18:26:23.835589+00:00"} {"global_step": 13057, "acc_step": 0, "speed/wps": 12905.302795297335, "speed/FLOPS": 202695514144549.0, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043238405138254166, "optim/lr": 0.0028439739070131023, "optim/total_tokens": 6845628416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.806030511856079, "created_at": "2025-01-16T18:26:33.995690+00:00"} {"global_step": 13058, "acc_step": 0, "speed/wps": 12900.564872544865, "speed/FLOPS": 202621098557133.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07672020047903061, "optim/lr": 0.0028439358553714197, "optim/total_tokens": 6846152704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8677725791931152, "created_at": "2025-01-16T18:26:44.161155+00:00"} {"global_step": 13059, "acc_step": 0, "speed/wps": 12900.891035733775, "speed/FLOPS": 202626221398210.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06827171891927719, "optim/lr": 0.0028438977993449103, "optim/total_tokens": 6846676992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8360564708709717, "created_at": "2025-01-16T18:26:54.329952+00:00"} {"global_step": 13060, "acc_step": 0, "speed/wps": 12907.35188328321, "speed/FLOPS": 202727697887109.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07897132635116577, "optim/lr": 0.0028438597389336983, "optim/total_tokens": 6847201280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 2.804506301879883, "created_at": "2025-01-16T18:27:04.491696+00:00"} {"global_step": 13061, "acc_step": 0, "speed/wps": 12905.482435864778, "speed/FLOPS": 202698335646512.7, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09630134701728821, "optim/lr": 0.0028438216741379085, "optim/total_tokens": 6847725568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.862417221069336, "created_at": "2025-01-16T18:27:14.651592+00:00"} {"global_step": 13062, "acc_step": 0, "speed/wps": 12906.483740814541, "speed/FLOPS": 202714062516685.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07766684144735336, "optim/lr": 0.0028437836049576646, "optim/total_tokens": 6848249856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 2.971139907836914, "created_at": "2025-01-16T18:27:24.814328+00:00"} {"global_step": 13063, "acc_step": 0, "speed/wps": 12907.612335184422, "speed/FLOPS": 202731788642116.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0737052857875824, "optim/lr": 0.0028437455313930905, "optim/total_tokens": 6848774144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.742929220199585, "created_at": "2025-01-16T18:27:34.975248+00:00"} {"global_step": 13064, "acc_step": 0, "speed/wps": 12904.36859942289, "speed/FLOPS": 202680841314621.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06223033741116524, "optim/lr": 0.0028437074534443108, "optim/total_tokens": 6849298432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.7858757972717285, "created_at": "2025-01-16T18:27:45.139163+00:00"} {"global_step": 13065, "acc_step": 0, "speed/wps": 12890.561122852136, "speed/FLOPS": 202463975921616.4, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06108161807060242, "optim/lr": 0.0028436693711114502, "optim/total_tokens": 6849822720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.860445261001587, "created_at": "2025-01-16T18:27:55.313274+00:00"} {"global_step": 13066, "acc_step": 0, "speed/wps": 12899.24876645396, "speed/FLOPS": 202600427302456.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06263888627290726, "optim/lr": 0.0028436312843946317, "optim/total_tokens": 6850347008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.747711658477783, "created_at": "2025-01-16T18:28:05.478933+00:00"} {"global_step": 13067, "acc_step": 0, "speed/wps": 12906.214882693912, "speed/FLOPS": 202709839730451.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047390419989824295, "optim/lr": 0.002843593193293981, "optim/total_tokens": 6850871296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.865025043487549, "created_at": "2025-01-16T18:28:15.638323+00:00"} {"global_step": 13068, "acc_step": 0, "speed/wps": 12904.435489288697, "speed/FLOPS": 202681891911879.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05572885274887085, "optim/lr": 0.0028435550978096213, "optim/total_tokens": 6851395584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.956571578979492, "created_at": "2025-01-16T18:28:25.799079+00:00"} {"global_step": 13069, "acc_step": 0, "speed/wps": 12900.641224618463, "speed/FLOPS": 202622297771367.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061443571001291275, "optim/lr": 0.0028435169979416776, "optim/total_tokens": 6851919872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.734161853790283, "created_at": "2025-01-16T18:28:35.963094+00:00"} {"global_step": 13070, "acc_step": 0, "speed/wps": 12894.185745021341, "speed/FLOPS": 202520905593536.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05144263058900833, "optim/lr": 0.002843478893690274, "optim/total_tokens": 6852444160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.912890672683716, "created_at": "2025-01-16T18:28:46.131942+00:00"} {"global_step": 13071, "acc_step": 0, "speed/wps": 12901.500043417154, "speed/FLOPS": 202635786700742.06, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05308770388364792, "optim/lr": 0.002843440785055534, "optim/total_tokens": 6852968448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346134, "loss/out": 2.8199350833892822, "created_at": "2025-01-16T18:28:56.296102+00:00"} {"global_step": 13072, "acc_step": 0, "speed/wps": 12904.338751706457, "speed/FLOPS": 202680372515219.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04849289730191231, "optim/lr": 0.0028434026720375837, "optim/total_tokens": 6853492736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.965963840484619, "created_at": "2025-01-16T18:29:06.460848+00:00"} {"global_step": 13073, "acc_step": 0, "speed/wps": 12905.137084266935, "speed/FLOPS": 202692911425105.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04969126358628273, "optim/lr": 0.002843364554636546, "optim/total_tokens": 6854017024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8740761280059814, "created_at": "2025-01-16T18:29:16.623745+00:00"} {"global_step": 13074, "acc_step": 0, "speed/wps": 12904.519843306127, "speed/FLOPS": 202683216807638.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050480615347623825, "optim/lr": 0.0028433264328525452, "optim/total_tokens": 6854541312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.784031391143799, "created_at": "2025-01-16T18:29:26.786483+00:00"} {"global_step": 13075, "acc_step": 0, "speed/wps": 12899.642722458235, "speed/FLOPS": 202606614922855.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04717369005084038, "optim/lr": 0.002843288306685707, "optim/total_tokens": 6855065600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.8985867500305176, "created_at": "2025-01-16T18:29:36.950990+00:00"} {"global_step": 13076, "acc_step": 0, "speed/wps": 12900.761225505921, "speed/FLOPS": 202624182550203.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05142553895711899, "optim/lr": 0.002843250176136154, "optim/total_tokens": 6855589888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428908, "loss/out": 2.788914442062378, "created_at": "2025-01-16T18:29:47.117324+00:00"} {"global_step": 13077, "acc_step": 0, "speed/wps": 12901.446758759337, "speed/FLOPS": 202634949791970.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04179322347044945, "optim/lr": 0.0028432120412040123, "optim/total_tokens": 6856114176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8670244216918945, "created_at": "2025-01-16T18:29:57.280843+00:00"} {"global_step": 13078, "acc_step": 0, "speed/wps": 12898.725152738878, "speed/FLOPS": 202592203229540.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04657992348074913, "optim/lr": 0.0028431739018894057, "optim/total_tokens": 6856638464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.889702320098877, "created_at": "2025-01-16T18:30:07.446185+00:00"} {"global_step": 13079, "acc_step": 0, "speed/wps": 12890.444436862776, "speed/FLOPS": 202462143207810.12, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05927422642707825, "optim/lr": 0.0028431357581924586, "optim/total_tokens": 6857162752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.919053316116333, "created_at": "2025-01-16T18:30:17.620460+00:00"} {"global_step": 13080, "acc_step": 0, "speed/wps": 12900.460904059684, "speed/FLOPS": 202619465589205.88, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0444297231733799, "optim/lr": 0.0028430976101132947, "optim/total_tokens": 6857687040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.0168402194976807, "created_at": "2025-01-16T18:30:27.788843+00:00"} {"global_step": 13081, "acc_step": 0, "speed/wps": 12895.772995906564, "speed/FLOPS": 202545835549800.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05016021803021431, "optim/lr": 0.0028430594576520393, "optim/total_tokens": 6858211328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.8761961460113525, "created_at": "2025-01-16T18:30:37.959010+00:00"} {"global_step": 13082, "acc_step": 0, "speed/wps": 12893.804857608015, "speed/FLOPS": 202514923233316.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.03981665521860123, "optim/lr": 0.0028430213008088165, "optim/total_tokens": 6858735616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8545331954956055, "created_at": "2025-01-16T18:30:48.128200+00:00"} {"global_step": 13083, "acc_step": 0, "speed/wps": 12892.770400308327, "speed/FLOPS": 202498675659931.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.0468815341591835, "optim/lr": 0.0028429831395837516, "optim/total_tokens": 6859259904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369341, "loss/out": 2.745084047317505, "created_at": "2025-01-16T18:30:58.299681+00:00"} {"global_step": 13084, "acc_step": 0, "speed/wps": 12898.81330853925, "speed/FLOPS": 202593587837523.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05419320985674858, "optim/lr": 0.002842944973976968, "optim/total_tokens": 6859784192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8273863792419434, "created_at": "2025-01-16T18:31:08.467831+00:00"} {"global_step": 13085, "acc_step": 0, "speed/wps": 12899.43830115953, "speed/FLOPS": 202603404205455.3, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05605682358145714, "optim/lr": 0.002842906803988591, "optim/total_tokens": 6860308480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.824368476867676, "created_at": "2025-01-16T18:31:18.632362+00:00"} {"global_step": 13086, "acc_step": 0, "speed/wps": 12906.951266757007, "speed/FLOPS": 202721405654060.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05188620463013649, "optim/lr": 0.002842868629618745, "optim/total_tokens": 6860832768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424126, "loss/out": 2.9347264766693115, "created_at": "2025-01-16T18:31:28.790963+00:00"} {"global_step": 13087, "acc_step": 0, "speed/wps": 12906.756140972948, "speed/FLOPS": 202718340935490.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05126509070396423, "optim/lr": 0.0028428304508675544, "optim/total_tokens": 6861357056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.8058366775512695, "created_at": "2025-01-16T18:31:38.957839+00:00"} {"global_step": 13088, "acc_step": 0, "speed/wps": 12900.000664803509, "speed/FLOPS": 202612236899252.62, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04939255490899086, "optim/lr": 0.0028427922677351438, "optim/total_tokens": 6861881344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.933906316757202, "created_at": "2025-01-16T18:31:49.128878+00:00"} {"global_step": 13089, "acc_step": 0, "speed/wps": 12896.58636487636, "speed/FLOPS": 202558610627157.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04512514919042587, "optim/lr": 0.0028427540802216378, "optim/total_tokens": 6862405632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 2.878267765045166, "created_at": "2025-01-16T18:31:59.296141+00:00"} {"global_step": 13090, "acc_step": 0, "speed/wps": 12899.924866580928, "speed/FLOPS": 202611046384009.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05900081619620323, "optim/lr": 0.0028427158883271605, "optim/total_tokens": 6862929920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.6734399795532227, "created_at": "2025-01-16T18:32:09.467426+00:00"} {"global_step": 13091, "acc_step": 0, "speed/wps": 12884.28107349622, "speed/FLOPS": 202365339116789.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05832192301750183, "optim/lr": 0.0028426776920518378, "optim/total_tokens": 6863454208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.834681272506714, "created_at": "2025-01-16T18:32:19.645874+00:00"} {"global_step": 13092, "acc_step": 0, "speed/wps": 12893.788120091194, "speed/FLOPS": 202514660347614.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04317973926663399, "optim/lr": 0.0028426394913957926, "optim/total_tokens": 6863978496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.7398719787597656, "created_at": "2025-01-16T18:32:29.817580+00:00"} {"global_step": 13093, "acc_step": 0, "speed/wps": 12891.710435092735, "speed/FLOPS": 202482027449675.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055871617048978806, "optim/lr": 0.002842601286359151, "optim/total_tokens": 6864502784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 2.8967740535736084, "created_at": "2025-01-16T18:32:39.990397+00:00"} {"global_step": 13094, "acc_step": 0, "speed/wps": 12886.193518115288, "speed/FLOPS": 202395376687506.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05485353618860245, "optim/lr": 0.0028425630769420367, "optim/total_tokens": 6865027072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.953747272491455, "created_at": "2025-01-16T18:32:50.166933+00:00"} {"global_step": 13095, "acc_step": 0, "speed/wps": 12896.51725543925, "speed/FLOPS": 202557525168480.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05253564566373825, "optim/lr": 0.0028425248631445753, "optim/total_tokens": 6865551360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.751311779022217, "created_at": "2025-01-16T18:33:00.337016+00:00"} {"global_step": 13096, "acc_step": 0, "speed/wps": 12895.217583139516, "speed/FLOPS": 202537112029076.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05353192612528801, "optim/lr": 0.00284248664496689, "optim/total_tokens": 6866075648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.983447790145874, "created_at": "2025-01-16T18:33:10.507663+00:00"} {"global_step": 13097, "acc_step": 0, "speed/wps": 12907.779942374573, "speed/FLOPS": 202734421143353.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05887632071971893, "optim/lr": 0.0028424484224091073, "optim/total_tokens": 6866599936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.837247848510742, "created_at": "2025-01-16T18:33:20.666321+00:00"} {"global_step": 13098, "acc_step": 0, "speed/wps": 12902.217275462504, "speed/FLOPS": 202647051815594.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0671866163611412, "optim/lr": 0.002842410195471351, "optim/total_tokens": 6867124224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.930880546569824, "created_at": "2025-01-16T18:33:30.831726+00:00"} {"global_step": 13099, "acc_step": 0, "speed/wps": 12905.952627897475, "speed/FLOPS": 202705720658497.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04134498909115791, "optim/lr": 0.0028423719641537453, "optim/total_tokens": 6867648512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.8406119346618652, "created_at": "2025-01-16T18:33:40.991811+00:00"} {"global_step": 13100, "acc_step": 0, "speed/wps": 12901.459240769815, "speed/FLOPS": 202635145839096.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05337808281183243, "optim/lr": 0.002842333728456416, "optim/total_tokens": 6868172800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.8315958976745605, "created_at": "2025-01-16T18:33:51.165307+00:00"} {"global_step": 13101, "acc_step": 0, "speed/wps": 12904.747250031442, "speed/FLOPS": 202686788542748.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055003486573696136, "optim/lr": 0.0028422954883794873, "optim/total_tokens": 6868697088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487384, "loss/out": 2.7053656578063965, "created_at": "2025-01-16T18:34:01.325996+00:00"} {"global_step": 13102, "acc_step": 0, "speed/wps": 12902.615803712986, "speed/FLOPS": 202653311249403.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06219162791967392, "optim/lr": 0.0028422572439230835, "optim/total_tokens": 6869221376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.869224786758423, "created_at": "2025-01-16T18:34:11.493002+00:00"} {"global_step": 13103, "acc_step": 0, "speed/wps": 12897.744685411843, "speed/FLOPS": 202576803642865.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07376225292682648, "optim/lr": 0.0028422189950873303, "optim/total_tokens": 6869745664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.884890079498291, "created_at": "2025-01-16T18:34:21.659028+00:00"} {"global_step": 13104, "acc_step": 0, "speed/wps": 12903.544728435814, "speed/FLOPS": 202667901288651.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07926250994205475, "optim/lr": 0.0028421807418723523, "optim/total_tokens": 6870269952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.878977060317993, "created_at": "2025-01-16T18:34:31.822301+00:00"} {"global_step": 13105, "acc_step": 0, "speed/wps": 12900.136681718077, "speed/FLOPS": 202614373231802.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04811960831284523, "optim/lr": 0.0028421424842782735, "optim/total_tokens": 6870794240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299229, "loss/out": 2.786425828933716, "created_at": "2025-01-16T18:34:41.989883+00:00"} {"global_step": 13106, "acc_step": 0, "speed/wps": 12903.328297591823, "speed/FLOPS": 202664501944839.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05757465213537216, "optim/lr": 0.00284210422230522, "optim/total_tokens": 6871318528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.9111294746398926, "created_at": "2025-01-16T18:34:52.154066+00:00"} {"global_step": 13107, "acc_step": 0, "speed/wps": 12910.050658408927, "speed/FLOPS": 202770085858964.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04482937231659889, "optim/lr": 0.0028420659559533156, "optim/total_tokens": 6871842816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369482, "loss/out": 2.789341449737549, "created_at": "2025-01-16T18:35:02.326654+00:00"} {"global_step": 13108, "acc_step": 0, "speed/wps": 12906.46274491393, "speed/FLOPS": 202713732747215.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06000284105539322, "optim/lr": 0.002842027685222685, "optim/total_tokens": 6872367104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.849470615386963, "created_at": "2025-01-16T18:35:12.486836+00:00"} {"global_step": 13109, "acc_step": 0, "speed/wps": 12898.877535237689, "speed/FLOPS": 202594596606079.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05282971262931824, "optim/lr": 0.002841989410113454, "optim/total_tokens": 6872891392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.923895835876465, "created_at": "2025-01-16T18:35:22.653763+00:00"} {"global_step": 13110, "acc_step": 0, "speed/wps": 12910.580784566719, "speed/FLOPS": 202778412218741.94, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05464183911681175, "optim/lr": 0.0028419511306257473, "optim/total_tokens": 6873415680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 2.9087867736816406, "created_at": "2025-01-16T18:35:32.810555+00:00"} {"global_step": 13111, "acc_step": 0, "speed/wps": 12903.16587038789, "speed/FLOPS": 202661950802403.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04854293167591095, "optim/lr": 0.002841912846759689, "optim/total_tokens": 6873939968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.902912139892578, "created_at": "2025-01-16T18:35:42.975408+00:00"} {"global_step": 13112, "acc_step": 0, "speed/wps": 12903.029702047872, "speed/FLOPS": 202659812091508.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057628754526376724, "optim/lr": 0.002841874558515405, "optim/total_tokens": 6874464256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8745367527008057, "created_at": "2025-01-16T18:35:53.139239+00:00"} {"global_step": 13113, "acc_step": 0, "speed/wps": 12898.615848076417, "speed/FLOPS": 202590486449616.06, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.045826178044080734, "optim/lr": 0.0028418362658930198, "optim/total_tokens": 6874988544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.8810455799102783, "created_at": "2025-01-16T18:36:03.306849+00:00"} {"global_step": 13114, "acc_step": 0, "speed/wps": 12905.303125462171, "speed/FLOPS": 202695519330241.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04633884131908417, "optim/lr": 0.002841797968892658, "optim/total_tokens": 6875512832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.818178653717041, "created_at": "2025-01-16T18:36:13.467858+00:00"} {"global_step": 13115, "acc_step": 0, "speed/wps": 12901.761480973313, "speed/FLOPS": 202639892936814.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05018948018550873, "optim/lr": 0.002841759667514445, "optim/total_tokens": 6876037120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.887542247772217, "created_at": "2025-01-16T18:36:23.631081+00:00"} {"global_step": 13116, "acc_step": 0, "speed/wps": 12900.154731678074, "speed/FLOPS": 202614656731227.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0643543154001236, "optim/lr": 0.0028417213617585053, "optim/total_tokens": 6876561408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.8937270641326904, "created_at": "2025-01-16T18:36:33.795292+00:00"} {"global_step": 13117, "acc_step": 0, "speed/wps": 12894.930356561248, "speed/FLOPS": 202532600740974.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0547940768301487, "optim/lr": 0.002841683051624965, "optim/total_tokens": 6877085696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.973604202270508, "created_at": "2025-01-16T18:36:43.963841+00:00"} {"global_step": 13118, "acc_step": 0, "speed/wps": 12906.411700032513, "speed/FLOPS": 202712931017209.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0458611436188221, "optim/lr": 0.002841644737113947, "optim/total_tokens": 6877609984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.829526424407959, "created_at": "2025-01-16T18:36:54.123019+00:00"} {"global_step": 13119, "acc_step": 0, "speed/wps": 12901.868435593473, "speed/FLOPS": 202641572806095.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05637691915035248, "optim/lr": 0.0028416064182255787, "optim/total_tokens": 6878134272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.8094897270202637, "created_at": "2025-01-16T18:37:04.285708+00:00"} {"global_step": 13120, "acc_step": 0, "speed/wps": 12902.39054682411, "speed/FLOPS": 202649773280428.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04454687982797623, "optim/lr": 0.0028415680949599835, "optim/total_tokens": 6878658560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503011, "loss/out": 2.7215335369110107, "created_at": "2025-01-16T18:37:14.451103+00:00"} {"global_step": 13121, "acc_step": 0, "speed/wps": 12904.685881634448, "speed/FLOPS": 202685824667745.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048378802835941315, "optim/lr": 0.002841529767317287, "optim/total_tokens": 6879182848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.87225341796875, "created_at": "2025-01-16T18:37:24.611580+00:00"} {"global_step": 13122, "acc_step": 0, "speed/wps": 12902.856529446926, "speed/FLOPS": 202657092177846.78, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04798891767859459, "optim/lr": 0.0028414914352976146, "optim/total_tokens": 6879707136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.861095905303955, "created_at": "2025-01-16T18:37:34.776204+00:00"} {"global_step": 13123, "acc_step": 0, "speed/wps": 12904.90841777234, "speed/FLOPS": 202689319903587.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045566484332084656, "optim/lr": 0.0028414530989010907, "optim/total_tokens": 6880231424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.850156784057617, "created_at": "2025-01-16T18:37:44.939629+00:00"} {"global_step": 13124, "acc_step": 0, "speed/wps": 12903.551372527083, "speed/FLOPS": 202668005643235.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04978640750050545, "optim/lr": 0.0028414147581278405, "optim/total_tokens": 6880755712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 2.8341994285583496, "created_at": "2025-01-16T18:37:55.103557+00:00"} {"global_step": 13125, "acc_step": 0, "speed/wps": 12906.652784034677, "speed/FLOPS": 202716717572747.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055522654205560684, "optim/lr": 0.0028413764129779895, "optim/total_tokens": 6881280000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.7558975219726562, "created_at": "2025-01-16T18:38:05.262863+00:00"} {"global_step": 13126, "acc_step": 0, "speed/wps": 12898.242530251011, "speed/FLOPS": 202584622980175.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0481589175760746, "optim/lr": 0.002841338063451662, "optim/total_tokens": 6881804288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8091115951538086, "created_at": "2025-01-16T18:38:15.428379+00:00"} {"global_step": 13127, "acc_step": 0, "speed/wps": 12907.993378196785, "speed/FLOPS": 202737773446233.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05751236528158188, "optim/lr": 0.0028412997095489845, "optim/total_tokens": 6882328576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.836843490600586, "created_at": "2025-01-16T18:38:25.587523+00:00"} {"global_step": 13128, "acc_step": 0, "speed/wps": 12910.930811117983, "speed/FLOPS": 202783909866716.44, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04658263549208641, "optim/lr": 0.0028412613512700806, "optim/total_tokens": 6882852864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.822744369506836, "created_at": "2025-01-16T18:38:35.746314+00:00"} {"global_step": 13129, "acc_step": 0, "speed/wps": 12903.49224319465, "speed/FLOPS": 202667076935817.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05903659388422966, "optim/lr": 0.0028412229886150765, "optim/total_tokens": 6883377152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411271, "loss/out": 2.8584957122802734, "created_at": "2025-01-16T18:38:45.909433+00:00"} {"global_step": 13130, "acc_step": 0, "speed/wps": 12906.25235293098, "speed/FLOPS": 202710428252014.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04796571284532547, "optim/lr": 0.0028411846215840976, "optim/total_tokens": 6883901440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.7978312969207764, "created_at": "2025-01-16T18:38:56.070133+00:00"} {"global_step": 13131, "acc_step": 0, "speed/wps": 12901.816828576239, "speed/FLOPS": 202640762246973.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04764880985021591, "optim/lr": 0.0028411462501772675, "optim/total_tokens": 6884425728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.838682174682617, "created_at": "2025-01-16T18:39:06.234100+00:00"} {"global_step": 13132, "acc_step": 0, "speed/wps": 12908.890242605363, "speed/FLOPS": 202751859934190.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054601896554231644, "optim/lr": 0.0028411078743947127, "optim/total_tokens": 6884950016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380023, "loss/out": 2.9907257556915283, "created_at": "2025-01-16T18:39:16.396240+00:00"} {"global_step": 13133, "acc_step": 0, "speed/wps": 12910.363924585838, "speed/FLOPS": 202775006134745.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056878462433815, "optim/lr": 0.0028410694942365582, "optim/total_tokens": 6885474304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8465631008148193, "created_at": "2025-01-16T18:39:26.555618+00:00"} {"global_step": 13134, "acc_step": 0, "speed/wps": 12903.154149287988, "speed/FLOPS": 202661766706423.16, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05143933370709419, "optim/lr": 0.0028410311097029294, "optim/total_tokens": 6885998592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9136788845062256, "created_at": "2025-01-16T18:39:36.717309+00:00"} {"global_step": 13135, "acc_step": 0, "speed/wps": 12903.57801048023, "speed/FLOPS": 202668424028893.66, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046194739639759064, "optim/lr": 0.0028409927207939507, "optim/total_tokens": 6886522880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.833479166030884, "created_at": "2025-01-16T18:39:46.880227+00:00"} {"global_step": 13136, "acc_step": 0, "speed/wps": 12909.010904970772, "speed/FLOPS": 202753755102446.97, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05316133052110672, "optim/lr": 0.002840954327509748, "optim/total_tokens": 6887047168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8340179920196533, "created_at": "2025-01-16T18:39:57.041283+00:00"} {"global_step": 13137, "acc_step": 0, "speed/wps": 12907.684517223395, "speed/FLOPS": 202732922360226.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0515836700797081, "optim/lr": 0.002840915929850447, "optim/total_tokens": 6887571456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455539, "loss/out": 2.84629487991333, "created_at": "2025-01-16T18:40:07.201146+00:00"} {"global_step": 13138, "acc_step": 0, "speed/wps": 12903.8605196271, "speed/FLOPS": 202672861223253.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04607100412249565, "optim/lr": 0.002840877527816172, "optim/total_tokens": 6888095744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.7886040210723877, "created_at": "2025-01-16T18:40:17.364091+00:00"} {"global_step": 13139, "acc_step": 0, "speed/wps": 12911.030974644118, "speed/FLOPS": 202785483072533.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05391288921236992, "optim/lr": 0.0028408391214070493, "optim/total_tokens": 6888620032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.9396934509277344, "created_at": "2025-01-16T18:40:27.520547+00:00"} {"global_step": 13140, "acc_step": 0, "speed/wps": 12913.894835145427, "speed/FLOPS": 202830463937063.28, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061791326850652695, "optim/lr": 0.002840800710623203, "optim/total_tokens": 6889144320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.927029609680176, "created_at": "2025-01-16T18:40:37.674135+00:00"} {"global_step": 13141, "acc_step": 0, "speed/wps": 12910.657434084838, "speed/FLOPS": 202779616104751.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04163588583469391, "optim/lr": 0.0028407622954647596, "optim/total_tokens": 6889668608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8600242137908936, "created_at": "2025-01-16T18:40:47.830852+00:00"} {"global_step": 13142, "acc_step": 0, "speed/wps": 12909.77242668877, "speed/FLOPS": 202765715847468.66, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052583955228328705, "optim/lr": 0.002840723875931844, "optim/total_tokens": 6890192896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 2.772958755493164, "created_at": "2025-01-16T18:40:57.992060+00:00"} {"global_step": 13143, "acc_step": 0, "speed/wps": 12901.72033249799, "speed/FLOPS": 202639246643465.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0648755207657814, "optim/lr": 0.002840685452024581, "optim/total_tokens": 6890717184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 2.968665599822998, "created_at": "2025-01-16T18:41:08.156191+00:00"} {"global_step": 13144, "acc_step": 0, "speed/wps": 12903.710055477155, "speed/FLOPS": 202670497977019.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06348994374275208, "optim/lr": 0.0028406470237430964, "optim/total_tokens": 6891241472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.027331829071045, "created_at": "2025-01-16T18:41:18.318313+00:00"} {"global_step": 13145, "acc_step": 0, "speed/wps": 12906.215442514862, "speed/FLOPS": 202709848523208.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03852440044283867, "optim/lr": 0.002840608591087516, "optim/total_tokens": 6891765760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285549, "loss/out": 2.7792820930480957, "created_at": "2025-01-16T18:41:28.481839+00:00"} {"global_step": 13146, "acc_step": 0, "speed/wps": 12907.718188797697, "speed/FLOPS": 202733451218570.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0546191968023777, "optim/lr": 0.002840570154057965, "optim/total_tokens": 6892290048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.845778226852417, "created_at": "2025-01-16T18:41:38.640820+00:00"} {"global_step": 13147, "acc_step": 0, "speed/wps": 12908.939587848012, "speed/FLOPS": 202752634969034.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04655147343873978, "optim/lr": 0.0028405317126545676, "optim/total_tokens": 6892814336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 2.73039174079895, "created_at": "2025-01-16T18:41:48.798692+00:00"} {"global_step": 13148, "acc_step": 0, "speed/wps": 12911.405234493055, "speed/FLOPS": 202791361337747.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049934256821870804, "optim/lr": 0.002840493266877451, "optim/total_tokens": 6893338624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9465479850769043, "created_at": "2025-01-16T18:41:58.956931+00:00"} {"global_step": 13149, "acc_step": 0, "speed/wps": 12906.779544203031, "speed/FLOPS": 202718708515378.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.047701217234134674, "optim/lr": 0.0028404548167267397, "optim/total_tokens": 6893862912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8367385864257812, "created_at": "2025-01-16T18:42:09.118524+00:00"} {"global_step": 13150, "acc_step": 0, "speed/wps": 12910.270958908595, "speed/FLOPS": 202773545981036.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05952977016568184, "optim/lr": 0.0028404163622025593, "optim/total_tokens": 6894387200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 2.8035364151000977, "created_at": "2025-01-16T18:42:19.279070+00:00"} {"global_step": 13151, "acc_step": 0, "speed/wps": 12908.090934195, "speed/FLOPS": 202739305697241.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06411448121070862, "optim/lr": 0.0028403779033050356, "optim/total_tokens": 6894911488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.916046619415283, "created_at": "2025-01-16T18:42:29.438692+00:00"} {"global_step": 13152, "acc_step": 0, "speed/wps": 12909.271156220479, "speed/FLOPS": 202757842705946.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04609279707074165, "optim/lr": 0.0028403394400342936, "optim/total_tokens": 6895435776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 3.0196633338928223, "created_at": "2025-01-16T18:42:39.595916+00:00"} {"global_step": 13153, "acc_step": 0, "speed/wps": 12908.951255522139, "speed/FLOPS": 202752818225889.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04912600666284561, "optim/lr": 0.0028403009723904585, "optim/total_tokens": 6895960064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.8805975914001465, "created_at": "2025-01-16T18:42:49.755367+00:00"} {"global_step": 13154, "acc_step": 0, "speed/wps": 12908.98570926628, "speed/FLOPS": 202753359369285.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.12950466573238373, "optim/lr": 0.002840262500373657, "optim/total_tokens": 6896484352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 2.8068504333496094, "created_at": "2025-01-16T18:42:59.912473+00:00"} {"global_step": 13155, "acc_step": 0, "speed/wps": 12907.19593868971, "speed/FLOPS": 202725248562977.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0829317569732666, "optim/lr": 0.0028402240239840133, "optim/total_tokens": 6897008640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 2.8928263187408447, "created_at": "2025-01-16T18:43:10.071673+00:00"} {"global_step": 13156, "acc_step": 0, "speed/wps": 12907.326275004225, "speed/FLOPS": 202727295673898.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05878252163529396, "optim/lr": 0.0028401855432216537, "optim/total_tokens": 6897532928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.7309093475341797, "created_at": "2025-01-16T18:43:20.233275+00:00"} {"global_step": 13157, "acc_step": 0, "speed/wps": 12911.050417210523, "speed/FLOPS": 202785788444755.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044421084225177765, "optim/lr": 0.002840147058086704, "optim/total_tokens": 6898057216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 2.875180959701538, "created_at": "2025-01-16T18:43:30.393490+00:00"} {"global_step": 13158, "acc_step": 0, "speed/wps": 12910.311138704717, "speed/FLOPS": 202774177059946.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05093088001012802, "optim/lr": 0.002840108568579289, "optim/total_tokens": 6898581504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 2.9072511196136475, "created_at": "2025-01-16T18:43:40.550415+00:00"} {"global_step": 13159, "acc_step": 0, "speed/wps": 12917.561443194872, "speed/FLOPS": 202888053054927.28, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055684976279735565, "optim/lr": 0.002840070074699535, "optim/total_tokens": 6899105792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.774426221847534, "created_at": "2025-01-16T18:43:50.702640+00:00"} {"global_step": 13160, "acc_step": 0, "speed/wps": 12907.039578832771, "speed/FLOPS": 202722792716563.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05029016360640526, "optim/lr": 0.002840031576447566, "optim/total_tokens": 6899630080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.858124017715454, "created_at": "2025-01-16T18:44:00.864713+00:00"} {"global_step": 13161, "acc_step": 0, "speed/wps": 12908.143138382351, "speed/FLOPS": 202740125635739.56, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046144064515829086, "optim/lr": 0.00283999307382351, "optim/total_tokens": 6900154368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9028797149658203, "created_at": "2025-01-16T18:44:11.025523+00:00"} {"global_step": 13162, "acc_step": 0, "speed/wps": 12902.7062164641, "speed/FLOPS": 202654731305899.78, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058251652866601944, "optim/lr": 0.002839954566827491, "optim/total_tokens": 6900678656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 2.8204424381256104, "created_at": "2025-01-16T18:44:21.190689+00:00"} {"global_step": 13163, "acc_step": 0, "speed/wps": 12907.976528055213, "speed/FLOPS": 202737508791605.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04270920902490616, "optim/lr": 0.0028399160554596356, "optim/total_tokens": 6901202944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.892415761947632, "created_at": "2025-01-16T18:44:31.349594+00:00"} {"global_step": 13164, "acc_step": 0, "speed/wps": 12911.531833355815, "speed/FLOPS": 202793349746852.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053140655159950256, "optim/lr": 0.0028398775397200683, "optim/total_tokens": 6901727232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503985, "loss/out": 2.8914437294006348, "created_at": "2025-01-16T18:44:41.507112+00:00"} {"global_step": 13165, "acc_step": 0, "speed/wps": 12906.10389707865, "speed/FLOPS": 202708096548853.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042251888662576675, "optim/lr": 0.002839839019608916, "optim/total_tokens": 6902251520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 2.802887439727783, "created_at": "2025-01-16T18:44:51.666342+00:00"} {"global_step": 13166, "acc_step": 0, "speed/wps": 12908.148806933654, "speed/FLOPS": 202740214668127.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04976903647184372, "optim/lr": 0.002839800495126303, "optim/total_tokens": 6902775808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.9126176834106445, "created_at": "2025-01-16T18:45:01.824767+00:00"} {"global_step": 13167, "acc_step": 0, "speed/wps": 12903.543039020535, "speed/FLOPS": 202667874754063.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04339365288615227, "optim/lr": 0.0028397619662723565, "optim/total_tokens": 6903300096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 2.8958637714385986, "created_at": "2025-01-16T18:45:11.986185+00:00"} {"global_step": 13168, "acc_step": 0, "speed/wps": 12918.070770885392, "speed/FLOPS": 202896052746196.47, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0891764834523201, "optim/lr": 0.002839723433047201, "optim/total_tokens": 6903824384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.869317054748535, "created_at": "2025-01-16T18:45:22.138876+00:00"} {"global_step": 13169, "acc_step": 0, "speed/wps": 12905.636368615093, "speed/FLOPS": 202700753371882.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06286238133907318, "optim/lr": 0.0028396848954509624, "optim/total_tokens": 6904348672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8405075073242188, "created_at": "2025-01-16T18:45:32.306597+00:00"} {"global_step": 13170, "acc_step": 0, "speed/wps": 12908.057782772208, "speed/FLOPS": 202738785008591.38, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05619402229785919, "optim/lr": 0.002839646353483768, "optim/total_tokens": 6904872960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.972015857696533, "created_at": "2025-01-16T18:45:42.464586+00:00"} {"global_step": 13171, "acc_step": 0, "speed/wps": 12895.319371045538, "speed/FLOPS": 202538710748011.97, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0506894625723362, "optim/lr": 0.002839607807145741, "optim/total_tokens": 6905397248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.9702401161193848, "created_at": "2025-01-16T18:45:52.634954+00:00"} {"global_step": 13172, "acc_step": 0, "speed/wps": 12909.22287887421, "speed/FLOPS": 202757084443883.16, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04824446141719818, "optim/lr": 0.002839569256437009, "optim/total_tokens": 6905921536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8516860008239746, "created_at": "2025-01-16T18:46:02.796312+00:00"} {"global_step": 13173, "acc_step": 0, "speed/wps": 12909.969240043869, "speed/FLOPS": 202768807071660.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05245992913842201, "optim/lr": 0.002839530701357697, "optim/total_tokens": 6906445824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.863234758377075, "created_at": "2025-01-16T18:46:12.953207+00:00"} {"global_step": 13174, "acc_step": 0, "speed/wps": 12913.004400602433, "speed/FLOPS": 202816478438980.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06376229971647263, "optim/lr": 0.002839492141907931, "optim/total_tokens": 6906970112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.891233444213867, "created_at": "2025-01-16T18:46:23.107850+00:00"} {"global_step": 13175, "acc_step": 0, "speed/wps": 12906.355373237297, "speed/FLOPS": 202712046327489.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04965386912226677, "optim/lr": 0.002839453578087837, "optim/total_tokens": 6907494400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318692, "loss/out": 2.765716552734375, "created_at": "2025-01-16T18:46:33.267783+00:00"} {"global_step": 13176, "acc_step": 0, "speed/wps": 12911.00007511962, "speed/FLOPS": 202784997753040.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04474300891160965, "optim/lr": 0.0028394150098975408, "optim/total_tokens": 6908018688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.9025464057922363, "created_at": "2025-01-16T18:46:43.423611+00:00"} {"global_step": 13177, "acc_step": 0, "speed/wps": 12910.106407565421, "speed/FLOPS": 202770961476074.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06136579439043999, "optim/lr": 0.002839376437337168, "optim/total_tokens": 6908542976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8776166439056396, "created_at": "2025-01-16T18:46:53.583449+00:00"} {"global_step": 13178, "acc_step": 0, "speed/wps": 12910.369762076347, "speed/FLOPS": 202775097820555.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09379368275403976, "optim/lr": 0.0028393378604068442, "optim/total_tokens": 6909067264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8694653511047363, "created_at": "2025-01-16T18:47:03.740264+00:00"} {"global_step": 13179, "acc_step": 0, "speed/wps": 12909.814099043922, "speed/FLOPS": 202766370369070.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060728494077920914, "optim/lr": 0.0028392992791066956, "optim/total_tokens": 6909591552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.932053327560425, "created_at": "2025-01-16T18:47:13.902041+00:00"} {"global_step": 13180, "acc_step": 0, "speed/wps": 12902.173887380504, "speed/FLOPS": 202646370346145.12, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06808420270681381, "optim/lr": 0.0028392606934368484, "optim/total_tokens": 6910115840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 2.8373422622680664, "created_at": "2025-01-16T18:47:24.064749+00:00"} {"global_step": 13181, "acc_step": 0, "speed/wps": 12908.513197019502, "speed/FLOPS": 202745937915149.88, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04902469739317894, "optim/lr": 0.002839222103397428, "optim/total_tokens": 6910640128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.934041976928711, "created_at": "2025-01-16T18:47:34.223215+00:00"} {"global_step": 13182, "acc_step": 0, "speed/wps": 12913.838759397184, "speed/FLOPS": 202829583190380.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05041328817605972, "optim/lr": 0.002839183508988561, "optim/total_tokens": 6911164416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 2.8447608947753906, "created_at": "2025-01-16T18:47:44.376870+00:00"} {"global_step": 13183, "acc_step": 0, "speed/wps": 12910.371391695251, "speed/FLOPS": 202775123415959.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05037981644272804, "optim/lr": 0.002839144910210372, "optim/total_tokens": 6911688704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.824634075164795, "created_at": "2025-01-16T18:47:54.537806+00:00"} {"global_step": 13184, "acc_step": 0, "speed/wps": 12907.850228570378, "speed/FLOPS": 202735525084642.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04925455152988434, "optim/lr": 0.002839106307062988, "optim/total_tokens": 6912212992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8084750175476074, "created_at": "2025-01-16T18:48:04.698260+00:00"} {"global_step": 13185, "acc_step": 0, "speed/wps": 12909.621558197467, "speed/FLOPS": 202763346250497.7, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0535697378218174, "optim/lr": 0.002839067699546535, "optim/total_tokens": 6912737280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 2.9727017879486084, "created_at": "2025-01-16T18:48:14.856540+00:00"} {"global_step": 13186, "acc_step": 0, "speed/wps": 12908.99049879885, "speed/FLOPS": 202753434595475.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051428135484457016, "optim/lr": 0.0028390290876611386, "optim/total_tokens": 6913261568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9656248092651367, "created_at": "2025-01-16T18:48:25.018340+00:00"} {"global_step": 13187, "acc_step": 0, "speed/wps": 12913.801051950453, "speed/FLOPS": 202828990943114.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06111445277929306, "optim/lr": 0.0028389904714069243, "optim/total_tokens": 6913785856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 2.7858333587646484, "created_at": "2025-01-16T18:48:35.175216+00:00"} {"global_step": 13188, "acc_step": 0, "speed/wps": 12909.449711295372, "speed/FLOPS": 202760647158757.9, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06045454367995262, "optim/lr": 0.0028389518507840193, "optim/total_tokens": 6914310144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.848694086074829, "created_at": "2025-01-16T18:48:45.333532+00:00"} {"global_step": 13189, "acc_step": 0, "speed/wps": 12906.543484639393, "speed/FLOPS": 202715000875551.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06212557479739189, "optim/lr": 0.0028389132257925485, "optim/total_tokens": 6914834432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 2.7700488567352295, "created_at": "2025-01-16T18:48:55.495583+00:00"} {"global_step": 13190, "acc_step": 0, "speed/wps": 12908.464019607918, "speed/FLOPS": 202745165516324.12, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054653000086545944, "optim/lr": 0.002838874596432638, "optim/total_tokens": 6915358720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9364829063415527, "created_at": "2025-01-16T18:49:05.653584+00:00"} {"global_step": 13191, "acc_step": 0, "speed/wps": 12911.881325104905, "speed/FLOPS": 202798838995024.12, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04604736715555191, "optim/lr": 0.0028388359627044145, "optim/total_tokens": 6915883008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.931184768676758, "created_at": "2025-01-16T18:49:15.808500+00:00"} {"global_step": 13192, "acc_step": 0, "speed/wps": 12907.71406196942, "speed/FLOPS": 202733386401061.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04812944680452347, "optim/lr": 0.002838797324608004, "optim/total_tokens": 6916407296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.8706116676330566, "created_at": "2025-01-16T18:49:25.969499+00:00"} {"global_step": 13193, "acc_step": 0, "speed/wps": 12902.134912636393, "speed/FLOPS": 202645758194231.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04648683965206146, "optim/lr": 0.002838758682143532, "optim/total_tokens": 6916931584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8341259956359863, "created_at": "2025-01-16T18:49:36.132442+00:00"} {"global_step": 13194, "acc_step": 0, "speed/wps": 12911.517944766734, "speed/FLOPS": 202793131607476.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05347307771444321, "optim/lr": 0.0028387200353111246, "optim/total_tokens": 6917455872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.8572936058044434, "created_at": "2025-01-16T18:49:46.287506+00:00"} {"global_step": 13195, "acc_step": 0, "speed/wps": 12908.941662310146, "speed/FLOPS": 202752667551312.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047889407724142075, "optim/lr": 0.002838681384110909, "optim/total_tokens": 6917980160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8044748306274414, "created_at": "2025-01-16T18:49:56.447771+00:00"} {"global_step": 13196, "acc_step": 0, "speed/wps": 12915.204445380454, "speed/FLOPS": 202851033165397.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04692741855978966, "optim/lr": 0.0028386427285430095, "optim/total_tokens": 6918504448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.894589424133301, "created_at": "2025-01-16T18:50:06.607521+00:00"} {"global_step": 13197, "acc_step": 0, "speed/wps": 12914.614428459976, "speed/FLOPS": 202841766138893.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0580688901245594, "optim/lr": 0.002838604068607554, "optim/total_tokens": 6919028736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.7009215354919434, "created_at": "2025-01-16T18:50:16.760542+00:00"} {"global_step": 13198, "acc_step": 0, "speed/wps": 12909.850966098575, "speed/FLOPS": 202766949416824.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053025636821985245, "optim/lr": 0.002838565404304668, "optim/total_tokens": 6919553024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.908007860183716, "created_at": "2025-01-16T18:50:26.920521+00:00"} {"global_step": 13199, "acc_step": 0, "speed/wps": 12911.52236542144, "speed/FLOPS": 202793201039933.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05279221013188362, "optim/lr": 0.0028385267356344766, "optim/total_tokens": 6920077312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 2.8209943771362305, "created_at": "2025-01-16T18:50:37.080042+00:00"} {"global_step": 13200, "acc_step": 0, "speed/wps": 12905.978665074399, "speed/FLOPS": 202706129608140.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04085056483745575, "optim/lr": 0.002838488062597107, "optim/total_tokens": 6920601600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.88627290725708, "created_at": "2025-01-16T18:50:47.242427+00:00"} {"global_step": 13201, "acc_step": 0, "speed/wps": 12899.84101956684, "speed/FLOPS": 202609729451435.22, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051980964839458466, "optim/lr": 0.0028384493851926853, "optim/total_tokens": 6921125888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.8745031356811523, "created_at": "2025-01-16T18:50:57.406700+00:00"} {"global_step": 13202, "acc_step": 0, "speed/wps": 12907.667070059006, "speed/FLOPS": 202732648328535.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04323701560497284, "optim/lr": 0.0028384107034213376, "optim/total_tokens": 6921650176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.8108763694763184, "created_at": "2025-01-16T18:51:07.567779+00:00"} {"global_step": 13203, "acc_step": 0, "speed/wps": 12910.620358713688, "speed/FLOPS": 202779033785100.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04339935630559921, "optim/lr": 0.00283837201728319, "optim/total_tokens": 6922174464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8975419998168945, "created_at": "2025-01-16T18:51:17.728854+00:00"} {"global_step": 13204, "acc_step": 0, "speed/wps": 12910.395318480081, "speed/FLOPS": 202775499218994.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0413566455245018, "optim/lr": 0.0028383333267783693, "optim/total_tokens": 6922698752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9703071117401123, "created_at": "2025-01-16T18:51:27.888443+00:00"} {"global_step": 13205, "acc_step": 0, "speed/wps": 12896.937958892386, "speed/FLOPS": 202564132894319.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04363558441400528, "optim/lr": 0.0028382946319070014, "optim/total_tokens": 6923223040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 2.6630401611328125, "created_at": "2025-01-16T18:51:38.058739+00:00"} {"global_step": 13206, "acc_step": 0, "speed/wps": 12908.885783495361, "speed/FLOPS": 202751789897740.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05152876302599907, "optim/lr": 0.002838255932669212, "optim/total_tokens": 6923747328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.926037311553955, "created_at": "2025-01-16T18:51:48.216041+00:00"} {"global_step": 13207, "acc_step": 0, "speed/wps": 12914.257953487855, "speed/FLOPS": 202836167209608.0, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05827885866165161, "optim/lr": 0.0028382172290651276, "optim/total_tokens": 6924271616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 2.878148317337036, "created_at": "2025-01-16T18:51:58.372602+00:00"} {"global_step": 13208, "acc_step": 0, "speed/wps": 12914.605950297331, "speed/FLOPS": 202841632977698.56, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05165698751807213, "optim/lr": 0.002838178521094875, "optim/total_tokens": 6924795904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 2.7865347862243652, "created_at": "2025-01-16T18:52:08.528661+00:00"} {"global_step": 13209, "acc_step": 0, "speed/wps": 12908.30006248534, "speed/FLOPS": 202742590344413.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09926167130470276, "optim/lr": 0.00283813980875858, "optim/total_tokens": 6925320192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 2.746913194656372, "created_at": "2025-01-16T18:52:18.686612+00:00"} {"global_step": 13210, "acc_step": 0, "speed/wps": 12907.32470889311, "speed/FLOPS": 202727271075971.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059448130428791046, "optim/lr": 0.002838101092056369, "optim/total_tokens": 6925844480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.910101890563965, "created_at": "2025-01-16T18:52:28.844954+00:00"} {"global_step": 13211, "acc_step": 0, "speed/wps": 12910.088089340188, "speed/FLOPS": 202770673763175.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04506266117095947, "optim/lr": 0.002838062370988368, "optim/total_tokens": 6926368768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9586522579193115, "created_at": "2025-01-16T18:52:39.003687+00:00"} {"global_step": 13212, "acc_step": 0, "speed/wps": 12904.545505528946, "speed/FLOPS": 202683619868111.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053123485296964645, "optim/lr": 0.0028380236455547046, "optim/total_tokens": 6926893056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380023, "loss/out": 2.898085117340088, "created_at": "2025-01-16T18:52:49.165982+00:00"} {"global_step": 13213, "acc_step": 0, "speed/wps": 12913.701153520526, "speed/FLOPS": 202827421900997.75, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04940245300531387, "optim/lr": 0.0028379849157555035, "optim/total_tokens": 6927417344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.784484624862671, "created_at": "2025-01-16T18:52:59.320382+00:00"} {"global_step": 13214, "acc_step": 0, "speed/wps": 12913.075942737058, "speed/FLOPS": 202817602106513.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04694155603647232, "optim/lr": 0.0028379461815908925, "optim/total_tokens": 6927941632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.8422648906707764, "created_at": "2025-01-16T18:53:09.474264+00:00"} {"global_step": 13215, "acc_step": 0, "speed/wps": 12907.586000199295, "speed/FLOPS": 202731375014986.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.04754309728741646, "optim/lr": 0.0028379074430609965, "optim/total_tokens": 6928465920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.7414708137512207, "created_at": "2025-01-16T18:53:19.634930+00:00"} {"global_step": 13216, "acc_step": 0, "speed/wps": 12906.4774393149, "speed/FLOPS": 202713963542974.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09318774938583374, "optim/lr": 0.0028378687001659434, "optim/total_tokens": 6928990208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.857809543609619, "created_at": "2025-01-16T18:53:29.793998+00:00"} {"global_step": 13217, "acc_step": 0, "speed/wps": 12909.962029195714, "speed/FLOPS": 202768693815381.34, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06795220077037811, "optim/lr": 0.0028378299529058584, "optim/total_tokens": 6929514496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 2.8955085277557373, "created_at": "2025-01-16T18:53:39.950758+00:00"} {"global_step": 13218, "acc_step": 0, "speed/wps": 12907.61689403862, "speed/FLOPS": 202731860245185.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06919198483228683, "optim/lr": 0.002837791201280869, "optim/total_tokens": 6930038784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.7961738109588623, "created_at": "2025-01-16T18:53:50.110803+00:00"} {"global_step": 13219, "acc_step": 0, "speed/wps": 12906.098816171774, "speed/FLOPS": 202708016746229.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05475849658250809, "optim/lr": 0.0028377524452911002, "optim/total_tokens": 6930563072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333702, "loss/out": 2.81002140045166, "created_at": "2025-01-16T18:54:00.270270+00:00"} {"global_step": 13220, "acc_step": 0, "speed/wps": 12907.721308189153, "speed/FLOPS": 202733500212899.97, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061340171843767166, "optim/lr": 0.0028377136849366802, "optim/total_tokens": 6931087360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.881770610809326, "created_at": "2025-01-16T18:54:10.432000+00:00"} {"global_step": 13221, "acc_step": 0, "speed/wps": 12908.159226520578, "speed/FLOPS": 202740378322057.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06810852885246277, "optim/lr": 0.002837674920217734, "optim/total_tokens": 6931611648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400297, "loss/out": 2.8948702812194824, "created_at": "2025-01-16T18:54:20.590543+00:00"} {"global_step": 13222, "acc_step": 0, "speed/wps": 12907.354276851021, "speed/FLOPS": 202727735481380.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06048913300037384, "optim/lr": 0.0028376361511343886, "optim/total_tokens": 6932135936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.923478603363037, "created_at": "2025-01-16T18:54:30.750320+00:00"} {"global_step": 13223, "acc_step": 0, "speed/wps": 12906.60818130617, "speed/FLOPS": 202716017025606.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0544159971177578, "optim/lr": 0.002837597377686771, "optim/total_tokens": 6932660224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8868184089660645, "created_at": "2025-01-16T18:54:40.909225+00:00"} {"global_step": 13224, "acc_step": 0, "speed/wps": 12905.653235133066, "speed/FLOPS": 202701018283723.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05873814597725868, "optim/lr": 0.002837558599875007, "optim/total_tokens": 6933184512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 2.8282508850097656, "created_at": "2025-01-16T18:54:51.072689+00:00"} {"global_step": 13225, "acc_step": 0, "speed/wps": 12904.752417437327, "speed/FLOPS": 202686869703959.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056413523852825165, "optim/lr": 0.002837519817699223, "optim/total_tokens": 6933708800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 2.8633615970611572, "created_at": "2025-01-16T18:55:01.235279+00:00"} {"global_step": 13226, "acc_step": 0, "speed/wps": 12915.107789236348, "speed/FLOPS": 202849515047834.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052401866763830185, "optim/lr": 0.002837481031159546, "optim/total_tokens": 6934233088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 2.8465850353240967, "created_at": "2025-01-16T18:55:11.387486+00:00"} {"global_step": 13227, "acc_step": 0, "speed/wps": 12907.668878936389, "speed/FLOPS": 202732676739439.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05451544374227524, "optim/lr": 0.0028374422402561027, "optim/total_tokens": 6934757376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.9519877433776855, "created_at": "2025-01-16T18:55:21.547904+00:00"} {"global_step": 13228, "acc_step": 0, "speed/wps": 12903.458924918168, "speed/FLOPS": 202666553626500.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04787972569465637, "optim/lr": 0.0028374034449890192, "optim/total_tokens": 6935281664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 2.830317258834839, "created_at": "2025-01-16T18:55:31.709681+00:00"} {"global_step": 13229, "acc_step": 0, "speed/wps": 12904.49498342091, "speed/FLOPS": 202682826348979.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05577266961336136, "optim/lr": 0.0028373646453584227, "optim/total_tokens": 6935805952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314047, "loss/out": 2.7991445064544678, "created_at": "2025-01-16T18:55:41.870213+00:00"} {"global_step": 13230, "acc_step": 0, "speed/wps": 12908.077370291447, "speed/FLOPS": 202739092657497.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048454929143190384, "optim/lr": 0.002837325841364439, "optim/total_tokens": 6936330240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9239730834960938, "created_at": "2025-01-16T18:55:52.028825+00:00"} {"global_step": 13231, "acc_step": 0, "speed/wps": 12906.838779421778, "speed/FLOPS": 202719638885886.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05340466648340225, "optim/lr": 0.0028372870330071946, "optim/total_tokens": 6936854528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.8670754432678223, "created_at": "2025-01-16T18:56:02.187666+00:00"} {"global_step": 13232, "acc_step": 0, "speed/wps": 12913.352569198385, "speed/FLOPS": 202821946905213.5, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05263199657201767, "optim/lr": 0.0028372482202868175, "optim/total_tokens": 6937378816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.0407333374023438, "created_at": "2025-01-16T18:56:12.341321+00:00"} {"global_step": 13233, "acc_step": 0, "speed/wps": 12906.502116278283, "speed/FLOPS": 202714351128594.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049343280494213104, "optim/lr": 0.0028372094032034327, "optim/total_tokens": 6937903104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390663, "loss/out": 2.9525644779205322, "created_at": "2025-01-16T18:56:22.500405+00:00"} {"global_step": 13234, "acc_step": 0, "speed/wps": 12912.826346674303, "speed/FLOPS": 202813681857367.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04877336323261261, "optim/lr": 0.0028371705817571684, "optim/total_tokens": 6938427392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.9504401683807373, "created_at": "2025-01-16T18:56:32.664102+00:00"} {"global_step": 13235, "acc_step": 0, "speed/wps": 12911.487588975458, "speed/FLOPS": 202792654828061.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04490848630666733, "optim/lr": 0.0028371317559481492, "optim/total_tokens": 6938951680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.778188943862915, "created_at": "2025-01-16T18:56:42.821501+00:00"} {"global_step": 13236, "acc_step": 0, "speed/wps": 12911.025225388365, "speed/FLOPS": 202785392772571.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04349694028496742, "optim/lr": 0.002837092925776504, "optim/total_tokens": 6939475968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 2.8420565128326416, "created_at": "2025-01-16T18:56:52.976966+00:00"} {"global_step": 13237, "acc_step": 0, "speed/wps": 12906.58690566589, "speed/FLOPS": 202715682862439.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04394768178462982, "optim/lr": 0.002837054091242358, "optim/total_tokens": 6940000256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.936549663543701, "created_at": "2025-01-16T18:57:03.136120+00:00"} {"global_step": 13238, "acc_step": 0, "speed/wps": 12905.306792778087, "speed/FLOPS": 202695576930477.3, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055940404534339905, "optim/lr": 0.0028370152523458387, "optim/total_tokens": 6940524544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8919546604156494, "created_at": "2025-01-16T18:57:13.298672+00:00"} {"global_step": 13239, "acc_step": 0, "speed/wps": 12910.727433022952, "speed/FLOPS": 202780715534263.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04902360588312149, "optim/lr": 0.002836976409087072, "optim/total_tokens": 6941048832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8427577018737793, "created_at": "2025-01-16T18:57:23.455039+00:00"} {"global_step": 13240, "acc_step": 0, "speed/wps": 12902.890011407668, "speed/FLOPS": 202657618058049.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0449276939034462, "optim/lr": 0.0028369375614661864, "optim/total_tokens": 6941573120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9410128593444824, "created_at": "2025-01-16T18:57:33.620321+00:00"} {"global_step": 13241, "acc_step": 0, "speed/wps": 12908.443883924465, "speed/FLOPS": 202744849257746.16, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05597347393631935, "optim/lr": 0.0028368987094833063, "optim/total_tokens": 6942097408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311795, "loss/out": 2.7577624320983887, "created_at": "2025-01-16T18:57:43.777894+00:00"} {"global_step": 13242, "acc_step": 0, "speed/wps": 12903.049910749829, "speed/FLOPS": 202660129496942.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.04744962602853775, "optim/lr": 0.00283685985313856, "optim/total_tokens": 6942621696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.8594446182250977, "created_at": "2025-01-16T18:57:53.939740+00:00"} {"global_step": 13243, "acc_step": 0, "speed/wps": 12907.620904791222, "speed/FLOPS": 202731923239566.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05236287787556648, "optim/lr": 0.0028368209924320735, "optim/total_tokens": 6943145984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.767690658569336, "created_at": "2025-01-16T18:58:04.099087+00:00"} {"global_step": 13244, "acc_step": 0, "speed/wps": 12912.658851275386, "speed/FLOPS": 202811051111964.47, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05160650610923767, "optim/lr": 0.002836782127363974, "optim/total_tokens": 6943670272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 2.7995176315307617, "created_at": "2025-01-16T18:58:14.253649+00:00"} {"global_step": 13245, "acc_step": 0, "speed/wps": 12908.081637051784, "speed/FLOPS": 202739159672831.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.046192772686481476, "optim/lr": 0.0028367432579343883, "optim/total_tokens": 6944194560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8884847164154053, "created_at": "2025-01-16T18:58:24.412642+00:00"} {"global_step": 13246, "acc_step": 0, "speed/wps": 12906.780053175698, "speed/FLOPS": 202718716509493.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04199773445725441, "optim/lr": 0.0028367043841434434, "optim/total_tokens": 6944718848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 2.8445961475372314, "created_at": "2025-01-16T18:58:34.575159+00:00"} {"global_step": 13247, "acc_step": 0, "speed/wps": 12902.330280773675, "speed/FLOPS": 202648826719291.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04654997959733009, "optim/lr": 0.0028366655059912657, "optim/total_tokens": 6945243136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.776095390319824, "created_at": "2025-01-16T18:58:44.738671+00:00"} {"global_step": 13248, "acc_step": 0, "speed/wps": 12906.458403473473, "speed/FLOPS": 202713664558927.28, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05078225955367088, "optim/lr": 0.0028366266234779824, "optim/total_tokens": 6945767424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.7270970344543457, "created_at": "2025-01-16T18:58:54.897872+00:00"} {"global_step": 13249, "acc_step": 0, "speed/wps": 12907.339072620336, "speed/FLOPS": 202727496678045.0, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044282469898462296, "optim/lr": 0.0028365877366037196, "optim/total_tokens": 6946291712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.828280210494995, "created_at": "2025-01-16T18:59:05.057888+00:00"} {"global_step": 13250, "acc_step": 0, "speed/wps": 12902.575412749633, "speed/FLOPS": 202652676853820.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050965316593647, "optim/lr": 0.0028365488453686053, "optim/total_tokens": 6946816000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9029133319854736, "created_at": "2025-01-16T18:59:15.220001+00:00"} {"global_step": 13251, "acc_step": 0, "speed/wps": 12892.066132845872, "speed/FLOPS": 202487614171668.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04233961179852486, "optim/lr": 0.0028365099497727653, "optim/total_tokens": 6947340288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.908952474594116, "created_at": "2025-01-16T18:59:25.391706+00:00"} {"global_step": 13252, "acc_step": 0, "speed/wps": 12908.534899323622, "speed/FLOPS": 202746278779658.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.001, "optim/grad_norm": 0.03771533817052841, "optim/lr": 0.002836471049816328, "optim/total_tokens": 6947864576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.860856533050537, "created_at": "2025-01-16T18:59:35.551041+00:00"} {"global_step": 13253, "acc_step": 0, "speed/wps": 12910.784756159499, "speed/FLOPS": 202781615872892.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041613444685935974, "optim/lr": 0.0028364321454994188, "optim/total_tokens": 6948388864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9193787574768066, "created_at": "2025-01-16T18:59:45.710711+00:00"} {"global_step": 13254, "acc_step": 0, "speed/wps": 12911.007223168044, "speed/FLOPS": 202785110022963.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0673980638384819, "optim/lr": 0.002836393236822165, "optim/total_tokens": 6948913152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.7720589637756348, "created_at": "2025-01-16T18:59:55.868143+00:00"} {"global_step": 13255, "acc_step": 0, "speed/wps": 12904.09555227299, "speed/FLOPS": 202676552733925.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05585312843322754, "optim/lr": 0.002836354323784694, "optim/total_tokens": 6949437440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.9011077880859375, "created_at": "2025-01-16T19:00:06.029255+00:00"} {"global_step": 13256, "acc_step": 0, "speed/wps": 12905.302630806618, "speed/FLOPS": 202695511560996.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06621103733778, "optim/lr": 0.002836315406387133, "optim/total_tokens": 6949961728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.799757242202759, "created_at": "2025-01-16T19:00:16.189501+00:00"} {"global_step": 13257, "acc_step": 0, "speed/wps": 12904.41106042162, "speed/FLOPS": 202681508222953.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05567331239581108, "optim/lr": 0.0028362764846296077, "optim/total_tokens": 6950486016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.863983154296875, "created_at": "2025-01-16T19:00:26.350513+00:00"} {"global_step": 13258, "acc_step": 0, "speed/wps": 12903.981915000206, "speed/FLOPS": 202674767904402.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05021089315414429, "optim/lr": 0.002836237558512246, "optim/total_tokens": 6951010304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.879979133605957, "created_at": "2025-01-16T19:00:36.512436+00:00"} {"global_step": 13259, "acc_step": 0, "speed/wps": 12905.188846836403, "speed/FLOPS": 202693724427388.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07477530837059021, "optim/lr": 0.002836198628035175, "optim/total_tokens": 6951534592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340231, "loss/out": 2.9609375, "created_at": "2025-01-16T19:00:46.673866+00:00"} {"global_step": 13260, "acc_step": 0, "speed/wps": 12907.941670078611, "speed/FLOPS": 202736961299182.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10447263717651367, "optim/lr": 0.0028361596931985216, "optim/total_tokens": 6952058880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.8186540603637695, "created_at": "2025-01-16T19:00:56.834349+00:00"} {"global_step": 13261, "acc_step": 0, "speed/wps": 12909.726230965272, "speed/FLOPS": 202764990280151.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8081231713294983, "optim/lr": 0.0028361207540024127, "optim/total_tokens": 6952583168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.779038190841675, "created_at": "2025-01-16T19:01:06.994131+00:00"} {"global_step": 13262, "acc_step": 0, "speed/wps": 12902.810761742554, "speed/FLOPS": 202656373333158.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10250041633844376, "optim/lr": 0.002836081810446976, "optim/total_tokens": 6953107456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 2.893714666366577, "created_at": "2025-01-16T19:01:17.159454+00:00"} {"global_step": 13263, "acc_step": 0, "speed/wps": 12910.71811194698, "speed/FLOPS": 202780569133956.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07385639101266861, "optim/lr": 0.0028360428625323368, "optim/total_tokens": 6953631744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8364129066467285, "created_at": "2025-01-16T19:01:27.315145+00:00"} {"global_step": 13264, "acc_step": 0, "speed/wps": 12907.008526773832, "speed/FLOPS": 202722305001308.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2809913456439972, "optim/lr": 0.0028360039102586234, "optim/total_tokens": 6954156032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.863762855529785, "created_at": "2025-01-16T19:01:37.476543+00:00"} {"global_step": 13265, "acc_step": 0, "speed/wps": 12904.188843147862, "speed/FLOPS": 202678017995310.84, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04698216915130615, "optim/lr": 0.0028359649536259636, "optim/total_tokens": 6954680320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.871429920196533, "created_at": "2025-01-16T19:01:47.639765+00:00"} {"global_step": 13266, "acc_step": 0, "speed/wps": 12892.618586070334, "speed/FLOPS": 202496291208708.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048965275287628174, "optim/lr": 0.0028359259926344833, "optim/total_tokens": 6955204608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.734006643295288, "created_at": "2025-01-16T19:01:57.811715+00:00"} {"global_step": 13267, "acc_step": 0, "speed/wps": 12906.263271286989, "speed/FLOPS": 202710599739798.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05258549004793167, "optim/lr": 0.0028358870272843094, "optim/total_tokens": 6955728896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 2.9080159664154053, "created_at": "2025-01-16T19:02:07.971135+00:00"} {"global_step": 13268, "acc_step": 0, "speed/wps": 12902.73677702466, "speed/FLOPS": 202655211301498.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041733238846063614, "optim/lr": 0.0028358480575755704, "optim/total_tokens": 6956253184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.866042137145996, "created_at": "2025-01-16T19:02:18.137567+00:00"} {"global_step": 13269, "acc_step": 0, "speed/wps": 12909.192937651746, "speed/FLOPS": 202756614175840.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044238001108169556, "optim/lr": 0.002835809083508393, "optim/total_tokens": 6956777472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.684359550476074, "created_at": "2025-01-16T19:02:28.297237+00:00"} {"global_step": 13270, "acc_step": 0, "speed/wps": 12910.582752964418, "speed/FLOPS": 202778443135132.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04958217218518257, "optim/lr": 0.002835770105082904, "optim/total_tokens": 6957301760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 2.834099531173706, "created_at": "2025-01-16T19:02:38.454660+00:00"} {"global_step": 13271, "acc_step": 0, "speed/wps": 12914.387521160013, "speed/FLOPS": 202838202247943.88, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041011177003383636, "optim/lr": 0.0028357311222992295, "optim/total_tokens": 6957826048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.896904468536377, "created_at": "2025-01-16T19:02:48.610908+00:00"} {"global_step": 13272, "acc_step": 0, "speed/wps": 12908.197796068485, "speed/FLOPS": 202740984109807.84, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04738987237215042, "optim/lr": 0.002835692135157499, "optim/total_tokens": 6958350336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.622605085372925, "created_at": "2025-01-16T19:02:58.768950+00:00"} {"global_step": 13273, "acc_step": 0, "speed/wps": 12913.422731255565, "speed/FLOPS": 202823048896734.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049317408353090286, "optim/lr": 0.002835653143657838, "optim/total_tokens": 6958874624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7467572689056396, "created_at": "2025-01-16T19:03:08.922605+00:00"} {"global_step": 13274, "acc_step": 0, "speed/wps": 12901.562989989996, "speed/FLOPS": 202636775363166.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042387571185827255, "optim/lr": 0.0028356141478003746, "optim/total_tokens": 6959398912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.781517267227173, "created_at": "2025-01-16T19:03:19.087499+00:00"} {"global_step": 13275, "acc_step": 0, "speed/wps": 12910.437788356665, "speed/FLOPS": 202776166266765.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045200709253549576, "optim/lr": 0.002835575147585235, "optim/total_tokens": 6959923200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 2.853132724761963, "created_at": "2025-01-16T19:03:29.246509+00:00"} {"global_step": 13276, "acc_step": 0, "speed/wps": 12912.312148800482, "speed/FLOPS": 202805605673176.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04696936905384064, "optim/lr": 0.0028355361430125476, "optim/total_tokens": 6960447488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.9052329063415527, "created_at": "2025-01-16T19:03:39.405895+00:00"} {"global_step": 13277, "acc_step": 0, "speed/wps": 12903.890663279399, "speed/FLOPS": 202673334670734.56, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04469899833202362, "optim/lr": 0.00283549713408244, "optim/total_tokens": 6960971776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.8775131702423096, "created_at": "2025-01-16T19:03:49.567531+00:00"} {"global_step": 13278, "acc_step": 0, "speed/wps": 12908.047931616218, "speed/FLOPS": 202738630282649.66, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049085259437561035, "optim/lr": 0.0028354581207950375, "optim/total_tokens": 6961496064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.83843994140625, "created_at": "2025-01-16T19:03:59.729308+00:00"} {"global_step": 13279, "acc_step": 0, "speed/wps": 12913.855044219226, "speed/FLOPS": 202829838965888.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052422162145376205, "optim/lr": 0.0028354191031504682, "optim/total_tokens": 6962020352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.791874885559082, "created_at": "2025-01-16T19:04:09.882716+00:00"} {"global_step": 13280, "acc_step": 0, "speed/wps": 12907.642293935443, "speed/FLOPS": 202732259185468.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051857929676771164, "optim/lr": 0.002835380081148861, "optim/total_tokens": 6962544640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.982783079147339, "created_at": "2025-01-16T19:04:20.040909+00:00"} {"global_step": 13281, "acc_step": 0, "speed/wps": 12905.434231821575, "speed/FLOPS": 202697578535774.88, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05001402646303177, "optim/lr": 0.002835341054790341, "optim/total_tokens": 6963068928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8789186477661133, "created_at": "2025-01-16T19:04:30.201315+00:00"} {"global_step": 13282, "acc_step": 0, "speed/wps": 12908.564796054223, "speed/FLOPS": 202746748348894.7, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05037964507937431, "optim/lr": 0.0028353020240750367, "optim/total_tokens": 6963593216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375182, "loss/out": 2.830052375793457, "created_at": "2025-01-16T19:04:40.359200+00:00"} {"global_step": 13283, "acc_step": 0, "speed/wps": 12907.888849545468, "speed/FLOPS": 202736131680126.84, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04203156754374504, "optim/lr": 0.0028352629890030755, "optim/total_tokens": 6964117504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 2.6852521896362305, "created_at": "2025-01-16T19:04:50.521414+00:00"} {"global_step": 13284, "acc_step": 0, "speed/wps": 12906.535469210208, "speed/FLOPS": 202714874982222.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06071634963154793, "optim/lr": 0.0028352239495745843, "optim/total_tokens": 6964641792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8134164810180664, "created_at": "2025-01-16T19:05:00.680655+00:00"} {"global_step": 13285, "acc_step": 0, "speed/wps": 12907.052966550371, "speed/FLOPS": 202723002989065.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059474702924489975, "optim/lr": 0.002835184905789691, "optim/total_tokens": 6965166080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340917, "loss/out": 2.884168863296509, "created_at": "2025-01-16T19:05:10.843878+00:00"} {"global_step": 13286, "acc_step": 0, "speed/wps": 12906.210509464276, "speed/FLOPS": 202709771042870.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054344117641448975, "optim/lr": 0.0028351458576485224, "optim/total_tokens": 6965690368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8040361404418945, "created_at": "2025-01-16T19:05:21.009079+00:00"} {"global_step": 13287, "acc_step": 0, "speed/wps": 12902.788656377015, "speed/FLOPS": 202656026138016.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.001, "optim/grad_norm": 0.060111626982688904, "optim/lr": 0.0028351068051512057, "optim/total_tokens": 6966214656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9779458045959473, "created_at": "2025-01-16T19:05:31.172766+00:00"} {"global_step": 13288, "acc_step": 0, "speed/wps": 12909.379400574633, "speed/FLOPS": 202759542832271.62, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05866221711039543, "optim/lr": 0.0028350677482978696, "optim/total_tokens": 6966738944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.9864039421081543, "created_at": "2025-01-16T19:05:41.329940+00:00"} {"global_step": 13289, "acc_step": 0, "speed/wps": 12908.403742981089, "speed/FLOPS": 202744218789071.0, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04803072661161423, "optim/lr": 0.00283502868708864, "optim/total_tokens": 6967263232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.845874786376953, "created_at": "2025-01-16T19:05:51.487638+00:00"} {"global_step": 13290, "acc_step": 0, "speed/wps": 12913.136938355612, "speed/FLOPS": 202818560126519.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05025837942957878, "optim/lr": 0.0028349896215236455, "optim/total_tokens": 6967787520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369865, "loss/out": 2.9327566623687744, "created_at": "2025-01-16T19:06:01.641561+00:00"} {"global_step": 13291, "acc_step": 0, "speed/wps": 12905.413265370395, "speed/FLOPS": 202697249228848.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04447704181075096, "optim/lr": 0.0028349505516030125, "optim/total_tokens": 6968311808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.8696906566619873, "created_at": "2025-01-16T19:06:11.801807+00:00"} {"global_step": 13292, "acc_step": 0, "speed/wps": 12910.183516873629, "speed/FLOPS": 202772172583718.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04701478034257889, "optim/lr": 0.00283491147732687, "optim/total_tokens": 6968836096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.842371940612793, "created_at": "2025-01-16T19:06:21.961208+00:00"} {"global_step": 13293, "acc_step": 0, "speed/wps": 12909.970914562888, "speed/FLOPS": 202768833372282.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04467051476240158, "optim/lr": 0.002834872398695344, "optim/total_tokens": 6969360384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 2.9054341316223145, "created_at": "2025-01-16T19:06:32.117579+00:00"} {"global_step": 13294, "acc_step": 0, "speed/wps": 12908.230273374078, "speed/FLOPS": 202741494210520.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04761213809251785, "optim/lr": 0.0028348333157085627, "optim/total_tokens": 6969884672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.787712812423706, "created_at": "2025-01-16T19:06:42.275402+00:00"} {"global_step": 13295, "acc_step": 0, "speed/wps": 12905.642305974692, "speed/FLOPS": 202700846626273.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041524577885866165, "optim/lr": 0.0028347942283666536, "optim/total_tokens": 6970408960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.829732894897461, "created_at": "2025-01-16T19:06:52.435077+00:00"} {"global_step": 13296, "acc_step": 0, "speed/wps": 12911.320695676088, "speed/FLOPS": 202790033539459.0, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044211938977241516, "optim/lr": 0.0028347551366697437, "optim/total_tokens": 6970933248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.876580238342285, "created_at": "2025-01-16T19:07:02.592525+00:00"} {"global_step": 13297, "acc_step": 0, "speed/wps": 12911.249263009502, "speed/FLOPS": 202788911591271.5, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042061563581228256, "optim/lr": 0.002834716040617961, "optim/total_tokens": 6971457536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8089404106140137, "created_at": "2025-01-16T19:07:12.747775+00:00"} {"global_step": 13298, "acc_step": 0, "speed/wps": 12898.861569532937, "speed/FLOPS": 202594345842747.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0454021580517292, "optim/lr": 0.002834676940211434, "optim/total_tokens": 6971981824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.781198024749756, "created_at": "2025-01-16T19:07:22.913142+00:00"} {"global_step": 13299, "acc_step": 0, "speed/wps": 12906.15751609847, "speed/FLOPS": 202708938709240.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0493125393986702, "optim/lr": 0.0028346378354502876, "optim/total_tokens": 6972506112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.792659044265747, "created_at": "2025-01-16T19:07:33.073695+00:00"} {"global_step": 13300, "acc_step": 0, "speed/wps": 12908.02588525478, "speed/FLOPS": 202738284014248.38, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046032313257455826, "optim/lr": 0.0028345987263346524, "optim/total_tokens": 6973030400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.910768508911133, "created_at": "2025-01-16T19:07:43.233508+00:00"} {"global_step": 13301, "acc_step": 0, "speed/wps": 12905.440483782699, "speed/FLOPS": 202697676731415.44, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05589143931865692, "optim/lr": 0.0028345596128646544, "optim/total_tokens": 6973554688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.795586109161377, "created_at": "2025-01-16T19:07:53.396610+00:00"} {"global_step": 13302, "acc_step": 0, "speed/wps": 12904.662467064109, "speed/FLOPS": 202685456909742.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04814305901527405, "optim/lr": 0.002834520495040421, "optim/total_tokens": 6974078976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461136, "loss/out": 2.78997540473938, "created_at": "2025-01-16T19:08:03.564431+00:00"} {"global_step": 13303, "acc_step": 0, "speed/wps": 12908.221446047082, "speed/FLOPS": 202741355565220.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04472820833325386, "optim/lr": 0.002834481372862081, "optim/total_tokens": 6974603264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.923275947570801, "created_at": "2025-01-16T19:08:13.724946+00:00"} {"global_step": 13304, "acc_step": 0, "speed/wps": 12899.067816851886, "speed/FLOPS": 202597585240305.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055743392556905746, "optim/lr": 0.002834442246329761, "optim/total_tokens": 6975127552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.794454574584961, "created_at": "2025-01-16T19:08:23.894098+00:00"} {"global_step": 13305, "acc_step": 0, "speed/wps": 12912.011341446736, "speed/FLOPS": 202800881080333.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04549001157283783, "optim/lr": 0.002834403115443589, "optim/total_tokens": 6975651840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420112, "loss/out": 2.8373823165893555, "created_at": "2025-01-16T19:08:34.050616+00:00"} {"global_step": 13306, "acc_step": 0, "speed/wps": 12905.247208323879, "speed/FLOPS": 202694641074746.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05219061300158501, "optim/lr": 0.002834363980203693, "optim/total_tokens": 6976176128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.7949118614196777, "created_at": "2025-01-16T19:08:44.210878+00:00"} {"global_step": 13307, "acc_step": 0, "speed/wps": 12908.301774457346, "speed/FLOPS": 202742617233286.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04501283913850784, "optim/lr": 0.0028343248406101995, "optim/total_tokens": 6976700416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.7603061199188232, "created_at": "2025-01-16T19:08:54.370002+00:00"} {"global_step": 13308, "acc_step": 0, "speed/wps": 12907.85759569621, "speed/FLOPS": 202735640795476.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05076514557003975, "optim/lr": 0.0028342856966632377, "optim/total_tokens": 6977224704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.903352975845337, "created_at": "2025-01-16T19:09:04.528173+00:00"} {"global_step": 13309, "acc_step": 0, "speed/wps": 12907.05253449716, "speed/FLOPS": 202722996203075.84, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05654124543070793, "optim/lr": 0.0028342465483629344, "optim/total_tokens": 6977748992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.907409191131592, "created_at": "2025-01-16T19:09:14.687352+00:00"} {"global_step": 13310, "acc_step": 0, "speed/wps": 12906.544522668974, "speed/FLOPS": 202715017179232.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04945379123091698, "optim/lr": 0.002834207395709418, "optim/total_tokens": 6978273280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392822, "loss/out": 2.90201473236084, "created_at": "2025-01-16T19:09:24.847018+00:00"} {"global_step": 13311, "acc_step": 0, "speed/wps": 12903.219977239034, "speed/FLOPS": 202662800624851.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06934366375207901, "optim/lr": 0.0028341682387028155, "optim/total_tokens": 6978797568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.879906177520752, "created_at": "2025-01-16T19:09:35.008902+00:00"} {"global_step": 13312, "acc_step": 0, "speed/wps": 12906.030680567419, "speed/FLOPS": 202706946582934.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052076589316129684, "optim/lr": 0.0028341290773432554, "optim/total_tokens": 6979321856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.7890477180480957, "created_at": "2025-01-16T19:09:45.168844+00:00"} {"global_step": 13313, "acc_step": 0, "speed/wps": 12897.378493062097, "speed/FLOPS": 202571052088812.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05370093137025833, "optim/lr": 0.002834089911630865, "optim/total_tokens": 6979846144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.962750196456909, "created_at": "2025-01-16T19:09:55.335235+00:00"} {"global_step": 13314, "acc_step": 0, "speed/wps": 12896.791681436234, "speed/FLOPS": 202561835405864.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05120232328772545, "optim/lr": 0.0028340507415657717, "optim/total_tokens": 6980370432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.85435152053833, "created_at": "2025-01-16T19:10:05.503982+00:00"} {"global_step": 13315, "acc_step": 0, "speed/wps": 12905.766258035907, "speed/FLOPS": 202702793463722.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046988870948553085, "optim/lr": 0.0028340115671481045, "optim/total_tokens": 6980894720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 2.8777780532836914, "created_at": "2025-01-16T19:10:15.667946+00:00"} {"global_step": 13316, "acc_step": 0, "speed/wps": 12907.489843967296, "speed/FLOPS": 202729864749230.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04543239250779152, "optim/lr": 0.0028339723883779896, "optim/total_tokens": 6981419008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8289246559143066, "created_at": "2025-01-16T19:10:25.835284+00:00"} {"global_step": 13317, "acc_step": 0, "speed/wps": 12908.68793993465, "speed/FLOPS": 202748682492754.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05277959257364273, "optim/lr": 0.0028339332052555563, "optim/total_tokens": 6981943296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.88735294342041, "created_at": "2025-01-16T19:10:35.993185+00:00"} {"global_step": 13318, "acc_step": 0, "speed/wps": 12901.976859812521, "speed/FLOPS": 202643275757445.7, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04852655902504921, "optim/lr": 0.0028338940177809317, "optim/total_tokens": 6982467584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.810436725616455, "created_at": "2025-01-16T19:10:46.157407+00:00"} {"global_step": 13319, "acc_step": 0, "speed/wps": 12905.392169996423, "speed/FLOPS": 202696917897011.88, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05240935832262039, "optim/lr": 0.0028338548259542445, "optim/total_tokens": 6982991872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.800550937652588, "created_at": "2025-01-16T19:10:56.320213+00:00"} {"global_step": 13320, "acc_step": 0, "speed/wps": 12905.757689708129, "speed/FLOPS": 202702658886360.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053730983287096024, "optim/lr": 0.002833815629775621, "optim/total_tokens": 6983516160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.808708667755127, "created_at": "2025-01-16T19:11:06.480128+00:00"} {"global_step": 13321, "acc_step": 0, "speed/wps": 12901.374284572752, "speed/FLOPS": 202633811485279.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055500201880931854, "optim/lr": 0.0028337764292451903, "optim/total_tokens": 6984040448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 2.8668954372406006, "created_at": "2025-01-16T19:11:16.646037+00:00"} {"global_step": 13322, "acc_step": 0, "speed/wps": 12905.973274186643, "speed/FLOPS": 202706044936840.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056680940091609955, "optim/lr": 0.002833737224363079, "optim/total_tokens": 6984564736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.714470386505127, "created_at": "2025-01-16T19:11:26.806968+00:00"} {"global_step": 13323, "acc_step": 0, "speed/wps": 12906.107531710739, "speed/FLOPS": 202708153635745.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06614751368761063, "optim/lr": 0.0028336980151294176, "optim/total_tokens": 6985089024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.9399030208587646, "created_at": "2025-01-16T19:11:36.967221+00:00"} {"global_step": 13324, "acc_step": 0, "speed/wps": 12905.087430836675, "speed/FLOPS": 202692131549752.53, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06189951300621033, "optim/lr": 0.0028336588015443315, "optim/total_tokens": 6985613312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.9207119941711426, "created_at": "2025-01-16T19:11:47.127809+00:00"} {"global_step": 13325, "acc_step": 0, "speed/wps": 12905.70803623596, "speed/FLOPS": 202701879010349.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04816097021102905, "optim/lr": 0.0028336195836079496, "optim/total_tokens": 6986137600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 2.882948160171509, "created_at": "2025-01-16T19:11:57.290200+00:00"} {"global_step": 13326, "acc_step": 0, "speed/wps": 12911.435364839914, "speed/FLOPS": 202791834576248.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05196365714073181, "optim/lr": 0.0028335803613204, "optim/total_tokens": 6986661888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302164, "loss/out": 2.8024260997772217, "created_at": "2025-01-16T19:12:07.449020+00:00"} {"global_step": 13327, "acc_step": 0, "speed/wps": 12910.017414021608, "speed/FLOPS": 202769563710178.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04658275842666626, "optim/lr": 0.0028335411346818106, "optim/total_tokens": 6987186176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.814678907394409, "created_at": "2025-01-16T19:12:17.606360+00:00"} {"global_step": 13328, "acc_step": 0, "speed/wps": 12902.394356792951, "speed/FLOPS": 202649833121224.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046621743589639664, "optim/lr": 0.0028335019036923085, "optim/total_tokens": 6987710464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.868269920349121, "created_at": "2025-01-16T19:12:27.769708+00:00"} {"global_step": 13329, "acc_step": 0, "speed/wps": 12908.364004829655, "speed/FLOPS": 202743594646796.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05190160125494003, "optim/lr": 0.002833462668352023, "optim/total_tokens": 6988234752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 2.862908363342285, "created_at": "2025-01-16T19:12:37.933125+00:00"} {"global_step": 13330, "acc_step": 0, "speed/wps": 12908.1451190598, "speed/FLOPS": 202740156745000.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043808192014694214, "optim/lr": 0.0028334234286610817, "optim/total_tokens": 6988759040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.840090751647949, "created_at": "2025-01-16T19:12:48.092111+00:00"} {"global_step": 13331, "acc_step": 0, "speed/wps": 12903.398226542426, "speed/FLOPS": 202665600275101.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05043431371450424, "optim/lr": 0.002833384184619612, "optim/total_tokens": 6989283328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8526978492736816, "created_at": "2025-01-16T19:12:58.253729+00:00"} {"global_step": 13332, "acc_step": 0, "speed/wps": 12909.293836786892, "speed/FLOPS": 202758198935408.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04574688896536827, "optim/lr": 0.002833344936227743, "optim/total_tokens": 6989807616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.864192008972168, "created_at": "2025-01-16T19:13:08.410797+00:00"} {"global_step": 13333, "acc_step": 0, "speed/wps": 12906.675696863673, "speed/FLOPS": 202717077450212.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.046706728637218475, "optim/lr": 0.002833305683485602, "optim/total_tokens": 6990331904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.849565029144287, "created_at": "2025-01-16T19:13:18.572173+00:00"} {"global_step": 13334, "acc_step": 0, "speed/wps": 12913.062807919447, "speed/FLOPS": 202817395806154.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04926416277885437, "optim/lr": 0.0028332664263933174, "optim/total_tokens": 6990856192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.986682415008545, "created_at": "2025-01-16T19:13:28.726421+00:00"} {"global_step": 13335, "acc_step": 0, "speed/wps": 12903.084658319389, "speed/FLOPS": 202660675255270.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07716403901576996, "optim/lr": 0.0028332271649510166, "optim/total_tokens": 6991380480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.972512722015381, "created_at": "2025-01-16T19:13:38.891239+00:00"} {"global_step": 13336, "acc_step": 0, "speed/wps": 12904.294792066952, "speed/FLOPS": 202679682068674.22, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0820455402135849, "optim/lr": 0.002833187899158828, "optim/total_tokens": 6991904768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8114824295043945, "created_at": "2025-01-16T19:13:49.052272+00:00"} {"global_step": 13337, "acc_step": 0, "speed/wps": 12898.2731074975, "speed/FLOPS": 202585103237849.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.059178777039051056, "optim/lr": 0.0028331486290168813, "optim/total_tokens": 6992429056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 2.7471728324890137, "created_at": "2025-01-16T19:13:59.217725+00:00"} {"global_step": 13338, "acc_step": 0, "speed/wps": 12908.016906670135, "speed/FLOPS": 202738142993238.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05135065317153931, "optim/lr": 0.0028331093545253022, "optim/total_tokens": 6992953344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.893141269683838, "created_at": "2025-01-16T19:14:09.381014+00:00"} {"global_step": 13339, "acc_step": 0, "speed/wps": 12909.021586444615, "speed/FLOPS": 202753922869671.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05041278526186943, "optim/lr": 0.00283307007568422, "optim/total_tokens": 6993477632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 2.8742363452911377, "created_at": "2025-01-16T19:14:19.538781+00:00"} {"global_step": 13340, "acc_step": 0, "speed/wps": 12905.182219992572, "speed/FLOPS": 202693620343699.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05779873579740524, "optim/lr": 0.0028330307924937627, "optim/total_tokens": 6994001920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.9650869369506836, "created_at": "2025-01-16T19:14:29.704051+00:00"} {"global_step": 13341, "acc_step": 0, "speed/wps": 12905.040966884857, "speed/FLOPS": 202691401769541.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062464796006679535, "optim/lr": 0.0028329915049540586, "optim/total_tokens": 6994526208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.7815499305725098, "created_at": "2025-01-16T19:14:39.864198+00:00"} {"global_step": 13342, "acc_step": 0, "speed/wps": 12904.093941985513, "speed/FLOPS": 202676527442148.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06974826008081436, "optim/lr": 0.0028329522130652366, "optim/total_tokens": 6995050496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 2.821294069290161, "created_at": "2025-01-16T19:14:50.027932+00:00"} {"global_step": 13343, "acc_step": 0, "speed/wps": 12910.175694697435, "speed/FLOPS": 202772049725693.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04938100650906563, "optim/lr": 0.0028329129168274234, "optim/total_tokens": 6995574784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.9726791381835938, "created_at": "2025-01-16T19:15:00.188769+00:00"} {"global_step": 13344, "acc_step": 0, "speed/wps": 12902.149704819825, "speed/FLOPS": 202645990525799.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051693301647901535, "optim/lr": 0.002832873616240748, "optim/total_tokens": 6996099072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.839193105697632, "created_at": "2025-01-16T19:15:10.351468+00:00"} {"global_step": 13345, "acc_step": 0, "speed/wps": 12905.87448859326, "speed/FLOPS": 202704493373350.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04855244979262352, "optim/lr": 0.002832834311305339, "optim/total_tokens": 6996623360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 2.737337827682495, "created_at": "2025-01-16T19:15:20.512867+00:00"} {"global_step": 13346, "acc_step": 0, "speed/wps": 12910.776773361986, "speed/FLOPS": 202781490492088.44, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048738911747932434, "optim/lr": 0.002832795002021323, "optim/total_tokens": 6997147648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8879170417785645, "created_at": "2025-01-16T19:15:30.669592+00:00"} {"global_step": 13347, "acc_step": 0, "speed/wps": 12903.805898665716, "speed/FLOPS": 202672003325997.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0491105355322361, "optim/lr": 0.002832755688388831, "optim/total_tokens": 6997671936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.826146364212036, "created_at": "2025-01-16T19:15:40.836340+00:00"} {"global_step": 13348, "acc_step": 0, "speed/wps": 12898.100151204384, "speed/FLOPS": 202582386721599.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0535174161195755, "optim/lr": 0.0028327163704079884, "optim/total_tokens": 6998196224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.930938243865967, "created_at": "2025-01-16T19:15:51.002001+00:00"} {"global_step": 13349, "acc_step": 0, "speed/wps": 12906.998338681837, "speed/FLOPS": 202722144983323.34, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05216575041413307, "optim/lr": 0.0028326770480789254, "optim/total_tokens": 6998720512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.8472061157226562, "created_at": "2025-01-16T19:16:01.164726+00:00"} {"global_step": 13350, "acc_step": 0, "speed/wps": 12908.034788083112, "speed/FLOPS": 202738423845401.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05211682245135307, "optim/lr": 0.0028326377214017693, "optim/total_tokens": 6999244800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.8005733489990234, "created_at": "2025-01-16T19:16:11.323692+00:00"} {"global_step": 13351, "acc_step": 0, "speed/wps": 12900.748993181856, "speed/FLOPS": 202623990424745.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04944176971912384, "optim/lr": 0.0028325983903766494, "optim/total_tokens": 6999769088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497719, "loss/out": 2.9085912704467773, "created_at": "2025-01-16T19:16:21.489199+00:00"} {"global_step": 13352, "acc_step": 0, "speed/wps": 12903.451041088836, "speed/FLOPS": 202666429800127.56, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06857660412788391, "optim/lr": 0.002832559055003693, "optim/total_tokens": 7000293376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416344, "loss/out": 2.8811514377593994, "created_at": "2025-01-16T19:16:31.651267+00:00"} {"global_step": 13353, "acc_step": 0, "speed/wps": 12901.711801475321, "speed/FLOPS": 202639112652031.66, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04482646659016609, "optim/lr": 0.002832519715283029, "optim/total_tokens": 7000817664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.7821619510650635, "created_at": "2025-01-16T19:16:41.814562+00:00"} {"global_step": 13354, "acc_step": 0, "speed/wps": 12899.835567591806, "speed/FLOPS": 202609643820675.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06806853413581848, "optim/lr": 0.0028324803712147855, "optim/total_tokens": 7001341952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.7989859580993652, "created_at": "2025-01-16T19:16:51.983879+00:00"} {"global_step": 13355, "acc_step": 0, "speed/wps": 12902.80077191073, "speed/FLOPS": 202656216429121.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05746198445558548, "optim/lr": 0.0028324410227990914, "optim/total_tokens": 7001866240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.877939462661743, "created_at": "2025-01-16T19:17:02.145833+00:00"} {"global_step": 13356, "acc_step": 0, "speed/wps": 12901.412007001418, "speed/FLOPS": 202634403967858.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04610765352845192, "optim/lr": 0.002832401670036074, "optim/total_tokens": 7002390528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.866576910018921, "created_at": "2025-01-16T19:17:12.308821+00:00"} {"global_step": 13357, "acc_step": 0, "speed/wps": 12900.096796983955, "speed/FLOPS": 202613746787246.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054557185620069504, "optim/lr": 0.002832362312925863, "optim/total_tokens": 7002914816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8147828578948975, "created_at": "2025-01-16T19:17:22.475564+00:00"} {"global_step": 13358, "acc_step": 0, "speed/wps": 12905.117028779183, "speed/FLOPS": 202692596426111.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0716434195637703, "optim/lr": 0.0028323229514685857, "optim/total_tokens": 7003439104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448320, "loss/out": 2.823831081390381, "created_at": "2025-01-16T19:17:32.635706+00:00"} {"global_step": 13359, "acc_step": 0, "speed/wps": 12907.825121449763, "speed/FLOPS": 202735130742812.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05534963682293892, "optim/lr": 0.0028322835856643713, "optim/total_tokens": 7003963392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480466, "loss/out": 2.8522443771362305, "created_at": "2025-01-16T19:17:42.799114+00:00"} {"global_step": 13360, "acc_step": 0, "speed/wps": 12905.064110659912, "speed/FLOPS": 202691765274331.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04502181336283684, "optim/lr": 0.0028322442155133474, "optim/total_tokens": 7004487680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.9269728660583496, "created_at": "2025-01-16T19:17:52.959479+00:00"} {"global_step": 13361, "acc_step": 0, "speed/wps": 12905.712670670098, "speed/FLOPS": 202701951800506.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052028387784957886, "optim/lr": 0.0028322048410156435, "optim/total_tokens": 7005011968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.8561787605285645, "created_at": "2025-01-16T19:18:03.126322+00:00"} {"global_step": 13362, "acc_step": 0, "speed/wps": 12910.944553944553, "speed/FLOPS": 202784125716692.34, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03955579921603203, "optim/lr": 0.0028321654621713877, "optim/total_tokens": 7005536256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.873805046081543, "created_at": "2025-01-16T19:18:13.292698+00:00"} {"global_step": 13363, "acc_step": 0, "speed/wps": 12904.248384699944, "speed/FLOPS": 202678953177204.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0477181114256382, "optim/lr": 0.0028321260789807077, "optim/total_tokens": 7006060544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9138541221618652, "created_at": "2025-01-16T19:18:23.453455+00:00"} {"global_step": 13364, "acc_step": 0, "speed/wps": 12902.527420863618, "speed/FLOPS": 202651923075303.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0450158417224884, "optim/lr": 0.002832086691443733, "optim/total_tokens": 7006584832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 2.861936330795288, "created_at": "2025-01-16T19:18:33.616981+00:00"} {"global_step": 13365, "acc_step": 0, "speed/wps": 12906.192746726027, "speed/FLOPS": 202709492054658.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.041518889367580414, "optim/lr": 0.002832047299560592, "optim/total_tokens": 7007109120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8036446571350098, "created_at": "2025-01-16T19:18:43.791144+00:00"} {"global_step": 13366, "acc_step": 0, "speed/wps": 12898.64193722788, "speed/FLOPS": 202590896215590.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048802994191646576, "optim/lr": 0.0028320079033314124, "optim/total_tokens": 7007633408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 2.8325648307800293, "created_at": "2025-01-16T19:18:53.959138+00:00"} {"global_step": 13367, "acc_step": 0, "speed/wps": 12904.114905333268, "speed/FLOPS": 202676856700331.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045540183782577515, "optim/lr": 0.002831968502756324, "optim/total_tokens": 7008157696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.846714735031128, "created_at": "2025-01-16T19:19:04.120104+00:00"} {"global_step": 13368, "acc_step": 0, "speed/wps": 12905.50609719605, "speed/FLOPS": 202698707280234.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05172570049762726, "optim/lr": 0.002831929097835454, "optim/total_tokens": 7008681984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.06296443939209, "created_at": "2025-01-16T19:19:14.285086+00:00"} {"global_step": 13369, "acc_step": 0, "speed/wps": 12905.944499585761, "speed/FLOPS": 202705592992192.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05192585289478302, "optim/lr": 0.0028318896885689317, "optim/total_tokens": 7009206272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.7362163066864014, "created_at": "2025-01-16T19:19:24.449822+00:00"} {"global_step": 13370, "acc_step": 0, "speed/wps": 12905.292794505986, "speed/FLOPS": 202695357068378.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05341516435146332, "optim/lr": 0.002831850274956886, "optim/total_tokens": 7009730560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.7620530128479004, "created_at": "2025-01-16T19:19:34.610715+00:00"} {"global_step": 13371, "acc_step": 0, "speed/wps": 12903.458892975863, "speed/FLOPS": 202666553124802.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05798323079943657, "optim/lr": 0.0028318108569994446, "optim/total_tokens": 7010254848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.8249034881591797, "created_at": "2025-01-16T19:19:44.776918+00:00"} {"global_step": 13372, "acc_step": 0, "speed/wps": 12907.749103395, "speed/FLOPS": 202733936774802.4, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04695260152220726, "optim/lr": 0.002831771434696737, "optim/total_tokens": 7010779136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 2.895085334777832, "created_at": "2025-01-16T19:19:54.940867+00:00"} {"global_step": 13373, "acc_step": 0, "speed/wps": 12911.338750888312, "speed/FLOPS": 202790317121377.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049609698355197906, "optim/lr": 0.0028317320080488905, "optim/total_tokens": 7011303424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.9089460372924805, "created_at": "2025-01-16T19:20:05.099113+00:00"} {"global_step": 13374, "acc_step": 0, "speed/wps": 12912.2426811748, "speed/FLOPS": 202804514588657.1, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05868019536137581, "optim/lr": 0.0028316925770560356, "optim/total_tokens": 7011827712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.895073413848877, "created_at": "2025-01-16T19:20:15.253790+00:00"} {"global_step": 13375, "acc_step": 0, "speed/wps": 12904.10840145858, "speed/FLOPS": 202676754548042.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05726698786020279, "optim/lr": 0.0028316531417183002, "optim/total_tokens": 7012352000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.887258529663086, "created_at": "2025-01-16T19:20:25.416608+00:00"} {"global_step": 13376, "acc_step": 0, "speed/wps": 12908.601532838726, "speed/FLOPS": 202747325350576.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05225870758295059, "optim/lr": 0.002831613702035812, "optim/total_tokens": 7012876288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.838609218597412, "created_at": "2025-01-16T19:20:35.575071+00:00"} {"global_step": 13377, "acc_step": 0, "speed/wps": 12907.448939500071, "speed/FLOPS": 202729222288364.12, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051115792244672775, "optim/lr": 0.002831574258008701, "optim/total_tokens": 7013400576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.8708009719848633, "created_at": "2025-01-16T19:20:45.733439+00:00"} {"global_step": 13378, "acc_step": 0, "speed/wps": 12902.328086603264, "speed/FLOPS": 202648792256830.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06573290377855301, "optim/lr": 0.002831534809637095, "optim/total_tokens": 7013924864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.83022403717041, "created_at": "2025-01-16T19:20:55.898458+00:00"} {"global_step": 13379, "acc_step": 0, "speed/wps": 12902.73040111097, "speed/FLOPS": 202655111159012.1, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048925988376140594, "optim/lr": 0.0028314953569211227, "optim/total_tokens": 7014449152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477322, "loss/out": 2.9906368255615234, "created_at": "2025-01-16T19:21:06.060874+00:00"} {"global_step": 13380, "acc_step": 0, "speed/wps": 12901.262542515484, "speed/FLOPS": 202632056422720.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0491221658885479, "optim/lr": 0.002831455899860914, "optim/total_tokens": 7014973440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 2.819850444793701, "created_at": "2025-01-16T19:21:16.227205+00:00"} {"global_step": 13381, "acc_step": 0, "speed/wps": 12906.115718221055, "speed/FLOPS": 202708282216139.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0478467121720314, "optim/lr": 0.0028314164384565963, "optim/total_tokens": 7015497728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.758073329925537, "created_at": "2025-01-16T19:21:26.386450+00:00"} {"global_step": 13382, "acc_step": 0, "speed/wps": 12908.430604613535, "speed/FLOPS": 202744640687920.3, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04870704934000969, "optim/lr": 0.0028313769727082994, "optim/total_tokens": 7016022016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392822, "loss/out": 2.86952805519104, "created_at": "2025-01-16T19:21:36.547200+00:00"} {"global_step": 13383, "acc_step": 0, "speed/wps": 12907.444202013847, "speed/FLOPS": 202729147879633.0, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053732845932245255, "optim/lr": 0.0028313375026161507, "optim/total_tokens": 7016546304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.808257818222046, "created_at": "2025-01-16T19:21:46.707234+00:00"} {"global_step": 13384, "acc_step": 0, "speed/wps": 12907.46847545264, "speed/FLOPS": 202729529127344.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05038444697856903, "optim/lr": 0.0028312980281802805, "optim/total_tokens": 7017070592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8271594047546387, "created_at": "2025-01-16T19:21:56.868109+00:00"} {"global_step": 13385, "acc_step": 0, "speed/wps": 12911.666378314243, "speed/FLOPS": 202795462960308.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06207314878702164, "optim/lr": 0.002831258549400816, "optim/total_tokens": 7017594880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.64562726020813, "created_at": "2025-01-16T19:22:07.025952+00:00"} {"global_step": 13386, "acc_step": 0, "speed/wps": 12902.296097930841, "speed/FLOPS": 202648289830773.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06761866807937622, "optim/lr": 0.002831219066277888, "optim/total_tokens": 7018119168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.913649559020996, "created_at": "2025-01-16T19:22:17.188687+00:00"} {"global_step": 13387, "acc_step": 0, "speed/wps": 12905.009867641806, "speed/FLOPS": 202690913313194.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05099267140030861, "optim/lr": 0.0028311795788116237, "optim/total_tokens": 7018643456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 2.859617233276367, "created_at": "2025-01-16T19:22:27.354133+00:00"} {"global_step": 13388, "acc_step": 0, "speed/wps": 12907.307154979779, "speed/FLOPS": 202726995367640.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04722938686609268, "optim/lr": 0.0028311400870021522, "optim/total_tokens": 7019167744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9162259101867676, "created_at": "2025-01-16T19:22:37.512704+00:00"} {"global_step": 13389, "acc_step": 0, "speed/wps": 12905.149465733526, "speed/FLOPS": 202693105893052.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.062222789973020554, "optim/lr": 0.0028311005908496028, "optim/total_tokens": 7019692032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8777687549591064, "created_at": "2025-01-16T19:22:47.680282+00:00"} {"global_step": 13390, "acc_step": 0, "speed/wps": 12905.009741025202, "speed/FLOPS": 202690911324506.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042997270822525024, "optim/lr": 0.0028310610903541046, "optim/total_tokens": 7020216320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392646, "loss/out": 2.9006412029266357, "created_at": "2025-01-16T19:22:57.843039+00:00"} {"global_step": 13391, "acc_step": 0, "speed/wps": 12905.572094709794, "speed/FLOPS": 202699743861877.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05567384138703346, "optim/lr": 0.0028310215855157856, "optim/total_tokens": 7020740608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9100754261016846, "created_at": "2025-01-16T19:23:08.005509+00:00"} {"global_step": 13392, "acc_step": 0, "speed/wps": 12904.612263520394, "speed/FLOPS": 202684668394102.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06020374223589897, "optim/lr": 0.002830982076334775, "optim/total_tokens": 7021264896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 2.9344234466552734, "created_at": "2025-01-16T19:23:18.167639+00:00"} {"global_step": 13393, "acc_step": 0, "speed/wps": 12908.606297222383, "speed/FLOPS": 202747400181769.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05529303103685379, "optim/lr": 0.002830942562811202, "optim/total_tokens": 7021789184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.7088632583618164, "created_at": "2025-01-16T19:23:28.325693+00:00"} {"global_step": 13394, "acc_step": 0, "speed/wps": 12905.831875945374, "speed/FLOPS": 202703824083159.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06058398261666298, "optim/lr": 0.002830903044945195, "optim/total_tokens": 7022313472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8686013221740723, "created_at": "2025-01-16T19:23:38.491137+00:00"} {"global_step": 13395, "acc_step": 0, "speed/wps": 12908.083897109944, "speed/FLOPS": 202739195170150.38, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06749019026756287, "optim/lr": 0.002830863522736884, "optim/total_tokens": 7022837760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419268, "loss/out": 2.760272264480591, "created_at": "2025-01-16T19:23:48.651965+00:00"} {"global_step": 13396, "acc_step": 0, "speed/wps": 12906.317053780505, "speed/FLOPS": 202711444467764.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04933555796742439, "optim/lr": 0.002830823996186397, "optim/total_tokens": 7023362048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.893965244293213, "created_at": "2025-01-16T19:23:58.814190+00:00"} {"global_step": 13397, "acc_step": 0, "speed/wps": 12904.806128174836, "speed/FLOPS": 202687713304898.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045041900128126144, "optim/lr": 0.0028307844652938627, "optim/total_tokens": 7023886336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.878502607345581, "created_at": "2025-01-16T19:24:08.976405+00:00"} {"global_step": 13398, "acc_step": 0, "speed/wps": 12906.015746929723, "speed/FLOPS": 202706712029633.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04158657044172287, "optim/lr": 0.0028307449300594115, "optim/total_tokens": 7024410624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 2.8790297508239746, "created_at": "2025-01-16T19:24:19.137962+00:00"} {"global_step": 13399, "acc_step": 0, "speed/wps": 12912.339117921883, "speed/FLOPS": 202806029260288.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047759078443050385, "optim/lr": 0.0028307053904831707, "optim/total_tokens": 7024934912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.8977890014648438, "created_at": "2025-01-16T19:24:29.294031+00:00"} {"global_step": 13400, "acc_step": 0, "speed/wps": 12901.903073899566, "speed/FLOPS": 202642116848290.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04392707720398903, "optim/lr": 0.0028306658465652706, "optim/total_tokens": 7025459200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.8063342571258545, "created_at": "2025-01-16T19:24:39.456617+00:00"} {"global_step": 13401, "acc_step": 0, "speed/wps": 12906.482856663111, "speed/FLOPS": 202714048629872.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04295268654823303, "optim/lr": 0.002830626298305839, "optim/total_tokens": 7025983488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.854182004928589, "created_at": "2025-01-16T19:24:49.616667+00:00"} {"global_step": 13402, "acc_step": 0, "speed/wps": 12906.320147639743, "speed/FLOPS": 202711493061075.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047413140535354614, "optim/lr": 0.0028305867457050064, "optim/total_tokens": 7026507776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9228289127349854, "created_at": "2025-01-16T19:24:59.775912+00:00"} {"global_step": 13403, "acc_step": 0, "speed/wps": 12908.35748483157, "speed/FLOPS": 202743492241267.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04918133467435837, "optim/lr": 0.002830547188762901, "optim/total_tokens": 7027032064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372699, "loss/out": 2.8365471363067627, "created_at": "2025-01-16T19:25:09.933731+00:00"} {"global_step": 13404, "acc_step": 0, "speed/wps": 12904.08816341128, "speed/FLOPS": 202676436681699.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043837983161211014, "optim/lr": 0.0028305076274796515, "optim/total_tokens": 7027556352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 2.9199726581573486, "created_at": "2025-01-16T19:25:20.097054+00:00"} {"global_step": 13405, "acc_step": 0, "speed/wps": 12907.80353399747, "speed/FLOPS": 202734791682209.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043589718639850616, "optim/lr": 0.0028304680618553876, "optim/total_tokens": 7028080640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460613, "loss/out": 2.8586201667785645, "created_at": "2025-01-16T19:25:30.255141+00:00"} {"global_step": 13406, "acc_step": 0, "speed/wps": 12911.444384900733, "speed/FLOPS": 202791976248698.84, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05717301741242409, "optim/lr": 0.002830428491890238, "optim/total_tokens": 7028604928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7864699363708496, "created_at": "2025-01-16T19:25:40.415279+00:00"} {"global_step": 13407, "acc_step": 0, "speed/wps": 12908.52793871306, "speed/FLOPS": 202746169453704.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04618922993540764, "optim/lr": 0.002830388917584332, "optim/total_tokens": 7029129216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.86356258392334, "created_at": "2025-01-16T19:25:50.574265+00:00"} {"global_step": 13408, "acc_step": 0, "speed/wps": 12903.683614554773, "speed/FLOPS": 202670082685999.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.045960914343595505, "optim/lr": 0.0028303493389377986, "optim/total_tokens": 7029653504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 2.8931915760040283, "created_at": "2025-01-16T19:26:00.735528+00:00"} {"global_step": 13409, "acc_step": 0, "speed/wps": 12909.434825399063, "speed/FLOPS": 202760413355301.0, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04090481996536255, "optim/lr": 0.002830309755950768, "optim/total_tokens": 7030177792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8253636360168457, "created_at": "2025-01-16T19:26:10.892396+00:00"} {"global_step": 13410, "acc_step": 0, "speed/wps": 12901.913495233339, "speed/FLOPS": 202642280529657.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04812617599964142, "optim/lr": 0.0028302701686233676, "optim/total_tokens": 7030702080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.878737449645996, "created_at": "2025-01-16T19:26:21.058203+00:00"} {"global_step": 13411, "acc_step": 0, "speed/wps": 12904.307865274885, "speed/FLOPS": 202679887401369.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059925418347120285, "optim/lr": 0.002830230576955727, "optim/total_tokens": 7031226368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.877939462661743, "created_at": "2025-01-16T19:26:31.221103+00:00"} {"global_step": 13412, "acc_step": 0, "speed/wps": 12899.129143649372, "speed/FLOPS": 202598548461931.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05633994936943054, "optim/lr": 0.0028301909809479767, "optim/total_tokens": 7031750656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.870901584625244, "created_at": "2025-01-16T19:26:41.390956+00:00"} {"global_step": 13413, "acc_step": 0, "speed/wps": 12903.077507210983, "speed/FLOPS": 202660562937286.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0559958890080452, "optim/lr": 0.002830151380600245, "optim/total_tokens": 7032274944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.81160044670105, "created_at": "2025-01-16T19:26:51.555734+00:00"} {"global_step": 13414, "acc_step": 0, "speed/wps": 12904.125379933666, "speed/FLOPS": 202677021218324.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08304660767316818, "optim/lr": 0.0028301117759126602, "optim/total_tokens": 7032799232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 2.9101266860961914, "created_at": "2025-01-16T19:27:01.716664+00:00"} {"global_step": 13415, "acc_step": 0, "speed/wps": 12903.544800602722, "speed/FLOPS": 202667902422132.22, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07775513082742691, "optim/lr": 0.002830072166885352, "optim/total_tokens": 7033323520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.817493438720703, "created_at": "2025-01-16T19:27:11.878227+00:00"} {"global_step": 13416, "acc_step": 0, "speed/wps": 12910.187109985445, "speed/FLOPS": 202772229018477.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05971045419573784, "optim/lr": 0.0028300325535184512, "optim/total_tokens": 7033847808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.9788527488708496, "created_at": "2025-01-16T19:27:22.034347+00:00"} {"global_step": 13417, "acc_step": 0, "speed/wps": 12905.431430689341, "speed/FLOPS": 202697534540144.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056247055530548096, "optim/lr": 0.0028299929358120853, "optim/total_tokens": 7034372096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.8863654136657715, "created_at": "2025-01-16T19:27:32.194465+00:00"} {"global_step": 13418, "acc_step": 0, "speed/wps": 12901.251116986197, "speed/FLOPS": 202631876969083.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059781648218631744, "optim/lr": 0.002829953313766384, "optim/total_tokens": 7034896384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.7550735473632812, "created_at": "2025-01-16T19:27:42.357564+00:00"} {"global_step": 13419, "acc_step": 0, "speed/wps": 12906.623428792884, "speed/FLOPS": 202716256508337.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05577107518911362, "optim/lr": 0.0028299136873814767, "optim/total_tokens": 7035420672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.8826546669006348, "created_at": "2025-01-16T19:27:52.518391+00:00"} {"global_step": 13420, "acc_step": 0, "speed/wps": 12906.313118403166, "speed/FLOPS": 202711382657255.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046642713248729706, "optim/lr": 0.002829874056657493, "optim/total_tokens": 7035944960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8042104244232178, "created_at": "2025-01-16T19:28:02.686867+00:00"} {"global_step": 13421, "acc_step": 0, "speed/wps": 12902.636845061226, "speed/FLOPS": 202653641732691.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04882071912288666, "optim/lr": 0.002829834421594561, "optim/total_tokens": 7036469248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8810603618621826, "created_at": "2025-01-16T19:28:12.849723+00:00"} {"global_step": 13422, "acc_step": 0, "speed/wps": 12903.483711011588, "speed/FLOPS": 202666942926157.56, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045208241790533066, "optim/lr": 0.002829794782192812, "optim/total_tokens": 7036993536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 3.0039472579956055, "created_at": "2025-01-16T19:28:23.011048+00:00"} {"global_step": 13423, "acc_step": 0, "speed/wps": 12904.181766539064, "speed/FLOPS": 202677906847445.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04382510855793953, "optim/lr": 0.002829755138452373, "optim/total_tokens": 7037517824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.7166850566864014, "created_at": "2025-01-16T19:28:33.171875+00:00"} {"global_step": 13424, "acc_step": 0, "speed/wps": 12905.80548313364, "speed/FLOPS": 202703409547783.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0512317456305027, "optim/lr": 0.0028297154903733756, "optim/total_tokens": 7038042112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456584, "loss/out": 2.864682674407959, "created_at": "2025-01-16T19:28:43.332085+00:00"} {"global_step": 13425, "acc_step": 0, "speed/wps": 12909.11040640823, "speed/FLOPS": 202755317909251.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0576491504907608, "optim/lr": 0.0028296758379559475, "optim/total_tokens": 7038566400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 2.882030963897705, "created_at": "2025-01-16T19:28:53.494559+00:00"} {"global_step": 13426, "acc_step": 0, "speed/wps": 12909.975245340045, "speed/FLOPS": 202768901393088.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 0.03806724399328232, "optim/lr": 0.002829636181200219, "optim/total_tokens": 7039090688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 2.8889920711517334, "created_at": "2025-01-16T19:29:03.652224+00:00"} {"global_step": 13427, "acc_step": 0, "speed/wps": 12901.930787280444, "speed/FLOPS": 202642552125019.12, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.05402761325240135, "optim/lr": 0.0028295965201063185, "optim/total_tokens": 7039614976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.0265824794769287, "created_at": "2025-01-16T19:29:13.814877+00:00"} {"global_step": 13428, "acc_step": 0, "speed/wps": 12898.255328773894, "speed/FLOPS": 202584823998564.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05047659948468208, "optim/lr": 0.002829556854674377, "optim/total_tokens": 7040139264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.7956879138946533, "created_at": "2025-01-16T19:29:23.984558+00:00"} {"global_step": 13429, "acc_step": 0, "speed/wps": 12906.597529865587, "speed/FLOPS": 202715849730095.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04630526900291443, "optim/lr": 0.0028295171849045224, "optim/total_tokens": 7040663552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.8160195350646973, "created_at": "2025-01-16T19:29:34.143626+00:00"} {"global_step": 13430, "acc_step": 0, "speed/wps": 12903.401379337363, "speed/FLOPS": 202665649794078.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04610452800989151, "optim/lr": 0.0028294775107968848, "optim/total_tokens": 7041187840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.8294620513916016, "created_at": "2025-01-16T19:29:44.305986+00:00"} {"global_step": 13431, "acc_step": 0, "speed/wps": 12905.751194828576, "speed/FLOPS": 202702556875352.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04199002683162689, "optim/lr": 0.0028294378323515936, "optim/total_tokens": 7041712128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8682708740234375, "created_at": "2025-01-16T19:29:54.473026+00:00"} {"global_step": 13432, "acc_step": 0, "speed/wps": 12907.5756253385, "speed/FLOPS": 202731212063541.25, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04643486812710762, "optim/lr": 0.0028293981495687787, "optim/total_tokens": 7042236416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.7628750801086426, "created_at": "2025-01-16T19:30:04.631318+00:00"} {"global_step": 13433, "acc_step": 0, "speed/wps": 12904.528770879904, "speed/FLOPS": 202683357027452.22, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.039693597704172134, "optim/lr": 0.002829358462448568, "optim/total_tokens": 7042760704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8606109619140625, "created_at": "2025-01-16T19:30:14.791894+00:00"} {"global_step": 13434, "acc_step": 0, "speed/wps": 12905.467993430333, "speed/FLOPS": 202698108808233.75, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04755260422825813, "optim/lr": 0.0028293187709910925, "optim/total_tokens": 7043284992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.963618278503418, "created_at": "2025-01-16T19:30:24.953825+00:00"} {"global_step": 13435, "acc_step": 0, "speed/wps": 12902.986351088208, "speed/FLOPS": 202659131205116.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05819733440876007, "optim/lr": 0.002829279075196482, "optim/total_tokens": 7043809280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 2.9137210845947266, "created_at": "2025-01-16T19:30:35.121212+00:00"} {"global_step": 13436, "acc_step": 0, "speed/wps": 12907.231179721739, "speed/FLOPS": 202725802071812.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04860123246908188, "optim/lr": 0.0028292393750648645, "optim/total_tokens": 7044333568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8247122764587402, "created_at": "2025-01-16T19:30:45.282798+00:00"} {"global_step": 13437, "acc_step": 0, "speed/wps": 12901.949350853189, "speed/FLOPS": 202642843691438.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05234045535326004, "optim/lr": 0.0028291996705963702, "optim/total_tokens": 7044857856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8793482780456543, "created_at": "2025-01-16T19:30:55.445468+00:00"} {"global_step": 13438, "acc_step": 0, "speed/wps": 12902.951190013775, "speed/FLOPS": 202658578952126.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05516180023550987, "optim/lr": 0.0028291599617911293, "optim/total_tokens": 7045382144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313963, "loss/out": 2.8743276596069336, "created_at": "2025-01-16T19:31:05.609435+00:00"} {"global_step": 13439, "acc_step": 0, "speed/wps": 12900.047125575478, "speed/FLOPS": 202612966629520.22, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06516940146684647, "optim/lr": 0.0028291202486492705, "optim/total_tokens": 7045906432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.831864356994629, "created_at": "2025-01-16T19:31:15.774031+00:00"} {"global_step": 13440, "acc_step": 0, "speed/wps": 12909.88766165081, "speed/FLOPS": 202767525770896.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05632110685110092, "optim/lr": 0.002829080531170924, "optim/total_tokens": 7046430720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 2.7976911067962646, "created_at": "2025-01-16T19:31:25.931084+00:00"} {"global_step": 13441, "acc_step": 0, "speed/wps": 12909.386266188512, "speed/FLOPS": 202759650666171.94, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06850330531597137, "optim/lr": 0.002829040809356219, "optim/total_tokens": 7046955008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.865567445755005, "created_at": "2025-01-16T19:31:36.092675+00:00"} {"global_step": 13442, "acc_step": 0, "speed/wps": 12910.998985442677, "speed/FLOPS": 202784980638167.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07821177691221237, "optim/lr": 0.0028290010832052845, "optim/total_tokens": 7047479296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8671317100524902, "created_at": "2025-01-16T19:31:46.256088+00:00"} {"global_step": 13443, "acc_step": 0, "speed/wps": 12910.13951266678, "speed/FLOPS": 202771481437182.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05161578580737114, "optim/lr": 0.0028289613527182514, "optim/total_tokens": 7048003584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376727, "loss/out": 2.9083781242370605, "created_at": "2025-01-16T19:31:56.412361+00:00"} {"global_step": 13444, "acc_step": 0, "speed/wps": 12906.595503501729, "speed/FLOPS": 202715817903266.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0944422110915184, "optim/lr": 0.0028289216178952487, "optim/total_tokens": 7048527872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8855202198028564, "created_at": "2025-01-16T19:32:06.572752+00:00"} {"global_step": 13445, "acc_step": 0, "speed/wps": 12906.258368972445, "speed/FLOPS": 202710522742212.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05931207165122032, "optim/lr": 0.002828881878736406, "optim/total_tokens": 7049052160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.882871627807617, "created_at": "2025-01-16T19:32:16.731942+00:00"} {"global_step": 13446, "acc_step": 0, "speed/wps": 12909.721106943201, "speed/FLOPS": 202764909800343.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07133135199546814, "optim/lr": 0.0028288421352418527, "optim/total_tokens": 7049576448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.772395372390747, "created_at": "2025-01-16T19:32:26.888414+00:00"} {"global_step": 13447, "acc_step": 0, "speed/wps": 12904.525182100253, "speed/FLOPS": 202683300660736.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06819654256105423, "optim/lr": 0.002828802387411719, "optim/total_tokens": 7050100736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8172426223754883, "created_at": "2025-01-16T19:32:37.049034+00:00"} {"global_step": 13448, "acc_step": 0, "speed/wps": 12908.037730044376, "speed/FLOPS": 202738470052945.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05300319567322731, "optim/lr": 0.0028287626352461346, "optim/total_tokens": 7050625024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.656473159790039, "created_at": "2025-01-16T19:32:47.207144+00:00"} {"global_step": 13449, "acc_step": 0, "speed/wps": 12901.076466553677, "speed/FLOPS": 202629133844045.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07659740746021271, "optim/lr": 0.002828722878745229, "optim/total_tokens": 7051149312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.915175199508667, "created_at": "2025-01-16T19:32:57.376641+00:00"} {"global_step": 13450, "acc_step": 0, "speed/wps": 12908.94000108271, "speed/FLOPS": 202752641459452.8, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05591967701911926, "optim/lr": 0.0028286831179091315, "optim/total_tokens": 7051673600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9613418579101562, "created_at": "2025-01-16T19:33:07.537104+00:00"} {"global_step": 13451, "acc_step": 0, "speed/wps": 12904.63880290988, "speed/FLOPS": 202685085231682.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07204627245664597, "optim/lr": 0.0028286433527379726, "optim/total_tokens": 7052197888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.789616823196411, "created_at": "2025-01-16T19:33:17.699931+00:00"} {"global_step": 13452, "acc_step": 0, "speed/wps": 12901.849150774091, "speed/FLOPS": 202641269911506.53, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05438343435525894, "optim/lr": 0.002828603583231881, "optim/total_tokens": 7052722176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.934749126434326, "created_at": "2025-01-16T19:33:27.866057+00:00"} {"global_step": 13453, "acc_step": 0, "speed/wps": 12898.991291880291, "speed/FLOPS": 202596383310469.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08507170528173447, "optim/lr": 0.0028285638093909875, "optim/total_tokens": 7053246464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370586, "loss/out": 2.8826231956481934, "created_at": "2025-01-16T19:33:38.031191+00:00"} {"global_step": 13454, "acc_step": 0, "speed/wps": 12903.51566889837, "speed/FLOPS": 202667444868684.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05230994522571564, "optim/lr": 0.0028285240312154215, "optim/total_tokens": 7053770752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.8561127185821533, "created_at": "2025-01-16T19:33:48.192658+00:00"} {"global_step": 13455, "acc_step": 0, "speed/wps": 12908.142302543754, "speed/FLOPS": 202740112507745.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047491274774074554, "optim/lr": 0.0028284842487053125, "optim/total_tokens": 7054295040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.938798666000366, "created_at": "2025-01-16T19:33:58.350296+00:00"} {"global_step": 13456, "acc_step": 0, "speed/wps": 12902.840717135614, "speed/FLOPS": 202656843823769.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04743170365691185, "optim/lr": 0.0028284444618607908, "optim/total_tokens": 7054819328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 2.8516499996185303, "created_at": "2025-01-16T19:34:08.515407+00:00"} {"global_step": 13457, "acc_step": 0, "speed/wps": 12902.227385038714, "speed/FLOPS": 202647210600380.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04868125915527344, "optim/lr": 0.0028284046706819857, "optim/total_tokens": 7055343616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.846637725830078, "created_at": "2025-01-16T19:34:18.679334+00:00"} {"global_step": 13458, "acc_step": 0, "speed/wps": 12905.684522096257, "speed/FLOPS": 202701509688473.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04070516303181648, "optim/lr": 0.0028283648751690274, "optim/total_tokens": 7055867904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464846, "loss/out": 2.951695442199707, "created_at": "2025-01-16T19:34:28.843524+00:00"} {"global_step": 13459, "acc_step": 0, "speed/wps": 12908.14666050794, "speed/FLOPS": 202740180955561.88, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04903357848525047, "optim/lr": 0.002828325075322046, "optim/total_tokens": 7056392192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8888823986053467, "created_at": "2025-01-16T19:34:39.001209+00:00"} {"global_step": 13460, "acc_step": 0, "speed/wps": 12897.621224250446, "speed/FLOPS": 202574864515672.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05313078686594963, "optim/lr": 0.0028282852711411705, "optim/total_tokens": 7056916480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.793262004852295, "created_at": "2025-01-16T19:34:49.170508+00:00"} {"global_step": 13461, "acc_step": 0, "speed/wps": 12910.656610948648, "speed/FLOPS": 202779603176266.53, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050899650901556015, "optim/lr": 0.0028282454626265318, "optim/total_tokens": 7057440768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408541, "loss/out": 2.832953929901123, "created_at": "2025-01-16T19:34:59.329209+00:00"} {"global_step": 13462, "acc_step": 0, "speed/wps": 12914.819272975317, "speed/FLOPS": 202844983503492.4, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.08350228518247604, "optim/lr": 0.002828205649778259, "optim/total_tokens": 7057965056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.9917070865631104, "created_at": "2025-01-16T19:35:09.481687+00:00"} {"global_step": 13463, "acc_step": 0, "speed/wps": 12903.159858412107, "speed/FLOPS": 202661856376062.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04829362779855728, "optim/lr": 0.0028281658325964814, "optim/total_tokens": 7058489344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.7897729873657227, "created_at": "2025-01-16T19:35:19.646924+00:00"} {"global_step": 13464, "acc_step": 0, "speed/wps": 12909.487691741917, "speed/FLOPS": 202761243693861.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06758469343185425, "optim/lr": 0.002828126011081331, "optim/total_tokens": 7059013632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 2.949821949005127, "created_at": "2025-01-16T19:35:29.804623+00:00"} {"global_step": 13465, "acc_step": 0, "speed/wps": 12912.188737927085, "speed/FLOPS": 202803667335825.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052116770297288895, "optim/lr": 0.002828086185232936, "optim/total_tokens": 7059537920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8576178550720215, "created_at": "2025-01-16T19:35:39.959787+00:00"} {"global_step": 13466, "acc_step": 0, "speed/wps": 12906.452197869954, "speed/FLOPS": 202713567091396.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05112513527274132, "optim/lr": 0.0028280463550514265, "optim/total_tokens": 7060062208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.9219155311584473, "created_at": "2025-01-16T19:35:50.120775+00:00"} {"global_step": 13467, "acc_step": 0, "speed/wps": 12906.23452379834, "speed/FLOPS": 202710148220986.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04489576816558838, "optim/lr": 0.0028280065205369327, "optim/total_tokens": 7060586496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.871727228164673, "created_at": "2025-01-16T19:36:00.280062+00:00"} {"global_step": 13468, "acc_step": 0, "speed/wps": 12907.210908203368, "speed/FLOPS": 202725483679760.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.049551401287317276, "optim/lr": 0.0028279666816895853, "optim/total_tokens": 7061110784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.8785295486450195, "created_at": "2025-01-16T19:36:10.439134+00:00"} {"global_step": 13469, "acc_step": 0, "speed/wps": 12904.474866034174, "speed/FLOPS": 202682510377777.0, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042585764080286026, "optim/lr": 0.0028279268385095133, "optim/total_tokens": 7061635072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.815225124359131, "created_at": "2025-01-16T19:36:20.600772+00:00"} {"global_step": 13470, "acc_step": 0, "speed/wps": 12899.1120258088, "speed/FLOPS": 202598279602722.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05086936801671982, "optim/lr": 0.002827886990996846, "optim/total_tokens": 7062159360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.796311855316162, "created_at": "2025-01-16T19:36:30.767216+00:00"} {"global_step": 13471, "acc_step": 0, "speed/wps": 12900.058142218166, "speed/FLOPS": 202613139661032.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04539340361952782, "optim/lr": 0.0028278471391517163, "optim/total_tokens": 7062683648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.719902992248535, "created_at": "2025-01-16T19:36:40.931351+00:00"} {"global_step": 13472, "acc_step": 0, "speed/wps": 12905.671996404999, "speed/FLOPS": 202701312955280.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047585126012563705, "optim/lr": 0.0028278072829742507, "optim/total_tokens": 7063207936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.804202079772949, "created_at": "2025-01-16T19:36:51.093652+00:00"} {"global_step": 13473, "acc_step": 0, "speed/wps": 12903.55863536901, "speed/FLOPS": 202668119716147.75, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05018176510930061, "optim/lr": 0.0028277674224645812, "optim/total_tokens": 7063732224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.791642665863037, "created_at": "2025-01-16T19:37:01.255426+00:00"} {"global_step": 13474, "acc_step": 0, "speed/wps": 12908.732330891855, "speed/FLOPS": 202749379713736.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04957130551338196, "optim/lr": 0.0028277275576228385, "optim/total_tokens": 7064256512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 2.8873493671417236, "created_at": "2025-01-16T19:37:11.415565+00:00"} {"global_step": 13475, "acc_step": 0, "speed/wps": 12906.524736220508, "speed/FLOPS": 202714706405870.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.056284740567207336, "optim/lr": 0.0028276876884491507, "optim/total_tokens": 7064780800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.765306234359741, "created_at": "2025-01-16T19:37:21.577407+00:00"} {"global_step": 13476, "acc_step": 0, "speed/wps": 12904.386631647036, "speed/FLOPS": 202681124535480.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049542345106601715, "optim/lr": 0.002827647814943649, "optim/total_tokens": 7065305088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.739765167236328, "created_at": "2025-01-16T19:37:31.740520+00:00"} {"global_step": 13477, "acc_step": 0, "speed/wps": 12907.20801751963, "speed/FLOPS": 202725438277599.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04634302854537964, "optim/lr": 0.002827607937106464, "optim/total_tokens": 7065829376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 2.759554147720337, "created_at": "2025-01-16T19:37:41.898920+00:00"} {"global_step": 13478, "acc_step": 0, "speed/wps": 12902.409105805624, "speed/FLOPS": 202650064774735.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06232993304729462, "optim/lr": 0.0028275680549377244, "optim/total_tokens": 7066353664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 2.713721752166748, "created_at": "2025-01-16T19:37:52.069751+00:00"} {"global_step": 13479, "acc_step": 0, "speed/wps": 12902.712917654555, "speed/FLOPS": 202654836557304.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04499531164765358, "optim/lr": 0.0028275281684375614, "optim/total_tokens": 7066877952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9226338863372803, "created_at": "2025-01-16T19:38:02.232533+00:00"} {"global_step": 13480, "acc_step": 0, "speed/wps": 12906.810060144531, "speed/FLOPS": 202719187810173.16, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047004345804452896, "optim/lr": 0.0028274882776061047, "optim/total_tokens": 7067402240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 2.845428228378296, "created_at": "2025-01-16T19:38:12.396516+00:00"} {"global_step": 13481, "acc_step": 0, "speed/wps": 12907.167803768596, "speed/FLOPS": 202724806665380.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045700620859861374, "optim/lr": 0.002827448382443485, "optim/total_tokens": 7067926528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.916391134262085, "created_at": "2025-01-16T19:38:22.560195+00:00"} {"global_step": 13482, "acc_step": 0, "speed/wps": 12909.698229483607, "speed/FLOPS": 202764550478403.9, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05999009311199188, "optim/lr": 0.0028274084829498316, "optim/total_tokens": 7068450816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8192577362060547, "created_at": "2025-01-16T19:38:32.720898+00:00"} {"global_step": 13483, "acc_step": 0, "speed/wps": 12907.843499549406, "speed/FLOPS": 202735419396121.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04392179846763611, "optim/lr": 0.0028273685791252753, "optim/total_tokens": 7068975104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.830326795578003, "created_at": "2025-01-16T19:38:42.883224+00:00"} {"global_step": 13484, "acc_step": 0, "speed/wps": 12908.71986085245, "speed/FLOPS": 202749183854632.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045629218220710754, "optim/lr": 0.0028273286709699463, "optim/total_tokens": 7069499392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8996481895446777, "created_at": "2025-01-16T19:38:53.044009+00:00"} {"global_step": 13485, "acc_step": 0, "speed/wps": 12902.628141306539, "speed/FLOPS": 202653505028263.8, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04275747016072273, "optim/lr": 0.0028272887584839743, "optim/total_tokens": 7070023680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 2.7839913368225098, "created_at": "2025-01-16T19:39:03.210664+00:00"} {"global_step": 13486, "acc_step": 0, "speed/wps": 12903.264977705161, "speed/FLOPS": 202663507419009.1, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05596264451742172, "optim/lr": 0.00282724884166749, "optim/total_tokens": 7070547968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.838106155395508, "created_at": "2025-01-16T19:39:13.375356+00:00"} {"global_step": 13487, "acc_step": 0, "speed/wps": 12909.35023067808, "speed/FLOPS": 202759084678962.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.041800789535045624, "optim/lr": 0.0028272089205206236, "optim/total_tokens": 7071072256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.870708703994751, "created_at": "2025-01-16T19:39:23.534785+00:00"} {"global_step": 13488, "acc_step": 0, "speed/wps": 12901.984062925989, "speed/FLOPS": 202643388892240.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06026971712708473, "optim/lr": 0.0028271689950435048, "optim/total_tokens": 7071596544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8094940185546875, "created_at": "2025-01-16T19:39:33.698289+00:00"} {"global_step": 13489, "acc_step": 0, "speed/wps": 12908.449825051557, "speed/FLOPS": 202744942571311.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05173134058713913, "optim/lr": 0.002827129065236265, "optim/total_tokens": 7072120832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 2.8970749378204346, "created_at": "2025-01-16T19:39:43.856654+00:00"} {"global_step": 13490, "acc_step": 0, "speed/wps": 12902.708044060806, "speed/FLOPS": 202654760010817.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04556490480899811, "optim/lr": 0.002827089131099033, "optim/total_tokens": 7072645120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422503, "loss/out": 2.7404236793518066, "created_at": "2025-01-16T19:39:54.022079+00:00"} {"global_step": 13491, "acc_step": 0, "speed/wps": 12902.967339738934, "speed/FLOPS": 202658832605752.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04739094153046608, "optim/lr": 0.0028270491926319402, "optim/total_tokens": 7073169408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.955385446548462, "created_at": "2025-01-16T19:40:04.183769+00:00"} {"global_step": 13492, "acc_step": 0, "speed/wps": 12909.052195874116, "speed/FLOPS": 202754403632823.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05541646108031273, "optim/lr": 0.002827009249835117, "optim/total_tokens": 7073693696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8781816959381104, "created_at": "2025-01-16T19:40:14.344950+00:00"} {"global_step": 13493, "acc_step": 0, "speed/wps": 12904.26485831863, "speed/FLOPS": 202679211918023.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06385302543640137, "optim/lr": 0.002826969302708693, "optim/total_tokens": 7074217984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315630, "loss/out": 2.8840548992156982, "created_at": "2025-01-16T19:40:24.508790+00:00"} {"global_step": 13494, "acc_step": 0, "speed/wps": 12908.318741474171, "speed/FLOPS": 202742883723600.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055555664002895355, "optim/lr": 0.0028269293512527983, "optim/total_tokens": 7074742272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.845337390899658, "created_at": "2025-01-16T19:40:34.669860+00:00"} {"global_step": 13495, "acc_step": 0, "speed/wps": 12901.392453897848, "speed/FLOPS": 202634096859497.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05812406912446022, "optim/lr": 0.0028268893954675645, "optim/total_tokens": 7075266560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.9069581031799316, "created_at": "2025-01-16T19:40:44.834054+00:00"} {"global_step": 13496, "acc_step": 0, "speed/wps": 12905.430949041462, "speed/FLOPS": 202697526975202.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04675241932272911, "optim/lr": 0.002826849435353121, "optim/total_tokens": 7075790848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 2.879657506942749, "created_at": "2025-01-16T19:40:54.994562+00:00"} {"global_step": 13497, "acc_step": 0, "speed/wps": 12901.075537021095, "speed/FLOPS": 202629119244459.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0572974868118763, "optim/lr": 0.002826809470909598, "optim/total_tokens": 7076315136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 2.8296432495117188, "created_at": "2025-01-16T19:41:05.157800+00:00"} {"global_step": 13498, "acc_step": 0, "speed/wps": 12907.661040865993, "speed/FLOPS": 202732553631773.75, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05092485994100571, "optim/lr": 0.002826769502137127, "optim/total_tokens": 7076839424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 2.7644262313842773, "created_at": "2025-01-16T19:41:15.318212+00:00"} {"global_step": 13499, "acc_step": 0, "speed/wps": 12905.3158646308, "speed/FLOPS": 202695719416390.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04967409744858742, "optim/lr": 0.0028267295290358373, "optim/total_tokens": 7077363712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8279173374176025, "created_at": "2025-01-16T19:41:25.478308+00:00"} {"global_step": 13500, "acc_step": 0, "speed/wps": 12902.622783972749, "speed/FLOPS": 202653420883974.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047770366072654724, "optim/lr": 0.00282668955160586, "optim/total_tokens": 7077888000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.7661991119384766, "created_at": "2025-01-16T19:41:35.644627+00:00"} {"global_step": 13501, "acc_step": 0, "speed/wps": 12903.673778355807, "speed/FLOPS": 202669928194978.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05203499272465706, "optim/lr": 0.0028266495698473252, "optim/total_tokens": 7078412288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 2.881375551223755, "created_at": "2025-01-16T19:41:45.810566+00:00"} {"global_step": 13502, "acc_step": 0, "speed/wps": 12902.899347208651, "speed/FLOPS": 202657764689632.44, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06202751398086548, "optim/lr": 0.0028266095837603633, "optim/total_tokens": 7078936576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8810577392578125, "created_at": "2025-01-16T19:41:55.972742+00:00"} {"global_step": 13503, "acc_step": 0, "speed/wps": 12907.723391730397, "speed/FLOPS": 202733532937778.28, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046011779457330704, "optim/lr": 0.002826569593345105, "optim/total_tokens": 7079460864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.8915677070617676, "created_at": "2025-01-16T19:42:06.133173+00:00"} {"global_step": 13504, "acc_step": 0, "speed/wps": 12900.344618189578, "speed/FLOPS": 202617639159822.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0648377537727356, "optim/lr": 0.0028265295986016803, "optim/total_tokens": 7079985152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 2.704735279083252, "created_at": "2025-01-16T19:42:16.297030+00:00"} {"global_step": 13505, "acc_step": 0, "speed/wps": 12904.023416284328, "speed/FLOPS": 202675419739098.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049165088683366776, "optim/lr": 0.0028264895995302203, "optim/total_tokens": 7080509440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 2.889913558959961, "created_at": "2025-01-16T19:42:26.461773+00:00"} {"global_step": 13506, "acc_step": 0, "speed/wps": 12902.083958756422, "speed/FLOPS": 202644957893529.03, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05887778848409653, "optim/lr": 0.0028264495961308556, "optim/total_tokens": 7081033728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.9037609100341797, "created_at": "2025-01-16T19:42:36.625609+00:00"} {"global_step": 13507, "acc_step": 0, "speed/wps": 12904.720742144445, "speed/FLOPS": 202686372199958.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04799830913543701, "optim/lr": 0.0028264095884037157, "optim/total_tokens": 7081558016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.657735824584961, "created_at": "2025-01-16T19:42:46.788482+00:00"} {"global_step": 13508, "acc_step": 0, "speed/wps": 12901.352580320745, "speed/FLOPS": 202633470590177.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05386364087462425, "optim/lr": 0.002826369576348932, "optim/total_tokens": 7082082304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.9875998497009277, "created_at": "2025-01-16T19:42:56.953914+00:00"} {"global_step": 13509, "acc_step": 0, "speed/wps": 12907.547203466738, "speed/FLOPS": 202730765658989.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04282552748918533, "optim/lr": 0.002826329559966635, "optim/total_tokens": 7082606592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326440, "loss/out": 2.817424774169922, "created_at": "2025-01-16T19:43:07.112213+00:00"} {"global_step": 13510, "acc_step": 0, "speed/wps": 12900.268611358759, "speed/FLOPS": 202616445368100.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04550430178642273, "optim/lr": 0.0028262895392569556, "optim/total_tokens": 7083130880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9529757499694824, "created_at": "2025-01-16T19:43:17.278460+00:00"} {"global_step": 13511, "acc_step": 0, "speed/wps": 12903.022264722053, "speed/FLOPS": 202659695278086.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05689656734466553, "optim/lr": 0.0028262495142200228, "optim/total_tokens": 7083655168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380023, "loss/out": 2.7790021896362305, "created_at": "2025-01-16T19:43:27.450573+00:00"} {"global_step": 13512, "acc_step": 0, "speed/wps": 12903.481497520017, "speed/FLOPS": 202666908160230.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0724758431315422, "optim/lr": 0.002826209484855969, "optim/total_tokens": 7084179456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.9649863243103027, "created_at": "2025-01-16T19:43:37.613444+00:00"} {"global_step": 13513, "acc_step": 0, "speed/wps": 12904.740573943223, "speed/FLOPS": 202686683685609.56, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05335752293467522, "optim/lr": 0.002826169451164924, "optim/total_tokens": 7084703744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.858680009841919, "created_at": "2025-01-16T19:43:47.777116+00:00"} {"global_step": 13514, "acc_step": 0, "speed/wps": 12906.118954008105, "speed/FLOPS": 202708333038621.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05560347065329552, "optim/lr": 0.002826129413147018, "optim/total_tokens": 7085228032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428074, "loss/out": 2.8363852500915527, "created_at": "2025-01-16T19:43:57.939072+00:00"} {"global_step": 13515, "acc_step": 0, "speed/wps": 12898.097737426244, "speed/FLOPS": 202582348809896.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05551014095544815, "optim/lr": 0.0028260893708023827, "optim/total_tokens": 7085752320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9247281551361084, "created_at": "2025-01-16T19:44:08.104977+00:00"} {"global_step": 13516, "acc_step": 0, "speed/wps": 12908.999894344637, "speed/FLOPS": 202753582165433.34, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05172508582472801, "optim/lr": 0.002826049324131148, "optim/total_tokens": 7086276608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.715355396270752, "created_at": "2025-01-16T19:44:18.264731+00:00"} {"global_step": 13517, "acc_step": 0, "speed/wps": 12909.35030646233, "speed/FLOPS": 202759085869258.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05299926921725273, "optim/lr": 0.002826009273133445, "optim/total_tokens": 7086800896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 2.8408331871032715, "created_at": "2025-01-16T19:44:28.426235+00:00"} {"global_step": 13518, "acc_step": 0, "speed/wps": 12905.381057853081, "speed/FLOPS": 202696743365531.0, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061300378292798996, "optim/lr": 0.0028259692178094035, "optim/total_tokens": 7087325184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.7490851879119873, "created_at": "2025-01-16T19:44:38.590269+00:00"} {"global_step": 13519, "acc_step": 0, "speed/wps": 12902.831225272283, "speed/FLOPS": 202656694741012.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04858603700995445, "optim/lr": 0.0028259291581591555, "optim/total_tokens": 7087849472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387505, "loss/out": 2.9265546798706055, "created_at": "2025-01-16T19:44:48.752175+00:00"} {"global_step": 13520, "acc_step": 0, "speed/wps": 12903.081962309185, "speed/FLOPS": 202660632910725.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05260805785655975, "optim/lr": 0.00282588909418283, "optim/total_tokens": 7088373760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.9169788360595703, "created_at": "2025-01-16T19:44:58.916486+00:00"} {"global_step": 13521, "acc_step": 0, "speed/wps": 12902.769938974956, "speed/FLOPS": 202655732155496.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05647794157266617, "optim/lr": 0.0028258490258805595, "optim/total_tokens": 7088898048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.921143054962158, "created_at": "2025-01-16T19:45:09.078590+00:00"} {"global_step": 13522, "acc_step": 0, "speed/wps": 12905.569065114694, "speed/FLOPS": 202699696277923.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04728199914097786, "optim/lr": 0.002825808953252474, "optim/total_tokens": 7089422336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 2.8988733291625977, "created_at": "2025-01-16T19:45:19.239682+00:00"} {"global_step": 13523, "acc_step": 0, "speed/wps": 12907.2319965025, "speed/FLOPS": 202725814900476.38, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056299977004528046, "optim/lr": 0.0028257688762987035, "optim/total_tokens": 7089946624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.863253593444824, "created_at": "2025-01-16T19:45:29.403796+00:00"} {"global_step": 13524, "acc_step": 0, "speed/wps": 12900.669093403683, "speed/FLOPS": 202622735488935.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054542552679777145, "optim/lr": 0.00282572879501938, "optim/total_tokens": 7090470912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.843052864074707, "created_at": "2025-01-16T19:45:39.570859+00:00"} {"global_step": 13525, "acc_step": 0, "speed/wps": 12906.95588551242, "speed/FLOPS": 202721478197960.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04686807468533516, "optim/lr": 0.0028256887094146336, "optim/total_tokens": 7090995200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 2.824448823928833, "created_at": "2025-01-16T19:45:49.729664+00:00"} {"global_step": 13526, "acc_step": 0, "speed/wps": 12903.880446981433, "speed/FLOPS": 202673174209736.34, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05315323919057846, "optim/lr": 0.002825648619484595, "optim/total_tokens": 7091519488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8022983074188232, "created_at": "2025-01-16T19:45:59.892648+00:00"} {"global_step": 13527, "acc_step": 0, "speed/wps": 12906.798385687745, "speed/FLOPS": 202719004446786.94, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052412621676921844, "optim/lr": 0.0028256085252293954, "optim/total_tokens": 7092043776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 2.8749611377716064, "created_at": "2025-01-16T19:46:10.058045+00:00"} {"global_step": 13528, "acc_step": 0, "speed/wps": 12905.753349925457, "speed/FLOPS": 202702590724110.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05537641793489456, "optim/lr": 0.0028255684266491653, "optim/total_tokens": 7092568064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.874472141265869, "created_at": "2025-01-16T19:46:20.220496+00:00"} {"global_step": 13529, "acc_step": 0, "speed/wps": 12904.357341140523, "speed/FLOPS": 202680664487826.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05014340952038765, "optim/lr": 0.002825528323744036, "optim/total_tokens": 7093092352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.7425546646118164, "created_at": "2025-01-16T19:46:30.381174+00:00"} {"global_step": 13530, "acc_step": 0, "speed/wps": 12912.298987104567, "speed/FLOPS": 202805398950656.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0552959181368351, "optim/lr": 0.002825488216514137, "optim/total_tokens": 7093616640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419023, "loss/out": 2.8948612213134766, "created_at": "2025-01-16T19:46:40.540970+00:00"} {"global_step": 13531, "acc_step": 0, "speed/wps": 12902.261158258709, "speed/FLOPS": 202647741055210.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48434510827064514, "optim/lr": 0.002825448104959601, "optim/total_tokens": 7094140928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 2.8629252910614014, "created_at": "2025-01-16T19:46:50.704580+00:00"} {"global_step": 13532, "acc_step": 0, "speed/wps": 12907.297501472363, "speed/FLOPS": 202726843746041.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25324419140815735, "optim/lr": 0.0028254079890805577, "optim/total_tokens": 7094665216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 2.854471206665039, "created_at": "2025-01-16T19:47:00.863657+00:00"} {"global_step": 13533, "acc_step": 0, "speed/wps": 12901.494072022297, "speed/FLOPS": 202635692911780.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.10948394984006882, "optim/lr": 0.0028253678688771387, "optim/total_tokens": 7095189504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 3.000805616378784, "created_at": "2025-01-16T19:47:11.026738+00:00"} {"global_step": 13534, "acc_step": 0, "speed/wps": 12902.490062150482, "speed/FLOPS": 202651336305377.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08885274827480316, "optim/lr": 0.0028253277443494743, "optim/total_tokens": 7095713792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.9825286865234375, "created_at": "2025-01-16T19:47:21.191594+00:00"} {"global_step": 13535, "acc_step": 0, "speed/wps": 12902.56998685228, "speed/FLOPS": 202652591632647.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0731443241238594, "optim/lr": 0.0028252876154976955, "optim/total_tokens": 7096238080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.72939395904541, "created_at": "2025-01-16T19:47:31.353972+00:00"} {"global_step": 13536, "acc_step": 0, "speed/wps": 12907.276274460613, "speed/FLOPS": 202726510346652.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14916592836380005, "optim/lr": 0.002825247482321933, "optim/total_tokens": 7096762368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.800682306289673, "created_at": "2025-01-16T19:47:41.517502+00:00"} {"global_step": 13537, "acc_step": 0, "speed/wps": 12907.065994447936, "speed/FLOPS": 202723207610098.72, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08483339846134186, "optim/lr": 0.002825207344822319, "optim/total_tokens": 7097286656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.868626594543457, "created_at": "2025-01-16T19:47:51.679575+00:00"} {"global_step": 13538, "acc_step": 0, "speed/wps": 12910.727378541402, "speed/FLOPS": 202780714678555.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055684901773929596, "optim/lr": 0.0028251672029989826, "optim/total_tokens": 7097810944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.764803886413574, "created_at": "2025-01-16T19:48:01.838838+00:00"} {"global_step": 13539, "acc_step": 0, "speed/wps": 12902.811156841015, "speed/FLOPS": 202656379538722.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052325330674648285, "optim/lr": 0.0028251270568520562, "optim/total_tokens": 7098335232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.6563820838928223, "created_at": "2025-01-16T19:48:12.000944+00:00"} {"global_step": 13540, "acc_step": 0, "speed/wps": 12905.935396078157, "speed/FLOPS": 202705450009096.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04848221689462662, "optim/lr": 0.00282508690638167, "optim/total_tokens": 7098859520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.886539936065674, "created_at": "2025-01-16T19:48:22.163406+00:00"} {"global_step": 13541, "acc_step": 0, "speed/wps": 12899.068594768014, "speed/FLOPS": 202597597458546.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04755052924156189, "optim/lr": 0.002825046751587956, "optim/total_tokens": 7099383808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 2.9036669731140137, "created_at": "2025-01-16T19:48:32.330595+00:00"} {"global_step": 13542, "acc_step": 0, "speed/wps": 12909.168823499953, "speed/FLOPS": 202756235429950.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.059739433228969574, "optim/lr": 0.0028250065924710433, "optim/total_tokens": 7099908096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.052664279937744, "created_at": "2025-01-16T19:48:42.491450+00:00"} {"global_step": 13543, "acc_step": 0, "speed/wps": 12908.68591765037, "speed/FLOPS": 202748650730000.7, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04976785555481911, "optim/lr": 0.0028249664290310654, "optim/total_tokens": 7100432384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.87640380859375, "created_at": "2025-01-16T19:48:52.649702+00:00"} {"global_step": 13544, "acc_step": 0, "speed/wps": 12908.541352014045, "speed/FLOPS": 202746380128028.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04671361669898033, "optim/lr": 0.002824926261268151, "optim/total_tokens": 7100956672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.86618971824646, "created_at": "2025-01-16T19:49:02.807557+00:00"} {"global_step": 13545, "acc_step": 0, "speed/wps": 12908.912005379736, "speed/FLOPS": 202752201748467.47, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047464463859796524, "optim/lr": 0.0028248860891824327, "optim/total_tokens": 7101480960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.803382396697998, "created_at": "2025-01-16T19:49:12.965572+00:00"} {"global_step": 13546, "acc_step": 0, "speed/wps": 12905.025208385865, "speed/FLOPS": 202691154260659.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0495426207780838, "optim/lr": 0.002824845912774041, "optim/total_tokens": 7102005248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 2.827543020248413, "created_at": "2025-01-16T19:49:23.127245+00:00"} {"global_step": 13547, "acc_step": 0, "speed/wps": 12904.647766146245, "speed/FLOPS": 202685226011626.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04799647629261017, "optim/lr": 0.0028248057320431072, "optim/total_tokens": 7102529536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8216333389282227, "created_at": "2025-01-16T19:49:33.289266+00:00"} {"global_step": 13548, "acc_step": 0, "speed/wps": 12905.311993766962, "speed/FLOPS": 202695658619154.5, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04988083988428116, "optim/lr": 0.002824765546989762, "optim/total_tokens": 7103053824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.827691078186035, "created_at": "2025-01-16T19:49:43.452100+00:00"} {"global_step": 13549, "acc_step": 0, "speed/wps": 12895.847202106186, "speed/FLOPS": 202547001060135.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04655800014734268, "optim/lr": 0.002824725357614137, "optim/total_tokens": 7103578112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.7724192142486572, "created_at": "2025-01-16T19:49:53.622860+00:00"} {"global_step": 13550, "acc_step": 0, "speed/wps": 12903.149577023765, "speed/FLOPS": 202661694892730.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.042038414627313614, "optim/lr": 0.0028246851639163635, "optim/total_tokens": 7104102400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.73014497756958, "created_at": "2025-01-16T19:50:03.789738+00:00"} {"global_step": 13551, "acc_step": 0, "speed/wps": 12917.41334594911, "speed/FLOPS": 202885726984175.62, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04694141820073128, "optim/lr": 0.0028246449658965713, "optim/total_tokens": 7104626688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 2.77022647857666, "created_at": "2025-01-16T19:50:13.943706+00:00"} {"global_step": 13552, "acc_step": 0, "speed/wps": 12913.632493279198, "speed/FLOPS": 202826343497558.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042355846613645554, "optim/lr": 0.002824604763554893, "optim/total_tokens": 7105150976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8983685970306396, "created_at": "2025-01-16T19:50:24.098347+00:00"} {"global_step": 13553, "acc_step": 0, "speed/wps": 12908.416187546643, "speed/FLOPS": 202744414248073.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054245129227638245, "optim/lr": 0.0028245645568914595, "optim/total_tokens": 7105675264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.976419687271118, "created_at": "2025-01-16T19:50:34.259297+00:00"} {"global_step": 13554, "acc_step": 0, "speed/wps": 12912.115281773384, "speed/FLOPS": 202802513605992.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05922522395849228, "optim/lr": 0.0028245243459064013, "optim/total_tokens": 7106199552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8193507194519043, "created_at": "2025-01-16T19:50:44.416953+00:00"} {"global_step": 13555, "acc_step": 0, "speed/wps": 12909.257851525175, "speed/FLOPS": 202757633737424.0, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06981494277715683, "optim/lr": 0.0028244841305998503, "optim/total_tokens": 7106723840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.744515895843506, "created_at": "2025-01-16T19:50:54.577734+00:00"} {"global_step": 13556, "acc_step": 0, "speed/wps": 12914.553407031875, "speed/FLOPS": 202840807713512.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04687695577740669, "optim/lr": 0.0028244439109719375, "optim/total_tokens": 7107248128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333415, "loss/out": 2.829483985900879, "created_at": "2025-01-16T19:51:04.732073+00:00"} {"global_step": 13557, "acc_step": 0, "speed/wps": 12911.242268674947, "speed/FLOPS": 202788801735636.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06050463765859604, "optim/lr": 0.0028244036870227938, "optim/total_tokens": 7107772416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 2.9741761684417725, "created_at": "2025-01-16T19:51:14.890773+00:00"} {"global_step": 13558, "acc_step": 0, "speed/wps": 12908.696895772659, "speed/FLOPS": 202748823156496.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04516289010643959, "optim/lr": 0.002824363458752551, "optim/total_tokens": 7108296704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 2.749643087387085, "created_at": "2025-01-16T19:51:25.053421+00:00"} {"global_step": 13559, "acc_step": 0, "speed/wps": 12904.68126805335, "speed/FLOPS": 202685752205114.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05342119187116623, "optim/lr": 0.00282432322616134, "optim/total_tokens": 7108820992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.896254539489746, "created_at": "2025-01-16T19:51:35.219854+00:00"} {"global_step": 13560, "acc_step": 0, "speed/wps": 12910.407034974452, "speed/FLOPS": 202775683242638.44, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04638189449906349, "optim/lr": 0.002824282989249292, "optim/total_tokens": 7109345280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.930870532989502, "created_at": "2025-01-16T19:51:45.376526+00:00"} {"global_step": 13561, "acc_step": 0, "speed/wps": 12913.49203621561, "speed/FLOPS": 202824137426366.78, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04625493288040161, "optim/lr": 0.0028242427480165383, "optim/total_tokens": 7109869568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.86185359954834, "created_at": "2025-01-16T19:51:55.532904+00:00"} {"global_step": 13562, "acc_step": 0, "speed/wps": 12910.713144659896, "speed/FLOPS": 202780491115887.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055281657725572586, "optim/lr": 0.002824202502463211, "optim/total_tokens": 7110393856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.829630136489868, "created_at": "2025-01-16T19:52:05.695089+00:00"} {"global_step": 13563, "acc_step": 0, "speed/wps": 12913.748745111827, "speed/FLOPS": 202828169392336.84, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043145813047885895, "optim/lr": 0.0028241622525894397, "optim/total_tokens": 7110918144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.807659149169922, "created_at": "2025-01-16T19:52:15.852580+00:00"} {"global_step": 13564, "acc_step": 0, "speed/wps": 12912.133262206411, "speed/FLOPS": 202802796013401.16, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054680924862623215, "optim/lr": 0.0028241219983953575, "optim/total_tokens": 7111442432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.727397918701172, "created_at": "2025-01-16T19:52:26.007496+00:00"} {"global_step": 13565, "acc_step": 0, "speed/wps": 12906.39209863075, "speed/FLOPS": 202712623150259.9, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05450674146413803, "optim/lr": 0.0028240817398810947, "optim/total_tokens": 7111966720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.9199061393737793, "created_at": "2025-01-16T19:52:36.170366+00:00"} {"global_step": 13566, "acc_step": 0, "speed/wps": 12916.373260950064, "speed/FLOPS": 202869391020039.06, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04111094027757645, "optim/lr": 0.002824041477046783, "optim/total_tokens": 7112491008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.8234505653381348, "created_at": "2025-01-16T19:52:46.326379+00:00"} {"global_step": 13567, "acc_step": 0, "speed/wps": 12907.919238139702, "speed/FLOPS": 202736608974757.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04812482371926308, "optim/lr": 0.0028240012098925535, "optim/total_tokens": 7113015296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 2.7792670726776123, "created_at": "2025-01-16T19:52:56.484281+00:00"} {"global_step": 13568, "acc_step": 0, "speed/wps": 12909.281472233079, "speed/FLOPS": 202758004733099.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.052825599908828735, "optim/lr": 0.0028239609384185377, "optim/total_tokens": 7113539584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.047330379486084, "created_at": "2025-01-16T19:53:06.642475+00:00"} {"global_step": 13569, "acc_step": 0, "speed/wps": 12908.344166628238, "speed/FLOPS": 202743283060582.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03956253454089165, "optim/lr": 0.002823920662624867, "optim/total_tokens": 7114063872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.8480255603790283, "created_at": "2025-01-16T19:53:16.803878+00:00"} {"global_step": 13570, "acc_step": 0, "speed/wps": 12910.927076646423, "speed/FLOPS": 202783851211708.88, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05702798441052437, "optim/lr": 0.0028238803825116738, "optim/total_tokens": 7114588160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.744685649871826, "created_at": "2025-01-16T19:53:26.959522+00:00"} {"global_step": 13571, "acc_step": 0, "speed/wps": 12905.972652844217, "speed/FLOPS": 202706035177803.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0597766637802124, "optim/lr": 0.0028238400980790876, "optim/total_tokens": 7115112448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.766781806945801, "created_at": "2025-01-16T19:53:37.123986+00:00"} {"global_step": 13572, "acc_step": 0, "speed/wps": 12908.026960223295, "speed/FLOPS": 202738300898106.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058569829910993576, "optim/lr": 0.002823799809327241, "optim/total_tokens": 7115636736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 2.6848514080047607, "created_at": "2025-01-16T19:53:47.318951+00:00"} {"global_step": 13573, "acc_step": 0, "speed/wps": 12902.041172427531, "speed/FLOPS": 202644285875439.22, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048055339604616165, "optim/lr": 0.0028237595162562655, "optim/total_tokens": 7116161024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8632869720458984, "created_at": "2025-01-16T19:53:57.483786+00:00"} {"global_step": 13574, "acc_step": 0, "speed/wps": 12908.909562688183, "speed/FLOPS": 202752163382640.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053390275686979294, "optim/lr": 0.0028237192188662924, "optim/total_tokens": 7116685312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447274, "loss/out": 2.8484883308410645, "created_at": "2025-01-16T19:54:07.645430+00:00"} {"global_step": 13575, "acc_step": 0, "speed/wps": 12904.314668701567, "speed/FLOPS": 202679994258534.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04080510511994362, "optim/lr": 0.002823678917157453, "optim/total_tokens": 7117209600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.899749279022217, "created_at": "2025-01-16T19:54:17.806948+00:00"} {"global_step": 13576, "acc_step": 0, "speed/wps": 12910.641761352928, "speed/FLOPS": 202779369942960.25, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04861348867416382, "optim/lr": 0.0028236386111298783, "optim/total_tokens": 7117733888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8495004177093506, "created_at": "2025-01-16T19:54:27.962679+00:00"} {"global_step": 13577, "acc_step": 0, "speed/wps": 12908.266258826803, "speed/FLOPS": 202742059411504.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04901224374771118, "optim/lr": 0.002823598300783701, "optim/total_tokens": 7118258176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 2.8356359004974365, "created_at": "2025-01-16T19:54:38.122466+00:00"} {"global_step": 13578, "acc_step": 0, "speed/wps": 12910.75412548485, "speed/FLOPS": 202781134776056.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04589581862092018, "optim/lr": 0.002823557986119052, "optim/total_tokens": 7118782464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.895101547241211, "created_at": "2025-01-16T19:54:48.281095+00:00"} {"global_step": 13579, "acc_step": 0, "speed/wps": 12909.39654922924, "speed/FLOPS": 202759812175456.94, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0603112131357193, "optim/lr": 0.0028235176671360628, "optim/total_tokens": 7119306752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 2.7988805770874023, "created_at": "2025-01-16T19:54:58.438042+00:00"} {"global_step": 13580, "acc_step": 0, "speed/wps": 12906.26905654371, "speed/FLOPS": 202710690605205.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05960813909769058, "optim/lr": 0.002823477343834865, "optim/total_tokens": 7119831040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.987643241882324, "created_at": "2025-01-16T19:55:08.599134+00:00"} {"global_step": 13581, "acc_step": 0, "speed/wps": 12911.56730773159, "speed/FLOPS": 202793906920671.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04892539232969284, "optim/lr": 0.0028234370162155908, "optim/total_tokens": 7120355328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8544578552246094, "created_at": "2025-01-16T19:55:18.758240+00:00"} {"global_step": 13582, "acc_step": 0, "speed/wps": 12908.434812409352, "speed/FLOPS": 202744706777135.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06613186001777649, "optim/lr": 0.0028233966842783702, "optim/total_tokens": 7120879616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.9427218437194824, "created_at": "2025-01-16T19:55:28.917709+00:00"} {"global_step": 13583, "acc_step": 0, "speed/wps": 12909.7952665166, "speed/FLOPS": 202766074578349.9, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06768251955509186, "optim/lr": 0.002823356348023336, "optim/total_tokens": 7121403904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.9537689685821533, "created_at": "2025-01-16T19:55:39.077993+00:00"} {"global_step": 13584, "acc_step": 0, "speed/wps": 12906.634893430513, "speed/FLOPS": 202716436576225.22, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05580366775393486, "optim/lr": 0.0028233160074506197, "optim/total_tokens": 7121928192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306165, "loss/out": 2.7580480575561523, "created_at": "2025-01-16T19:55:49.239337+00:00"} {"global_step": 13585, "acc_step": 0, "speed/wps": 12908.409951645368, "speed/FLOPS": 202744316304675.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.10022848099470139, "optim/lr": 0.002823275662560352, "optim/total_tokens": 7122452480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8670575618743896, "created_at": "2025-01-16T19:55:59.398309+00:00"} {"global_step": 13586, "acc_step": 0, "speed/wps": 12908.025776337276, "speed/FLOPS": 202738282303549.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04976191744208336, "optim/lr": 0.0028232353133526666, "optim/total_tokens": 7122976768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464032, "loss/out": 2.844212293624878, "created_at": "2025-01-16T19:56:09.560978+00:00"} {"global_step": 13587, "acc_step": 0, "speed/wps": 12903.722793867048, "speed/FLOPS": 202670698050936.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.07187813520431519, "optim/lr": 0.0028231949598276932, "optim/total_tokens": 7123501056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.812822103500366, "created_at": "2025-01-16T19:56:19.724098+00:00"} {"global_step": 13588, "acc_step": 0, "speed/wps": 12908.50519571924, "speed/FLOPS": 202745812243734.34, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04992610216140747, "optim/lr": 0.0028231546019855643, "optim/total_tokens": 7124025344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.7957568168640137, "created_at": "2025-01-16T19:56:29.882145+00:00"} {"global_step": 13589, "acc_step": 0, "speed/wps": 12905.426089961922, "speed/FLOPS": 202697450656680.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05499985069036484, "optim/lr": 0.002823114239826411, "optim/total_tokens": 7124549632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.8106276988983154, "created_at": "2025-01-16T19:56:40.042971+00:00"} {"global_step": 13590, "acc_step": 0, "speed/wps": 12910.573443940886, "speed/FLOPS": 202778296924126.16, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050608906894922256, "optim/lr": 0.0028230738733503657, "optim/total_tokens": 7125073920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8947982788085938, "created_at": "2025-01-16T19:56:50.199406+00:00"} {"global_step": 13591, "acc_step": 0, "speed/wps": 12911.682108023508, "speed/FLOPS": 202795710017007.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06873725354671478, "optim/lr": 0.0028230335025575593, "optim/total_tokens": 7125598208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 2.858015537261963, "created_at": "2025-01-16T19:57:00.358115+00:00"} {"global_step": 13592, "acc_step": 0, "speed/wps": 12904.295181340296, "speed/FLOPS": 202679688182747.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1046883836388588, "optim/lr": 0.0028229931274481247, "optim/total_tokens": 7126122496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.8261382579803467, "created_at": "2025-01-16T19:57:10.520317+00:00"} {"global_step": 13593, "acc_step": 0, "speed/wps": 12904.45587869582, "speed/FLOPS": 202682212155536.84, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051557525992393494, "optim/lr": 0.002822952748022193, "optim/total_tokens": 7126646784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8814573287963867, "created_at": "2025-01-16T19:57:20.683643+00:00"} {"global_step": 13594, "acc_step": 0, "speed/wps": 12906.815884952372, "speed/FLOPS": 202719279296784.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09390731155872345, "optim/lr": 0.0028229123642798943, "optim/total_tokens": 7127171072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485284, "loss/out": 2.9284324645996094, "created_at": "2025-01-16T19:57:30.849246+00:00"} {"global_step": 13595, "acc_step": 0, "speed/wps": 12909.464805636391, "speed/FLOPS": 202760884236125.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05425126850605011, "optim/lr": 0.002822871976221363, "optim/total_tokens": 7127695360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.811582088470459, "created_at": "2025-01-16T19:57:41.009425+00:00"} {"global_step": 13596, "acc_step": 0, "speed/wps": 12909.542409366035, "speed/FLOPS": 202762103109338.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08750685304403305, "optim/lr": 0.0028228315838467305, "optim/total_tokens": 7128219648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.902878522872925, "created_at": "2025-01-16T19:57:51.171266+00:00"} {"global_step": 13597, "acc_step": 0, "speed/wps": 12907.879851025855, "speed/FLOPS": 202735990346011.06, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052035409957170486, "optim/lr": 0.002822791187156127, "optim/total_tokens": 7128743936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 2.8484201431274414, "created_at": "2025-01-16T19:58:01.331587+00:00"} {"global_step": 13598, "acc_step": 0, "speed/wps": 12907.535414031981, "speed/FLOPS": 202730580489716.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07344713062047958, "optim/lr": 0.0028227507861496855, "optim/total_tokens": 7129268224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.7144908905029297, "created_at": "2025-01-16T19:58:11.489783+00:00"} {"global_step": 13599, "acc_step": 0, "speed/wps": 12905.072110061534, "speed/FLOPS": 202691890915926.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04907584935426712, "optim/lr": 0.0028227103808275374, "optim/total_tokens": 7129792512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424894, "loss/out": 2.845093250274658, "created_at": "2025-01-16T19:58:21.655799+00:00"} {"global_step": 13600, "acc_step": 0, "speed/wps": 12909.679123746626, "speed/FLOPS": 202764250396550.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04652497172355652, "optim/lr": 0.0028226699711898146, "optim/total_tokens": 7130316800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.9047837257385254, "created_at": "2025-01-16T19:58:31.815015+00:00"} {"global_step": 13601, "acc_step": 0, "speed/wps": 12900.838813322023, "speed/FLOPS": 202625401173472.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04399721324443817, "optim/lr": 0.002822629557236649, "optim/total_tokens": 7130841088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8718318939208984, "created_at": "2025-01-16T19:58:41.980580+00:00"} {"global_step": 13602, "acc_step": 0, "speed/wps": 12902.821940435933, "speed/FLOPS": 202656548909898.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0478534959256649, "optim/lr": 0.0028225891389681724, "optim/total_tokens": 7131365376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7971889972686768, "created_at": "2025-01-16T19:58:52.146105+00:00"} {"global_step": 13603, "acc_step": 0, "speed/wps": 12909.595592764463, "speed/FLOPS": 202762938427692.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04791830852627754, "optim/lr": 0.002822548716384516, "optim/total_tokens": 7131889664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.76070499420166, "created_at": "2025-01-16T19:59:02.304264+00:00"} {"global_step": 13604, "acc_step": 0, "speed/wps": 12909.116246317482, "speed/FLOPS": 202755409633051.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04931024834513664, "optim/lr": 0.0028225082894858134, "optim/total_tokens": 7132413952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 2.740927219390869, "created_at": "2025-01-16T19:59:12.465479+00:00"} {"global_step": 13605, "acc_step": 0, "speed/wps": 12906.982782548695, "speed/FLOPS": 202721900652876.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05321802943944931, "optim/lr": 0.0028224678582721954, "optim/total_tokens": 7132938240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.776841402053833, "created_at": "2025-01-16T19:59:22.624993+00:00"} {"global_step": 13606, "acc_step": 0, "speed/wps": 12912.573877835, "speed/FLOPS": 202809716487317.47, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05177410691976547, "optim/lr": 0.0028224274227437937, "optim/total_tokens": 7133462528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.7882637977600098, "created_at": "2025-01-16T19:59:32.779496+00:00"} {"global_step": 13607, "acc_step": 0, "speed/wps": 12903.564418201267, "speed/FLOPS": 202668210543474.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04611044004559517, "optim/lr": 0.002822386982900741, "optim/total_tokens": 7133986816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 2.807278633117676, "created_at": "2025-01-16T19:59:42.942314+00:00"} {"global_step": 13608, "acc_step": 0, "speed/wps": 12904.396870044498, "speed/FLOPS": 202681285343581.44, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050333309918642044, "optim/lr": 0.002822346538743168, "optim/total_tokens": 7134511104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8865702152252197, "created_at": "2025-01-16T19:59:53.102945+00:00"} {"global_step": 13609, "acc_step": 0, "speed/wps": 12901.02115486776, "speed/FLOPS": 202628265098012.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050692059099674225, "optim/lr": 0.0028223060902712077, "optim/total_tokens": 7135035392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.8567168712615967, "created_at": "2025-01-16T20:00:03.266878+00:00"} {"global_step": 13610, "acc_step": 0, "speed/wps": 12901.386274452958, "speed/FLOPS": 202633999802824.44, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06662825495004654, "optim/lr": 0.0028222656374849923, "optim/total_tokens": 7135559680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 2.7696728706359863, "created_at": "2025-01-16T20:00:13.430681+00:00"} {"global_step": 13611, "acc_step": 0, "speed/wps": 12907.00159148325, "speed/FLOPS": 202722196073039.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0898951068520546, "optim/lr": 0.0028222251803846523, "optim/total_tokens": 7136083968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.828052282333374, "created_at": "2025-01-16T20:00:23.589360+00:00"} {"global_step": 13612, "acc_step": 0, "speed/wps": 12904.167464264288, "speed/FLOPS": 202677682210566.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0710153579711914, "optim/lr": 0.002822184718970321, "optim/total_tokens": 7136608256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 2.809096336364746, "created_at": "2025-01-16T20:00:33.753461+00:00"} {"global_step": 13613, "acc_step": 0, "speed/wps": 12910.72950924412, "speed/FLOPS": 202780748144169.66, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0561550036072731, "optim/lr": 0.0028221442532421308, "optim/total_tokens": 7137132544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.8413448333740234, "created_at": "2025-01-16T20:00:43.911479+00:00"} {"global_step": 13614, "acc_step": 0, "speed/wps": 12913.645791589404, "speed/FLOPS": 202826552365794.22, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049645476043224335, "optim/lr": 0.0028221037832002124, "optim/total_tokens": 7137656832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.8804726600646973, "created_at": "2025-01-16T20:00:54.066237+00:00"} {"global_step": 13615, "acc_step": 0, "speed/wps": 12905.633921245813, "speed/FLOPS": 202700714932584.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057234156876802444, "optim/lr": 0.002822063308844699, "optim/total_tokens": 7138181120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 2.9305038452148438, "created_at": "2025-01-16T20:01:04.229337+00:00"} {"global_step": 13616, "acc_step": 0, "speed/wps": 12907.55437728736, "speed/FLOPS": 202730878333700.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04318545386195183, "optim/lr": 0.0028220228301757216, "optim/total_tokens": 7138705408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.85038423538208, "created_at": "2025-01-16T20:01:14.392717+00:00"} {"global_step": 13617, "acc_step": 0, "speed/wps": 12909.855765747001, "speed/FLOPS": 202767024801898.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052246902137994766, "optim/lr": 0.0028219823471934127, "optim/total_tokens": 7139229696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 2.790320634841919, "created_at": "2025-01-16T20:01:24.549492+00:00"} {"global_step": 13618, "acc_step": 0, "speed/wps": 12906.723031915413, "speed/FLOPS": 202717820912245.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06016344949603081, "optim/lr": 0.0028219418598979047, "optim/total_tokens": 7139753984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.913299083709717, "created_at": "2025-01-16T20:01:34.710885+00:00"} {"global_step": 13619, "acc_step": 0, "speed/wps": 12909.741648055515, "speed/FLOPS": 202765232426739.1, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06524335592985153, "optim/lr": 0.0028219013682893295, "optim/total_tokens": 7140278272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.879091262817383, "created_at": "2025-01-16T20:01:44.869998+00:00"} {"global_step": 13620, "acc_step": 0, "speed/wps": 12912.116698598415, "speed/FLOPS": 202802535859176.78, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04713471233844757, "optim/lr": 0.002821860872367819, "optim/total_tokens": 7140802560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 2.8545868396759033, "created_at": "2025-01-16T20:01:55.028840+00:00"} {"global_step": 13621, "acc_step": 0, "speed/wps": 12917.132302374792, "speed/FLOPS": 202881312808646.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05764105170965195, "optim/lr": 0.0028218203721335055, "optim/total_tokens": 7141326848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.8168866634368896, "created_at": "2025-01-16T20:02:05.179859+00:00"} {"global_step": 13622, "acc_step": 0, "speed/wps": 12910.682328378434, "speed/FLOPS": 202780007103840.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06123923882842064, "optim/lr": 0.0028217798675865212, "optim/total_tokens": 7141851136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 2.7823891639709473, "created_at": "2025-01-16T20:02:15.339611+00:00"} {"global_step": 13623, "acc_step": 0, "speed/wps": 12911.865419590871, "speed/FLOPS": 202798589177069.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044542767107486725, "optim/lr": 0.002821739358726999, "optim/total_tokens": 7142375424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.737086296081543, "created_at": "2025-01-16T20:02:25.494870+00:00"} {"global_step": 13624, "acc_step": 0, "speed/wps": 12905.265915163238, "speed/FLOPS": 202694934891364.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07873446494340897, "optim/lr": 0.002821698845555069, "optim/total_tokens": 7142899712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419875, "loss/out": 2.807091236114502, "created_at": "2025-01-16T20:02:35.655881+00:00"} {"global_step": 13625, "acc_step": 0, "speed/wps": 12912.47496629203, "speed/FLOPS": 202808162945616.4, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05014030262827873, "optim/lr": 0.0028216583280708655, "optim/total_tokens": 7143424000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 509028, "loss/out": 3.0142199993133545, "created_at": "2025-01-16T20:02:45.812192+00:00"} {"global_step": 13626, "acc_step": 0, "speed/wps": 12908.555467444832, "speed/FLOPS": 202746601830265.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051717013120651245, "optim/lr": 0.002821617806274519, "optim/total_tokens": 7143948288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.877432107925415, "created_at": "2025-01-16T20:02:55.972932+00:00"} {"global_step": 13627, "acc_step": 0, "speed/wps": 12907.6250954858, "speed/FLOPS": 202731989060183.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05895949900150299, "optim/lr": 0.0028215772801661633, "optim/total_tokens": 7144472576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 2.7768702507019043, "created_at": "2025-01-16T20:03:06.132548+00:00"} {"global_step": 13628, "acc_step": 0, "speed/wps": 12907.613926225355, "speed/FLOPS": 202731813631600.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04577884450554848, "optim/lr": 0.0028215367497459293, "optim/total_tokens": 7144996864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.827639102935791, "created_at": "2025-01-16T20:03:16.294845+00:00"} {"global_step": 13629, "acc_step": 0, "speed/wps": 12909.358385778776, "speed/FLOPS": 202759212766025.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07091153413057327, "optim/lr": 0.0028214962150139504, "optim/total_tokens": 7145521152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.7457070350646973, "created_at": "2025-01-16T20:03:26.458595+00:00"} {"global_step": 13630, "acc_step": 0, "speed/wps": 12906.801534226175, "speed/FLOPS": 202719053898909.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050173621624708176, "optim/lr": 0.002821455675970358, "optim/total_tokens": 7146045440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.756012439727783, "created_at": "2025-01-16T20:03:36.619699+00:00"} {"global_step": 13631, "acc_step": 0, "speed/wps": 12907.477343182309, "speed/FLOPS": 202729668407224.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07487708330154419, "optim/lr": 0.0028214151326152845, "optim/total_tokens": 7146569728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.933685779571533, "created_at": "2025-01-16T20:03:46.778054+00:00"} {"global_step": 13632, "acc_step": 0, "speed/wps": 12905.590698348391, "speed/FLOPS": 202700036057586.06, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05212102457880974, "optim/lr": 0.002821374584948862, "optim/total_tokens": 7147094016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 2.803149700164795, "created_at": "2025-01-16T20:03:56.941373+00:00"} {"global_step": 13633, "acc_step": 0, "speed/wps": 12908.547801158973, "speed/FLOPS": 202746481420711.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05086831375956535, "optim/lr": 0.0028213340329712234, "optim/total_tokens": 7147618304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8700289726257324, "created_at": "2025-01-16T20:04:07.099612+00:00"} {"global_step": 13634, "acc_step": 0, "speed/wps": 12906.585781225102, "speed/FLOPS": 202715665201551.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06723315268754959, "optim/lr": 0.0028212934766825015, "optim/total_tokens": 7148142592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8156824111938477, "created_at": "2025-01-16T20:04:17.258642+00:00"} {"global_step": 13635, "acc_step": 0, "speed/wps": 12906.754268434132, "speed/FLOPS": 202718311524695.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05194932594895363, "optim/lr": 0.002821252916082827, "optim/total_tokens": 7148666880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 2.9117112159729004, "created_at": "2025-01-16T20:04:27.418060+00:00"} {"global_step": 13636, "acc_step": 0, "speed/wps": 12910.663012463801, "speed/FLOPS": 202779703720858.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05272328853607178, "optim/lr": 0.002821212351172333, "optim/total_tokens": 7149191168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.7375712394714355, "created_at": "2025-01-16T20:04:37.577572+00:00"} {"global_step": 13637, "acc_step": 0, "speed/wps": 12913.035854731803, "speed/FLOPS": 202816972469304.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05220490321516991, "optim/lr": 0.002821171781951152, "optim/total_tokens": 7149715456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305632, "loss/out": 2.7876076698303223, "created_at": "2025-01-16T20:04:47.733614+00:00"} {"global_step": 13638, "acc_step": 0, "speed/wps": 12910.164973416684, "speed/FLOPS": 202771881333246.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.056552622467279434, "optim/lr": 0.0028211312084194164, "optim/total_tokens": 7150239744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.864931344985962, "created_at": "2025-01-16T20:04:57.894068+00:00"} {"global_step": 13639, "acc_step": 0, "speed/wps": 12903.659781259943, "speed/FLOPS": 202669708351353.88, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05811328813433647, "optim/lr": 0.0028210906305772584, "optim/total_tokens": 7150764032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.834348678588867, "created_at": "2025-01-16T20:05:08.056226+00:00"} {"global_step": 13640, "acc_step": 0, "speed/wps": 12902.265934522842, "speed/FLOPS": 202647816073001.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058892279863357544, "optim/lr": 0.0028210500484248105, "optim/total_tokens": 7151288320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.8448307514190674, "created_at": "2025-01-16T20:05:18.220112+00:00"} {"global_step": 13641, "acc_step": 0, "speed/wps": 12909.788566256973, "speed/FLOPS": 202765969341565.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0612436980009079, "optim/lr": 0.0028210094619622052, "optim/total_tokens": 7151812608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8870980739593506, "created_at": "2025-01-16T20:05:28.377844+00:00"} {"global_step": 13642, "acc_step": 0, "speed/wps": 12906.462528315276, "speed/FLOPS": 202713729345235.94, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.062354110181331635, "optim/lr": 0.002820968871189575, "optim/total_tokens": 7152336896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 2.805891513824463, "created_at": "2025-01-16T20:05:38.537434+00:00"} {"global_step": 13643, "acc_step": 0, "speed/wps": 12913.793109265456, "speed/FLOPS": 202828866192332.34, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05487601086497307, "optim/lr": 0.0028209282761070516, "optim/total_tokens": 7152861184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.748141288757324, "created_at": "2025-01-16T20:05:48.693798+00:00"} {"global_step": 13644, "acc_step": 0, "speed/wps": 12912.01914570898, "speed/FLOPS": 202801003656996.5, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05383310467004776, "optim/lr": 0.002820887676714768, "optim/total_tokens": 7153385472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 2.7852725982666016, "created_at": "2025-01-16T20:05:58.848581+00:00"} {"global_step": 13645, "acc_step": 0, "speed/wps": 12908.208524737463, "speed/FLOPS": 202741152618296.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06744463741779327, "optim/lr": 0.0028208470730128566, "optim/total_tokens": 7153909760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.9137797355651855, "created_at": "2025-01-16T20:06:09.018952+00:00"} {"global_step": 13646, "acc_step": 0, "speed/wps": 12907.54015158516, "speed/FLOPS": 202730654899499.0, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04922759532928467, "optim/lr": 0.0028208064650014505, "optim/total_tokens": 7154434048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.7877678871154785, "created_at": "2025-01-16T20:06:19.179873+00:00"} {"global_step": 13647, "acc_step": 0, "speed/wps": 12904.750507614748, "speed/FLOPS": 202686839707571.16, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0760577991604805, "optim/lr": 0.002820765852680681, "optim/total_tokens": 7154958336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 2.7862002849578857, "created_at": "2025-01-16T20:06:29.342471+00:00"} {"global_step": 13648, "acc_step": 0, "speed/wps": 12905.97586252231, "speed/FLOPS": 202706085590208.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046006664633750916, "optim/lr": 0.0028207252360506812, "optim/total_tokens": 7155482624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.799546718597412, "created_at": "2025-01-16T20:06:39.501870+00:00"} {"global_step": 13649, "acc_step": 0, "speed/wps": 12907.22760719493, "speed/FLOPS": 202725745960369.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.0640844851732254, "optim/lr": 0.0028206846151115836, "optim/total_tokens": 7156006912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.7463607788085938, "created_at": "2025-01-16T20:06:49.666545+00:00"} {"global_step": 13650, "acc_step": 0, "speed/wps": 12907.65513716331, "speed/FLOPS": 202732460906010.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051233161240816116, "optim/lr": 0.0028206439898635213, "optim/total_tokens": 7156531200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 2.658233880996704, "created_at": "2025-01-16T20:06:59.826039+00:00"} {"global_step": 13651, "acc_step": 0, "speed/wps": 12914.613367800215, "speed/FLOPS": 202841749479774.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04601789265871048, "optim/lr": 0.0028206033603066258, "optim/total_tokens": 7157055488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 2.754459857940674, "created_at": "2025-01-16T20:07:09.978927+00:00"} {"global_step": 13652, "acc_step": 0, "speed/wps": 12908.726092317098, "speed/FLOPS": 202749281728347.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05156705155968666, "optim/lr": 0.0028205627264410297, "optim/total_tokens": 7157579776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 2.9438652992248535, "created_at": "2025-01-16T20:07:20.141846+00:00"} {"global_step": 13653, "acc_step": 0, "speed/wps": 12903.581803404339, "speed/FLOPS": 202668483601978.78, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05585779249668121, "optim/lr": 0.0028205220882668667, "optim/total_tokens": 7158104064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.817000150680542, "created_at": "2025-01-16T20:07:30.304109+00:00"} {"global_step": 13654, "acc_step": 0, "speed/wps": 12910.7332293888, "speed/FLOPS": 202780806574153.88, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04964539408683777, "optim/lr": 0.0028204814457842694, "optim/total_tokens": 7158628352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8994874954223633, "created_at": "2025-01-16T20:07:40.461772+00:00"} {"global_step": 13655, "acc_step": 0, "speed/wps": 12901.5337702524, "speed/FLOPS": 202636316427035.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060637086629867554, "optim/lr": 0.002820440798993369, "optim/total_tokens": 7159152640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348781, "loss/out": 2.905040740966797, "created_at": "2025-01-16T20:07:50.627116+00:00"} {"global_step": 13656, "acc_step": 0, "speed/wps": 12911.664281654332, "speed/FLOPS": 202795430029383.94, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04816054180264473, "optim/lr": 0.002820400147894298, "optim/total_tokens": 7159676928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.899986743927002, "created_at": "2025-01-16T20:08:00.782123+00:00"} {"global_step": 13657, "acc_step": 0, "speed/wps": 12907.27615135078, "speed/FLOPS": 202726508413043.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05688600242137909, "optim/lr": 0.002820359492487191, "optim/total_tokens": 7160201216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.8285365104675293, "created_at": "2025-01-16T20:08:10.940942+00:00"} {"global_step": 13658, "acc_step": 0, "speed/wps": 12908.222676141952, "speed/FLOPS": 202741374885550.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048588287085294724, "optim/lr": 0.0028203188327721794, "optim/total_tokens": 7160725504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.7809042930603027, "created_at": "2025-01-16T20:08:21.099008+00:00"} {"global_step": 13659, "acc_step": 0, "speed/wps": 12899.814012767083, "speed/FLOPS": 202609305272533.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.065916508436203, "optim/lr": 0.0028202781687493954, "optim/total_tokens": 7161249792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8003530502319336, "created_at": "2025-01-16T20:08:31.265606+00:00"} {"global_step": 13660, "acc_step": 0, "speed/wps": 12913.3965953114, "speed/FLOPS": 202822638395816.6, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0662011131644249, "optim/lr": 0.0028202375004189725, "optim/total_tokens": 7161774080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 2.7706236839294434, "created_at": "2025-01-16T20:08:41.424876+00:00"} {"global_step": 13661, "acc_step": 0, "speed/wps": 12907.44879271134, "speed/FLOPS": 202729219982845.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.0508902408182621, "optim/lr": 0.0028201968277810434, "optim/total_tokens": 7162298368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8597145080566406, "created_at": "2025-01-16T20:08:51.583071+00:00"} {"global_step": 13662, "acc_step": 0, "speed/wps": 12910.527615553201, "speed/FLOPS": 202777577126322.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05053093656897545, "optim/lr": 0.0028201561508357405, "optim/total_tokens": 7162822656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482614, "loss/out": 2.8873870372772217, "created_at": "2025-01-16T20:09:01.738998+00:00"} {"global_step": 13663, "acc_step": 0, "speed/wps": 12910.51645429673, "speed/FLOPS": 202777401823452.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05900036171078682, "optim/lr": 0.0028201154695831968, "optim/total_tokens": 7163346944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8186323642730713, "created_at": "2025-01-16T20:09:11.895011+00:00"} {"global_step": 13664, "acc_step": 0, "speed/wps": 12907.25228821076, "speed/FLOPS": 202726133609638.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04237852245569229, "optim/lr": 0.0028200747840235445, "optim/total_tokens": 7163871232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478894, "loss/out": 2.79897141456604, "created_at": "2025-01-16T20:09:22.057264+00:00"} {"global_step": 13665, "acc_step": 0, "speed/wps": 12906.864971684508, "speed/FLOPS": 202720050271364.84, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06432506442070007, "optim/lr": 0.002820034094156916, "optim/total_tokens": 7164395520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.773663282394409, "created_at": "2025-01-16T20:09:32.215998+00:00"} {"global_step": 13666, "acc_step": 0, "speed/wps": 12905.742443083393, "speed/FLOPS": 202702419417168.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06800185143947601, "optim/lr": 0.0028199933999834455, "optim/total_tokens": 7164919808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8530702590942383, "created_at": "2025-01-16T20:09:42.377768+00:00"} {"global_step": 13667, "acc_step": 0, "speed/wps": 12910.97044900098, "speed/FLOPS": 202784532434137.06, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04830063879489899, "optim/lr": 0.002819952701503265, "optim/total_tokens": 7165444096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 2.7717065811157227, "created_at": "2025-01-16T20:09:52.535723+00:00"} {"global_step": 13668, "acc_step": 0, "speed/wps": 12904.254997569618, "speed/FLOPS": 202679057041410.25, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07209000736474991, "optim/lr": 0.0028199119987165067, "optim/total_tokens": 7165968384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7468724250793457, "created_at": "2025-01-16T20:10:02.696957+00:00"} {"global_step": 13669, "acc_step": 0, "speed/wps": 12904.181604443284, "speed/FLOPS": 202677904301508.44, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044786229729652405, "optim/lr": 0.002819871291623304, "optim/total_tokens": 7166492672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.8516881465911865, "created_at": "2025-01-16T20:10:12.861176+00:00"} {"global_step": 13670, "acc_step": 0, "speed/wps": 12902.987948086684, "speed/FLOPS": 202659156288172.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05531306564807892, "optim/lr": 0.00281983058022379, "optim/total_tokens": 7167016960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 2.789806365966797, "created_at": "2025-01-16T20:10:23.024793+00:00"} {"global_step": 13671, "acc_step": 0, "speed/wps": 12910.056121390493, "speed/FLOPS": 202770171662596.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06061005964875221, "optim/lr": 0.0028197898645180974, "optim/total_tokens": 7167541248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 2.7530808448791504, "created_at": "2025-01-16T20:10:33.183688+00:00"} {"global_step": 13672, "acc_step": 0, "speed/wps": 12898.490607669242, "speed/FLOPS": 202588519376923.03, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04229014739394188, "optim/lr": 0.002819749144506358, "optim/total_tokens": 7168065536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.9186601638793945, "created_at": "2025-01-16T20:10:43.349446+00:00"} {"global_step": 13673, "acc_step": 0, "speed/wps": 12905.68205814033, "speed/FLOPS": 202701470988659.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050311144441366196, "optim/lr": 0.002819708420188706, "optim/total_tokens": 7168589824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 2.8656601905822754, "created_at": "2025-01-16T20:10:53.509621+00:00"} {"global_step": 13674, "acc_step": 0, "speed/wps": 12903.279249587102, "speed/FLOPS": 202663731578526.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04923860728740692, "optim/lr": 0.002819667691565274, "optim/total_tokens": 7169114112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9102272987365723, "created_at": "2025-01-16T20:11:03.672604+00:00"} {"global_step": 13675, "acc_step": 0, "speed/wps": 12902.262930136372, "speed/FLOPS": 202647768884983.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05864645913243294, "optim/lr": 0.0028196269586361947, "optim/total_tokens": 7169638400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8003993034362793, "created_at": "2025-01-16T20:11:13.838846+00:00"} {"global_step": 13676, "acc_step": 0, "speed/wps": 12900.528464109497, "speed/FLOPS": 202620526712625.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0494871661067009, "optim/lr": 0.002819586221401601, "optim/total_tokens": 7170162688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 2.8330187797546387, "created_at": "2025-01-16T20:11:24.005934+00:00"} {"global_step": 13677, "acc_step": 0, "speed/wps": 12901.54785380405, "speed/FLOPS": 202636537628566.62, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04275035858154297, "optim/lr": 0.0028195454798616245, "optim/total_tokens": 7170686976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.884751558303833, "created_at": "2025-01-16T20:11:34.176460+00:00"} {"global_step": 13678, "acc_step": 0, "speed/wps": 12904.138342763257, "speed/FLOPS": 202677224817375.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05311231315135956, "optim/lr": 0.0028195047340164006, "optim/total_tokens": 7171211264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8353564739227295, "created_at": "2025-01-16T20:11:44.338451+00:00"} {"global_step": 13679, "acc_step": 0, "speed/wps": 12897.517131822613, "speed/FLOPS": 202573229601055.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04881806671619415, "optim/lr": 0.0028194639838660606, "optim/total_tokens": 7171735552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.9974164962768555, "created_at": "2025-01-16T20:11:54.505809+00:00"} {"global_step": 13680, "acc_step": 0, "speed/wps": 12903.780996573092, "speed/FLOPS": 202671612204414.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04665180668234825, "optim/lr": 0.0028194232294107384, "optim/total_tokens": 7172259840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.772005558013916, "created_at": "2025-01-16T20:12:04.675165+00:00"} {"global_step": 13681, "acc_step": 0, "speed/wps": 12915.629760032845, "speed/FLOPS": 202857713316456.8, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04274078458547592, "optim/lr": 0.002819382470650566, "optim/total_tokens": 7172784128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 3.0335140228271484, "created_at": "2025-01-16T20:12:14.827065+00:00"} {"global_step": 13682, "acc_step": 0, "speed/wps": 12903.964160762109, "speed/FLOPS": 202674489049696.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05066681280732155, "optim/lr": 0.0028193417075856766, "optim/total_tokens": 7173308416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.937159538269043, "created_at": "2025-01-16T20:12:24.991227+00:00"} {"global_step": 13683, "acc_step": 0, "speed/wps": 12909.78260613438, "speed/FLOPS": 202765875729649.97, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04244042932987213, "optim/lr": 0.0028193009402162038, "optim/total_tokens": 7173832704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.865729808807373, "created_at": "2025-01-16T20:12:35.149408+00:00"} {"global_step": 13684, "acc_step": 0, "speed/wps": 12904.52513832023, "speed/FLOPS": 202683299973110.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04967178404331207, "optim/lr": 0.0028192601685422805, "optim/total_tokens": 7174356992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 2.8180525302886963, "created_at": "2025-01-16T20:12:45.316041+00:00"} {"global_step": 13685, "acc_step": 0, "speed/wps": 12911.298811068784, "speed/FLOPS": 202789689811628.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057977691292762756, "optim/lr": 0.002819219392564039, "optim/total_tokens": 7174881280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8612616062164307, "created_at": "2025-01-16T20:12:55.471455+00:00"} {"global_step": 13686, "acc_step": 0, "speed/wps": 12909.400093353359, "speed/FLOPS": 202759867840796.38, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045103225857019424, "optim/lr": 0.0028191786122816124, "optim/total_tokens": 7175405568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483664, "loss/out": 2.740645170211792, "created_at": "2025-01-16T20:13:05.629819+00:00"} {"global_step": 13687, "acc_step": 0, "speed/wps": 12901.598875594793, "speed/FLOPS": 202637338995901.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05819040909409523, "optim/lr": 0.002819137827695135, "optim/total_tokens": 7175929856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.9396133422851562, "created_at": "2025-01-16T20:13:15.792998+00:00"} {"global_step": 13688, "acc_step": 0, "speed/wps": 12901.375207052388, "speed/FLOPS": 202633825974090.28, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05597914755344391, "optim/lr": 0.0028190970388047385, "optim/total_tokens": 7176454144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8876543045043945, "created_at": "2025-01-16T20:13:25.960040+00:00"} {"global_step": 13689, "acc_step": 0, "speed/wps": 12908.264729189395, "speed/FLOPS": 202742035386447.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04555319994688034, "optim/lr": 0.002819056245610557, "optim/total_tokens": 7176978432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447274, "loss/out": 2.7400307655334473, "created_at": "2025-01-16T20:13:36.117681+00:00"} {"global_step": 13690, "acc_step": 0, "speed/wps": 12907.368743631647, "speed/FLOPS": 202727962702049.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.057780880481004715, "optim/lr": 0.0028190154481127225, "optim/total_tokens": 7177502720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.7782018184661865, "created_at": "2025-01-16T20:13:46.278331+00:00"} {"global_step": 13691, "acc_step": 0, "speed/wps": 12909.734284699885, "speed/FLOPS": 202765116775120.62, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05020248144865036, "optim/lr": 0.0028189746463113687, "optim/total_tokens": 7178027008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382133, "loss/out": 2.8435401916503906, "created_at": "2025-01-16T20:13:56.434890+00:00"} {"global_step": 13692, "acc_step": 0, "speed/wps": 12911.404813993475, "speed/FLOPS": 202791354733224.25, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05874915048480034, "optim/lr": 0.0028189338402066296, "optim/total_tokens": 7178551296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.9177112579345703, "created_at": "2025-01-16T20:14:06.591063+00:00"} {"global_step": 13693, "acc_step": 0, "speed/wps": 12900.859651385403, "speed/FLOPS": 202625728463892.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.051573753356933594, "optim/lr": 0.0028188930297986365, "optim/total_tokens": 7179075584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8455638885498047, "created_at": "2025-01-16T20:14:16.755222+00:00"} {"global_step": 13694, "acc_step": 0, "speed/wps": 12911.737400997965, "speed/FLOPS": 202796578469152.2, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07065045088529587, "optim/lr": 0.002818852215087524, "optim/total_tokens": 7179599872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 2.85119891166687, "created_at": "2025-01-16T20:14:26.910692+00:00"} {"global_step": 13695, "acc_step": 0, "speed/wps": 12906.149360307212, "speed/FLOPS": 202708810611331.34, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052631523460149765, "optim/lr": 0.0028188113960734246, "optim/total_tokens": 7180124160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.802821159362793, "created_at": "2025-01-16T20:14:37.077798+00:00"} {"global_step": 13696, "acc_step": 0, "speed/wps": 12904.67231187007, "speed/FLOPS": 202685611535948.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057366859167814255, "optim/lr": 0.0028187705727564715, "optim/total_tokens": 7180648448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 2.809905767440796, "created_at": "2025-01-16T20:14:47.239726+00:00"} {"global_step": 13697, "acc_step": 0, "speed/wps": 12905.081860824852, "speed/FLOPS": 202692044065062.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0441536009311676, "optim/lr": 0.002818729745136799, "optim/total_tokens": 7181172736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.7751522064208984, "created_at": "2025-01-16T20:14:57.401248+00:00"} {"global_step": 13698, "acc_step": 0, "speed/wps": 12907.69942867906, "speed/FLOPS": 202733156565128.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06108585000038147, "optim/lr": 0.0028186889132145385, "optim/total_tokens": 7181697024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.7869009971618652, "created_at": "2025-01-16T20:15:07.563043+00:00"} {"global_step": 13699, "acc_step": 0, "speed/wps": 12905.732078286806, "speed/FLOPS": 202702256623795.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04269338399171829, "optim/lr": 0.0028186480769898242, "optim/total_tokens": 7182221312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 2.936784267425537, "created_at": "2025-01-16T20:15:17.724053+00:00"} {"global_step": 13700, "acc_step": 0, "speed/wps": 12898.638571604379, "speed/FLOPS": 202590843353848.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05982579290866852, "optim/lr": 0.0028186072364627893, "optim/total_tokens": 7182745600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8202850818634033, "created_at": "2025-01-16T20:15:27.890293+00:00"} {"global_step": 13701, "acc_step": 0, "speed/wps": 12906.103652087242, "speed/FLOPS": 202708092700927.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04992048814892769, "optim/lr": 0.0028185663916335665, "optim/total_tokens": 7183269888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8119966983795166, "created_at": "2025-01-16T20:15:38.051965+00:00"} {"global_step": 13702, "acc_step": 0, "speed/wps": 12899.996046294951, "speed/FLOPS": 202612164359229.25, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05175956338644028, "optim/lr": 0.00281852554250229, "optim/total_tokens": 7183794176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335952, "loss/out": 2.8170816898345947, "created_at": "2025-01-16T20:15:48.216101+00:00"} {"global_step": 13703, "acc_step": 0, "speed/wps": 12903.721533869157, "speed/FLOPS": 202670678260938.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04545069485902786, "optim/lr": 0.0028184846890690926, "optim/total_tokens": 7184318464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7870349884033203, "created_at": "2025-01-16T20:15:58.394654+00:00"} {"global_step": 13704, "acc_step": 0, "speed/wps": 12909.301191351251, "speed/FLOPS": 202758314448947.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05665552243590355, "optim/lr": 0.002818443831334108, "optim/total_tokens": 7184842752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7933855056762695, "created_at": "2025-01-16T20:16:08.551699+00:00"} {"global_step": 13705, "acc_step": 0, "speed/wps": 12900.989897606625, "speed/FLOPS": 202627774159773.9, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0533173531293869, "optim/lr": 0.0028184029692974685, "optim/total_tokens": 7185367040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.903578758239746, "created_at": "2025-01-16T20:16:18.718780+00:00"} {"global_step": 13706, "acc_step": 0, "speed/wps": 12907.22838491183, "speed/FLOPS": 202725758175482.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05080105736851692, "optim/lr": 0.0028183621029593087, "optim/total_tokens": 7185891328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.9568138122558594, "created_at": "2025-01-16T20:16:28.883228+00:00"} {"global_step": 13707, "acc_step": 0, "speed/wps": 12912.649580723419, "speed/FLOPS": 202810905505206.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045512571930885315, "optim/lr": 0.0028183212323197606, "optim/total_tokens": 7186415616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.908474922180176, "created_at": "2025-01-16T20:16:39.045359+00:00"} {"global_step": 13708, "acc_step": 0, "speed/wps": 12910.97485862095, "speed/FLOPS": 202784601693278.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.050901349633932114, "optim/lr": 0.002818280357378958, "optim/total_tokens": 7186939904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.902158737182617, "created_at": "2025-01-16T20:16:49.204165+00:00"} {"global_step": 13709, "acc_step": 0, "speed/wps": 12902.238117979245, "speed/FLOPS": 202647379175960.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05186128616333008, "optim/lr": 0.0028182394781370353, "optim/total_tokens": 7187464192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.7995898723602295, "created_at": "2025-01-16T20:16:59.370612+00:00"} {"global_step": 13710, "acc_step": 0, "speed/wps": 12904.914246780472, "speed/FLOPS": 202689411456170.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04695400223135948, "optim/lr": 0.0028181985945941244, "optim/total_tokens": 7187988480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.842982530593872, "created_at": "2025-01-16T20:17:09.531872+00:00"} {"global_step": 13711, "acc_step": 0, "speed/wps": 12907.873331516852, "speed/FLOPS": 202735887948163.44, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05962233990430832, "optim/lr": 0.0028181577067503596, "optim/total_tokens": 7188512768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 2.7535505294799805, "created_at": "2025-01-16T20:17:19.691363+00:00"} {"global_step": 13712, "acc_step": 0, "speed/wps": 12908.219325643482, "speed/FLOPS": 202741322261367.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.08539541810750961, "optim/lr": 0.002818116814605874, "optim/total_tokens": 7189037056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.949517250061035, "created_at": "2025-01-16T20:17:29.850066+00:00"} {"global_step": 13713, "acc_step": 0, "speed/wps": 12901.95230896551, "speed/FLOPS": 202642890152657.0, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07298560440540314, "optim/lr": 0.0028180759181608006, "optim/total_tokens": 7189561344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8581619262695312, "created_at": "2025-01-16T20:17:40.012783+00:00"} {"global_step": 13714, "acc_step": 0, "speed/wps": 12907.620459678059, "speed/FLOPS": 202731916248452.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05602174997329712, "optim/lr": 0.002818035017415274, "optim/total_tokens": 7190085632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.7664685249328613, "created_at": "2025-01-16T20:17:50.171539+00:00"} {"global_step": 13715, "acc_step": 0, "speed/wps": 12901.44878232411, "speed/FLOPS": 202634981574835.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06431564688682556, "optim/lr": 0.0028179941123694266, "optim/total_tokens": 7190609920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.7501683235168457, "created_at": "2025-01-16T20:18:00.336762+00:00"} {"global_step": 13716, "acc_step": 0, "speed/wps": 12909.503657711259, "speed/FLOPS": 202761494461349.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052946168929338455, "optim/lr": 0.0028179532030233926, "optim/total_tokens": 7191134208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.7008962631225586, "created_at": "2025-01-16T20:18:10.497853+00:00"} {"global_step": 13717, "acc_step": 0, "speed/wps": 12900.631810490318, "speed/FLOPS": 202622149909549.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05715928226709366, "optim/lr": 0.0028179122893773048, "optim/total_tokens": 7191658496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497150, "loss/out": 2.8767266273498535, "created_at": "2025-01-16T20:18:20.665422+00:00"} {"global_step": 13718, "acc_step": 0, "speed/wps": 12900.525462897158, "speed/FLOPS": 202620479574461.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054850541055202484, "optim/lr": 0.0028178713714312967, "optim/total_tokens": 7192182784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 2.83176326751709, "created_at": "2025-01-16T20:18:30.830527+00:00"} {"global_step": 13719, "acc_step": 0, "speed/wps": 12906.037257391241, "speed/FLOPS": 202707049880990.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053146764636039734, "optim/lr": 0.002817830449185502, "optim/total_tokens": 7192707072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.868431329727173, "created_at": "2025-01-16T20:18:40.991664+00:00"} {"global_step": 13720, "acc_step": 0, "speed/wps": 12908.346753547083, "speed/FLOPS": 202743323691698.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05028475821018219, "optim/lr": 0.0028177895226400547, "optim/total_tokens": 7193231360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8573508262634277, "created_at": "2025-01-16T20:18:51.149528+00:00"} {"global_step": 13721, "acc_step": 0, "speed/wps": 12905.185460045008, "speed/FLOPS": 202693671233175.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05230117589235306, "optim/lr": 0.002817748591795088, "optim/total_tokens": 7193755648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 2.916414976119995, "created_at": "2025-01-16T20:19:01.309685+00:00"} {"global_step": 13722, "acc_step": 0, "speed/wps": 12910.197575497545, "speed/FLOPS": 202772393393726.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06277680397033691, "optim/lr": 0.002817707656650735, "optim/total_tokens": 7194279936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.893524408340454, "created_at": "2025-01-16T20:19:11.486992+00:00"} {"global_step": 13723, "acc_step": 0, "speed/wps": 12905.922848596918, "speed/FLOPS": 202705252933661.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05215618386864662, "optim/lr": 0.0028176667172071295, "optim/total_tokens": 7194804224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474161, "loss/out": 2.9259731769561768, "created_at": "2025-01-16T20:19:21.647696+00:00"} {"global_step": 13724, "acc_step": 0, "speed/wps": 12908.352255340862, "speed/FLOPS": 202743410104929.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.051949869841337204, "optim/lr": 0.0028176257734644056, "optim/total_tokens": 7195328512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9071640968322754, "created_at": "2025-01-16T20:19:31.807169+00:00"} {"global_step": 13725, "acc_step": 0, "speed/wps": 12908.51411223249, "speed/FLOPS": 202745952289827.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05127852410078049, "optim/lr": 0.0028175848254226965, "optim/total_tokens": 7195852800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.865448474884033, "created_at": "2025-01-16T20:19:41.967931+00:00"} {"global_step": 13726, "acc_step": 0, "speed/wps": 12907.127886198437, "speed/FLOPS": 202724179705088.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049613066017627716, "optim/lr": 0.0028175438730821353, "optim/total_tokens": 7196377088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 2.8243260383605957, "created_at": "2025-01-16T20:19:52.130432+00:00"} {"global_step": 13727, "acc_step": 0, "speed/wps": 12909.05017229163, "speed/FLOPS": 202754371849680.0, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07799473404884338, "optim/lr": 0.002817502916442856, "optim/total_tokens": 7196901376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.942052125930786, "created_at": "2025-01-16T20:20:02.287641+00:00"} {"global_step": 13728, "acc_step": 0, "speed/wps": 12908.202373095024, "speed/FLOPS": 202741055998299.12, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04524047672748566, "optim/lr": 0.002817461955504993, "optim/total_tokens": 7197425664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.8728854656219482, "created_at": "2025-01-16T20:20:12.446661+00:00"} {"global_step": 13729, "acc_step": 0, "speed/wps": 12904.452174000124, "speed/FLOPS": 202682153968200.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07703905552625656, "optim/lr": 0.002817420990268679, "optim/total_tokens": 7197949952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 2.9173836708068848, "created_at": "2025-01-16T20:20:22.610855+00:00"} {"global_step": 13730, "acc_step": 0, "speed/wps": 12911.597584533454, "speed/FLOPS": 202794382459449.1, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04702102765440941, "optim/lr": 0.0028173800207340477, "optim/total_tokens": 7198474240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 2.8619353771209717, "created_at": "2025-01-16T20:20:32.769228+00:00"} {"global_step": 13731, "acc_step": 0, "speed/wps": 12905.843390066695, "speed/FLOPS": 202704004928256.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046179208904504776, "optim/lr": 0.002817339046901233, "optim/total_tokens": 7198998528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 2.8079633712768555, "created_at": "2025-01-16T20:20:42.933012+00:00"} {"global_step": 13732, "acc_step": 0, "speed/wps": 12904.014054010497, "speed/FLOPS": 202675272691723.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04558618366718292, "optim/lr": 0.0028172980687703685, "optim/total_tokens": 7199522816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8317575454711914, "created_at": "2025-01-16T20:20:53.096451+00:00"} {"global_step": 13733, "acc_step": 0, "speed/wps": 12907.958111692806, "speed/FLOPS": 202737219537326.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04845011606812477, "optim/lr": 0.0028172570863415885, "optim/total_tokens": 7200047104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.7669668197631836, "created_at": "2025-01-16T20:21:03.257029+00:00"} {"global_step": 13734, "acc_step": 0, "speed/wps": 12908.443881556539, "speed/FLOPS": 202744849220554.62, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04929746314883232, "optim/lr": 0.0028172160996150254, "optim/total_tokens": 7200571392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.613525867462158, "created_at": "2025-01-16T20:21:13.414968+00:00"} {"global_step": 13735, "acc_step": 0, "speed/wps": 12909.367598319615, "speed/FLOPS": 202759357461638.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045521579682826996, "optim/lr": 0.002817175108590814, "optim/total_tokens": 7201095680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.870293378829956, "created_at": "2025-01-16T20:21:23.573452+00:00"} {"global_step": 13736, "acc_step": 0, "speed/wps": 12902.68104418155, "speed/FLOPS": 202654335940612.03, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04524708911776543, "optim/lr": 0.002817134113269088, "optim/total_tokens": 7201619968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.812162399291992, "created_at": "2025-01-16T20:21:33.736773+00:00"} {"global_step": 13737, "acc_step": 0, "speed/wps": 12910.097998086676, "speed/FLOPS": 202770829393654.72, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047846682369709015, "optim/lr": 0.0028170931136499806, "optim/total_tokens": 7202144256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8490114212036133, "created_at": "2025-01-16T20:21:43.894107+00:00"} {"global_step": 13738, "acc_step": 0, "speed/wps": 12904.992930622326, "speed/FLOPS": 202690647294029.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048616744577884674, "optim/lr": 0.0028170521097336263, "optim/total_tokens": 7202668544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8778748512268066, "created_at": "2025-01-16T20:21:54.058689+00:00"} {"global_step": 13739, "acc_step": 0, "speed/wps": 12902.83363490775, "speed/FLOPS": 202656732587648.62, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047303225845098495, "optim/lr": 0.002817011101520158, "optim/total_tokens": 7203192832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304284, "loss/out": 2.8939852714538574, "created_at": "2025-01-16T20:22:04.220556+00:00"} {"global_step": 13740, "acc_step": 0, "speed/wps": 12904.665678457095, "speed/FLOPS": 202685507349082.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04877374693751335, "optim/lr": 0.00281697008900971, "optim/total_tokens": 7203717120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.757167339324951, "created_at": "2025-01-16T20:22:14.384656+00:00"} {"global_step": 13741, "acc_step": 0, "speed/wps": 12894.309056542283, "speed/FLOPS": 202522842370420.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05164555460214615, "optim/lr": 0.0028169290722024165, "optim/total_tokens": 7204241408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.841357946395874, "created_at": "2025-01-16T20:22:24.553603+00:00"} {"global_step": 13742, "acc_step": 0, "speed/wps": 12908.05131162298, "speed/FLOPS": 202738683370300.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04517707601189613, "optim/lr": 0.0028168880510984103, "optim/total_tokens": 7204765696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 2.833700656890869, "created_at": "2025-01-16T20:22:34.711428+00:00"} {"global_step": 13743, "acc_step": 0, "speed/wps": 12902.538460669135, "speed/FLOPS": 202652096470619.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05504641681909561, "optim/lr": 0.002816847025697826, "optim/total_tokens": 7205289984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464534, "loss/out": 2.8177645206451416, "created_at": "2025-01-16T20:22:44.882385+00:00"} {"global_step": 13744, "acc_step": 0, "speed/wps": 12909.042392931278, "speed/FLOPS": 202754249664135.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05290204659104347, "optim/lr": 0.0028168059960007974, "optim/total_tokens": 7205814272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385799, "loss/out": 2.8259010314941406, "created_at": "2025-01-16T20:22:55.042862+00:00"} {"global_step": 13745, "acc_step": 0, "speed/wps": 12902.70808664559, "speed/FLOPS": 202654760679670.03, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05938122794032097, "optim/lr": 0.0028167649620074583, "optim/total_tokens": 7206338560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.893251419067383, "created_at": "2025-01-16T20:23:05.208075+00:00"} {"global_step": 13746, "acc_step": 0, "speed/wps": 12908.48643449683, "speed/FLOPS": 202745517572955.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055457837879657745, "optim/lr": 0.002816723923717942, "optim/total_tokens": 7206862848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.8747949600219727, "created_at": "2025-01-16T20:23:15.366670+00:00"} {"global_step": 13747, "acc_step": 0, "speed/wps": 12904.660601049096, "speed/FLOPS": 202685427601413.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05842917039990425, "optim/lr": 0.0028166828811323837, "optim/total_tokens": 7207387136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373317, "loss/out": 2.782287836074829, "created_at": "2025-01-16T20:23:25.529594+00:00"} {"global_step": 13748, "acc_step": 0, "speed/wps": 12905.342436499239, "speed/FLOPS": 202696136764097.6, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05931951850652695, "optim/lr": 0.0028166418342509157, "optim/total_tokens": 7207911424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.882122039794922, "created_at": "2025-01-16T20:23:35.689570+00:00"} {"global_step": 13749, "acc_step": 0, "speed/wps": 12906.972188490749, "speed/FLOPS": 202721734258638.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05442902073264122, "optim/lr": 0.0028166007830736734, "optim/total_tokens": 7208435712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.942744016647339, "created_at": "2025-01-16T20:23:45.848221+00:00"} {"global_step": 13750, "acc_step": 0, "speed/wps": 12906.445924807422, "speed/FLOPS": 202713468564329.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08055387437343597, "optim/lr": 0.0028165597276007897, "optim/total_tokens": 7208960000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.793602228164673, "created_at": "2025-01-16T20:23:56.007416+00:00"} {"global_step": 13751, "acc_step": 0, "speed/wps": 12906.32038080364, "speed/FLOPS": 202711496723234.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07205376029014587, "optim/lr": 0.0028165186678323993, "optim/total_tokens": 7209484288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.74280047416687, "created_at": "2025-01-16T20:24:06.170515+00:00"} {"global_step": 13752, "acc_step": 0, "speed/wps": 12905.874072004326, "speed/FLOPS": 202704486830248.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06370741128921509, "optim/lr": 0.0028164776037686353, "optim/total_tokens": 7210008576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 2.873779773712158, "created_at": "2025-01-16T20:24:16.332138+00:00"} {"global_step": 13753, "acc_step": 0, "speed/wps": 12908.290761489068, "speed/FLOPS": 202742444259485.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050892896950244904, "optim/lr": 0.0028164365354096327, "optim/total_tokens": 7210532864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.841959238052368, "created_at": "2025-01-16T20:24:26.490169+00:00"} {"global_step": 13754, "acc_step": 0, "speed/wps": 12907.492777393645, "speed/FLOPS": 202729910822722.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06316015869379044, "optim/lr": 0.0028163954627555246, "optim/total_tokens": 7211057152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.872955083847046, "created_at": "2025-01-16T20:24:36.651565+00:00"} {"global_step": 13755, "acc_step": 0, "speed/wps": 12906.556586076307, "speed/FLOPS": 202715206651621.8, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06722554564476013, "optim/lr": 0.0028163543858064455, "optim/total_tokens": 7211581440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 2.8583502769470215, "created_at": "2025-01-16T20:24:46.810700+00:00"} {"global_step": 13756, "acc_step": 0, "speed/wps": 12905.642426686465, "speed/FLOPS": 202700848522217.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04772695526480675, "optim/lr": 0.0028163133045625293, "optim/total_tokens": 7212105728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.753465175628662, "created_at": "2025-01-16T20:24:56.980502+00:00"} {"global_step": 13757, "acc_step": 0, "speed/wps": 12909.213324333929, "speed/FLOPS": 202756934376698.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04312252625823021, "optim/lr": 0.00281627221902391, "optim/total_tokens": 7212630016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 2.8653945922851562, "created_at": "2025-01-16T20:25:07.137909+00:00"} {"global_step": 13758, "acc_step": 0, "speed/wps": 12912.388788076254, "speed/FLOPS": 202806809398316.38, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05336201563477516, "optim/lr": 0.0028162311291907217, "optim/total_tokens": 7213154304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.9253530502319336, "created_at": "2025-01-16T20:25:17.292877+00:00"} {"global_step": 13759, "acc_step": 0, "speed/wps": 12907.24365280384, "speed/FLOPS": 202725997978706.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044549841433763504, "optim/lr": 0.0028161900350630984, "optim/total_tokens": 7213678592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.0485587120056152, "created_at": "2025-01-16T20:25:27.451581+00:00"} {"global_step": 13760, "acc_step": 0, "speed/wps": 12911.49813951271, "speed/FLOPS": 202792820538746.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04422403499484062, "optim/lr": 0.0028161489366411744, "optim/total_tokens": 7214202880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8258554935455322, "created_at": "2025-01-16T20:25:37.609316+00:00"} {"global_step": 13761, "acc_step": 0, "speed/wps": 12912.29129386287, "speed/FLOPS": 202805278117724.1, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045598484575748444, "optim/lr": 0.0028161078339250835, "optim/total_tokens": 7214727168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 2.7528109550476074, "created_at": "2025-01-16T20:25:47.766091+00:00"} {"global_step": 13762, "acc_step": 0, "speed/wps": 12903.308877092684, "speed/FLOPS": 202664196919214.22, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04479501396417618, "optim/lr": 0.00281606672691496, "optim/total_tokens": 7215251456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381609, "loss/out": 2.8831424713134766, "created_at": "2025-01-16T20:25:57.932271+00:00"} {"global_step": 13763, "acc_step": 0, "speed/wps": 12911.355105213546, "speed/FLOPS": 202790573988528.44, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04935142770409584, "optim/lr": 0.002816025615610938, "optim/total_tokens": 7215775744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 2.860912799835205, "created_at": "2025-01-16T20:26:08.091736+00:00"} {"global_step": 13764, "acc_step": 0, "speed/wps": 12905.988353298048, "speed/FLOPS": 202706281775005.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05627095326781273, "optim/lr": 0.0028159845000131512, "optim/total_tokens": 7216300032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.7822439670562744, "created_at": "2025-01-16T20:26:18.254264+00:00"} {"global_step": 13765, "acc_step": 0, "speed/wps": 12906.670214253525, "speed/FLOPS": 202716991338286.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04434702545404434, "optim/lr": 0.0028159433801217344, "optim/total_tokens": 7216824320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.917855978012085, "created_at": "2025-01-16T20:26:28.413267+00:00"} {"global_step": 13766, "acc_step": 0, "speed/wps": 12906.796874153113, "speed/FLOPS": 202718980706058.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05235721543431282, "optim/lr": 0.0028159022559368216, "optim/total_tokens": 7217348608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 2.9583969116210938, "created_at": "2025-01-16T20:26:38.574178+00:00"} {"global_step": 13767, "acc_step": 0, "speed/wps": 12913.043961173435, "speed/FLOPS": 202817099792109.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0564223937690258, "optim/lr": 0.0028158611274585466, "optim/total_tokens": 7217872896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456884, "loss/out": 2.913830280303955, "created_at": "2025-01-16T20:26:48.732470+00:00"} {"global_step": 13768, "acc_step": 0, "speed/wps": 12908.022235335491, "speed/FLOPS": 202738226687250.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05122225731611252, "optim/lr": 0.002815819994687044, "optim/total_tokens": 7218397184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.876253128051758, "created_at": "2025-01-16T20:26:58.890962+00:00"} {"global_step": 13769, "acc_step": 0, "speed/wps": 12904.341423394468, "speed/FLOPS": 202680414477751.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05204516276717186, "optim/lr": 0.002815778857622447, "optim/total_tokens": 7218921472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8186874389648438, "created_at": "2025-01-16T20:27:09.052102+00:00"} {"global_step": 13770, "acc_step": 0, "speed/wps": 12904.3556929249, "speed/FLOPS": 202680638600334.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06229447200894356, "optim/lr": 0.002815737716264892, "optim/total_tokens": 7219445760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.694889545440674, "created_at": "2025-01-16T20:27:19.213132+00:00"} {"global_step": 13771, "acc_step": 0, "speed/wps": 12910.343245297458, "speed/FLOPS": 202774681338105.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057889752089977264, "optim/lr": 0.0028156965706145113, "optim/total_tokens": 7219970048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.8382835388183594, "created_at": "2025-01-16T20:27:29.374309+00:00"} {"global_step": 13772, "acc_step": 0, "speed/wps": 12903.038545934938, "speed/FLOPS": 202659950996907.88, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046066634356975555, "optim/lr": 0.00281565542067144, "optim/total_tokens": 7220494336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 2.887362480163574, "created_at": "2025-01-16T20:27:39.536960+00:00"} {"global_step": 13773, "acc_step": 0, "speed/wps": 12905.093140488372, "speed/FLOPS": 202692221227678.34, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04776440188288689, "optim/lr": 0.002815614266435812, "optim/total_tokens": 7221018624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.7347359657287598, "created_at": "2025-01-16T20:27:49.697802+00:00"} {"global_step": 13774, "acc_step": 0, "speed/wps": 12910.39274495192, "speed/FLOPS": 202775458798198.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0426519475877285, "optim/lr": 0.0028155731079077613, "optim/total_tokens": 7221542912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.840318202972412, "created_at": "2025-01-16T20:27:59.853861+00:00"} {"global_step": 13775, "acc_step": 0, "speed/wps": 12909.181211519934, "speed/FLOPS": 202756430000827.28, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05852629989385605, "optim/lr": 0.0028155319450874224, "optim/total_tokens": 7222067200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318033, "loss/out": 2.805619239807129, "created_at": "2025-01-16T20:28:10.012183+00:00"} {"global_step": 13776, "acc_step": 0, "speed/wps": 12911.992574769736, "speed/FLOPS": 202800586323883.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0579080805182457, "optim/lr": 0.00281549077797493, "optim/total_tokens": 7222591488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.873325824737549, "created_at": "2025-01-16T20:28:20.166869+00:00"} {"global_step": 13777, "acc_step": 0, "speed/wps": 12906.582535734995, "speed/FLOPS": 202715614226670.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06307050585746765, "optim/lr": 0.0028154496065704183, "optim/total_tokens": 7223115776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8199052810668945, "created_at": "2025-01-16T20:28:30.328957+00:00"} {"global_step": 13778, "acc_step": 0, "speed/wps": 12911.132524353165, "speed/FLOPS": 202787078050258.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05259254202246666, "optim/lr": 0.002815408430874022, "optim/total_tokens": 7223640064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.85132098197937, "created_at": "2025-01-16T20:28:40.489745+00:00"} {"global_step": 13779, "acc_step": 0, "speed/wps": 12905.512265203635, "speed/FLOPS": 202698804157269.25, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05554131418466568, "optim/lr": 0.002815367250885874, "optim/total_tokens": 7224164352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.8873329162597656, "created_at": "2025-01-16T20:28:50.649581+00:00"} {"global_step": 13780, "acc_step": 0, "speed/wps": 12912.594651939473, "speed/FLOPS": 202810042773174.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06241261586546898, "optim/lr": 0.0028153260666061103, "optim/total_tokens": 7224688640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.9108357429504395, "created_at": "2025-01-16T20:29:00.807201+00:00"} {"global_step": 13781, "acc_step": 0, "speed/wps": 12912.063826121183, "speed/FLOPS": 202801705424267.44, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04823543503880501, "optim/lr": 0.002815284878034865, "optim/total_tokens": 7225212928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.908764123916626, "created_at": "2025-01-16T20:29:10.964633+00:00"} {"global_step": 13782, "acc_step": 0, "speed/wps": 12904.865467161622, "speed/FLOPS": 202688645305225.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05044488608837128, "optim/lr": 0.0028152436851722707, "optim/total_tokens": 7225737216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372460, "loss/out": 2.7996826171875, "created_at": "2025-01-16T20:29:21.126378+00:00"} {"global_step": 13783, "acc_step": 0, "speed/wps": 12909.50768384638, "speed/FLOPS": 202761557697334.1, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05173620581626892, "optim/lr": 0.002815202488018464, "optim/total_tokens": 7226261504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8063478469848633, "created_at": "2025-01-16T20:29:31.283603+00:00"} {"global_step": 13784, "acc_step": 0, "speed/wps": 12908.127002928048, "speed/FLOPS": 202739872206257.3, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05113239586353302, "optim/lr": 0.0028151612865735787, "optim/total_tokens": 7226785792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375761, "loss/out": 2.840834856033325, "created_at": "2025-01-16T20:29:41.445269+00:00"} {"global_step": 13785, "acc_step": 0, "speed/wps": 12909.258849731605, "speed/FLOPS": 202757649415627.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04976153373718262, "optim/lr": 0.0028151200808377483, "optim/total_tokens": 7227310080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.9250035285949707, "created_at": "2025-01-16T20:29:51.602812+00:00"} {"global_step": 13786, "acc_step": 0, "speed/wps": 12904.438773928852, "speed/FLOPS": 202681943501667.2, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07204623520374298, "optim/lr": 0.0028150788708111085, "optim/total_tokens": 7227834368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8970742225646973, "created_at": "2025-01-16T20:30:01.765327+00:00"} {"global_step": 13787, "acc_step": 0, "speed/wps": 12904.640361269685, "speed/FLOPS": 202685109707865.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.058678481727838516, "optim/lr": 0.0028150376564937928, "optim/total_tokens": 7228358656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341575, "loss/out": 2.925321578979492, "created_at": "2025-01-16T20:30:11.928104+00:00"} {"global_step": 13788, "acc_step": 0, "speed/wps": 12908.444519712417, "speed/FLOPS": 202744859243669.6, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0545375719666481, "optim/lr": 0.002814996437885936, "optim/total_tokens": 7228882944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.899122714996338, "created_at": "2025-01-16T20:30:22.085912+00:00"} {"global_step": 13789, "acc_step": 0, "speed/wps": 12906.84735983815, "speed/FLOPS": 202719773653116.6, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.041283514350652695, "optim/lr": 0.002814955214987673, "optim/total_tokens": 7229407232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.7780799865722656, "created_at": "2025-01-16T20:30:32.248523+00:00"} {"global_step": 13790, "acc_step": 0, "speed/wps": 12905.20306498948, "speed/FLOPS": 202693947743020.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048629771918058395, "optim/lr": 0.0028149139877991377, "optim/total_tokens": 7229931520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9964945316314697, "created_at": "2025-01-16T20:30:42.413813+00:00"} {"global_step": 13791, "acc_step": 0, "speed/wps": 12904.66215349782, "speed/FLOPS": 202685451984753.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05075780674815178, "optim/lr": 0.002814872756320465, "optim/total_tokens": 7230455808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8051252365112305, "created_at": "2025-01-16T20:30:52.578507+00:00"} {"global_step": 13792, "acc_step": 0, "speed/wps": 12906.142650812599, "speed/FLOPS": 202708705229498.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04744546487927437, "optim/lr": 0.0028148315205517895, "optim/total_tokens": 7230980096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 2.914241313934326, "created_at": "2025-01-16T20:31:02.744249+00:00"} {"global_step": 13793, "acc_step": 0, "speed/wps": 12905.969102317213, "speed/FLOPS": 202705979411897.97, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04398699477314949, "optim/lr": 0.002814790280493245, "optim/total_tokens": 7231504384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.825747013092041, "created_at": "2025-01-16T20:31:12.908364+00:00"} {"global_step": 13794, "acc_step": 0, "speed/wps": 12909.082606523372, "speed/FLOPS": 202754881273858.8, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0546075738966465, "optim/lr": 0.002814749036144967, "optim/total_tokens": 7232028672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.9729130268096924, "created_at": "2025-01-16T20:31:23.067807+00:00"} {"global_step": 13795, "acc_step": 0, "speed/wps": 12904.829787026243, "speed/FLOPS": 202688084899670.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04852735996246338, "optim/lr": 0.0028147077875070895, "optim/total_tokens": 7232552960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460830, "loss/out": 2.9040253162384033, "created_at": "2025-01-16T20:31:33.228593+00:00"} {"global_step": 13796, "acc_step": 0, "speed/wps": 12911.864400843262, "speed/FLOPS": 202798573176238.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07277030497789383, "optim/lr": 0.002814666534579747, "optim/total_tokens": 7233077248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.791851043701172, "created_at": "2025-01-16T20:31:43.385306+00:00"} {"global_step": 13797, "acc_step": 0, "speed/wps": 12908.454822564254, "speed/FLOPS": 202745021064115.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06700216233730316, "optim/lr": 0.0028146252773630746, "optim/total_tokens": 7233601536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.8489224910736084, "created_at": "2025-01-16T20:31:53.544205+00:00"} {"global_step": 13798, "acc_step": 0, "speed/wps": 12903.438167185357, "speed/FLOPS": 202666227597782.84, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05093742534518242, "optim/lr": 0.0028145840158572064, "optim/total_tokens": 7234125824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.8964474201202393, "created_at": "2025-01-16T20:32:03.709027+00:00"} {"global_step": 13799, "acc_step": 0, "speed/wps": 12904.879839615614, "speed/FLOPS": 202688871044365.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.058709193021059036, "optim/lr": 0.0028145427500622777, "optim/total_tokens": 7234650112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8415324687957764, "created_at": "2025-01-16T20:32:13.870528+00:00"} {"global_step": 13800, "acc_step": 0, "speed/wps": 12905.034950794945, "speed/FLOPS": 202691307278582.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048500075936317444, "optim/lr": 0.0028145014799784225, "optim/total_tokens": 7235174400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370151, "loss/out": 2.8545117378234863, "created_at": "2025-01-16T20:32:24.030850+00:00"} {"global_step": 13801, "acc_step": 0, "speed/wps": 12908.406052878181, "speed/FLOPS": 202744255069178.97, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05230789631605148, "optim/lr": 0.0028144602056057754, "optim/total_tokens": 7235698688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.665220022201538, "created_at": "2025-01-16T20:32:34.195466+00:00"} {"global_step": 13802, "acc_step": 0, "speed/wps": 12908.772254834634, "speed/FLOPS": 202750006774119.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04989970847964287, "optim/lr": 0.0028144189269444712, "optim/total_tokens": 7236222976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8707380294799805, "created_at": "2025-01-16T20:32:44.357564+00:00"} {"global_step": 13803, "acc_step": 0, "speed/wps": 12909.198202162566, "speed/FLOPS": 202756696862217.4, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059360701590776443, "optim/lr": 0.0028143776439946452, "optim/total_tokens": 7236747264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.8567376136779785, "created_at": "2025-01-16T20:32:54.519331+00:00"} {"global_step": 13804, "acc_step": 0, "speed/wps": 12911.013325369084, "speed/FLOPS": 202785205866416.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057935845106840134, "optim/lr": 0.0028143363567564314, "optim/total_tokens": 7237271552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8980050086975098, "created_at": "2025-01-16T20:33:04.677483+00:00"} {"global_step": 13805, "acc_step": 0, "speed/wps": 12905.360999240223, "speed/FLOPS": 202696428317452.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053650181740522385, "optim/lr": 0.0028142950652299647, "optim/total_tokens": 7237795840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341862, "loss/out": 2.898423433303833, "created_at": "2025-01-16T20:33:14.838274+00:00"} {"global_step": 13806, "acc_step": 0, "speed/wps": 12908.148615140484, "speed/FLOPS": 202740211655751.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04848442226648331, "optim/lr": 0.0028142537694153796, "optim/total_tokens": 7238320128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8447513580322266, "created_at": "2025-01-16T20:33:24.996738+00:00"} {"global_step": 13807, "acc_step": 0, "speed/wps": 12905.902211937057, "speed/FLOPS": 202704928806560.6, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05323801189661026, "optim/lr": 0.002814212469312811, "optim/total_tokens": 7238844416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.969973087310791, "created_at": "2025-01-16T20:33:35.156203+00:00"} {"global_step": 13808, "acc_step": 0, "speed/wps": 12906.945626482262, "speed/FLOPS": 202721317065794.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06650766730308533, "optim/lr": 0.0028141711649223937, "optim/total_tokens": 7239368704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302164, "loss/out": 2.8589258193969727, "created_at": "2025-01-16T20:33:45.315952+00:00"} {"global_step": 13809, "acc_step": 0, "speed/wps": 12901.027249998111, "speed/FLOPS": 202628360830409.75, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048694461584091187, "optim/lr": 0.002814129856244263, "optim/total_tokens": 7239892992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.859691619873047, "created_at": "2025-01-16T20:33:55.482267+00:00"} {"global_step": 13810, "acc_step": 0, "speed/wps": 12905.678150369698, "speed/FLOPS": 202701409611752.38, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05810835212469101, "optim/lr": 0.002814088543278552, "optim/total_tokens": 7240417280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351470, "loss/out": 2.8339221477508545, "created_at": "2025-01-16T20:34:05.641956+00:00"} {"global_step": 13811, "acc_step": 0, "speed/wps": 12907.866185754243, "speed/FLOPS": 202735775714142.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053020089864730835, "optim/lr": 0.0028140472260253973, "optim/total_tokens": 7240941568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.881899118423462, "created_at": "2025-01-16T20:34:15.801579+00:00"} {"global_step": 13812, "acc_step": 0, "speed/wps": 12905.950247868044, "speed/FLOPS": 202705683276865.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058300815522670746, "optim/lr": 0.002814005904484933, "optim/total_tokens": 7241465856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.788011074066162, "created_at": "2025-01-16T20:34:25.961085+00:00"} {"global_step": 13813, "acc_step": 0, "speed/wps": 12906.512759276126, "speed/FLOPS": 202714518291500.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0534355454146862, "optim/lr": 0.002813964578657294, "optim/total_tokens": 7241990144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 2.8495423793792725, "created_at": "2025-01-16T20:34:36.121925+00:00"} {"global_step": 13814, "acc_step": 0, "speed/wps": 12905.761628297163, "speed/FLOPS": 202702720747313.9, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06521300971508026, "optim/lr": 0.0028139232485426147, "optim/total_tokens": 7242514432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.7386226654052734, "created_at": "2025-01-16T20:34:46.282568+00:00"} {"global_step": 13815, "acc_step": 0, "speed/wps": 12908.255547820774, "speed/FLOPS": 202741891180436.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05231054499745369, "optim/lr": 0.002813881914141031, "optim/total_tokens": 7243038720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.7870047092437744, "created_at": "2025-01-16T20:34:56.444631+00:00"} {"global_step": 13816, "acc_step": 0, "speed/wps": 12910.818991102622, "speed/FLOPS": 202782153579718.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06086438149213791, "optim/lr": 0.0028138405754526766, "optim/total_tokens": 7243563008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.8238892555236816, "created_at": "2025-01-16T20:35:06.601087+00:00"} {"global_step": 13817, "acc_step": 0, "speed/wps": 12916.200002635102, "speed/FLOPS": 202866669760120.7, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05029762536287308, "optim/lr": 0.0028137992324776865, "optim/total_tokens": 7244087296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468286, "loss/out": 2.87453031539917, "created_at": "2025-01-16T20:35:16.753638+00:00"} {"global_step": 13818, "acc_step": 0, "speed/wps": 12909.592130240952, "speed/FLOPS": 202762884044002.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0609477236866951, "optim/lr": 0.0028137578852161964, "optim/total_tokens": 7244611584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474863, "loss/out": 2.885645866394043, "created_at": "2025-01-16T20:35:26.910529+00:00"} {"global_step": 13819, "acc_step": 0, "speed/wps": 12910.11323605601, "speed/FLOPS": 202771068726902.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049248144030570984, "optim/lr": 0.002813716533668341, "optim/total_tokens": 7245135872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9901785850524902, "created_at": "2025-01-16T20:35:37.070753+00:00"} {"global_step": 13820, "acc_step": 0, "speed/wps": 12905.678990623026, "speed/FLOPS": 202701422809085.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06362468749284744, "optim/lr": 0.0028136751778342546, "optim/total_tokens": 7245660160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 2.7353920936584473, "created_at": "2025-01-16T20:35:47.230854+00:00"} {"global_step": 13821, "acc_step": 0, "speed/wps": 12909.140475129616, "speed/FLOPS": 202755790179839.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06128155440092087, "optim/lr": 0.0028136338177140725, "optim/total_tokens": 7246184448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8669750690460205, "created_at": "2025-01-16T20:35:57.389361+00:00"} {"global_step": 13822, "acc_step": 0, "speed/wps": 12915.88337962073, "speed/FLOPS": 202861696760594.34, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049161531031131744, "optim/lr": 0.00281359245330793, "optim/total_tokens": 7246708736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 2.906116008758545, "created_at": "2025-01-16T20:36:07.545569+00:00"} {"global_step": 13823, "acc_step": 0, "speed/wps": 12909.709467482251, "speed/FLOPS": 202764726986615.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059891849756240845, "optim/lr": 0.0028135510846159614, "optim/total_tokens": 7247233024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.926659107208252, "created_at": "2025-01-16T20:36:17.702959+00:00"} {"global_step": 13824, "acc_step": 0, "speed/wps": 12910.102508957661, "speed/FLOPS": 202770900243082.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05309899151325226, "optim/lr": 0.0028135097116383017, "optim/total_tokens": 7247757312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8296101093292236, "created_at": "2025-01-16T20:36:27.859574+00:00"} {"global_step": 13825, "acc_step": 0, "speed/wps": 12909.997648898136, "speed/FLOPS": 202769253271754.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06032165512442589, "optim/lr": 0.0028134683343750866, "optim/total_tokens": 7248281600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.7089810371398926, "created_at": "2025-01-16T20:36:38.016192+00:00"} {"global_step": 13826, "acc_step": 0, "speed/wps": 12901.858000139433, "speed/FLOPS": 202641408902949.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04720626771450043, "optim/lr": 0.002813426952826451, "optim/total_tokens": 7248805888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.7973411083221436, "created_at": "2025-01-16T20:36:48.178989+00:00"} {"global_step": 13827, "acc_step": 0, "speed/wps": 12909.230303220025, "speed/FLOPS": 202757201053436.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051414087414741516, "optim/lr": 0.0028133855669925287, "optim/total_tokens": 7249330176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468042, "loss/out": 2.8388733863830566, "created_at": "2025-01-16T20:36:58.339360+00:00"} {"global_step": 13828, "acc_step": 0, "speed/wps": 12916.130147041373, "speed/FLOPS": 202865572582029.53, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06015409901738167, "optim/lr": 0.002813344176873456, "optim/total_tokens": 7249854464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.7722158432006836, "created_at": "2025-01-16T20:37:08.493854+00:00"} {"global_step": 13829, "acc_step": 0, "speed/wps": 12915.233399956818, "speed/FLOPS": 202851487936806.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05699581652879715, "optim/lr": 0.0028133027824693677, "optim/total_tokens": 7250378752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376816, "loss/out": 2.9162096977233887, "created_at": "2025-01-16T20:37:18.647163+00:00"} {"global_step": 13830, "acc_step": 0, "speed/wps": 12912.82704568667, "speed/FLOPS": 202813692836317.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05269410088658333, "optim/lr": 0.0028132613837803986, "optim/total_tokens": 7250903040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.844115734100342, "created_at": "2025-01-16T20:37:28.803996+00:00"} {"global_step": 13831, "acc_step": 0, "speed/wps": 12915.09798302022, "speed/FLOPS": 202849361027735.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06068392097949982, "optim/lr": 0.0028132199808066837, "optim/total_tokens": 7251427328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.7718441486358643, "created_at": "2025-01-16T20:37:38.957794+00:00"} {"global_step": 13832, "acc_step": 0, "speed/wps": 12904.38346653843, "speed/FLOPS": 202681074823100.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044834524393081665, "optim/lr": 0.0028131785735483585, "optim/total_tokens": 7251951616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8939218521118164, "created_at": "2025-01-16T20:37:49.121437+00:00"} {"global_step": 13833, "acc_step": 0, "speed/wps": 12913.88160268007, "speed/FLOPS": 202830256103011.53, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07979724556207657, "optim/lr": 0.0028131371620055577, "optim/total_tokens": 7252475904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8324038982391357, "created_at": "2025-01-16T20:37:59.276426+00:00"} {"global_step": 13834, "acc_step": 0, "speed/wps": 12914.728198563165, "speed/FLOPS": 202843553054700.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.045653264969587326, "optim/lr": 0.002813095746178416, "optim/total_tokens": 7253000192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 2.7579970359802246, "created_at": "2025-01-16T20:38:09.430290+00:00"} {"global_step": 13835, "acc_step": 0, "speed/wps": 12906.101286204503, "speed/FLOPS": 202708055541487.53, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.073736771941185, "optim/lr": 0.00281305432606707, "optim/total_tokens": 7253524480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 2.8472275733947754, "created_at": "2025-01-16T20:38:19.591028+00:00"} {"global_step": 13836, "acc_step": 0, "speed/wps": 12910.350408014667, "speed/FLOPS": 202774793838421.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053956255316734314, "optim/lr": 0.002813012901671654, "optim/total_tokens": 7254048768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 2.9430699348449707, "created_at": "2025-01-16T20:38:29.748403+00:00"} {"global_step": 13837, "acc_step": 0, "speed/wps": 12905.783629053574, "speed/FLOPS": 202703066299425.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0768929272890091, "optim/lr": 0.0028129714729923025, "optim/total_tokens": 7254573056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.9171719551086426, "created_at": "2025-01-16T20:38:39.908032+00:00"} {"global_step": 13838, "acc_step": 0, "speed/wps": 12901.913171155744, "speed/FLOPS": 202642275439573.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0737076997756958, "optim/lr": 0.0028129300400291514, "optim/total_tokens": 7255097344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.913832664489746, "created_at": "2025-01-16T20:38:50.073026+00:00"} {"global_step": 13839, "acc_step": 0, "speed/wps": 12906.559870612724, "speed/FLOPS": 202715258239779.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07647693902254105, "optim/lr": 0.0028128886027823357, "optim/total_tokens": 7255621632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 2.699824810028076, "created_at": "2025-01-16T20:39:00.236305+00:00"} {"global_step": 13840, "acc_step": 0, "speed/wps": 12912.078340184335, "speed/FLOPS": 202801933387573.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07362806797027588, "optim/lr": 0.002812847161251991, "optim/total_tokens": 7256145920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.673362970352173, "created_at": "2025-01-16T20:39:10.391079+00:00"} {"global_step": 13841, "acc_step": 0, "speed/wps": 12902.793295812253, "speed/FLOPS": 202656099006722.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05559930205345154, "optim/lr": 0.002812805715438252, "optim/total_tokens": 7256670208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.928463935852051, "created_at": "2025-01-16T20:39:20.556849+00:00"} {"global_step": 13842, "acc_step": 0, "speed/wps": 12908.835062815973, "speed/FLOPS": 202750993259773.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06979101896286011, "optim/lr": 0.0028127642653412535, "optim/total_tokens": 7257194496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.927856206893921, "created_at": "2025-01-16T20:39:30.715992+00:00"} {"global_step": 13843, "acc_step": 0, "speed/wps": 12908.359591073195, "speed/FLOPS": 202743525322686.44, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04094164818525314, "optim/lr": 0.002812722810961132, "optim/total_tokens": 7257718784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.81668758392334, "created_at": "2025-01-16T20:39:40.876099+00:00"} {"global_step": 13844, "acc_step": 0, "speed/wps": 12903.616065227823, "speed/FLOPS": 202669021730997.3, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.07287784665822983, "optim/lr": 0.002812681352298022, "optim/total_tokens": 7258243072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 2.9271156787872314, "created_at": "2025-01-16T20:39:51.038465+00:00"} {"global_step": 13845, "acc_step": 0, "speed/wps": 12912.639615931543, "speed/FLOPS": 202810748994457.28, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05341758206486702, "optim/lr": 0.0028126398893520586, "optim/total_tokens": 7258767360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.883701801300049, "created_at": "2025-01-16T20:40:01.193349+00:00"} {"global_step": 13846, "acc_step": 0, "speed/wps": 12911.13001449168, "speed/FLOPS": 202787038629434.47, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0527711883187294, "optim/lr": 0.0028125984221233774, "optim/total_tokens": 7259291648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 3.013171911239624, "created_at": "2025-01-16T20:40:11.351212+00:00"} {"global_step": 13847, "acc_step": 0, "speed/wps": 12913.633874889765, "speed/FLOPS": 202826365197651.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06268474459648132, "optim/lr": 0.0028125569506121137, "optim/total_tokens": 7259815936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.8937196731567383, "created_at": "2025-01-16T20:40:21.504568+00:00"} {"global_step": 13848, "acc_step": 0, "speed/wps": 12914.085652514941, "speed/FLOPS": 202833460986062.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04500139504671097, "optim/lr": 0.0028125154748184023, "optim/total_tokens": 7260340224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.7734498977661133, "created_at": "2025-01-16T20:40:31.663282+00:00"} {"global_step": 13849, "acc_step": 0, "speed/wps": 12909.38618329903, "speed/FLOPS": 202759649364278.72, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050355829298496246, "optim/lr": 0.0028124739947423794, "optim/total_tokens": 7260864512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 2.9128265380859375, "created_at": "2025-01-16T20:40:41.823951+00:00"} {"global_step": 13850, "acc_step": 0, "speed/wps": 12908.495695520034, "speed/FLOPS": 202745663030051.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042317990213632584, "optim/lr": 0.0028124325103841795, "optim/total_tokens": 7261388800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.985208034515381, "created_at": "2025-01-16T20:40:51.981562+00:00"} {"global_step": 13851, "acc_step": 0, "speed/wps": 12908.10602535869, "speed/FLOPS": 202739542724704.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059588510543107986, "optim/lr": 0.0028123910217439383, "optim/total_tokens": 7261913088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 2.859405517578125, "created_at": "2025-01-16T20:41:02.140275+00:00"} {"global_step": 13852, "acc_step": 0, "speed/wps": 12908.485735954326, "speed/FLOPS": 202745506601385.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05043168365955353, "optim/lr": 0.0028123495288217915, "optim/total_tokens": 7262437376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 2.8678126335144043, "created_at": "2025-01-16T20:41:12.300229+00:00"} {"global_step": 13853, "acc_step": 0, "speed/wps": 12908.952018054042, "speed/FLOPS": 202752830202500.62, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05294480174779892, "optim/lr": 0.0028123080316178745, "optim/total_tokens": 7262961664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.871018409729004, "created_at": "2025-01-16T20:41:22.458470+00:00"} {"global_step": 13854, "acc_step": 0, "speed/wps": 12906.513343978695, "speed/FLOPS": 202714527475057.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05514907464385033, "optim/lr": 0.0028122665301323216, "optim/total_tokens": 7263485952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 2.8357229232788086, "created_at": "2025-01-16T20:41:32.617531+00:00"} {"global_step": 13855, "acc_step": 0, "speed/wps": 12900.519076161128, "speed/FLOPS": 202620379261995.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05008113011717796, "optim/lr": 0.0028122250243652696, "optim/total_tokens": 7264010240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 2.878037929534912, "created_at": "2025-01-16T20:41:42.786014+00:00"} {"global_step": 13856, "acc_step": 0, "speed/wps": 12891.417506570513, "speed/FLOPS": 202477426604708.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04977084696292877, "optim/lr": 0.0028121835143168528, "optim/total_tokens": 7264534528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 2.8495113849639893, "created_at": "2025-01-16T20:41:52.958408+00:00"} {"global_step": 13857, "acc_step": 0, "speed/wps": 12902.478729076365, "speed/FLOPS": 202651158303875.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051610853523015976, "optim/lr": 0.0028121419999872074, "optim/total_tokens": 7265058816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.8799166679382324, "created_at": "2025-01-16T20:42:03.121790+00:00"} {"global_step": 13858, "acc_step": 0, "speed/wps": 12901.628278947757, "speed/FLOPS": 202637800815964.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05785970389842987, "optim/lr": 0.0028121004813764683, "optim/total_tokens": 7265583104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.7906060218811035, "created_at": "2025-01-16T20:42:13.286099+00:00"} {"global_step": 13859, "acc_step": 0, "speed/wps": 12902.853945911107, "speed/FLOPS": 202657051599866.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047925449907779694, "optim/lr": 0.0028120589584847712, "optim/total_tokens": 7266107392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 2.956653594970703, "created_at": "2025-01-16T20:42:23.451484+00:00"} {"global_step": 13860, "acc_step": 0, "speed/wps": 12909.28179076036, "speed/FLOPS": 202758009736008.53, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05833744630217552, "optim/lr": 0.0028120174313122515, "optim/total_tokens": 7266631680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.9002633094787598, "created_at": "2025-01-16T20:42:33.609391+00:00"} {"global_step": 13861, "acc_step": 0, "speed/wps": 12904.892581498518, "speed/FLOPS": 202689071173145.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05102405697107315, "optim/lr": 0.0028119758998590453, "optim/total_tokens": 7267155968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.79697322845459, "created_at": "2025-01-16T20:42:43.774575+00:00"} {"global_step": 13862, "acc_step": 0, "speed/wps": 12902.810488485846, "speed/FLOPS": 202656369041286.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060179829597473145, "optim/lr": 0.0028119343641252875, "optim/total_tokens": 7267680256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 2.899570941925049, "created_at": "2025-01-16T20:42:53.937380+00:00"} {"global_step": 13863, "acc_step": 0, "speed/wps": 12907.779178797293, "speed/FLOPS": 202734409150323.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05628013610839844, "optim/lr": 0.0028118928241111134, "optim/total_tokens": 7268204544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.7339110374450684, "created_at": "2025-01-16T20:43:04.098098+00:00"} {"global_step": 13864, "acc_step": 0, "speed/wps": 12901.775932854169, "speed/FLOPS": 202640119923462.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1415075957775116, "optim/lr": 0.002811851279816659, "optim/total_tokens": 7268728832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 2.891329288482666, "created_at": "2025-01-16T20:43:14.260936+00:00"} {"global_step": 13865, "acc_step": 0, "speed/wps": 12898.389649190045, "speed/FLOPS": 202586933685275.75, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05986126884818077, "optim/lr": 0.0028118097312420597, "optim/total_tokens": 7269253120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8272318840026855, "created_at": "2025-01-16T20:43:24.427523+00:00"} {"global_step": 13866, "acc_step": 0, "speed/wps": 12901.028662025637, "speed/FLOPS": 202628383008242.38, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0529562272131443, "optim/lr": 0.002811768178387451, "optim/total_tokens": 7269777408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.743753433227539, "created_at": "2025-01-16T20:43:34.592777+00:00"} {"global_step": 13867, "acc_step": 0, "speed/wps": 12908.288497806667, "speed/FLOPS": 202742408705243.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06285028159618378, "optim/lr": 0.0028117266212529684, "optim/total_tokens": 7270301696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 2.977506637573242, "created_at": "2025-01-16T20:43:44.750898+00:00"} {"global_step": 13868, "acc_step": 0, "speed/wps": 12906.335151812867, "speed/FLOPS": 202711728722231.94, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05399063974618912, "optim/lr": 0.002811685059838748, "optim/total_tokens": 7270825984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8115954399108887, "created_at": "2025-01-16T20:43:54.910732+00:00"} {"global_step": 13869, "acc_step": 0, "speed/wps": 12909.096847466553, "speed/FLOPS": 202755104947440.6, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.057221755385398865, "optim/lr": 0.002811643494144925, "optim/total_tokens": 7271350272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 2.8083841800689697, "created_at": "2025-01-16T20:44:05.068947+00:00"} {"global_step": 13870, "acc_step": 0, "speed/wps": 12902.651230295009, "speed/FLOPS": 202653867672555.3, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04343356192111969, "optim/lr": 0.002811601924171635, "optim/total_tokens": 7271874560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.9038896560668945, "created_at": "2025-01-16T20:44:15.233319+00:00"} {"global_step": 13871, "acc_step": 0, "speed/wps": 12912.262656935207, "speed/FLOPS": 202804828335423.88, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05070202052593231, "optim/lr": 0.002811560349919013, "optim/total_tokens": 7272398848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.832700252532959, "created_at": "2025-01-16T20:44:25.388363+00:00"} {"global_step": 13872, "acc_step": 0, "speed/wps": 12912.707114043196, "speed/FLOPS": 202811809145052.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04473811015486717, "optim/lr": 0.0028115187713871957, "optim/total_tokens": 7272923136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.7728354930877686, "created_at": "2025-01-16T20:44:35.546482+00:00"} {"global_step": 13873, "acc_step": 0, "speed/wps": 12903.81477556091, "speed/FLOPS": 202672142749835.03, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052309975028038025, "optim/lr": 0.0028114771885763183, "optim/total_tokens": 7273447424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.8557941913604736, "created_at": "2025-01-16T20:44:45.710315+00:00"} {"global_step": 13874, "acc_step": 0, "speed/wps": 12909.781526136598, "speed/FLOPS": 202765858766800.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3087650239467621, "optim/lr": 0.0028114356014865166, "optim/total_tokens": 7273971712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.939725875854492, "created_at": "2025-01-16T20:44:55.868464+00:00"} {"global_step": 13875, "acc_step": 0, "speed/wps": 12906.585074602937, "speed/FLOPS": 202715654103079.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07791721075773239, "optim/lr": 0.002811394010117926, "optim/total_tokens": 7274496000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.7531681060791016, "created_at": "2025-01-16T20:45:06.028739+00:00"} {"global_step": 13876, "acc_step": 0, "speed/wps": 12913.567582798983, "speed/FLOPS": 202825323989270.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08220407366752625, "optim/lr": 0.002811352414470683, "optim/total_tokens": 7275020288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497719, "loss/out": 2.906874656677246, "created_at": "2025-01-16T20:45:16.183314+00:00"} {"global_step": 13877, "acc_step": 0, "speed/wps": 12909.685421267914, "speed/FLOPS": 202764349307776.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05768272653222084, "optim/lr": 0.0028113108145449224, "optim/total_tokens": 7275544576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.899810791015625, "created_at": "2025-01-16T20:45:26.340057+00:00"} {"global_step": 13878, "acc_step": 0, "speed/wps": 12907.507593703032, "speed/FLOPS": 202730143533220.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05267614126205444, "optim/lr": 0.0028112692103407803, "optim/total_tokens": 7276068864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8819031715393066, "created_at": "2025-01-16T20:45:36.503495+00:00"} {"global_step": 13879, "acc_step": 0, "speed/wps": 12910.077237936091, "speed/FLOPS": 202770503326962.38, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050052739679813385, "optim/lr": 0.0028112276018583925, "optim/total_tokens": 7276593152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418499, "loss/out": 2.893125534057617, "created_at": "2025-01-16T20:45:46.663187+00:00"} {"global_step": 13880, "acc_step": 0, "speed/wps": 12909.43690002037, "speed/FLOPS": 202760445940079.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04889840632677078, "optim/lr": 0.002811185989097894, "optim/total_tokens": 7277117440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.7648391723632812, "created_at": "2025-01-16T20:45:56.820574+00:00"} {"global_step": 13881, "acc_step": 0, "speed/wps": 12911.68429353099, "speed/FLOPS": 202795744343406.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0556405708193779, "optim/lr": 0.002811144372059422, "optim/total_tokens": 7277641728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8709840774536133, "created_at": "2025-01-16T20:46:06.978954+00:00"} {"global_step": 13882, "acc_step": 0, "speed/wps": 12910.684576319278, "speed/FLOPS": 202780042410840.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06265158951282501, "optim/lr": 0.002811102750743111, "optim/total_tokens": 7278166016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.819925308227539, "created_at": "2025-01-16T20:46:17.134651+00:00"} {"global_step": 13883, "acc_step": 0, "speed/wps": 12906.99696085815, "speed/FLOPS": 202722123342709.03, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.046623215079307556, "optim/lr": 0.002811061125149098, "optim/total_tokens": 7278690304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.879404067993164, "created_at": "2025-01-16T20:46:27.293293+00:00"} {"global_step": 13884, "acc_step": 0, "speed/wps": 12909.674566988157, "speed/FLOPS": 202764178826397.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04439504072070122, "optim/lr": 0.002811019495277517, "optim/total_tokens": 7279214592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.769972324371338, "created_at": "2025-01-16T20:46:37.451707+00:00"} {"global_step": 13885, "acc_step": 0, "speed/wps": 12905.07882199315, "speed/FLOPS": 202691996336035.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041776690632104874, "optim/lr": 0.002810977861128506, "optim/total_tokens": 7279738880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.7680091857910156, "created_at": "2025-01-16T20:46:47.612633+00:00"} {"global_step": 13886, "acc_step": 0, "speed/wps": 12907.261548646951, "speed/FLOPS": 202726279057513.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04388485476374626, "optim/lr": 0.002810936222702199, "optim/total_tokens": 7280263168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.6851296424865723, "created_at": "2025-01-16T20:46:57.774014+00:00"} {"global_step": 13887, "acc_step": 0, "speed/wps": 12910.319694130865, "speed/FLOPS": 202774311434671.0, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048299115151166916, "optim/lr": 0.002810894579998733, "optim/total_tokens": 7280787456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.842319965362549, "created_at": "2025-01-16T20:47:07.932041+00:00"} {"global_step": 13888, "acc_step": 0, "speed/wps": 12904.420932058165, "speed/FLOPS": 202681663270570.75, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04897996783256531, "optim/lr": 0.002810852933018244, "optim/total_tokens": 7281311744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.8160486221313477, "created_at": "2025-01-16T20:47:18.097037+00:00"} {"global_step": 13889, "acc_step": 0, "speed/wps": 12907.713213164276, "speed/FLOPS": 202733373069410.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04677148535847664, "optim/lr": 0.0028108112817608667, "optim/total_tokens": 7281836032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 2.8257083892822266, "created_at": "2025-01-16T20:47:28.255591+00:00"} {"global_step": 13890, "acc_step": 0, "speed/wps": 12916.623986384799, "speed/FLOPS": 202873329007524.4, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04084618762135506, "optim/lr": 0.0028107696262267374, "optim/total_tokens": 7282360320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.8432698249816895, "created_at": "2025-01-16T20:47:38.407204+00:00"} {"global_step": 13891, "acc_step": 0, "speed/wps": 12907.036754513225, "speed/FLOPS": 202722748356744.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048762235790491104, "optim/lr": 0.002810727966415993, "optim/total_tokens": 7282884608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347841, "loss/out": 2.8518481254577637, "created_at": "2025-01-16T20:47:48.565765+00:00"} {"global_step": 13892, "acc_step": 0, "speed/wps": 12907.27220828604, "speed/FLOPS": 202726446481793.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05944392457604408, "optim/lr": 0.002810686302328768, "optim/total_tokens": 7283408896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.879873514175415, "created_at": "2025-01-16T20:47:58.725639+00:00"} {"global_step": 13893, "acc_step": 0, "speed/wps": 12912.645712567333, "speed/FLOPS": 202810844750500.03, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047528769820928574, "optim/lr": 0.0028106446339651995, "optim/total_tokens": 7283933184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.918139934539795, "created_at": "2025-01-16T20:48:08.880250+00:00"} {"global_step": 13894, "acc_step": 0, "speed/wps": 12915.047713655656, "speed/FLOPS": 202848571478287.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059059467166662216, "optim/lr": 0.002810602961325423, "optim/total_tokens": 7284457472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 2.858112335205078, "created_at": "2025-01-16T20:48:19.032522+00:00"} {"global_step": 13895, "acc_step": 0, "speed/wps": 12912.656125198595, "speed/FLOPS": 202811008295182.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0499434731900692, "optim/lr": 0.002810561284409574, "optim/total_tokens": 7284981760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.8788280487060547, "created_at": "2025-01-16T20:48:29.187807+00:00"} {"global_step": 13896, "acc_step": 0, "speed/wps": 12908.710813813685, "speed/FLOPS": 202749041758456.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05066097527742386, "optim/lr": 0.0028105196032177896, "optim/total_tokens": 7285506048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 2.83023738861084, "created_at": "2025-01-16T20:48:39.345293+00:00"} {"global_step": 13897, "acc_step": 0, "speed/wps": 12911.562253320008, "speed/FLOPS": 202793827534191.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04192109778523445, "optim/lr": 0.002810477917750205, "optim/total_tokens": 7286030336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.8989675045013428, "created_at": "2025-01-16T20:48:49.500716+00:00"} {"global_step": 13898, "acc_step": 0, "speed/wps": 12910.136885948308, "speed/FLOPS": 202771440180959.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05681876465678215, "optim/lr": 0.0028104362280069557, "optim/total_tokens": 7286554624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348010, "loss/out": 2.832334518432617, "created_at": "2025-01-16T20:48:59.659722+00:00"} {"global_step": 13899, "acc_step": 0, "speed/wps": 12906.822803472072, "speed/FLOPS": 202719387961643.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051013655960559845, "optim/lr": 0.0028103945339881787, "optim/total_tokens": 7287078912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8651976585388184, "created_at": "2025-01-16T20:49:09.824478+00:00"} {"global_step": 13900, "acc_step": 0, "speed/wps": 12906.875202178255, "speed/FLOPS": 202720210955327.3, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054516974836587906, "optim/lr": 0.00281035283569401, "optim/total_tokens": 7287603200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.825282573699951, "created_at": "2025-01-16T20:49:19.983160+00:00"} {"global_step": 13901, "acc_step": 0, "speed/wps": 12909.193927559436, "speed/FLOPS": 202756629723701.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048655666410923004, "optim/lr": 0.0028103111331245854, "optim/total_tokens": 7288127488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307800, "loss/out": 2.80790376663208, "created_at": "2025-01-16T20:49:30.141959+00:00"} {"global_step": 13902, "acc_step": 0, "speed/wps": 12906.14508772296, "speed/FLOPS": 202708743504524.5, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05131823569536209, "optim/lr": 0.00281026942628004, "optim/total_tokens": 7288651776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.95976185798645, "created_at": "2025-01-16T20:49:40.306599+00:00"} {"global_step": 13903, "acc_step": 0, "speed/wps": 12909.184208472327, "speed/FLOPS": 202756477072082.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054763007909059525, "optim/lr": 0.0028102277151605118, "optim/total_tokens": 7289176064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 2.826106548309326, "created_at": "2025-01-16T20:49:50.466102+00:00"} {"global_step": 13904, "acc_step": 0, "speed/wps": 12912.780950129723, "speed/FLOPS": 202812968842253.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048817191272974014, "optim/lr": 0.002810185999766135, "optim/total_tokens": 7289700352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.7865381240844727, "created_at": "2025-01-16T20:50:00.625758+00:00"} {"global_step": 13905, "acc_step": 0, "speed/wps": 12911.180757944961, "speed/FLOPS": 202787835625097.06, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06073296070098877, "optim/lr": 0.002810144280097047, "optim/total_tokens": 7290224640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7965452671051025, "created_at": "2025-01-16T20:50:10.788006+00:00"} {"global_step": 13906, "acc_step": 0, "speed/wps": 12910.57931004537, "speed/FLOPS": 202778389059357.97, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06728252023458481, "optim/lr": 0.0028101025561533834, "optim/total_tokens": 7290748928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.7770891189575195, "created_at": "2025-01-16T20:50:20.943731+00:00"} {"global_step": 13907, "acc_step": 0, "speed/wps": 12909.17960943572, "speed/FLOPS": 202756404837893.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062083106487989426, "optim/lr": 0.0028100608279352807, "optim/total_tokens": 7291273216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8382058143615723, "created_at": "2025-01-16T20:50:31.101060+00:00"} {"global_step": 13908, "acc_step": 0, "speed/wps": 12907.587212415976, "speed/FLOPS": 202731394054515.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04979817196726799, "optim/lr": 0.002810019095442874, "optim/total_tokens": 7291797504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410019, "loss/out": 2.845611572265625, "created_at": "2025-01-16T20:50:41.259948+00:00"} {"global_step": 13909, "acc_step": 0, "speed/wps": 12910.374134572188, "speed/FLOPS": 202775166496611.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05246003344655037, "optim/lr": 0.0028099773586763013, "optim/total_tokens": 7292321792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 2.797480821609497, "created_at": "2025-01-16T20:50:51.421987+00:00"} {"global_step": 13910, "acc_step": 0, "speed/wps": 12911.28775389797, "speed/FLOPS": 202789516143566.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.058012958616018295, "optim/lr": 0.002809935617635697, "optim/total_tokens": 7292846080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.95339298248291, "created_at": "2025-01-16T20:51:01.579271+00:00"} {"global_step": 13911, "acc_step": 0, "speed/wps": 12910.352896255425, "speed/FLOPS": 202774832919661.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055429887026548386, "optim/lr": 0.002809893872321198, "optim/total_tokens": 7293370368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.917647361755371, "created_at": "2025-01-16T20:51:11.742149+00:00"} {"global_step": 13912, "acc_step": 0, "speed/wps": 12913.226207141492, "speed/FLOPS": 202819962215470.3, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06820041686296463, "optim/lr": 0.002809852122732941, "optim/total_tokens": 7293894656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.910092830657959, "created_at": "2025-01-16T20:51:21.899337+00:00"} {"global_step": 13913, "acc_step": 0, "speed/wps": 12914.93634734039, "speed/FLOPS": 202846822317272.9, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04877495393157005, "optim/lr": 0.0028098103688710607, "optim/total_tokens": 7294418944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.862542152404785, "created_at": "2025-01-16T20:51:32.054111+00:00"} {"global_step": 13914, "acc_step": 0, "speed/wps": 12909.571623945356, "speed/FLOPS": 202762561964451.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055892471224069595, "optim/lr": 0.0028097686107356953, "optim/total_tokens": 7294943232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.8536925315856934, "created_at": "2025-01-16T20:51:42.213027+00:00"} {"global_step": 13915, "acc_step": 0, "speed/wps": 12905.970265706346, "speed/FLOPS": 202705997684522.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05103178694844246, "optim/lr": 0.0028097268483269796, "optim/total_tokens": 7295467520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461122, "loss/out": 2.872434616088867, "created_at": "2025-01-16T20:51:52.372574+00:00"} {"global_step": 13916, "acc_step": 0, "speed/wps": 12916.779046890324, "speed/FLOPS": 202875764445839.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05338974669575691, "optim/lr": 0.0028096850816450502, "optim/total_tokens": 7295991808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.838597297668457, "created_at": "2025-01-16T20:52:02.526035+00:00"} {"global_step": 13917, "acc_step": 0, "speed/wps": 12908.434689277414, "speed/FLOPS": 202744704843179.44, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04462362825870514, "optim/lr": 0.0028096433106900437, "optim/total_tokens": 7296516096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460613, "loss/out": 2.9363720417022705, "created_at": "2025-01-16T20:52:12.687077+00:00"} {"global_step": 13918, "acc_step": 0, "speed/wps": 12908.320494889032, "speed/FLOPS": 202742911263390.38, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.050442345440387726, "optim/lr": 0.002809601535462096, "optim/total_tokens": 7297040384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.942044496536255, "created_at": "2025-01-16T20:52:22.850750+00:00"} {"global_step": 13919, "acc_step": 0, "speed/wps": 12906.627715897406, "speed/FLOPS": 202716323843204.8, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04499619081616402, "optim/lr": 0.002809559755961344, "optim/total_tokens": 7297564672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.850216865539551, "created_at": "2025-01-16T20:52:33.012380+00:00"} {"global_step": 13920, "acc_step": 0, "speed/wps": 12913.170480800462, "speed/FLOPS": 202819086956708.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04851975664496422, "optim/lr": 0.0028095179721879236, "optim/total_tokens": 7298088960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409931, "loss/out": 2.8019678592681885, "created_at": "2025-01-16T20:52:43.169237+00:00"} {"global_step": 13921, "acc_step": 0, "speed/wps": 12909.148325653692, "speed/FLOPS": 202755913483108.03, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03845313563942909, "optim/lr": 0.0028094761841419708, "optim/total_tokens": 7298613248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.81793212890625, "created_at": "2025-01-16T20:52:53.327641+00:00"} {"global_step": 13922, "acc_step": 0, "speed/wps": 12908.145295462064, "speed/FLOPS": 202740159515640.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048924852162599564, "optim/lr": 0.0028094343918236226, "optim/total_tokens": 7299137536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439286, "loss/out": 2.952899217605591, "created_at": "2025-01-16T20:53:03.487153+00:00"} {"global_step": 13923, "acc_step": 0, "speed/wps": 12912.74743103402, "speed/FLOPS": 202812442378795.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04000239446759224, "optim/lr": 0.0028093925952330143, "optim/total_tokens": 7299661824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.889212131500244, "created_at": "2025-01-16T20:53:13.642488+00:00"} {"global_step": 13924, "acc_step": 0, "speed/wps": 12906.834593972437, "speed/FLOPS": 202719573147653.4, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050854455679655075, "optim/lr": 0.0028093507943702835, "optim/total_tokens": 7300186112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8663880825042725, "created_at": "2025-01-16T20:53:23.805937+00:00"} {"global_step": 13925, "acc_step": 0, "speed/wps": 12906.329297849556, "speed/FLOPS": 202711636777693.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04968659207224846, "optim/lr": 0.0028093089892355662, "optim/total_tokens": 7300710400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 2.8075132369995117, "created_at": "2025-01-16T20:53:33.968693+00:00"} {"global_step": 13926, "acc_step": 0, "speed/wps": 12908.625067081186, "speed/FLOPS": 202747694988194.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052441567182540894, "optim/lr": 0.0028092671798289983, "optim/total_tokens": 7301234688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.7360219955444336, "created_at": "2025-01-16T20:53:44.127670+00:00"} {"global_step": 13927, "acc_step": 0, "speed/wps": 12910.64028681765, "speed/FLOPS": 202779346783357.47, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04884334281086922, "optim/lr": 0.002809225366150717, "optim/total_tokens": 7301758976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.7761168479919434, "created_at": "2025-01-16T20:53:54.283706+00:00"} {"global_step": 13928, "acc_step": 0, "speed/wps": 12911.801146408367, "speed/FLOPS": 202797579678417.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05922174081206322, "optim/lr": 0.0028091835482008582, "optim/total_tokens": 7302283264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8650221824645996, "created_at": "2025-01-16T20:54:04.439290+00:00"} {"global_step": 13929, "acc_step": 0, "speed/wps": 12903.81994813562, "speed/FLOPS": 202672223992228.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04951487109065056, "optim/lr": 0.0028091417259795584, "optim/total_tokens": 7302807552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8521316051483154, "created_at": "2025-01-16T20:54:14.601877+00:00"} {"global_step": 13930, "acc_step": 0, "speed/wps": 12910.536042172498, "speed/FLOPS": 202777709477958.22, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05119415000081062, "optim/lr": 0.002809099899486954, "optim/total_tokens": 7303331840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.7977330684661865, "created_at": "2025-01-16T20:54:24.759239+00:00"} {"global_step": 13931, "acc_step": 0, "speed/wps": 12908.680243891293, "speed/FLOPS": 202748561615817.8, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056691575795412064, "optim/lr": 0.002809058068723181, "optim/total_tokens": 7303856128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8945798873901367, "created_at": "2025-01-16T20:54:34.918745+00:00"} {"global_step": 13932, "acc_step": 0, "speed/wps": 12904.99140649392, "speed/FLOPS": 202690623355498.3, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052702538669109344, "optim/lr": 0.0028090162336883774, "optim/total_tokens": 7304380416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.825444221496582, "created_at": "2025-01-16T20:54:45.079168+00:00"} {"global_step": 13933, "acc_step": 0, "speed/wps": 12913.253866106586, "speed/FLOPS": 202820396637524.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07987720519304276, "optim/lr": 0.002808974394382678, "optim/total_tokens": 7304904704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.8279175758361816, "created_at": "2025-01-16T20:54:55.233287+00:00"} {"global_step": 13934, "acc_step": 0, "speed/wps": 12908.626584964033, "speed/FLOPS": 202747718828630.1, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.03905598446726799, "optim/lr": 0.002808932550806221, "optim/total_tokens": 7305428992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.9048616886138916, "created_at": "2025-01-16T20:55:05.395633+00:00"} {"global_step": 13935, "acc_step": 0, "speed/wps": 12910.962476633658, "speed/FLOPS": 202784407217153.1, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05643361061811447, "optim/lr": 0.0028088907029591415, "optim/total_tokens": 7305953280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.828839063644409, "created_at": "2025-01-16T20:55:15.552094+00:00"} {"global_step": 13936, "acc_step": 0, "speed/wps": 12911.510799706362, "speed/FLOPS": 202793019384484.78, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04586763679981232, "optim/lr": 0.002808848850841576, "optim/total_tokens": 7306477568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 2.8477630615234375, "created_at": "2025-01-16T20:55:25.709543+00:00"} {"global_step": 13937, "acc_step": 0, "speed/wps": 12909.847884756957, "speed/FLOPS": 202766901020119.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0558137521147728, "optim/lr": 0.002808806994453662, "optim/total_tokens": 7307001856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335952, "loss/out": 2.727755069732666, "created_at": "2025-01-16T20:55:35.867559+00:00"} {"global_step": 13938, "acc_step": 0, "speed/wps": 12910.780236521001, "speed/FLOPS": 202781544885759.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049206074327230453, "optim/lr": 0.0028087651337955354, "optim/total_tokens": 7307526144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.798584461212158, "created_at": "2025-01-16T20:55:46.027709+00:00"} {"global_step": 13939, "acc_step": 0, "speed/wps": 12905.316929680872, "speed/FLOPS": 202695736144465.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07846130430698395, "optim/lr": 0.002808723268867333, "optim/total_tokens": 7308050432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.781285524368286, "created_at": "2025-01-16T20:55:56.190538+00:00"} {"global_step": 13940, "acc_step": 0, "speed/wps": 12910.81185984418, "speed/FLOPS": 202782041573505.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04050654172897339, "optim/lr": 0.0028086813996691918, "optim/total_tokens": 7308574720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287670, "loss/out": 2.7637500762939453, "created_at": "2025-01-16T20:56:06.346108+00:00"} {"global_step": 13941, "acc_step": 0, "speed/wps": 12904.573716579793, "speed/FLOPS": 202684062961430.84, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06591475754976273, "optim/lr": 0.002808639526201248, "optim/total_tokens": 7309099008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.8173625469207764, "created_at": "2025-01-16T20:56:16.508276+00:00"} {"global_step": 13942, "acc_step": 0, "speed/wps": 12907.794413660931, "speed/FLOPS": 202734648434791.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04259253293275833, "optim/lr": 0.0028085976484636373, "optim/total_tokens": 7309623296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.8014371395111084, "created_at": "2025-01-16T20:56:26.667383+00:00"} {"global_step": 13943, "acc_step": 0, "speed/wps": 12923.07683483594, "speed/FLOPS": 202974679859592.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0679042711853981, "optim/lr": 0.002808555766456498, "optim/total_tokens": 7310147584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.727238178253174, "created_at": "2025-01-16T20:56:36.814317+00:00"} {"global_step": 13944, "acc_step": 0, "speed/wps": 12907.57634864238, "speed/FLOPS": 202731223424022.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06716594099998474, "optim/lr": 0.0028085138801799657, "optim/total_tokens": 7310671872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8024649620056152, "created_at": "2025-01-16T20:56:46.976794+00:00"} {"global_step": 13945, "acc_step": 0, "speed/wps": 12906.909378513119, "speed/FLOPS": 202720747741629.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05141705647110939, "optim/lr": 0.002808471989634178, "optim/total_tokens": 7311196160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.806844711303711, "created_at": "2025-01-16T20:56:57.137306+00:00"} {"global_step": 13946, "acc_step": 0, "speed/wps": 12907.975552540742, "speed/FLOPS": 202737493469810.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05533270165324211, "optim/lr": 0.0028084300948192703, "optim/total_tokens": 7311720448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.755875825881958, "created_at": "2025-01-16T20:57:07.295739+00:00"} {"global_step": 13947, "acc_step": 0, "speed/wps": 12909.1269137566, "speed/FLOPS": 202755577179841.2, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05529019981622696, "optim/lr": 0.00280838819573538, "optim/total_tokens": 7312244736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 2.71437931060791, "created_at": "2025-01-16T20:57:17.454835+00:00"} {"global_step": 13948, "acc_step": 0, "speed/wps": 12910.99760676472, "speed/FLOPS": 202784958984135.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0688769593834877, "optim/lr": 0.002808346292382644, "optim/total_tokens": 7312769024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.816117763519287, "created_at": "2025-01-16T20:57:27.610399+00:00"} {"global_step": 13949, "acc_step": 0, "speed/wps": 12909.323628093656, "speed/FLOPS": 202758666848818.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04657473415136337, "optim/lr": 0.0028083043847611975, "optim/total_tokens": 7313293312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347598, "loss/out": 2.831664562225342, "created_at": "2025-01-16T20:57:37.771109+00:00"} {"global_step": 13950, "acc_step": 0, "speed/wps": 12904.97403050887, "speed/FLOPS": 202690350441776.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06553361564874649, "optim/lr": 0.00280826247287118, "optim/total_tokens": 7313817600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.8663723468780518, "created_at": "2025-01-16T20:57:47.932738+00:00"} {"global_step": 13951, "acc_step": 0, "speed/wps": 12909.234268795155, "speed/FLOPS": 202757263338242.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06007694825530052, "optim/lr": 0.0028082205567127264, "optim/total_tokens": 7314341888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.863917827606201, "created_at": "2025-01-16T20:57:58.094345+00:00"} {"global_step": 13952, "acc_step": 0, "speed/wps": 12908.925846921307, "speed/FLOPS": 202752419148898.12, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060363996773958206, "optim/lr": 0.0028081786362859733, "optim/total_tokens": 7314866176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340231, "loss/out": 2.8459482192993164, "created_at": "2025-01-16T20:58:08.251918+00:00"} {"global_step": 13953, "acc_step": 0, "speed/wps": 12903.335161502979, "speed/FLOPS": 202664609751996.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05947304144501686, "optim/lr": 0.0028081367115910584, "optim/total_tokens": 7315390464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506365, "loss/out": 2.9956910610198975, "created_at": "2025-01-16T20:58:18.416265+00:00"} {"global_step": 13954, "acc_step": 0, "speed/wps": 12904.749239133149, "speed/FLOPS": 202686819784324.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05719917267560959, "optim/lr": 0.0028080947826281174, "optim/total_tokens": 7315914752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.738969326019287, "created_at": "2025-01-16T20:58:28.582068+00:00"} {"global_step": 13955, "acc_step": 0, "speed/wps": 12909.60195651806, "speed/FLOPS": 202763038379187.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04917806759476662, "optim/lr": 0.0028080528493972883, "optim/total_tokens": 7316439040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.934678792953491, "created_at": "2025-01-16T20:58:38.740861+00:00"} {"global_step": 13956, "acc_step": 0, "speed/wps": 12914.882747793312, "speed/FLOPS": 202845980462732.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04272330179810524, "optim/lr": 0.0028080109118987077, "optim/total_tokens": 7316963328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8853273391723633, "created_at": "2025-01-16T20:58:48.895086+00:00"} {"global_step": 13957, "acc_step": 0, "speed/wps": 12911.413164764917, "speed/FLOPS": 202791485893564.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05822712182998657, "optim/lr": 0.002807968970132511, "optim/total_tokens": 7317487616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 2.7294392585754395, "created_at": "2025-01-16T20:58:59.052529+00:00"} {"global_step": 13958, "acc_step": 0, "speed/wps": 12904.125376384154, "speed/FLOPS": 202677021162574.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04741595312952995, "optim/lr": 0.0028079270240988366, "optim/total_tokens": 7318011904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8698205947875977, "created_at": "2025-01-16T20:59:09.214917+00:00"} {"global_step": 13959, "acc_step": 0, "speed/wps": 12904.45184861175, "speed/FLOPS": 202682148857528.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.050138454884290695, "optim/lr": 0.002807885073797821, "optim/total_tokens": 7318536192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.811246395111084, "created_at": "2025-01-16T20:59:19.378283+00:00"} {"global_step": 13960, "acc_step": 0, "speed/wps": 12911.119291625839, "speed/FLOPS": 202786870212091.66, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.042802080512046814, "optim/lr": 0.002807843119229601, "optim/total_tokens": 7319060480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323796, "loss/out": 2.846463918685913, "created_at": "2025-01-16T20:59:29.537078+00:00"} {"global_step": 13961, "acc_step": 0, "speed/wps": 12910.057568554388, "speed/FLOPS": 202770194392294.16, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05846543237566948, "optim/lr": 0.0028078011603943137, "optim/total_tokens": 7319584768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.788674831390381, "created_at": "2025-01-16T20:59:39.696423+00:00"} {"global_step": 13962, "acc_step": 0, "speed/wps": 12911.139664257047, "speed/FLOPS": 202787190192259.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04860978573560715, "optim/lr": 0.0028077591972920954, "optim/total_tokens": 7320109056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 2.7830753326416016, "created_at": "2025-01-16T20:59:49.852016+00:00"} {"global_step": 13963, "acc_step": 0, "speed/wps": 12909.481280734857, "speed/FLOPS": 202761143000185.97, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054472919553518295, "optim/lr": 0.0028077172299230827, "optim/total_tokens": 7320633344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 2.844878911972046, "created_at": "2025-01-16T21:00:00.009139+00:00"} {"global_step": 13964, "acc_step": 0, "speed/wps": 12911.922133286902, "speed/FLOPS": 202799479943598.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055002354085445404, "optim/lr": 0.002807675258287414, "optim/total_tokens": 7321157632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.699533462524414, "created_at": "2025-01-16T21:00:10.166378+00:00"} {"global_step": 13965, "acc_step": 0, "speed/wps": 12909.68425010375, "speed/FLOPS": 202764330913033.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05882399529218674, "optim/lr": 0.0028076332823852254, "optim/total_tokens": 7321681920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.764237642288208, "created_at": "2025-01-16T21:00:20.325463+00:00"} {"global_step": 13966, "acc_step": 0, "speed/wps": 12906.961929434672, "speed/FLOPS": 202721573126065.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05121380090713501, "optim/lr": 0.0028075913022166535, "optim/total_tokens": 7322206208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 2.9447124004364014, "created_at": "2025-01-16T21:00:30.488164+00:00"} {"global_step": 13967, "acc_step": 0, "speed/wps": 12913.67234800464, "speed/FLOPS": 202826969470787.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04189828410744667, "optim/lr": 0.0028075493177818355, "optim/total_tokens": 7322730496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349715, "loss/out": 2.861077070236206, "created_at": "2025-01-16T21:00:40.645895+00:00"} {"global_step": 13968, "acc_step": 0, "speed/wps": 12899.384604355368, "speed/FLOPS": 202602560823358.8, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06330905109643936, "optim/lr": 0.002807507329080909, "optim/total_tokens": 7323254784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.8554186820983887, "created_at": "2025-01-16T21:00:50.812349+00:00"} {"global_step": 13969, "acc_step": 0, "speed/wps": 12908.605477899695, "speed/FLOPS": 202747387313180.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05943678319454193, "optim/lr": 0.0028074653361140106, "optim/total_tokens": 7323779072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481564, "loss/out": 2.786271572113037, "created_at": "2025-01-16T21:01:00.971727+00:00"} {"global_step": 13970, "acc_step": 0, "speed/wps": 12905.430218877547, "speed/FLOPS": 202697515506974.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.03905607759952545, "optim/lr": 0.002807423338881277, "optim/total_tokens": 7324303360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.753509044647217, "created_at": "2025-01-16T21:01:11.132728+00:00"} {"global_step": 13971, "acc_step": 0, "speed/wps": 12913.725491971007, "speed/FLOPS": 202827804169807.1, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057799916714429855, "optim/lr": 0.0028073813373828455, "optim/total_tokens": 7324827648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.9359469413757324, "created_at": "2025-01-16T21:01:21.286034+00:00"} {"global_step": 13972, "acc_step": 0, "speed/wps": 12902.643647908924, "speed/FLOPS": 202653748580762.56, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05442852899432182, "optim/lr": 0.0028073393316188524, "optim/total_tokens": 7325351936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 2.862703800201416, "created_at": "2025-01-16T21:01:31.450029+00:00"} {"global_step": 13973, "acc_step": 0, "speed/wps": 12908.24745212579, "speed/FLOPS": 202741764026422.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04758942499756813, "optim/lr": 0.002807297321589436, "optim/total_tokens": 7325876224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.9343175888061523, "created_at": "2025-01-16T21:01:41.607608+00:00"} {"global_step": 13974, "acc_step": 0, "speed/wps": 12911.854361449556, "speed/FLOPS": 202798415493764.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05066429823637009, "optim/lr": 0.002807255307294733, "optim/total_tokens": 7326400512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9705541133880615, "created_at": "2025-01-16T21:01:51.763592+00:00"} {"global_step": 13975, "acc_step": 0, "speed/wps": 12903.132961927891, "speed/FLOPS": 202661433929818.38, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05685004964470863, "optim/lr": 0.00280721328873488, "optim/total_tokens": 7326924800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 2.869095802307129, "created_at": "2025-01-16T21:02:01.925546+00:00"} {"global_step": 13976, "acc_step": 0, "speed/wps": 12910.185181973977, "speed/FLOPS": 202772198736408.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0538504421710968, "optim/lr": 0.0028071712659100143, "optim/total_tokens": 7327449088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.820134162902832, "created_at": "2025-01-16T21:02:12.084577+00:00"} {"global_step": 13977, "acc_step": 0, "speed/wps": 12906.513881336974, "speed/FLOPS": 202714535915008.12, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0491451695561409, "optim/lr": 0.0028071292388202732, "optim/total_tokens": 7327973376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.886619806289673, "created_at": "2025-01-16T21:02:22.247168+00:00"} {"global_step": 13978, "acc_step": 0, "speed/wps": 12906.452154076855, "speed/FLOPS": 202713566403565.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07099797576665878, "optim/lr": 0.002807087207465794, "optim/total_tokens": 7328497664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.808469295501709, "created_at": "2025-01-16T21:02:32.406294+00:00"} {"global_step": 13979, "acc_step": 0, "speed/wps": 12904.227875263194, "speed/FLOPS": 202678631048317.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04495438188314438, "optim/lr": 0.002807045171846713, "optim/total_tokens": 7329021952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.8581955432891846, "created_at": "2025-01-16T21:02:42.567362+00:00"} {"global_step": 13980, "acc_step": 0, "speed/wps": 12911.435424065487, "speed/FLOPS": 202791835506467.84, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06735225021839142, "optim/lr": 0.002807003131963168, "optim/total_tokens": 7329546240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.86704158782959, "created_at": "2025-01-16T21:02:52.726113+00:00"} {"global_step": 13981, "acc_step": 0, "speed/wps": 12909.745052633367, "speed/FLOPS": 202765285900312.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054125282913446426, "optim/lr": 0.0028069610878152964, "optim/total_tokens": 7330070528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.877626419067383, "created_at": "2025-01-16T21:03:02.890351+00:00"} {"global_step": 13982, "acc_step": 0, "speed/wps": 12903.521861030822, "speed/FLOPS": 202667542124633.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05296872928738594, "optim/lr": 0.0028069190394032343, "optim/total_tokens": 7330594816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340231, "loss/out": 2.740233898162842, "created_at": "2025-01-16T21:03:13.051790+00:00"} {"global_step": 13983, "acc_step": 0, "speed/wps": 12908.415303130432, "speed/FLOPS": 202744400357101.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046458203345537186, "optim/lr": 0.0028068769867271204, "optim/total_tokens": 7331119104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.890660047531128, "created_at": "2025-01-16T21:03:23.210212+00:00"} {"global_step": 13984, "acc_step": 0, "speed/wps": 12906.92665672975, "speed/FLOPS": 202721019119764.53, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05714574083685875, "optim/lr": 0.0028068349297870904, "optim/total_tokens": 7331643392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8603785037994385, "created_at": "2025-01-16T21:03:33.369407+00:00"} {"global_step": 13985, "acc_step": 0, "speed/wps": 12904.967357735983, "speed/FLOPS": 202690245636708.44, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05024817958474159, "optim/lr": 0.0028067928685832823, "optim/total_tokens": 7332167680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 2.7277021408081055, "created_at": "2025-01-16T21:03:43.530059+00:00"} {"global_step": 13986, "acc_step": 0, "speed/wps": 12907.014429885035, "speed/FLOPS": 202722397717781.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0689157247543335, "optim/lr": 0.002806750803115834, "optim/total_tokens": 7332691968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8690242767333984, "created_at": "2025-01-16T21:03:53.689036+00:00"} {"global_step": 13987, "acc_step": 0, "speed/wps": 12910.36918886747, "speed/FLOPS": 202775088817522.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04894084483385086, "optim/lr": 0.0028067087333848812, "optim/total_tokens": 7333216256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438756, "loss/out": 2.743622303009033, "created_at": "2025-01-16T21:04:03.849266+00:00"} {"global_step": 13988, "acc_step": 0, "speed/wps": 12905.802664088069, "speed/FLOPS": 202703365270799.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04815923795104027, "optim/lr": 0.0028066666593905623, "optim/total_tokens": 7333740544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.820763111114502, "created_at": "2025-01-16T21:04:14.009097+00:00"} {"global_step": 13989, "acc_step": 0, "speed/wps": 12901.333364455002, "speed/FLOPS": 202633168778600.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048908088356256485, "optim/lr": 0.0028066245811330143, "optim/total_tokens": 7334264832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.7737650871276855, "created_at": "2025-01-16T21:04:24.180025+00:00"} {"global_step": 13990, "acc_step": 0, "speed/wps": 12907.699100759253, "speed/FLOPS": 202733151414697.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05589577928185463, "optim/lr": 0.0028065824986123747, "optim/total_tokens": 7334789120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.8391242027282715, "created_at": "2025-01-16T21:04:34.342536+00:00"} {"global_step": 13991, "acc_step": 0, "speed/wps": 12909.08967785167, "speed/FLOPS": 202754992338786.78, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07544256746768951, "optim/lr": 0.00280654041182878, "optim/total_tokens": 7335313408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.893841028213501, "created_at": "2025-01-16T21:04:44.500715+00:00"} {"global_step": 13992, "acc_step": 0, "speed/wps": 12905.050981485992, "speed/FLOPS": 202691559062613.4, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04792996123433113, "optim/lr": 0.0028064983207823683, "optim/total_tokens": 7335837696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.904860258102417, "created_at": "2025-01-16T21:04:54.664090+00:00"} {"global_step": 13993, "acc_step": 0, "speed/wps": 12908.356574376578, "speed/FLOPS": 202743477941320.7, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08219479024410248, "optim/lr": 0.0028064562254732765, "optim/total_tokens": 7336361984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328312, "loss/out": 2.8579208850860596, "created_at": "2025-01-16T21:05:04.822753+00:00"} {"global_step": 13994, "acc_step": 0, "speed/wps": 12909.36323005793, "speed/FLOPS": 202759288852086.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0635741651058197, "optim/lr": 0.0028064141259016425, "optim/total_tokens": 7336886272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.840853214263916, "created_at": "2025-01-16T21:05:14.979531+00:00"} {"global_step": 13995, "acc_step": 0, "speed/wps": 12902.58881368288, "speed/FLOPS": 202652887333892.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05303145572543144, "optim/lr": 0.002806372022067603, "optim/total_tokens": 7337410560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 2.7423532009124756, "created_at": "2025-01-16T21:05:25.142480+00:00"} {"global_step": 13996, "acc_step": 0, "speed/wps": 12910.381368382754, "speed/FLOPS": 202775280113546.22, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07215432822704315, "optim/lr": 0.002806329913971296, "optim/total_tokens": 7337934848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9082555770874023, "created_at": "2025-01-16T21:05:35.299495+00:00"} {"global_step": 13997, "acc_step": 0, "speed/wps": 12904.824008976255, "speed/FLOPS": 202687994147455.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042724139988422394, "optim/lr": 0.002806287801612858, "optim/total_tokens": 7338459136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.9091641902923584, "created_at": "2025-01-16T21:05:45.459936+00:00"} {"global_step": 13998, "acc_step": 0, "speed/wps": 12900.87797299575, "speed/FLOPS": 202626016229959.56, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06210489571094513, "optim/lr": 0.0028062456849924267, "optim/total_tokens": 7338983424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.848879814147949, "created_at": "2025-01-16T21:05:55.624733+00:00"} {"global_step": 13999, "acc_step": 0, "speed/wps": 12909.058137561175, "speed/FLOPS": 202754496955183.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047427888959646225, "optim/lr": 0.0028062035641101403, "optim/total_tokens": 7339507712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.84907603263855, "created_at": "2025-01-16T21:06:05.787020+00:00"} {"global_step": 14000, "acc_step": 0, "speed/wps": 12907.83148345798, "speed/FLOPS": 202735230666893.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06519196182489395, "optim/lr": 0.0028061614389661357, "optim/total_tokens": 7340032000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.7963805198669434, "created_at": "2025-01-16T21:06:15.948685+00:00"} {"global_step": 14001, "acc_step": 0, "speed/wps": 12610.196595365735, "speed/FLOPS": 198060465756209.22, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.2446, "optim/grad_norm": 0.04971528425812721, "optim/lr": 0.00280611930956055, "optim/total_tokens": 7340556288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.9582793712615967, "created_at": "2025-01-16T21:06:26.347602+00:00"} {"global_step": 14002, "acc_step": 0, "speed/wps": 12922.193005988585, "speed/FLOPS": 202960798113036.75, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055049050599336624, "optim/lr": 0.0028060771758935212, "optim/total_tokens": 7341080576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.7722084522247314, "created_at": "2025-01-16T21:06:36.494279+00:00"} {"global_step": 14003, "acc_step": 0, "speed/wps": 12908.697679586676, "speed/FLOPS": 202748835467372.97, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05102641135454178, "optim/lr": 0.0028060350379651867, "optim/total_tokens": 7341604864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 2.8174679279327393, "created_at": "2025-01-16T21:06:46.654212+00:00"} {"global_step": 14004, "acc_step": 0, "speed/wps": 12912.136799540876, "speed/FLOPS": 202802851572099.72, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047960538417100906, "optim/lr": 0.0028059928957756834, "optim/total_tokens": 7342129152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.81217622756958, "created_at": "2025-01-16T21:06:56.815061+00:00"} {"global_step": 14005, "acc_step": 0, "speed/wps": 12910.397754628024, "speed/FLOPS": 202775537482045.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044283948838710785, "optim/lr": 0.00280595074932515, "optim/total_tokens": 7342653440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8122849464416504, "created_at": "2025-01-16T21:07:06.972250+00:00"} {"global_step": 14006, "acc_step": 0, "speed/wps": 12910.131469085005, "speed/FLOPS": 202771355101677.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04948609694838524, "optim/lr": 0.002805908598613722, "optim/total_tokens": 7343177728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.740286350250244, "created_at": "2025-01-16T21:07:17.128401+00:00"} {"global_step": 14007, "acc_step": 0, "speed/wps": 12916.139957195724, "speed/FLOPS": 202865726663984.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04832742363214493, "optim/lr": 0.002805866443641539, "optim/total_tokens": 7343702016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.968816041946411, "created_at": "2025-01-16T21:07:27.280020+00:00"} {"global_step": 14008, "acc_step": 0, "speed/wps": 12914.867459453215, "speed/FLOPS": 202845740338342.94, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04619870334863663, "optim/lr": 0.002805824284408737, "optim/total_tokens": 7344226304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.7876930236816406, "created_at": "2025-01-16T21:07:37.432697+00:00"} {"global_step": 14009, "acc_step": 0, "speed/wps": 12909.85239426894, "speed/FLOPS": 202766971848202.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051143329590559006, "optim/lr": 0.002805782120915455, "optim/total_tokens": 7344750592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 2.9335389137268066, "created_at": "2025-01-16T21:07:47.595953+00:00"} {"global_step": 14010, "acc_step": 0, "speed/wps": 12912.59771326739, "speed/FLOPS": 202810090855536.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05023181438446045, "optim/lr": 0.0028057399531618296, "optim/total_tokens": 7345274880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.83660888671875, "created_at": "2025-01-16T21:07:57.752538+00:00"} {"global_step": 14011, "acc_step": 0, "speed/wps": 12912.039016531004, "speed/FLOPS": 202801315755561.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06795670092105865, "optim/lr": 0.0028056977811479985, "optim/total_tokens": 7345799168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8294456005096436, "created_at": "2025-01-16T21:08:07.909946+00:00"} {"global_step": 14012, "acc_step": 0, "speed/wps": 12912.080714182099, "speed/FLOPS": 202801970674470.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05488467216491699, "optim/lr": 0.0028056556048741, "optim/total_tokens": 7346323456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 2.852219581604004, "created_at": "2025-01-16T21:08:18.064925+00:00"} {"global_step": 14013, "acc_step": 0, "speed/wps": 12914.524474465623, "speed/FLOPS": 202840353287802.06, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04825746268033981, "optim/lr": 0.0028056134243402697, "optim/total_tokens": 7346847744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.848766326904297, "created_at": "2025-01-16T21:08:28.218427+00:00"} {"global_step": 14014, "acc_step": 0, "speed/wps": 12908.17332454089, "speed/FLOPS": 202740599750839.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051903776824474335, "optim/lr": 0.002805571239546648, "optim/total_tokens": 7347372032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 2.8138089179992676, "created_at": "2025-01-16T21:08:38.376421+00:00"} {"global_step": 14015, "acc_step": 0, "speed/wps": 12913.53483335675, "speed/FLOPS": 202824809614277.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05565465986728668, "optim/lr": 0.0028055290504933705, "optim/total_tokens": 7347896320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.7489264011383057, "created_at": "2025-01-16T21:08:48.532113+00:00"} {"global_step": 14016, "acc_step": 0, "speed/wps": 12911.459531287283, "speed/FLOPS": 202792214143512.94, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061322491616010666, "optim/lr": 0.0028054868571805755, "optim/total_tokens": 7348420608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8829712867736816, "created_at": "2025-01-16T21:08:58.690376+00:00"} {"global_step": 14017, "acc_step": 0, "speed/wps": 12890.74984385844, "speed/FLOPS": 202466940044350.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052324000746011734, "optim/lr": 0.0028054446596084007, "optim/total_tokens": 7348944896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8915679454803467, "created_at": "2025-01-16T21:09:08.865314+00:00"} {"global_step": 14018, "acc_step": 0, "speed/wps": 12897.208087667079, "speed/FLOPS": 202568375637921.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05919082090258598, "optim/lr": 0.0028054024577769837, "optim/total_tokens": 7349469184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.883474826812744, "created_at": "2025-01-16T21:09:19.036186+00:00"} {"global_step": 14019, "acc_step": 0, "speed/wps": 12896.290968253332, "speed/FLOPS": 202553971017274.84, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04622097685933113, "optim/lr": 0.002805360251686462, "optim/total_tokens": 7349993472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289355, "loss/out": 2.868337869644165, "created_at": "2025-01-16T21:09:29.206982+00:00"} {"global_step": 14020, "acc_step": 0, "speed/wps": 12904.524121913833, "speed/FLOPS": 202683284009051.75, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05340363457798958, "optim/lr": 0.002805318041336974, "optim/total_tokens": 7350517760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.7589967250823975, "created_at": "2025-01-16T21:09:39.369049+00:00"} {"global_step": 14021, "acc_step": 0, "speed/wps": 12908.569469243866, "speed/FLOPS": 202746821747759.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04883452132344246, "optim/lr": 0.0028052758267286568, "optim/total_tokens": 7351042048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.868490695953369, "created_at": "2025-01-16T21:09:49.530698+00:00"} {"global_step": 14022, "acc_step": 0, "speed/wps": 12912.90260226127, "speed/FLOPS": 202814879556146.75, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05396384000778198, "optim/lr": 0.002805233607861648, "optim/total_tokens": 7351566336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.944150924682617, "created_at": "2025-01-16T21:09:59.687569+00:00"} {"global_step": 14023, "acc_step": 0, "speed/wps": 12907.830132684961, "speed/FLOPS": 202735209451147.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058844588696956635, "optim/lr": 0.0028051913847360857, "optim/total_tokens": 7352090624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.7358548641204834, "created_at": "2025-01-16T21:10:09.850265+00:00"} {"global_step": 14024, "acc_step": 0, "speed/wps": 12905.498170628576, "speed/FLOPS": 202698582782599.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059564296156167984, "optim/lr": 0.002805149157352108, "optim/total_tokens": 7352614912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 2.7988667488098145, "created_at": "2025-01-16T21:10:20.010123+00:00"} {"global_step": 14025, "acc_step": 0, "speed/wps": 12911.003453119314, "speed/FLOPS": 202785050809167.9, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04897588491439819, "optim/lr": 0.002805106925709852, "optim/total_tokens": 7353139200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.817136287689209, "created_at": "2025-01-16T21:10:30.170897+00:00"} {"global_step": 14026, "acc_step": 0, "speed/wps": 12911.160701435154, "speed/FLOPS": 202787520610050.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04956747218966484, "optim/lr": 0.0028050646898094556, "optim/total_tokens": 7353663488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 2.825571060180664, "created_at": "2025-01-16T21:10:40.326884+00:00"} {"global_step": 14027, "acc_step": 0, "speed/wps": 12908.362420707488, "speed/FLOPS": 202743569765981.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04510711878538132, "optim/lr": 0.002805022449651057, "optim/total_tokens": 7354187776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.7399301528930664, "created_at": "2025-01-16T21:10:50.484760+00:00"} {"global_step": 14028, "acc_step": 0, "speed/wps": 12911.030006958974, "speed/FLOPS": 202785467873708.56, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04561946168541908, "optim/lr": 0.0028049802052347935, "optim/total_tokens": 7354712064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 2.830824851989746, "created_at": "2025-01-16T21:11:00.644131+00:00"} {"global_step": 14029, "acc_step": 0, "speed/wps": 12907.150021823061, "speed/FLOPS": 202724527375490.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04867580905556679, "optim/lr": 0.002804937956560803, "optim/total_tokens": 7355236352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.8850464820861816, "created_at": "2025-01-16T21:11:10.813144+00:00"} {"global_step": 14030, "acc_step": 0, "speed/wps": 12906.643949396028, "speed/FLOPS": 202716578812607.8, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046843208372592926, "optim/lr": 0.0028048957036292243, "optim/total_tokens": 7355760640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 2.8108718395233154, "created_at": "2025-01-16T21:11:20.972473+00:00"} {"global_step": 14031, "acc_step": 0, "speed/wps": 12912.690304906106, "speed/FLOPS": 202811545134456.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04919963330030441, "optim/lr": 0.0028048534464401947, "optim/total_tokens": 7356284928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.793776512145996, "created_at": "2025-01-16T21:11:31.127736+00:00"} {"global_step": 14032, "acc_step": 0, "speed/wps": 12907.689715393652, "speed/FLOPS": 202733004004633.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041725706309080124, "optim/lr": 0.0028048111849938505, "optim/total_tokens": 7356809216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.786170482635498, "created_at": "2025-01-16T21:11:41.317429+00:00"} {"global_step": 14033, "acc_step": 0, "speed/wps": 12910.690282634163, "speed/FLOPS": 202780132036357.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05359311029314995, "optim/lr": 0.002804768919290332, "optim/total_tokens": 7357333504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.7987475395202637, "created_at": "2025-01-16T21:11:51.474358+00:00"} {"global_step": 14034, "acc_step": 0, "speed/wps": 12908.581772050416, "speed/FLOPS": 202747014980242.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052373118698596954, "optim/lr": 0.0028047266493297757, "optim/total_tokens": 7357857792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.8627395629882812, "created_at": "2025-01-16T21:12:01.635213+00:00"} {"global_step": 14035, "acc_step": 0, "speed/wps": 12910.306824282461, "speed/FLOPS": 202774109296016.8, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0508924201130867, "optim/lr": 0.0028046843751123197, "optim/total_tokens": 7358382080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8870303630828857, "created_at": "2025-01-16T21:12:11.792553+00:00"} {"global_step": 14036, "acc_step": 0, "speed/wps": 12909.74158766128, "speed/FLOPS": 202765231478164.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05719074606895447, "optim/lr": 0.0028046420966381025, "optim/total_tokens": 7358906368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.860992908477783, "created_at": "2025-01-16T21:12:21.952230+00:00"} {"global_step": 14037, "acc_step": 0, "speed/wps": 12911.486075157787, "speed/FLOPS": 202792631051474.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07222914695739746, "optim/lr": 0.002804599813907262, "optim/total_tokens": 7359430656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.779297351837158, "created_at": "2025-01-16T21:12:32.112646+00:00"} {"global_step": 14038, "acc_step": 0, "speed/wps": 12912.581565489823, "speed/FLOPS": 202809837232500.4, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.046284038573503494, "optim/lr": 0.002804557526919935, "optim/total_tokens": 7359954944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363445, "loss/out": 2.805048704147339, "created_at": "2025-01-16T21:12:42.266855+00:00"} {"global_step": 14039, "acc_step": 0, "speed/wps": 12916.379017340583, "speed/FLOPS": 202869481432062.25, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0763879269361496, "optim/lr": 0.0028045152356762607, "optim/total_tokens": 7360479232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9777064323425293, "created_at": "2025-01-16T21:12:52.418522+00:00"} {"global_step": 14040, "acc_step": 0, "speed/wps": 12908.550491165886, "speed/FLOPS": 202746523670966.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06732042878866196, "optim/lr": 0.002804472940176377, "optim/total_tokens": 7361003520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424393, "loss/out": 2.9050936698913574, "created_at": "2025-01-16T21:13:02.578783+00:00"} {"global_step": 14041, "acc_step": 0, "speed/wps": 12911.184375301204, "speed/FLOPS": 202787892440647.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05237734317779541, "optim/lr": 0.0028044306404204204, "optim/total_tokens": 7361527808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.8480873107910156, "created_at": "2025-01-16T21:13:12.734946+00:00"} {"global_step": 14042, "acc_step": 0, "speed/wps": 12908.865915273165, "speed/FLOPS": 202751477840009.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05458638444542885, "optim/lr": 0.0028043883364085305, "optim/total_tokens": 7362052096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304284, "loss/out": 2.7667908668518066, "created_at": "2025-01-16T21:13:22.892802+00:00"} {"global_step": 14043, "acc_step": 0, "speed/wps": 12906.886901631222, "speed/FLOPS": 202720394711312.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04534484073519707, "optim/lr": 0.0028043460281408457, "optim/total_tokens": 7362576384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 2.8143503665924072, "created_at": "2025-01-16T21:13:33.052230+00:00"} {"global_step": 14044, "acc_step": 0, "speed/wps": 12906.056291990382, "speed/FLOPS": 202707348845526.25, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044673528522253036, "optim/lr": 0.0028043037156175025, "optim/total_tokens": 7363100672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.702733039855957, "created_at": "2025-01-16T21:13:43.212354+00:00"} {"global_step": 14045, "acc_step": 0, "speed/wps": 12904.448412511523, "speed/FLOPS": 202682094888852.84, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04567573964595795, "optim/lr": 0.00280426139883864, "optim/total_tokens": 7363624960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.908629894256592, "created_at": "2025-01-16T21:13:53.374789+00:00"} {"global_step": 14046, "acc_step": 0, "speed/wps": 12903.766733046039, "speed/FLOPS": 202671388176121.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057847969233989716, "optim/lr": 0.0028042190778043955, "optim/total_tokens": 7364149248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 2.8217296600341797, "created_at": "2025-01-16T21:14:03.542649+00:00"} {"global_step": 14047, "acc_step": 0, "speed/wps": 12913.009498795984, "speed/FLOPS": 202816558513115.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.059482280164957047, "optim/lr": 0.002804176752514908, "optim/total_tokens": 7364673536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406769, "loss/out": 2.8390135765075684, "created_at": "2025-01-16T21:14:13.696719+00:00"} {"global_step": 14048, "acc_step": 0, "speed/wps": 12905.20836056428, "speed/FLOPS": 202694030917299.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06252289563417435, "optim/lr": 0.0028041344229703144, "optim/total_tokens": 7365197824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.823556900024414, "created_at": "2025-01-16T21:14:23.862814+00:00"} {"global_step": 14049, "acc_step": 0, "speed/wps": 12911.256927766537, "speed/FLOPS": 202789031976813.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04130612313747406, "optim/lr": 0.0028040920891707543, "optim/total_tokens": 7365722112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383767, "loss/out": 2.9642794132232666, "created_at": "2025-01-16T21:14:34.018726+00:00"} {"global_step": 14050, "acc_step": 0, "speed/wps": 12907.44329405524, "speed/FLOPS": 202729133618895.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06112752482295036, "optim/lr": 0.0028040497511163647, "optim/total_tokens": 7366246400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8207650184631348, "created_at": "2025-01-16T21:14:44.177291+00:00"} {"global_step": 14051, "acc_step": 0, "speed/wps": 12907.372845380029, "speed/FLOPS": 202728027125644.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04292771592736244, "optim/lr": 0.002804007408807284, "optim/total_tokens": 7366770688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.840775966644287, "created_at": "2025-01-16T21:14:54.337664+00:00"} {"global_step": 14052, "acc_step": 0, "speed/wps": 12908.33766085751, "speed/FLOPS": 202743180878513.84, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07242362946271896, "optim/lr": 0.0028039650622436506, "optim/total_tokens": 7367294976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9249327182769775, "created_at": "2025-01-16T21:15:04.495835+00:00"} {"global_step": 14053, "acc_step": 0, "speed/wps": 12904.361651587249, "speed/FLOPS": 202680732189315.1, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0504898726940155, "optim/lr": 0.0028039227114256024, "optim/total_tokens": 7367819264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.895590305328369, "created_at": "2025-01-16T21:15:14.658863+00:00"} {"global_step": 14054, "acc_step": 0, "speed/wps": 12905.885826450094, "speed/FLOPS": 202704671449971.84, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0683416798710823, "optim/lr": 0.0028038803563532775, "optim/total_tokens": 7368343552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375098, "loss/out": 2.8923473358154297, "created_at": "2025-01-16T21:15:24.822351+00:00"} {"global_step": 14055, "acc_step": 0, "speed/wps": 12911.571005852906, "speed/FLOPS": 202793965004748.16, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05959466099739075, "optim/lr": 0.0028038379970268144, "optim/total_tokens": 7368867840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.8027777671813965, "created_at": "2025-01-16T21:15:34.977437+00:00"} {"global_step": 14056, "acc_step": 0, "speed/wps": 12907.820188352436, "speed/FLOPS": 202735053261739.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055382419377565384, "optim/lr": 0.0028037956334463513, "optim/total_tokens": 7369392128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333941, "loss/out": 2.815673589706421, "created_at": "2025-01-16T21:15:45.137282+00:00"} {"global_step": 14057, "acc_step": 0, "speed/wps": 12909.407037142442, "speed/FLOPS": 202759976902545.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04623511806130409, "optim/lr": 0.0028037532656120263, "optim/total_tokens": 7369916416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.7969233989715576, "created_at": "2025-01-16T21:15:55.294008+00:00"} {"global_step": 14058, "acc_step": 0, "speed/wps": 12907.108253028282, "speed/FLOPS": 202723871339172.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048403721302747726, "optim/lr": 0.002803710893523977, "optim/total_tokens": 7370440704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.7831006050109863, "created_at": "2025-01-16T21:16:05.454218+00:00"} {"global_step": 14059, "acc_step": 0, "speed/wps": 12910.63208035398, "speed/FLOPS": 202779217889568.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06327986717224121, "optim/lr": 0.002803668517182343, "optim/total_tokens": 7370964992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.6762266159057617, "created_at": "2025-01-16T21:16:15.610665+00:00"} {"global_step": 14060, "acc_step": 0, "speed/wps": 12913.349001576194, "speed/FLOPS": 202821890870804.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.062068283557891846, "optim/lr": 0.0028036261365872615, "optim/total_tokens": 7371489280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 2.914706230163574, "created_at": "2025-01-16T21:16:25.767901+00:00"} {"global_step": 14061, "acc_step": 0, "speed/wps": 12907.030524679467, "speed/FLOPS": 202722650508644.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049555107951164246, "optim/lr": 0.0028035837517388714, "optim/total_tokens": 7372013568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 2.8527307510375977, "created_at": "2025-01-16T21:16:35.927654+00:00"} {"global_step": 14062, "acc_step": 0, "speed/wps": 12914.433424357252, "speed/FLOPS": 202838923220734.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05126341059803963, "optim/lr": 0.00280354136263731, "optim/total_tokens": 7372537856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.0020484924316406, "created_at": "2025-01-16T21:16:46.083098+00:00"} {"global_step": 14063, "acc_step": 0, "speed/wps": 12908.306481790738, "speed/FLOPS": 202742691168425.66, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08516541868448257, "optim/lr": 0.0028034989692827168, "optim/total_tokens": 7373062144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 2.851546287536621, "created_at": "2025-01-16T21:16:56.243674+00:00"} {"global_step": 14064, "acc_step": 0, "speed/wps": 12911.163680355337, "speed/FLOPS": 202787567398084.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07268451154232025, "optim/lr": 0.0028034565716752295, "optim/total_tokens": 7373586432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.7743139266967773, "created_at": "2025-01-16T21:17:06.399435+00:00"} {"global_step": 14065, "acc_step": 0, "speed/wps": 12910.075057709662, "speed/FLOPS": 202770469083510.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04498724639415741, "optim/lr": 0.0028034141698149865, "optim/total_tokens": 7374110720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8169901371002197, "created_at": "2025-01-16T21:17:16.559419+00:00"} {"global_step": 14066, "acc_step": 0, "speed/wps": 12910.860975791365, "speed/FLOPS": 202782813006948.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058757536113262177, "optim/lr": 0.002803371763702126, "optim/total_tokens": 7374635008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.791574716567993, "created_at": "2025-01-16T21:17:26.715184+00:00"} {"global_step": 14067, "acc_step": 0, "speed/wps": 12908.330197284393, "speed/FLOPS": 202743063652842.0, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06376047432422638, "optim/lr": 0.0028033293533367867, "optim/total_tokens": 7375159296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.966993808746338, "created_at": "2025-01-16T21:17:36.874753+00:00"} {"global_step": 14068, "acc_step": 0, "speed/wps": 12905.7133783802, "speed/FLOPS": 202701962916065.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045203957706689835, "optim/lr": 0.002803286938719107, "optim/total_tokens": 7375683584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.91141676902771, "created_at": "2025-01-16T21:17:47.036068+00:00"} {"global_step": 14069, "acc_step": 0, "speed/wps": 12915.990901591551, "speed/FLOPS": 202863385540896.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.057979315519332886, "optim/lr": 0.0028032445198492247, "optim/total_tokens": 7376207872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 2.800147771835327, "created_at": "2025-01-16T21:17:57.187802+00:00"} {"global_step": 14070, "acc_step": 0, "speed/wps": 12908.499684323308, "speed/FLOPS": 202745725679687.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04211373254656792, "optim/lr": 0.002803202096727279, "optim/total_tokens": 7376732160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8620717525482178, "created_at": "2025-01-16T21:18:07.348248+00:00"} {"global_step": 14071, "acc_step": 0, "speed/wps": 12910.478018933323, "speed/FLOPS": 202776798143254.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0480031743645668, "optim/lr": 0.0028031596693534077, "optim/total_tokens": 7377256448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 2.8080995082855225, "created_at": "2025-01-16T21:18:17.507422+00:00"} {"global_step": 14072, "acc_step": 0, "speed/wps": 12908.776141982798, "speed/FLOPS": 202750067827122.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04485594853758812, "optim/lr": 0.0028031172377277493, "optim/total_tokens": 7377780736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344503, "loss/out": 2.8429980278015137, "created_at": "2025-01-16T21:18:27.665625+00:00"} {"global_step": 14073, "acc_step": 0, "speed/wps": 12909.278391163309, "speed/FLOPS": 202757956340665.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0483902208507061, "optim/lr": 0.0028030748018504424, "optim/total_tokens": 7378305024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8227758407592773, "created_at": "2025-01-16T21:18:37.822550+00:00"} {"global_step": 14074, "acc_step": 0, "speed/wps": 12911.112026192563, "speed/FLOPS": 202786756098478.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07326193898916245, "optim/lr": 0.0028030323617216257, "optim/total_tokens": 7378829312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403804, "loss/out": 2.8252546787261963, "created_at": "2025-01-16T21:18:47.981262+00:00"} {"global_step": 14075, "acc_step": 0, "speed/wps": 12910.881808342803, "speed/FLOPS": 202783140210795.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045480187982320786, "optim/lr": 0.002802989917341437, "optim/total_tokens": 7379353600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.7893104553222656, "created_at": "2025-01-16T21:18:58.140964+00:00"} {"global_step": 14076, "acc_step": 0, "speed/wps": 12912.867055385137, "speed/FLOPS": 202814321243610.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06308142840862274, "optim/lr": 0.0028029474687100153, "optim/total_tokens": 7379877888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360021, "loss/out": 2.953214168548584, "created_at": "2025-01-16T21:19:08.298855+00:00"} {"global_step": 14077, "acc_step": 0, "speed/wps": 12907.460648279668, "speed/FLOPS": 202729406190837.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059381525963544846, "optim/lr": 0.0028029050158274987, "optim/total_tokens": 7380402176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.747215509414673, "created_at": "2025-01-16T21:19:18.464236+00:00"} {"global_step": 14078, "acc_step": 0, "speed/wps": 12910.08810947268, "speed/FLOPS": 202770674079383.53, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062349721789360046, "optim/lr": 0.0028028625586940265, "optim/total_tokens": 7380926464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.731332302093506, "created_at": "2025-01-16T21:19:28.626139+00:00"} {"global_step": 14079, "acc_step": 0, "speed/wps": 12909.823865263936, "speed/FLOPS": 202766523760975.62, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04987119510769844, "optim/lr": 0.0028028200973097364, "optim/total_tokens": 7381450752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.8209149837493896, "created_at": "2025-01-16T21:19:38.782822+00:00"} {"global_step": 14080, "acc_step": 0, "speed/wps": 12904.980757721722, "speed/FLOPS": 202690456101897.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06474151462316513, "optim/lr": 0.002802777631674768, "optim/total_tokens": 7381975040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8339974880218506, "created_at": "2025-01-16T21:19:48.943176+00:00"} {"global_step": 14081, "acc_step": 0, "speed/wps": 12903.917994765912, "speed/FLOPS": 202673763949287.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09591119736433029, "optim/lr": 0.002802735161789258, "optim/total_tokens": 7382499328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.828655481338501, "created_at": "2025-01-16T21:19:59.104158+00:00"} {"global_step": 14082, "acc_step": 0, "speed/wps": 12907.232852346799, "speed/FLOPS": 202725828342687.12, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07699449360370636, "optim/lr": 0.0028026926876533464, "optim/total_tokens": 7383023616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.850131034851074, "created_at": "2025-01-16T21:20:09.265378+00:00"} {"global_step": 14083, "acc_step": 0, "speed/wps": 12905.95360547322, "speed/FLOPS": 202705736012668.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06348006427288055, "optim/lr": 0.0028026502092671716, "optim/total_tokens": 7383547904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413950, "loss/out": 2.8523335456848145, "created_at": "2025-01-16T21:20:19.424877+00:00"} {"global_step": 14084, "acc_step": 0, "speed/wps": 12905.83728920121, "speed/FLOPS": 202703909105780.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05925435572862625, "optim/lr": 0.0028026077266308717, "optim/total_tokens": 7384072192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.9148192405700684, "created_at": "2025-01-16T21:20:29.587287+00:00"} {"global_step": 14085, "acc_step": 0, "speed/wps": 12907.640283816923, "speed/FLOPS": 202732227613794.94, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054050832986831665, "optim/lr": 0.002802565239744586, "optim/total_tokens": 7384596480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.8110642433166504, "created_at": "2025-01-16T21:20:39.747253+00:00"} {"global_step": 14086, "acc_step": 0, "speed/wps": 12909.31760919421, "speed/FLOPS": 202758572313731.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046271227300167084, "optim/lr": 0.0028025227486084526, "optim/total_tokens": 7385120768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 2.7813870906829834, "created_at": "2025-01-16T21:20:49.905064+00:00"} {"global_step": 14087, "acc_step": 0, "speed/wps": 12913.72106151304, "speed/FLOPS": 202827734583376.66, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04816922917962074, "optim/lr": 0.0028024802532226104, "optim/total_tokens": 7385645056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.841032028198242, "created_at": "2025-01-16T21:21:00.061813+00:00"} {"global_step": 14088, "acc_step": 0, "speed/wps": 12912.343834123762, "speed/FLOPS": 202806103334719.28, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05717067793011665, "optim/lr": 0.002802437753587198, "optim/total_tokens": 7386169344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 2.713535785675049, "created_at": "2025-01-16T21:21:10.216160+00:00"} {"global_step": 14089, "acc_step": 0, "speed/wps": 12905.53118109482, "speed/FLOPS": 202699101257333.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045415591448545456, "optim/lr": 0.002802395249702354, "optim/total_tokens": 7386693632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.812326431274414, "created_at": "2025-01-16T21:21:20.376046+00:00"} {"global_step": 14090, "acc_step": 0, "speed/wps": 12905.639928426612, "speed/FLOPS": 202700809283613.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050184693187475204, "optim/lr": 0.002802352741568217, "optim/total_tokens": 7387217920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370300, "loss/out": 2.823380708694458, "created_at": "2025-01-16T21:21:30.539827+00:00"} {"global_step": 14091, "acc_step": 0, "speed/wps": 12906.519594618181, "speed/FLOPS": 202714625649940.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.044218696653842926, "optim/lr": 0.0028023102291849255, "optim/total_tokens": 7387742208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.7892918586730957, "created_at": "2025-01-16T21:21:40.702126+00:00"} {"global_step": 14092, "acc_step": 0, "speed/wps": 12909.13430720571, "speed/FLOPS": 202755693304119.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05447908863425255, "optim/lr": 0.0028022677125526187, "optim/total_tokens": 7388266496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8135323524475098, "created_at": "2025-01-16T21:21:50.863258+00:00"} {"global_step": 14093, "acc_step": 0, "speed/wps": 12914.35843291071, "speed/FLOPS": 202837745377017.16, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053288161754608154, "optim/lr": 0.002802225191671435, "optim/total_tokens": 7388790784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 2.8994598388671875, "created_at": "2025-01-16T21:22:01.020342+00:00"} {"global_step": 14094, "acc_step": 0, "speed/wps": 12912.422342035507, "speed/FLOPS": 202807336409354.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0416780561208725, "optim/lr": 0.0028021826665415137, "optim/total_tokens": 7389315072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.7934141159057617, "created_at": "2025-01-16T21:22:11.175492+00:00"} {"global_step": 14095, "acc_step": 0, "speed/wps": 12910.728690836268, "speed/FLOPS": 202780735289949.72, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0458921417593956, "optim/lr": 0.0028021401371629926, "optim/total_tokens": 7389839360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9417576789855957, "created_at": "2025-01-16T21:22:21.331212+00:00"} {"global_step": 14096, "acc_step": 0, "speed/wps": 12910.381120860287, "speed/FLOPS": 202775276225865.75, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041277121752500534, "optim/lr": 0.0028020976035360106, "optim/total_tokens": 7390363648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.711724281311035, "created_at": "2025-01-16T21:22:31.487176+00:00"} {"global_step": 14097, "acc_step": 0, "speed/wps": 12910.495150373075, "speed/FLOPS": 202777067216057.06, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04901619628071785, "optim/lr": 0.0028020550656607073, "optim/total_tokens": 7390887936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 2.8478739261627197, "created_at": "2025-01-16T21:22:41.643741+00:00"} {"global_step": 14098, "acc_step": 0, "speed/wps": 12912.443648765764, "speed/FLOPS": 202807671060831.62, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05252853035926819, "optim/lr": 0.0028020125235372214, "optim/total_tokens": 7391412224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8930273056030273, "created_at": "2025-01-16T21:22:51.800330+00:00"} {"global_step": 14099, "acc_step": 0, "speed/wps": 12912.899385582594, "speed/FLOPS": 202814829033787.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05140409618616104, "optim/lr": 0.0028019699771656905, "optim/total_tokens": 7391936512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.753309726715088, "created_at": "2025-01-16T21:23:01.955287+00:00"} {"global_step": 14100, "acc_step": 0, "speed/wps": 12906.404605598842, "speed/FLOPS": 202712819589380.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047775428742170334, "optim/lr": 0.0028019274265462545, "optim/total_tokens": 7392460800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.821303129196167, "created_at": "2025-01-16T21:23:12.117334+00:00"} {"global_step": 14101, "acc_step": 0, "speed/wps": 12907.40973282223, "speed/FLOPS": 202728606493613.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05331538990139961, "optim/lr": 0.0028018848716790524, "optim/total_tokens": 7392985088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 2.9054818153381348, "created_at": "2025-01-16T21:23:22.277662+00:00"} {"global_step": 14102, "acc_step": 0, "speed/wps": 12913.390142505492, "speed/FLOPS": 202822537045632.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05554148182272911, "optim/lr": 0.0028018423125642216, "optim/total_tokens": 7393509376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.7877819538116455, "created_at": "2025-01-16T21:23:32.433162+00:00"} {"global_step": 14103, "acc_step": 0, "speed/wps": 12911.71532781709, "speed/FLOPS": 202796231779514.8, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06281796097755432, "optim/lr": 0.0028017997492019026, "optim/total_tokens": 7394033664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 2.8075780868530273, "created_at": "2025-01-16T21:23:42.588131+00:00"} {"global_step": 14104, "acc_step": 0, "speed/wps": 12908.10136197579, "speed/FLOPS": 202739469479867.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06957165151834488, "optim/lr": 0.0028017571815922336, "optim/total_tokens": 7394557952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.958777904510498, "created_at": "2025-01-16T21:23:52.750707+00:00"} {"global_step": 14105, "acc_step": 0, "speed/wps": 12913.116036915315, "speed/FLOPS": 202818231840678.44, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04581087827682495, "optim/lr": 0.002801714609735353, "optim/total_tokens": 7395082240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.821223497390747, "created_at": "2025-01-16T21:24:02.906316+00:00"} {"global_step": 14106, "acc_step": 0, "speed/wps": 12907.608421509512, "speed/FLOPS": 202731727172473.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059192877262830734, "optim/lr": 0.0028016720336314005, "optim/total_tokens": 7395606528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.872899293899536, "created_at": "2025-01-16T21:24:13.064800+00:00"} {"global_step": 14107, "acc_step": 0, "speed/wps": 12911.594843505718, "speed/FLOPS": 202794339407841.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05524621903896332, "optim/lr": 0.002801629453280515, "optim/total_tokens": 7396130816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.847507953643799, "created_at": "2025-01-16T21:24:23.220071+00:00"} {"global_step": 14108, "acc_step": 0, "speed/wps": 12912.178468207334, "speed/FLOPS": 202803506035765.2, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06541021168231964, "optim/lr": 0.002801586868682835, "optim/total_tokens": 7396655104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.830875873565674, "created_at": "2025-01-16T21:24:33.378541+00:00"} {"global_step": 14109, "acc_step": 0, "speed/wps": 12912.776547569265, "speed/FLOPS": 202812899693991.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05561158433556557, "optim/lr": 0.0028015442798384996, "optim/total_tokens": 7397179392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.8633060455322266, "created_at": "2025-01-16T21:24:43.532552+00:00"} {"global_step": 14110, "acc_step": 0, "speed/wps": 12909.514713010658, "speed/FLOPS": 202761668100018.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05736560747027397, "optim/lr": 0.0028015016867476475, "optim/total_tokens": 7397703680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.828409194946289, "created_at": "2025-01-16T21:24:53.690743+00:00"} {"global_step": 14111, "acc_step": 0, "speed/wps": 12901.973249979372, "speed/FLOPS": 202643219060055.66, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05161609128117561, "optim/lr": 0.0028014590894104184, "optim/total_tokens": 7398227968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.864468574523926, "created_at": "2025-01-16T21:25:03.856997+00:00"} {"global_step": 14112, "acc_step": 0, "speed/wps": 12911.916493405231, "speed/FLOPS": 202799391361507.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0596294030547142, "optim/lr": 0.00280141648782695, "optim/total_tokens": 7398752256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 3.0115840435028076, "created_at": "2025-01-16T21:25:14.012230+00:00"} {"global_step": 14113, "acc_step": 0, "speed/wps": 12912.977006915115, "speed/FLOPS": 202816048183478.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058670759201049805, "optim/lr": 0.0028013738819973825, "optim/total_tokens": 7399276544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 3.0011813640594482, "created_at": "2025-01-16T21:25:24.168298+00:00"} {"global_step": 14114, "acc_step": 0, "speed/wps": 12916.716448468726, "speed/FLOPS": 202874781251608.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04554006829857826, "optim/lr": 0.002801331271921854, "optim/total_tokens": 7399800832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 2.7934088706970215, "created_at": "2025-01-16T21:25:34.319209+00:00"} {"global_step": 14115, "acc_step": 0, "speed/wps": 12908.849695129244, "speed/FLOPS": 202751223080360.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04893157258629799, "optim/lr": 0.002801288657600505, "optim/total_tokens": 7400325120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8579840660095215, "created_at": "2025-01-16T21:25:44.480419+00:00"} {"global_step": 14116, "acc_step": 0, "speed/wps": 12913.14787547205, "speed/FLOPS": 202818731908962.75, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046118929982185364, "optim/lr": 0.002801246039033473, "optim/total_tokens": 7400849408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.819941759109497, "created_at": "2025-01-16T21:25:54.637435+00:00"} {"global_step": 14117, "acc_step": 0, "speed/wps": 12905.52562963477, "speed/FLOPS": 202699014064024.88, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044899262487888336, "optim/lr": 0.002801203416220897, "optim/total_tokens": 7401373696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8745014667510986, "created_at": "2025-01-16T21:26:04.807148+00:00"} {"global_step": 14118, "acc_step": 0, "speed/wps": 12911.010813185077, "speed/FLOPS": 202785166409114.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055978525429964066, "optim/lr": 0.0028011607891629173, "optim/total_tokens": 7401897984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.9525949954986572, "created_at": "2025-01-16T21:26:14.963361+00:00"} {"global_step": 14119, "acc_step": 0, "speed/wps": 12913.357447282133, "speed/FLOPS": 202822023522221.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0427025705575943, "optim/lr": 0.002801118157859672, "optim/total_tokens": 7402422272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8598129749298096, "created_at": "2025-01-16T21:26:25.117889+00:00"} {"global_step": 14120, "acc_step": 0, "speed/wps": 12911.687930122316, "speed/FLOPS": 202795801461070.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07085079699754715, "optim/lr": 0.0028010755223113015, "optim/total_tokens": 7402946560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.7129693031311035, "created_at": "2025-01-16T21:26:35.275169+00:00"} {"global_step": 14121, "acc_step": 0, "speed/wps": 12908.939532197495, "speed/FLOPS": 202752634094966.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06595934182405472, "optim/lr": 0.0028010328825179426, "optim/total_tokens": 7403470848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9028568267822266, "created_at": "2025-01-16T21:26:45.433350+00:00"} {"global_step": 14122, "acc_step": 0, "speed/wps": 12908.90667005949, "speed/FLOPS": 202752117949931.53, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05990287661552429, "optim/lr": 0.002800990238479736, "optim/total_tokens": 7403995136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.838021755218506, "created_at": "2025-01-16T21:26:55.593402+00:00"} {"global_step": 14123, "acc_step": 0, "speed/wps": 12912.972418208388, "speed/FLOPS": 202815976111533.38, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07540310174226761, "optim/lr": 0.002800947590196821, "optim/total_tokens": 7404519424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.871894359588623, "created_at": "2025-01-16T21:27:05.749227+00:00"} {"global_step": 14124, "acc_step": 0, "speed/wps": 12907.479875299585, "speed/FLOPS": 202729708177605.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04159965366125107, "optim/lr": 0.0028009049376693366, "optim/total_tokens": 7405043712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.7830352783203125, "created_at": "2025-01-16T21:27:15.911493+00:00"} {"global_step": 14125, "acc_step": 0, "speed/wps": 12910.150892362532, "speed/FLOPS": 202771660170941.4, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0654531717300415, "optim/lr": 0.002800862280897421, "optim/total_tokens": 7405568000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 2.7879772186279297, "created_at": "2025-01-16T21:27:26.067642+00:00"} {"global_step": 14126, "acc_step": 0, "speed/wps": 12907.439990128716, "speed/FLOPS": 202729081726189.28, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052772682160139084, "optim/lr": 0.0028008196198812145, "optim/total_tokens": 7406092288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.8057780265808105, "created_at": "2025-01-16T21:27:36.229881+00:00"} {"global_step": 14127, "acc_step": 0, "speed/wps": 12904.541581881113, "speed/FLOPS": 202683558241830.38, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05368189513683319, "optim/lr": 0.0028007769546208555, "optim/total_tokens": 7406616576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9799389839172363, "created_at": "2025-01-16T21:27:46.391796+00:00"} {"global_step": 14128, "acc_step": 0, "speed/wps": 12903.15493597802, "speed/FLOPS": 202661779062471.16, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06742162257432938, "optim/lr": 0.002800734285116484, "optim/total_tokens": 7407140864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.7237143516540527, "created_at": "2025-01-16T21:27:56.558386+00:00"} {"global_step": 14129, "acc_step": 0, "speed/wps": 12916.486618690058, "speed/FLOPS": 202871171459115.5, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04642969369888306, "optim/lr": 0.0028006916113682383, "optim/total_tokens": 7407665152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.855879306793213, "created_at": "2025-01-16T21:28:06.709552+00:00"} {"global_step": 14130, "acc_step": 0, "speed/wps": 12908.008923743353, "speed/FLOPS": 202738017610403.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05290544405579567, "optim/lr": 0.0028006489333762584, "optim/total_tokens": 7408189440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8527603149414062, "created_at": "2025-01-16T21:28:16.869964+00:00"} {"global_step": 14131, "acc_step": 0, "speed/wps": 12908.057027449297, "speed/FLOPS": 202738773145207.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048068542033433914, "optim/lr": 0.0028006062511406835, "optim/total_tokens": 7408713728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.9197778701782227, "created_at": "2025-01-16T21:28:27.027792+00:00"} {"global_step": 14132, "acc_step": 0, "speed/wps": 12911.509505020374, "speed/FLOPS": 202792999049662.25, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07031980156898499, "optim/lr": 0.0028005635646616523, "optim/total_tokens": 7409238016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 2.8735275268554688, "created_at": "2025-01-16T21:28:37.183093+00:00"} {"global_step": 14133, "acc_step": 0, "speed/wps": 12904.061489040181, "speed/FLOPS": 202676017724046.84, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0458449125289917, "optim/lr": 0.002800520873939304, "optim/total_tokens": 7409762304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 2.9527645111083984, "created_at": "2025-01-16T21:28:47.344607+00:00"} {"global_step": 14134, "acc_step": 0, "speed/wps": 12914.778274635342, "speed/FLOPS": 202844339568225.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0655771866440773, "optim/lr": 0.0028004781789737793, "optim/total_tokens": 7410286592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 2.8873605728149414, "created_at": "2025-01-16T21:28:57.497231+00:00"} {"global_step": 14135, "acc_step": 0, "speed/wps": 12909.117262263306, "speed/FLOPS": 202755425589876.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06490383297204971, "optim/lr": 0.0028004354797652154, "optim/total_tokens": 7410810880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.9081454277038574, "created_at": "2025-01-16T21:29:07.654428+00:00"} {"global_step": 14136, "acc_step": 0, "speed/wps": 12910.468777567867, "speed/FLOPS": 202776652994911.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06290523707866669, "optim/lr": 0.0028003927763137534, "optim/total_tokens": 7411335168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 2.7602195739746094, "created_at": "2025-01-16T21:29:17.812012+00:00"} {"global_step": 14137, "acc_step": 0, "speed/wps": 12906.691376538913, "speed/FLOPS": 202717323721058.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048757970333099365, "optim/lr": 0.002800350068619532, "optim/total_tokens": 7411859456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.8588898181915283, "created_at": "2025-01-16T21:29:27.971519+00:00"} {"global_step": 14138, "acc_step": 0, "speed/wps": 12907.94783094354, "speed/FLOPS": 202737058064032.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0773385539650917, "optim/lr": 0.0028003073566826896, "optim/total_tokens": 7412383744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.888154983520508, "created_at": "2025-01-16T21:29:38.132064+00:00"} {"global_step": 14139, "acc_step": 0, "speed/wps": 12908.15783187429, "speed/FLOPS": 202740356417221.47, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06960815191268921, "optim/lr": 0.0028002646405033667, "optim/total_tokens": 7412908032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8741884231567383, "created_at": "2025-01-16T21:29:48.290080+00:00"} {"global_step": 14140, "acc_step": 0, "speed/wps": 12907.583601810322, "speed/FLOPS": 202731337344992.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06429433077573776, "optim/lr": 0.0028002219200817025, "optim/total_tokens": 7413432320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.7659552097320557, "created_at": "2025-01-16T21:29:58.448737+00:00"} {"global_step": 14141, "acc_step": 0, "speed/wps": 12913.237454837772, "speed/FLOPS": 202820138875996.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054713353514671326, "optim/lr": 0.0028001791954178363, "optim/total_tokens": 7413956608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.720947742462158, "created_at": "2025-01-16T21:30:08.602985+00:00"} {"global_step": 14142, "acc_step": 0, "speed/wps": 12906.705547106294, "speed/FLOPS": 202717546289291.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06370936334133148, "optim/lr": 0.0028001364665119076, "optim/total_tokens": 7414480896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.7947964668273926, "created_at": "2025-01-16T21:30:18.765273+00:00"} {"global_step": 14143, "acc_step": 0, "speed/wps": 12902.735378818186, "speed/FLOPS": 202655189340744.0, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.056852277368307114, "optim/lr": 0.002800093733364056, "optim/total_tokens": 7415005184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.749485969543457, "created_at": "2025-01-16T21:30:28.928451+00:00"} {"global_step": 14144, "acc_step": 0, "speed/wps": 12907.239626897037, "speed/FLOPS": 202725934746307.8, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053372759371995926, "optim/lr": 0.00280005099597442, "optim/total_tokens": 7415529472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.7845497131347656, "created_at": "2025-01-16T21:30:39.088757+00:00"} {"global_step": 14145, "acc_step": 0, "speed/wps": 12906.963757053483, "speed/FLOPS": 202721601831329.94, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05559229105710983, "optim/lr": 0.00280000825434314, "optim/total_tokens": 7416053760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411556, "loss/out": 2.8529016971588135, "created_at": "2025-01-16T21:30:49.250774+00:00"} {"global_step": 14146, "acc_step": 0, "speed/wps": 12911.456572367339, "speed/FLOPS": 202792167669609.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06496542692184448, "optim/lr": 0.0027999655084703545, "optim/total_tokens": 7416578048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.8366622924804688, "created_at": "2025-01-16T21:30:59.406223+00:00"} {"global_step": 14147, "acc_step": 0, "speed/wps": 12908.490534587669, "speed/FLOPS": 202745581970516.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04827598109841347, "optim/lr": 0.0027999227583562036, "optim/total_tokens": 7417102336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8455092906951904, "created_at": "2025-01-16T21:31:09.564230+00:00"} {"global_step": 14148, "acc_step": 0, "speed/wps": 12910.083217278949, "speed/FLOPS": 202770597240758.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05475683882832527, "optim/lr": 0.0027998800040008273, "optim/total_tokens": 7417626624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.720831871032715, "created_at": "2025-01-16T21:31:19.725333+00:00"} {"global_step": 14149, "acc_step": 0, "speed/wps": 12907.455589980103, "speed/FLOPS": 202729326743291.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056505464017391205, "optim/lr": 0.0027998372454043643, "optim/total_tokens": 7418150912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.887373447418213, "created_at": "2025-01-16T21:31:29.887017+00:00"} {"global_step": 14150, "acc_step": 0, "speed/wps": 12905.054717292383, "speed/FLOPS": 202691617738586.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05080302804708481, "optim/lr": 0.002799794482566954, "optim/total_tokens": 7418675200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.7980380058288574, "created_at": "2025-01-16T21:31:40.047627+00:00"} {"global_step": 14151, "acc_step": 0, "speed/wps": 12903.63291226853, "speed/FLOPS": 202669286336921.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04774824157357216, "optim/lr": 0.0027997517154887365, "optim/total_tokens": 7419199488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8419666290283203, "created_at": "2025-01-16T21:31:50.211614+00:00"} {"global_step": 14152, "acc_step": 0, "speed/wps": 12912.028497094532, "speed/FLOPS": 202801150533356.75, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04755174741148949, "optim/lr": 0.0027997089441698514, "optim/total_tokens": 7419723776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 2.8776330947875977, "created_at": "2025-01-16T21:32:00.370162+00:00"} {"global_step": 14153, "acc_step": 0, "speed/wps": 12911.304119938886, "speed/FLOPS": 202789773194728.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05351179838180542, "optim/lr": 0.0027996661686104374, "optim/total_tokens": 7420248064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8253908157348633, "created_at": "2025-01-16T21:32:10.528325+00:00"} {"global_step": 14154, "acc_step": 0, "speed/wps": 12908.993706425988, "speed/FLOPS": 202753484975667.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.039543844759464264, "optim/lr": 0.0027996233888106345, "optim/total_tokens": 7420772352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.8287103176116943, "created_at": "2025-01-16T21:32:20.687348+00:00"} {"global_step": 14155, "acc_step": 0, "speed/wps": 12914.073501559358, "speed/FLOPS": 202833270138607.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0531802624464035, "optim/lr": 0.002799580604770583, "optim/total_tokens": 7421296640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.883079767227173, "created_at": "2025-01-16T21:32:30.845214+00:00"} {"global_step": 14156, "acc_step": 0, "speed/wps": 12910.423191525682, "speed/FLOPS": 202775937003476.94, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046683359891176224, "optim/lr": 0.0027995378164904217, "optim/total_tokens": 7421820928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 2.874046802520752, "created_at": "2025-01-16T21:32:41.003486+00:00"} {"global_step": 14157, "acc_step": 0, "speed/wps": 12907.169904887085, "speed/FLOPS": 202724839666333.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06694579124450684, "optim/lr": 0.0027994950239702897, "optim/total_tokens": 7422345216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.7460391521453857, "created_at": "2025-01-16T21:32:51.163808+00:00"} {"global_step": 14158, "acc_step": 0, "speed/wps": 12911.171958562614, "speed/FLOPS": 202787697418705.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053702183067798615, "optim/lr": 0.0027994522272103283, "optim/total_tokens": 7422869504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.8449606895446777, "created_at": "2025-01-16T21:33:01.322320+00:00"} {"global_step": 14159, "acc_step": 0, "speed/wps": 12908.954945040694, "speed/FLOPS": 202752876174848.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06482835114002228, "optim/lr": 0.0027994094262106754, "optim/total_tokens": 7423393792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.7941384315490723, "created_at": "2025-01-16T21:33:11.479741+00:00"} {"global_step": 14160, "acc_step": 0, "speed/wps": 12910.062312696238, "speed/FLOPS": 202770268905561.03, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054590705782175064, "optim/lr": 0.0027993666209714715, "optim/total_tokens": 7423918080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8489980697631836, "created_at": "2025-01-16T21:33:21.638876+00:00"} {"global_step": 14161, "acc_step": 0, "speed/wps": 12912.672315813226, "speed/FLOPS": 202811262591033.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0656484067440033, "optim/lr": 0.0027993238114928567, "optim/total_tokens": 7424442368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.748429298400879, "created_at": "2025-01-16T21:33:31.794383+00:00"} {"global_step": 14162, "acc_step": 0, "speed/wps": 12912.12183399944, "speed/FLOPS": 202802616517706.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07153919339179993, "optim/lr": 0.0027992809977749693, "optim/total_tokens": 7424966656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.903524875640869, "created_at": "2025-01-16T21:33:41.949333+00:00"} {"global_step": 14163, "acc_step": 0, "speed/wps": 12901.030237252691, "speed/FLOPS": 202628407749347.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050686806440353394, "optim/lr": 0.0027992381798179504, "optim/total_tokens": 7425490944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 2.735883951187134, "created_at": "2025-01-16T21:33:52.114851+00:00"} {"global_step": 14164, "acc_step": 0, "speed/wps": 12908.939495491835, "speed/FLOPS": 202752633518453.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06343895196914673, "optim/lr": 0.0027991953576219386, "optim/total_tokens": 7426015232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.7942302227020264, "created_at": "2025-01-16T21:34:02.277709+00:00"} {"global_step": 14165, "acc_step": 0, "speed/wps": 12913.217761607224, "speed/FLOPS": 202819829566749.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06867296248674393, "optim/lr": 0.0027991525311870744, "optim/total_tokens": 7426539520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8150839805603027, "created_at": "2025-01-16T21:34:12.431463+00:00"} {"global_step": 14166, "acc_step": 0, "speed/wps": 12910.39681783014, "speed/FLOPS": 202775522768347.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05785108730196953, "optim/lr": 0.002799109700513497, "optim/total_tokens": 7427063808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8635470867156982, "created_at": "2025-01-16T21:34:22.590106+00:00"} {"global_step": 14167, "acc_step": 0, "speed/wps": 12909.156140664518, "speed/FLOPS": 202756036228592.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.08155474066734314, "optim/lr": 0.002799066865601347, "optim/total_tokens": 7427588096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8199803829193115, "created_at": "2025-01-16T21:34:32.750217+00:00"} {"global_step": 14168, "acc_step": 0, "speed/wps": 12904.977561555164, "speed/FLOPS": 202690405901710.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05826755240559578, "optim/lr": 0.0027990240264507624, "optim/total_tokens": 7428112384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8617260456085205, "created_at": "2025-01-16T21:34:42.914496+00:00"} {"global_step": 14169, "acc_step": 0, "speed/wps": 12910.535144437972, "speed/FLOPS": 202777695377803.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08402042090892792, "optim/lr": 0.0027989811830618854, "optim/total_tokens": 7428636672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.8126447200775146, "created_at": "2025-01-16T21:34:53.070449+00:00"} {"global_step": 14170, "acc_step": 0, "speed/wps": 12909.241995101862, "speed/FLOPS": 202757384690506.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.054795969277620316, "optim/lr": 0.002798938335434853, "optim/total_tokens": 7429160960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.9222922325134277, "created_at": "2025-01-16T21:35:03.229321+00:00"} {"global_step": 14171, "acc_step": 0, "speed/wps": 12909.010545014527, "speed/FLOPS": 202753749448839.53, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07098579406738281, "optim/lr": 0.002798895483569808, "optim/total_tokens": 7429685248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 2.8158419132232666, "created_at": "2025-01-16T21:35:13.389745+00:00"} {"global_step": 14172, "acc_step": 0, "speed/wps": 12902.446183711498, "speed/FLOPS": 202650647134197.75, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06254896521568298, "optim/lr": 0.0027988526274668876, "optim/total_tokens": 7430209536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 2.779414176940918, "created_at": "2025-01-16T21:35:23.552042+00:00"} {"global_step": 14173, "acc_step": 0, "speed/wps": 12910.039188886476, "speed/FLOPS": 202769905714353.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04431803897023201, "optim/lr": 0.002798809767126233, "optim/total_tokens": 7430733824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8071188926696777, "created_at": "2025-01-16T21:35:33.711440+00:00"} {"global_step": 14174, "acc_step": 0, "speed/wps": 12912.190635736122, "speed/FLOPS": 202803697143524.75, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04555101320147514, "optim/lr": 0.002798766902547984, "optim/total_tokens": 7431258112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332063, "loss/out": 2.9419682025909424, "created_at": "2025-01-16T21:35:43.866487+00:00"} {"global_step": 14175, "acc_step": 0, "speed/wps": 12910.986526421219, "speed/FLOPS": 202784784952114.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04643024504184723, "optim/lr": 0.0027987240337322803, "optim/total_tokens": 7431782400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8257224559783936, "created_at": "2025-01-16T21:35:54.025595+00:00"} {"global_step": 14176, "acc_step": 0, "speed/wps": 12910.70188477936, "speed/FLOPS": 202780314263990.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045428045094013214, "optim/lr": 0.0027986811606792615, "optim/total_tokens": 7432306688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 2.8055193424224854, "created_at": "2025-01-16T21:36:04.181308+00:00"} {"global_step": 14177, "acc_step": 0, "speed/wps": 12905.185329874865, "speed/FLOPS": 202693669188674.7, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04543803632259369, "optim/lr": 0.0027986382833890674, "optim/total_tokens": 7432830976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.794588088989258, "created_at": "2025-01-16T21:36:14.344988+00:00"} {"global_step": 14178, "acc_step": 0, "speed/wps": 12910.100685186917, "speed/FLOPS": 202770871598256.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04213941842317581, "optim/lr": 0.002798595401861838, "optim/total_tokens": 7433355264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8878087997436523, "created_at": "2025-01-16T21:36:24.501765+00:00"} {"global_step": 14179, "acc_step": 0, "speed/wps": 12910.93345947908, "speed/FLOPS": 202783951462866.7, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04704425111413002, "optim/lr": 0.002798552516097714, "optim/total_tokens": 7433879552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.913560152053833, "created_at": "2025-01-16T21:36:34.658430+00:00"} {"global_step": 14180, "acc_step": 0, "speed/wps": 12906.693746188925, "speed/FLOPS": 202717360939668.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.041124630719423294, "optim/lr": 0.0027985096260968348, "optim/total_tokens": 7434403840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.7987451553344727, "created_at": "2025-01-16T21:36:44.824855+00:00"} {"global_step": 14181, "acc_step": 0, "speed/wps": 12910.419130474073, "speed/FLOPS": 202775873219080.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04009934142231941, "optim/lr": 0.0027984667318593395, "optim/total_tokens": 7434928128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.7954792976379395, "created_at": "2025-01-16T21:36:54.981999+00:00"} {"global_step": 14182, "acc_step": 0, "speed/wps": 12906.245858737093, "speed/FLOPS": 202710326251775.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04805823415517807, "optim/lr": 0.0027984238333853694, "optim/total_tokens": 7435452416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.7882866859436035, "created_at": "2025-01-16T21:37:05.144297+00:00"} {"global_step": 14183, "acc_step": 0, "speed/wps": 12907.341885234586, "speed/FLOPS": 202727540854016.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05422171577811241, "optim/lr": 0.0027983809306750635, "optim/total_tokens": 7435976704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 2.7567124366760254, "created_at": "2025-01-16T21:37:15.308085+00:00"} {"global_step": 14184, "acc_step": 0, "speed/wps": 12908.618872417908, "speed/FLOPS": 202747597692495.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053257983177900314, "optim/lr": 0.0027983380237285626, "optim/total_tokens": 7436500992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.9188156127929688, "created_at": "2025-01-16T21:37:25.469558+00:00"} {"global_step": 14185, "acc_step": 0, "speed/wps": 12906.079461894382, "speed/FLOPS": 202707712760707.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060019705444574356, "optim/lr": 0.0027982951125460056, "optim/total_tokens": 7437025280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.9090576171875, "created_at": "2025-01-16T21:37:35.631479+00:00"} {"global_step": 14186, "acc_step": 0, "speed/wps": 12906.805595369555, "speed/FLOPS": 202719117684746.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0565757155418396, "optim/lr": 0.0027982521971275333, "optim/total_tokens": 7437549568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.751732110977173, "created_at": "2025-01-16T21:37:45.791559+00:00"} {"global_step": 14187, "acc_step": 0, "speed/wps": 12909.875422671535, "speed/FLOPS": 202767333540909.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04618263617157936, "optim/lr": 0.0027982092774732855, "optim/total_tokens": 7438073856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.823131561279297, "created_at": "2025-01-16T21:37:55.950418+00:00"} {"global_step": 14188, "acc_step": 0, "speed/wps": 12910.41306436441, "speed/FLOPS": 202775777942492.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04734570533037186, "optim/lr": 0.002798166353583402, "optim/total_tokens": 7438598144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.850531578063965, "created_at": "2025-01-16T21:38:06.110874+00:00"} {"global_step": 14189, "acc_step": 0, "speed/wps": 12907.302255423443, "speed/FLOPS": 202726918413375.66, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04863712564110756, "optim/lr": 0.0027981234254580234, "optim/total_tokens": 7439122432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8891732692718506, "created_at": "2025-01-16T21:38:16.271423+00:00"} {"global_step": 14190, "acc_step": 0, "speed/wps": 12907.392733865643, "speed/FLOPS": 202728339501640.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05240819975733757, "optim/lr": 0.0027980804930972897, "optim/total_tokens": 7439646720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297773, "loss/out": 2.7980291843414307, "created_at": "2025-01-16T21:38:26.433456+00:00"} {"global_step": 14191, "acc_step": 0, "speed/wps": 12910.992835877982, "speed/FLOPS": 202784884050803.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05110754445195198, "optim/lr": 0.002798037556501341, "optim/total_tokens": 7440171008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.734029769897461, "created_at": "2025-01-16T21:38:36.591762+00:00"} {"global_step": 14192, "acc_step": 0, "speed/wps": 12914.783669326014, "speed/FLOPS": 202844424299255.56, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04663806036114693, "optim/lr": 0.002797994615670316, "optim/total_tokens": 7440695296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445168, "loss/out": 2.8027760982513428, "created_at": "2025-01-16T21:38:46.744548+00:00"} {"global_step": 14193, "acc_step": 0, "speed/wps": 12909.089546418674, "speed/FLOPS": 202754990274450.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053200338035821915, "optim/lr": 0.002797951670604357, "optim/total_tokens": 7441219584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8709702491760254, "created_at": "2025-01-16T21:38:56.903186+00:00"} {"global_step": 14194, "acc_step": 0, "speed/wps": 12912.71877549662, "speed/FLOPS": 202811992304203.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049183107912540436, "optim/lr": 0.0027979087213036028, "optim/total_tokens": 7441743872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 2.847201347351074, "created_at": "2025-01-16T21:39:07.061753+00:00"} {"global_step": 14195, "acc_step": 0, "speed/wps": 12908.41382199994, "speed/FLOPS": 202744377093912.1, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05455787479877472, "optim/lr": 0.0027978657677681937, "optim/total_tokens": 7442268160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.716723918914795, "created_at": "2025-01-16T21:39:17.220647+00:00"} {"global_step": 14196, "acc_step": 0, "speed/wps": 12910.303649163228, "speed/FLOPS": 202774059426406.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04605870321393013, "optim/lr": 0.0027978228099982696, "optim/total_tokens": 7442792448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.7367894649505615, "created_at": "2025-01-16T21:39:27.376642+00:00"} {"global_step": 14197, "acc_step": 0, "speed/wps": 12912.103954292137, "speed/FLOPS": 202802335692334.2, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0638815388083458, "optim/lr": 0.0027977798479939713, "optim/total_tokens": 7443316736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.906888008117676, "created_at": "2025-01-16T21:39:37.531422+00:00"} {"global_step": 14198, "acc_step": 0, "speed/wps": 12910.517485859416, "speed/FLOPS": 202777418025562.22, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06572726368904114, "optim/lr": 0.0027977368817554385, "optim/total_tokens": 7443841024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8830270767211914, "created_at": "2025-01-16T21:39:47.687975+00:00"} {"global_step": 14199, "acc_step": 0, "speed/wps": 12912.879046495262, "speed/FLOPS": 202814509580472.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051208678632974625, "optim/lr": 0.0027976939112828118, "optim/total_tokens": 7444365312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.87066650390625, "created_at": "2025-01-16T21:39:57.843569+00:00"} {"global_step": 14200, "acc_step": 0, "speed/wps": 12908.424700206866, "speed/FLOPS": 202744547951100.06, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04991232603788376, "optim/lr": 0.002797650936576231, "optim/total_tokens": 7444889600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.938030481338501, "created_at": "2025-01-16T21:40:08.003136+00:00"} {"global_step": 14201, "acc_step": 0, "speed/wps": 12912.954524297464, "speed/FLOPS": 202815695063073.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06756075471639633, "optim/lr": 0.002797607957635836, "optim/total_tokens": 7445413888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.860255718231201, "created_at": "2025-01-16T21:40:18.157173+00:00"} {"global_step": 14202, "acc_step": 0, "speed/wps": 12911.715636987987, "speed/FLOPS": 202796236635468.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0610334649682045, "optim/lr": 0.0027975649744617684, "optim/total_tokens": 7445938176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.822577476501465, "created_at": "2025-01-16T21:40:28.313385+00:00"} {"global_step": 14203, "acc_step": 0, "speed/wps": 12901.454290066162, "speed/FLOPS": 202635068081493.03, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0724484845995903, "optim/lr": 0.002797521987054167, "optim/total_tokens": 7446462464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 2.8109521865844727, "created_at": "2025-01-16T21:40:38.476658+00:00"} {"global_step": 14204, "acc_step": 0, "speed/wps": 12907.516376253394, "speed/FLOPS": 202730281475242.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.076371930539608, "optim/lr": 0.0027974789954131724, "optim/total_tokens": 7446986752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.8222708702087402, "created_at": "2025-01-16T21:40:48.636900+00:00"} {"global_step": 14205, "acc_step": 0, "speed/wps": 12914.237581679929, "speed/FLOPS": 202835847242370.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05049138888716698, "optim/lr": 0.0027974359995389256, "optim/total_tokens": 7447511040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 2.8363633155822754, "created_at": "2025-01-16T21:40:58.790670+00:00"} {"global_step": 14206, "acc_step": 0, "speed/wps": 12911.09039931788, "speed/FLOPS": 202786416418693.0, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.066900834441185, "optim/lr": 0.0027973929994315663, "optim/total_tokens": 7448035328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.808912754058838, "created_at": "2025-01-16T21:41:08.948296+00:00"} {"global_step": 14207, "acc_step": 0, "speed/wps": 12907.273832388066, "speed/FLOPS": 202726471990547.72, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044482264667749405, "optim/lr": 0.0027973499950912344, "optim/total_tokens": 7448559616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397090, "loss/out": 2.921355724334717, "created_at": "2025-01-16T21:41:19.106784+00:00"} {"global_step": 14208, "acc_step": 0, "speed/wps": 12905.39134634808, "speed/FLOPS": 202696904960482.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05645252764225006, "optim/lr": 0.0027973069865180706, "optim/total_tokens": 7449083904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.857574224472046, "created_at": "2025-01-16T21:41:29.270500+00:00"} {"global_step": 14209, "acc_step": 0, "speed/wps": 12906.79558751428, "speed/FLOPS": 202718960497627.56, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043138787150382996, "optim/lr": 0.0027972639737122154, "optim/total_tokens": 7449608192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 2.7561333179473877, "created_at": "2025-01-16T21:41:39.429245+00:00"} {"global_step": 14210, "acc_step": 0, "speed/wps": 12909.748017874115, "speed/FLOPS": 202765332473493.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0578809529542923, "optim/lr": 0.0027972209566738087, "optim/total_tokens": 7450132480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8315768241882324, "created_at": "2025-01-16T21:41:49.587301+00:00"} {"global_step": 14211, "acc_step": 0, "speed/wps": 12910.427320086457, "speed/FLOPS": 202776001848197.16, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07120061665773392, "optim/lr": 0.0027971779354029913, "optim/total_tokens": 7450656768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.7875542640686035, "created_at": "2025-01-16T21:41:59.743185+00:00"} {"global_step": 14212, "acc_step": 0, "speed/wps": 12906.517507914721, "speed/FLOPS": 202714592875395.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0682491585612297, "optim/lr": 0.0027971349098999034, "optim/total_tokens": 7451181056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 2.805352210998535, "created_at": "2025-01-16T21:42:09.904694+00:00"} {"global_step": 14213, "acc_step": 0, "speed/wps": 12904.090803050096, "speed/FLOPS": 202676478140854.22, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06105170398950577, "optim/lr": 0.0027970918801646857, "optim/total_tokens": 7451705344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.9639298915863037, "created_at": "2025-01-16T21:42:20.072190+00:00"} {"global_step": 14214, "acc_step": 0, "speed/wps": 12907.660573257503, "speed/FLOPS": 202732546287339.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06819120049476624, "optim/lr": 0.0027970488461974783, "optim/total_tokens": 7452229632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373224, "loss/out": 2.9447765350341797, "created_at": "2025-01-16T21:42:30.233492+00:00"} {"global_step": 14215, "acc_step": 0, "speed/wps": 12908.731888070066, "speed/FLOPS": 202749372758612.0, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0445089116692543, "optim/lr": 0.002797005807998421, "optim/total_tokens": 7452753920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.9405813217163086, "created_at": "2025-01-16T21:42:40.390950+00:00"} {"global_step": 14216, "acc_step": 0, "speed/wps": 12906.35442755733, "speed/FLOPS": 202712031474286.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05582081899046898, "optim/lr": 0.0027969627655676553, "optim/total_tokens": 7453278208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 2.8006114959716797, "created_at": "2025-01-16T21:42:50.550650+00:00"} {"global_step": 14217, "acc_step": 0, "speed/wps": 12914.489094050126, "speed/FLOPS": 202839797589758.88, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05911624804139137, "optim/lr": 0.002796919718905321, "optim/total_tokens": 7453802496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.968947649002075, "created_at": "2025-01-16T21:43:00.714010+00:00"} {"global_step": 14218, "acc_step": 0, "speed/wps": 12910.98459224883, "speed/FLOPS": 202784754573279.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05568443238735199, "optim/lr": 0.0027968766680115586, "optim/total_tokens": 7454326784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8355579376220703, "created_at": "2025-01-16T21:43:10.871025+00:00"} {"global_step": 14219, "acc_step": 0, "speed/wps": 12906.638534279846, "speed/FLOPS": 202716493760767.03, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05266961827874184, "optim/lr": 0.0027968336128865088, "optim/total_tokens": 7454851072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411033, "loss/out": 2.804457664489746, "created_at": "2025-01-16T21:43:21.032392+00:00"} {"global_step": 14220, "acc_step": 0, "speed/wps": 12911.979792800917, "speed/FLOPS": 202800385565498.8, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05894285440444946, "optim/lr": 0.002796790553530312, "optim/total_tokens": 7455375360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 2.7373461723327637, "created_at": "2025-01-16T21:43:31.189910+00:00"} {"global_step": 14221, "acc_step": 0, "speed/wps": 12909.470727572714, "speed/FLOPS": 202760977248272.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050630323588848114, "optim/lr": 0.0027967474899431083, "optim/total_tokens": 7455899648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 2.7754569053649902, "created_at": "2025-01-16T21:43:41.350802+00:00"} {"global_step": 14222, "acc_step": 0, "speed/wps": 12911.403946935272, "speed/FLOPS": 202791341114883.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05777815729379654, "optim/lr": 0.0027967044221250386, "optim/total_tokens": 7456423936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.868009328842163, "created_at": "2025-01-16T21:43:51.507994+00:00"} {"global_step": 14223, "acc_step": 0, "speed/wps": 12909.395697834527, "speed/FLOPS": 202759798803133.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04659585654735565, "optim/lr": 0.0027966613500762434, "optim/total_tokens": 7456948224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8759870529174805, "created_at": "2025-01-16T21:44:01.669459+00:00"} {"global_step": 14224, "acc_step": 0, "speed/wps": 12911.120865764644, "speed/FLOPS": 202786894936104.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05692831799387932, "optim/lr": 0.002796618273796863, "optim/total_tokens": 7457472512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8535547256469727, "created_at": "2025-01-16T21:44:11.824849+00:00"} {"global_step": 14225, "acc_step": 0, "speed/wps": 12904.229948211263, "speed/FLOPS": 202678663606815.84, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047082655131816864, "optim/lr": 0.002796575193287038, "optim/total_tokens": 7457996800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.926851272583008, "created_at": "2025-01-16T21:44:21.988292+00:00"} {"global_step": 14226, "acc_step": 0, "speed/wps": 12909.45488009963, "speed/FLOPS": 202760728341931.47, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04692936688661575, "optim/lr": 0.0027965321085469098, "optim/total_tokens": 7458521088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.8221468925476074, "created_at": "2025-01-16T21:44:32.145268+00:00"} {"global_step": 14227, "acc_step": 0, "speed/wps": 12907.542519179258, "speed/FLOPS": 202730692085817.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050566647201776505, "optim/lr": 0.0027964890195766173, "optim/total_tokens": 7459045376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8210320472717285, "created_at": "2025-01-16T21:44:42.303445+00:00"} {"global_step": 14228, "acc_step": 0, "speed/wps": 12909.745137895861, "speed/FLOPS": 202765287239477.28, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053132619708776474, "optim/lr": 0.0027964459263763028, "optim/total_tokens": 7459569664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 2.9068896770477295, "created_at": "2025-01-16T21:44:52.464428+00:00"} {"global_step": 14229, "acc_step": 0, "speed/wps": 12909.489503497929, "speed/FLOPS": 202761272149979.28, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06209293380379677, "optim/lr": 0.0027964028289461057, "optim/total_tokens": 7460093952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 2.6661124229431152, "created_at": "2025-01-16T21:45:02.624104+00:00"} {"global_step": 14230, "acc_step": 0, "speed/wps": 12907.012427064266, "speed/FLOPS": 202722366260728.88, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04853934794664383, "optim/lr": 0.002796359727286167, "optim/total_tokens": 7460618240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431299, "loss/out": 2.756667375564575, "created_at": "2025-01-16T21:45:12.787064+00:00"} {"global_step": 14231, "acc_step": 0, "speed/wps": 12911.738075016536, "speed/FLOPS": 202796589055540.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04609324783086777, "optim/lr": 0.0027963166213966274, "optim/total_tokens": 7461142528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.785926342010498, "created_at": "2025-01-16T21:45:22.944336+00:00"} {"global_step": 14232, "acc_step": 0, "speed/wps": 12913.501498744663, "speed/FLOPS": 202824286048388.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04873516410589218, "optim/lr": 0.002796273511277627, "optim/total_tokens": 7461666816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.7908077239990234, "created_at": "2025-01-16T21:45:33.102057+00:00"} {"global_step": 14233, "acc_step": 0, "speed/wps": 12907.813854767683, "speed/FLOPS": 202734953784088.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05517036095261574, "optim/lr": 0.002796230396929308, "optim/total_tokens": 7462191104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 2.856717109680176, "created_at": "2025-01-16T21:45:43.261907+00:00"} {"global_step": 14234, "acc_step": 0, "speed/wps": 12913.429428178768, "speed/FLOPS": 202823154081115.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04642637073993683, "optim/lr": 0.0027961872783518094, "optim/total_tokens": 7462715392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 2.810560464859009, "created_at": "2025-01-16T21:45:53.416362+00:00"} {"global_step": 14235, "acc_step": 0, "speed/wps": 12905.760305177142, "speed/FLOPS": 202702699965895.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06138024851679802, "optim/lr": 0.002796144155545273, "optim/total_tokens": 7463239680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 2.846069812774658, "created_at": "2025-01-16T21:46:03.579494+00:00"} {"global_step": 14236, "acc_step": 0, "speed/wps": 12904.716225567292, "speed/FLOPS": 202686301260907.38, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04989788681268692, "optim/lr": 0.0027961010285098386, "optim/total_tokens": 7463763968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8748934268951416, "created_at": "2025-01-16T21:46:13.740976+00:00"} {"global_step": 14237, "acc_step": 0, "speed/wps": 12913.11554047583, "speed/FLOPS": 202818224043414.16, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056200843304395676, "optim/lr": 0.0027960578972456465, "optim/total_tokens": 7464288256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 2.8150320053100586, "created_at": "2025-01-16T21:46:23.899909+00:00"} {"global_step": 14238, "acc_step": 0, "speed/wps": 12913.777276062185, "speed/FLOPS": 202828617510117.84, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048882417380809784, "optim/lr": 0.0027960147617528395, "optim/total_tokens": 7464812544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.8819565773010254, "created_at": "2025-01-16T21:46:34.055906+00:00"} {"global_step": 14239, "acc_step": 0, "speed/wps": 12903.903956913804, "speed/FLOPS": 202673543465530.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04659003019332886, "optim/lr": 0.0027959716220315564, "optim/total_tokens": 7465336832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 2.822166919708252, "created_at": "2025-01-16T21:46:44.220296+00:00"} {"global_step": 14240, "acc_step": 0, "speed/wps": 12906.479845578031, "speed/FLOPS": 202714001336643.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06289972364902496, "optim/lr": 0.0027959284780819387, "optim/total_tokens": 7465861120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.7612204551696777, "created_at": "2025-01-16T21:46:54.379208+00:00"} {"global_step": 14241, "acc_step": 0, "speed/wps": 12907.726910075431, "speed/FLOPS": 202733588198221.75, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04903697967529297, "optim/lr": 0.002795885329904127, "optim/total_tokens": 7466385408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.841839075088501, "created_at": "2025-01-16T21:47:04.539847+00:00"} {"global_step": 14242, "acc_step": 0, "speed/wps": 12909.61896958346, "speed/FLOPS": 202763305592757.6, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05218677222728729, "optim/lr": 0.002795842177498263, "optim/total_tokens": 7466909696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.7745823860168457, "created_at": "2025-01-16T21:47:14.698282+00:00"} {"global_step": 14243, "acc_step": 0, "speed/wps": 12906.60884176945, "speed/FLOPS": 202716027399089.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044700562953948975, "optim/lr": 0.0027957990208644858, "optim/total_tokens": 7467433984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.8393783569335938, "created_at": "2025-01-16T21:47:24.859100+00:00"} {"global_step": 14244, "acc_step": 0, "speed/wps": 12905.068149409817, "speed/FLOPS": 202691828708448.28, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05191987380385399, "optim/lr": 0.002795755860002937, "optim/total_tokens": 7467958272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.831587791442871, "created_at": "2025-01-16T21:47:35.023735+00:00"} {"global_step": 14245, "acc_step": 0, "speed/wps": 12904.464738736826, "speed/FLOPS": 202682351314655.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.039595067501068115, "optim/lr": 0.0027957126949137577, "optim/total_tokens": 7468482560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.838441848754883, "created_at": "2025-01-16T21:47:45.189852+00:00"} {"global_step": 14246, "acc_step": 0, "speed/wps": 12905.790017452948, "speed/FLOPS": 202703166638016.28, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055630456656217575, "optim/lr": 0.0027956695255970886, "optim/total_tokens": 7469006848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378681, "loss/out": 2.847771644592285, "created_at": "2025-01-16T21:47:55.351494+00:00"} {"global_step": 14247, "acc_step": 0, "speed/wps": 12907.961714220864, "speed/FLOPS": 202737276119980.5, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07811298221349716, "optim/lr": 0.00279562635205307, "optim/total_tokens": 7469531136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.89406156539917, "created_at": "2025-01-16T21:48:05.517833+00:00"} {"global_step": 14248, "acc_step": 0, "speed/wps": 12912.179578222396, "speed/FLOPS": 202803523470077.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060132384300231934, "optim/lr": 0.002795583174281844, "optim/total_tokens": 7470055424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 2.895186424255371, "created_at": "2025-01-16T21:48:15.672424+00:00"} {"global_step": 14249, "acc_step": 0, "speed/wps": 12908.076465795444, "speed/FLOPS": 202739078451144.47, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04618094861507416, "optim/lr": 0.00279553999228355, "optim/total_tokens": 7470579712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.771982431411743, "created_at": "2025-01-16T21:48:25.834057+00:00"} {"global_step": 14250, "acc_step": 0, "speed/wps": 12909.094527849571, "speed/FLOPS": 202755068514668.38, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050386782735586166, "optim/lr": 0.0027954968060583305, "optim/total_tokens": 7471104000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8525071144104004, "created_at": "2025-01-16T21:48:35.997440+00:00"} {"global_step": 14251, "acc_step": 0, "speed/wps": 12905.541519533257, "speed/FLOPS": 202699263636715.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04839557409286499, "optim/lr": 0.0027954536156063244, "optim/total_tokens": 7471628288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 2.8902087211608887, "created_at": "2025-01-16T21:48:46.157735+00:00"} {"global_step": 14252, "acc_step": 0, "speed/wps": 12905.806518677557, "speed/FLOPS": 202703425812423.8, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06713718920946121, "optim/lr": 0.0027954104209276746, "optim/total_tokens": 7472152576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8964526653289795, "created_at": "2025-01-16T21:48:56.319034+00:00"} {"global_step": 14253, "acc_step": 0, "speed/wps": 12911.11070908196, "speed/FLOPS": 202786735411446.4, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.064655601978302, "optim/lr": 0.00279536722202252, "optim/total_tokens": 7472676864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 2.8786239624023438, "created_at": "2025-01-16T21:49:06.474740+00:00"} {"global_step": 14254, "acc_step": 0, "speed/wps": 12908.540195265214, "speed/FLOPS": 202746361959698.25, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05135255306959152, "optim/lr": 0.002795324018891004, "optim/total_tokens": 7473201152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.7842369079589844, "created_at": "2025-01-16T21:49:16.635550+00:00"} {"global_step": 14255, "acc_step": 0, "speed/wps": 12909.099278388156, "speed/FLOPS": 202755143128404.78, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053892262279987335, "optim/lr": 0.0027952808115332653, "optim/total_tokens": 7473725440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 2.785094976425171, "created_at": "2025-01-16T21:49:26.795755+00:00"} {"global_step": 14256, "acc_step": 0, "speed/wps": 12908.217592382765, "speed/FLOPS": 202741295038126.6, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042100682854652405, "optim/lr": 0.002795237599949447, "optim/total_tokens": 7474249728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.8802695274353027, "created_at": "2025-01-16T21:49:36.953395+00:00"} {"global_step": 14257, "acc_step": 0, "speed/wps": 12907.51790334586, "speed/FLOPS": 202730305460328.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043394673615694046, "optim/lr": 0.0027951943841396875, "optim/total_tokens": 7474774016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430246, "loss/out": 2.9049160480499268, "created_at": "2025-01-16T21:49:47.113850+00:00"} {"global_step": 14258, "acc_step": 0, "speed/wps": 12908.473277047056, "speed/FLOPS": 202745310917126.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06348372250795364, "optim/lr": 0.0027951511641041296, "optim/total_tokens": 7475298304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 3.0413479804992676, "created_at": "2025-01-16T21:49:57.271677+00:00"} {"global_step": 14259, "acc_step": 0, "speed/wps": 12910.26517361649, "speed/FLOPS": 202773455115073.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0784364715218544, "optim/lr": 0.0027951079398429147, "optim/total_tokens": 7475822592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8255581855773926, "created_at": "2025-01-16T21:50:07.427761+00:00"} {"global_step": 14260, "acc_step": 0, "speed/wps": 12908.949571795181, "speed/FLOPS": 202752791780643.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045381810516119, "optim/lr": 0.0027950647113561817, "optim/total_tokens": 7476346880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8824210166931152, "created_at": "2025-01-16T21:50:17.590466+00:00"} {"global_step": 14261, "acc_step": 0, "speed/wps": 12911.815899166391, "speed/FLOPS": 202797811390755.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059497907757759094, "optim/lr": 0.002795021478644074, "optim/total_tokens": 7476871168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8469126224517822, "created_at": "2025-01-16T21:50:27.748164+00:00"} {"global_step": 14262, "acc_step": 0, "speed/wps": 12910.860781548778, "speed/FLOPS": 202782809956101.34, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07638441026210785, "optim/lr": 0.002794978241706731, "optim/total_tokens": 7477395456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.934680938720703, "created_at": "2025-01-16T21:50:37.907756+00:00"} {"global_step": 14263, "acc_step": 0, "speed/wps": 12908.940094622947, "speed/FLOPS": 202752642928630.78, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048548370599746704, "optim/lr": 0.0027949350005442948, "optim/total_tokens": 7477919744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.797163963317871, "created_at": "2025-01-16T21:50:48.066706+00:00"} {"global_step": 14264, "acc_step": 0, "speed/wps": 12907.717435882187, "speed/FLOPS": 202733439392997.97, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04442286491394043, "optim/lr": 0.002794891755156906, "optim/total_tokens": 7478444032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 2.696991443634033, "created_at": "2025-01-16T21:50:58.227565+00:00"} {"global_step": 14265, "acc_step": 0, "speed/wps": 12915.391322611016, "speed/FLOPS": 202853968329099.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05010364204645157, "optim/lr": 0.002794848505544706, "optim/total_tokens": 7478968320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8632352352142334, "created_at": "2025-01-16T21:51:08.382106+00:00"} {"global_step": 14266, "acc_step": 0, "speed/wps": 12911.661999020276, "speed/FLOPS": 202795394177479.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04244555905461311, "optim/lr": 0.002794805251707835, "optim/total_tokens": 7479492608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422788, "loss/out": 2.884767532348633, "created_at": "2025-01-16T21:51:18.537102+00:00"} {"global_step": 14267, "acc_step": 0, "speed/wps": 12903.986651128354, "speed/FLOPS": 202674842291803.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050457440316677094, "optim/lr": 0.0027947619936464357, "optim/total_tokens": 7480016896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 2.9039788246154785, "created_at": "2025-01-16T21:51:28.704195+00:00"} {"global_step": 14268, "acc_step": 0, "speed/wps": 12908.975301369686, "speed/FLOPS": 202753195898967.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04695696756243706, "optim/lr": 0.0027947187313606477, "optim/total_tokens": 7480541184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393108, "loss/out": 2.824807643890381, "created_at": "2025-01-16T21:51:38.869116+00:00"} {"global_step": 14269, "acc_step": 0, "speed/wps": 12912.575673871277, "speed/FLOPS": 202809744696535.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047321878373622894, "optim/lr": 0.0027946754648506132, "optim/total_tokens": 7481065472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.829927921295166, "created_at": "2025-01-16T21:51:49.025076+00:00"} {"global_step": 14270, "acc_step": 0, "speed/wps": 12913.035432941198, "speed/FLOPS": 202816965844503.5, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04316135495901108, "optim/lr": 0.002794632194116473, "optim/total_tokens": 7481589760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8843681812286377, "created_at": "2025-01-16T21:51:59.179090+00:00"} {"global_step": 14271, "acc_step": 0, "speed/wps": 12906.721733454837, "speed/FLOPS": 202717800518137.9, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.043321747332811356, "optim/lr": 0.0027945889191583672, "optim/total_tokens": 7482114048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.791092872619629, "created_at": "2025-01-16T21:52:09.337898+00:00"} {"global_step": 14272, "acc_step": 0, "speed/wps": 12913.90304220258, "speed/FLOPS": 202830592840173.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048521336168050766, "optim/lr": 0.002794545639976439, "optim/total_tokens": 7482638336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9066531658172607, "created_at": "2025-01-16T21:52:19.494753+00:00"} {"global_step": 14273, "acc_step": 0, "speed/wps": 12912.311288728444, "speed/FLOPS": 202805592164563.75, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052809618413448334, "optim/lr": 0.0027945023565708288, "optim/total_tokens": 7483162624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 2.7800660133361816, "created_at": "2025-01-16T21:52:29.650557+00:00"} {"global_step": 14274, "acc_step": 0, "speed/wps": 12908.302747653253, "speed/FLOPS": 202742632518665.22, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04700412601232529, "optim/lr": 0.002794459068941677, "optim/total_tokens": 7483686912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 2.854854106903076, "created_at": "2025-01-16T21:52:39.808756+00:00"} {"global_step": 14275, "acc_step": 0, "speed/wps": 12910.533235272087, "speed/FLOPS": 202777665391730.2, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04776446521282196, "optim/lr": 0.002794415777089126, "optim/total_tokens": 7484211200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.834108591079712, "created_at": "2025-01-16T21:52:49.969408+00:00"} {"global_step": 14276, "acc_step": 0, "speed/wps": 12907.346533863603, "speed/FLOPS": 202727613867123.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06108028069138527, "optim/lr": 0.0027943724810133162, "optim/total_tokens": 7484735488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.7981133460998535, "created_at": "2025-01-16T21:53:00.128841+00:00"} {"global_step": 14277, "acc_step": 0, "speed/wps": 12910.35987186171, "speed/FLOPS": 202774942481143.9, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06360499560832977, "optim/lr": 0.0027943291807143895, "optim/total_tokens": 7485259776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 2.8286585807800293, "created_at": "2025-01-16T21:53:10.284932+00:00"} {"global_step": 14278, "acc_step": 0, "speed/wps": 12908.60818687453, "speed/FLOPS": 202747429861352.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0653541088104248, "optim/lr": 0.0027942858761924867, "optim/total_tokens": 7485784064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8776116371154785, "created_at": "2025-01-16T21:53:20.443902+00:00"} {"global_step": 14279, "acc_step": 0, "speed/wps": 12908.13406847384, "speed/FLOPS": 202739983180363.06, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054396845400333405, "optim/lr": 0.0027942425674477495, "optim/total_tokens": 7486308352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.917285203933716, "created_at": "2025-01-16T21:53:30.601634+00:00"} {"global_step": 14280, "acc_step": 0, "speed/wps": 12910.657636611888, "speed/FLOPS": 202779619285717.1, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05570675805211067, "optim/lr": 0.0027941992544803186, "optim/total_tokens": 7486832640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.846874713897705, "created_at": "2025-01-16T21:53:40.759673+00:00"} {"global_step": 14281, "acc_step": 0, "speed/wps": 12906.017049982325, "speed/FLOPS": 202706732495865.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0629739984869957, "optim/lr": 0.0027941559372903356, "optim/total_tokens": 7487356928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373755, "loss/out": 2.8935067653656006, "created_at": "2025-01-16T21:53:50.919206+00:00"} {"global_step": 14282, "acc_step": 0, "speed/wps": 12912.177219588697, "speed/FLOPS": 202803486424493.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04957958310842514, "optim/lr": 0.002794112615877942, "optim/total_tokens": 7487881216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.796719551086426, "created_at": "2025-01-16T21:54:01.075014+00:00"} {"global_step": 14283, "acc_step": 0, "speed/wps": 12907.472951340302, "speed/FLOPS": 202729599427310.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0510028675198555, "optim/lr": 0.0027940692902432797, "optim/total_tokens": 7488405504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.8708102703094482, "created_at": "2025-01-16T21:54:11.234490+00:00"} {"global_step": 14284, "acc_step": 0, "speed/wps": 12905.868023183399, "speed/FLOPS": 202704391825204.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049332395195961, "optim/lr": 0.002794025960386489, "optim/total_tokens": 7488929792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.826096773147583, "created_at": "2025-01-16T21:54:21.394089+00:00"} {"global_step": 14285, "acc_step": 0, "speed/wps": 12909.39234791122, "speed/FLOPS": 202759746187984.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.046735428273677826, "optim/lr": 0.0027939826263077114, "optim/total_tokens": 7489454080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.9140472412109375, "created_at": "2025-01-16T21:54:31.550843+00:00"} {"global_step": 14286, "acc_step": 0, "speed/wps": 12909.694295596608, "speed/FLOPS": 202764488691302.84, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044234298169612885, "optim/lr": 0.0027939392880070887, "optim/total_tokens": 7489978368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.745896100997925, "created_at": "2025-01-16T21:54:41.709064+00:00"} {"global_step": 14287, "acc_step": 0, "speed/wps": 12910.097628595677, "speed/FLOPS": 202770823590290.84, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04934043064713478, "optim/lr": 0.0027938959454847623, "optim/total_tokens": 7490502656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.7891461849212646, "created_at": "2025-01-16T21:54:51.866835+00:00"} {"global_step": 14288, "acc_step": 0, "speed/wps": 12908.592747627587, "speed/FLOPS": 202747187366763.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054204169660806656, "optim/lr": 0.002793852598740873, "optim/total_tokens": 7491026944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.7132949829101562, "created_at": "2025-01-16T21:55:02.024327+00:00"} {"global_step": 14289, "acc_step": 0, "speed/wps": 12908.380873742137, "speed/FLOPS": 202743859596247.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05730649083852768, "optim/lr": 0.0027938092477755635, "optim/total_tokens": 7491551232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.8258254528045654, "created_at": "2025-01-16T21:55:12.182904+00:00"} {"global_step": 14290, "acc_step": 0, "speed/wps": 12908.883603671335, "speed/FLOPS": 202751755660609.0, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05035795643925667, "optim/lr": 0.0027937658925889742, "optim/total_tokens": 7492075520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.9117929935455322, "created_at": "2025-01-16T21:55:22.340083+00:00"} {"global_step": 14291, "acc_step": 0, "speed/wps": 12912.379902934592, "speed/FLOPS": 202806669844957.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06823766976594925, "optim/lr": 0.0027937225331812468, "optim/total_tokens": 7492599808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 2.7960801124572754, "created_at": "2025-01-16T21:55:32.494549+00:00"} {"global_step": 14292, "acc_step": 0, "speed/wps": 12916.4311973106, "speed/FLOPS": 202870300990194.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0609448179602623, "optim/lr": 0.0027936791695525225, "optim/total_tokens": 7493124096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347072, "loss/out": 2.8724708557128906, "created_at": "2025-01-16T21:55:42.646827+00:00"} {"global_step": 14293, "acc_step": 0, "speed/wps": 12905.992209181968, "speed/FLOPS": 202706342336961.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060974229127168655, "optim/lr": 0.0027936358017029432, "optim/total_tokens": 7493648384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.740062713623047, "created_at": "2025-01-16T21:55:52.809486+00:00"} {"global_step": 14294, "acc_step": 0, "speed/wps": 12902.883685026134, "speed/FLOPS": 202657518693533.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07449489086866379, "optim/lr": 0.0027935924296326503, "optim/total_tokens": 7494172672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 2.882986545562744, "created_at": "2025-01-16T21:56:02.972113+00:00"} {"global_step": 14295, "acc_step": 0, "speed/wps": 12910.467782727314, "speed/FLOPS": 202776637369573.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04865036904811859, "optim/lr": 0.0027935490533417856, "optim/total_tokens": 7494696960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.751784324645996, "created_at": "2025-01-16T21:56:13.131470+00:00"} {"global_step": 14296, "acc_step": 0, "speed/wps": 12901.661167905848, "speed/FLOPS": 202638317382245.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06831254810094833, "optim/lr": 0.00279350567283049, "optim/total_tokens": 7495221248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.7388532161712646, "created_at": "2025-01-16T21:56:23.297787+00:00"} {"global_step": 14297, "acc_step": 0, "speed/wps": 12907.812606993028, "speed/FLOPS": 202734934186072.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06555336713790894, "optim/lr": 0.002793462288098905, "optim/total_tokens": 7495745536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.888918876647949, "created_at": "2025-01-16T21:56:33.455869+00:00"} {"global_step": 14298, "acc_step": 0, "speed/wps": 12911.038399885381, "speed/FLOPS": 202785599696150.88, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05357249826192856, "optim/lr": 0.0027934188991471727, "optim/total_tokens": 7496269824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.775967836380005, "created_at": "2025-01-16T21:56:43.613518+00:00"} {"global_step": 14299, "acc_step": 0, "speed/wps": 12907.307298214026, "speed/FLOPS": 202726997617331.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07742484658956528, "optim/lr": 0.002793375505975434, "optim/total_tokens": 7496794112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348536, "loss/out": 2.698112964630127, "created_at": "2025-01-16T21:56:53.774916+00:00"} {"global_step": 14300, "acc_step": 0, "speed/wps": 12911.252592575996, "speed/FLOPS": 202788963886688.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043076034635305405, "optim/lr": 0.0027933321085838315, "optim/total_tokens": 7497318400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.813737154006958, "created_at": "2025-01-16T21:57:03.930150+00:00"} {"global_step": 14301, "acc_step": 0, "speed/wps": 12906.08119221389, "speed/FLOPS": 202707739937752.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06313518434762955, "optim/lr": 0.002793288706972506, "optim/total_tokens": 7497842688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.8983676433563232, "created_at": "2025-01-16T21:57:14.095072+00:00"} {"global_step": 14302, "acc_step": 0, "speed/wps": 12901.943349473144, "speed/FLOPS": 202642749431517.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054732974618673325, "optim/lr": 0.0027932453011415997, "optim/total_tokens": 7498366976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7534990310668945, "created_at": "2025-01-16T21:57:24.260449+00:00"} {"global_step": 14303, "acc_step": 0, "speed/wps": 12907.561482455516, "speed/FLOPS": 202730989930129.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.061166878789663315, "optim/lr": 0.0027932018910912532, "optim/total_tokens": 7498891264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.899008274078369, "created_at": "2025-01-16T21:57:34.418652+00:00"} {"global_step": 14304, "acc_step": 0, "speed/wps": 12898.703710385102, "speed/FLOPS": 202591866447909.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051224976778030396, "optim/lr": 0.0027931584768216096, "optim/total_tokens": 7499415552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.7931954860687256, "created_at": "2025-01-16T21:57:44.585977+00:00"} {"global_step": 14305, "acc_step": 0, "speed/wps": 12909.044862906776, "speed/FLOPS": 202754288458495.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05409012362360954, "optim/lr": 0.0027931150583328087, "optim/total_tokens": 7499939840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.861694574356079, "created_at": "2025-01-16T21:57:54.744000+00:00"} {"global_step": 14306, "acc_step": 0, "speed/wps": 12915.094830432748, "speed/FLOPS": 202849311512017.0, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05598164722323418, "optim/lr": 0.002793071635624994, "optim/total_tokens": 7500464128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 2.8184797763824463, "created_at": "2025-01-16T21:58:04.897190+00:00"} {"global_step": 14307, "acc_step": 0, "speed/wps": 12906.81836591256, "speed/FLOPS": 202719318263673.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055530864745378494, "optim/lr": 0.0027930282086983055, "optim/total_tokens": 7500988416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.792224407196045, "created_at": "2025-01-16T21:58:15.058766+00:00"} {"global_step": 14308, "acc_step": 0, "speed/wps": 12910.416810381867, "speed/FLOPS": 202775836778844.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05389581248164177, "optim/lr": 0.0027929847775528863, "optim/total_tokens": 7501512704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.811100959777832, "created_at": "2025-01-16T21:58:25.214662+00:00"} {"global_step": 14309, "acc_step": 0, "speed/wps": 12910.489265395661, "speed/FLOPS": 202776974784399.97, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04683580994606018, "optim/lr": 0.0027929413421888772, "optim/total_tokens": 7502036992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.736536741256714, "created_at": "2025-01-16T21:58:35.374365+00:00"} {"global_step": 14310, "acc_step": 0, "speed/wps": 12913.061296101994, "speed/FLOPS": 202817372060983.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04659377783536911, "optim/lr": 0.002792897902606421, "optim/total_tokens": 7502561280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 2.8583292961120605, "created_at": "2025-01-16T21:58:45.530544+00:00"} {"global_step": 14311, "acc_step": 0, "speed/wps": 12910.943780517442, "speed/FLOPS": 202784113568956.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044342052191495895, "optim/lr": 0.002792854458805658, "optim/total_tokens": 7503085568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8232364654541016, "created_at": "2025-01-16T21:58:55.688595+00:00"} {"global_step": 14312, "acc_step": 0, "speed/wps": 12910.725076104121, "speed/FLOPS": 202780678515614.22, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05439572408795357, "optim/lr": 0.0027928110107867305, "optim/total_tokens": 7503609856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.7682065963745117, "created_at": "2025-01-16T21:59:05.846497+00:00"} {"global_step": 14313, "acc_step": 0, "speed/wps": 12909.229664986453, "speed/FLOPS": 202757191029101.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06645477563142776, "optim/lr": 0.002792767558549781, "optim/total_tokens": 7504134144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8575923442840576, "created_at": "2025-01-16T21:59:16.005976+00:00"} {"global_step": 14314, "acc_step": 0, "speed/wps": 12910.021138492933, "speed/FLOPS": 202769622208118.84, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050404030829668045, "optim/lr": 0.0027927241020949505, "optim/total_tokens": 7504658432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405621, "loss/out": 2.855203151702881, "created_at": "2025-01-16T21:59:26.165007+00:00"} {"global_step": 14315, "acc_step": 0, "speed/wps": 12910.558294895793, "speed/FLOPS": 202778058987556.0, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05222423002123833, "optim/lr": 0.00279268064142238, "optim/total_tokens": 7505182720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.8837664127349854, "created_at": "2025-01-16T21:59:36.324260+00:00"} {"global_step": 14316, "acc_step": 0, "speed/wps": 12908.373702554625, "speed/FLOPS": 202743746962893.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.048386719077825546, "optim/lr": 0.0027926371765322135, "optim/total_tokens": 7505707008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.765688419342041, "created_at": "2025-01-16T21:59:46.481809+00:00"} {"global_step": 14317, "acc_step": 0, "speed/wps": 12915.582682022436, "speed/FLOPS": 202856973891611.34, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04725661501288414, "optim/lr": 0.002792593707424591, "optim/total_tokens": 7506231296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8844995498657227, "created_at": "2025-01-16T21:59:56.637002+00:00"} {"global_step": 14318, "acc_step": 0, "speed/wps": 12910.480168503116, "speed/FLOPS": 202776831905202.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04946018382906914, "optim/lr": 0.0027925502340996546, "optim/total_tokens": 7506755584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.912363052368164, "created_at": "2025-01-16T22:00:06.793325+00:00"} {"global_step": 14319, "acc_step": 0, "speed/wps": 12907.544698550391, "speed/FLOPS": 202730726315836.16, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04290775582194328, "optim/lr": 0.0027925067565575472, "optim/total_tokens": 7507279872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 2.8468847274780273, "created_at": "2025-01-16T22:00:16.955028+00:00"} {"global_step": 14320, "acc_step": 0, "speed/wps": 12911.01636936516, "speed/FLOPS": 202785253676557.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06054438650608063, "optim/lr": 0.0027924632747984087, "optim/total_tokens": 7507804160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8106579780578613, "created_at": "2025-01-16T22:00:27.113020+00:00"} {"global_step": 14321, "acc_step": 0, "speed/wps": 12913.269857980891, "speed/FLOPS": 202820647811885.62, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0632089376449585, "optim/lr": 0.002792419788822383, "optim/total_tokens": 7508328448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 2.747255563735962, "created_at": "2025-01-16T22:00:37.266753+00:00"} {"global_step": 14322, "acc_step": 0, "speed/wps": 12911.815707264232, "speed/FLOPS": 202797808376668.0, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04812351241707802, "optim/lr": 0.002792376298629611, "optim/total_tokens": 7508852736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7713804244995117, "created_at": "2025-01-16T22:00:47.422496+00:00"} {"global_step": 14323, "acc_step": 0, "speed/wps": 12909.792297701517, "speed/FLOPS": 202766027949029.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05472550541162491, "optim/lr": 0.002792332804220235, "optim/total_tokens": 7509377024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 2.759464979171753, "created_at": "2025-01-16T22:00:57.581455+00:00"} {"global_step": 14324, "acc_step": 0, "speed/wps": 12916.36243809628, "speed/FLOPS": 202869221032248.2, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05813268572092056, "optim/lr": 0.0027922893055943956, "optim/total_tokens": 7509901312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.7783851623535156, "created_at": "2025-01-16T22:01:07.737818+00:00"} {"global_step": 14325, "acc_step": 0, "speed/wps": 12909.660765287419, "speed/FLOPS": 202763962051721.56, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04566795378923416, "optim/lr": 0.0027922458027522367, "optim/total_tokens": 7510425600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.891101837158203, "created_at": "2025-01-16T22:01:17.895065+00:00"} {"global_step": 14326, "acc_step": 0, "speed/wps": 12907.289597562942, "speed/FLOPS": 202726719604282.75, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06584810465574265, "optim/lr": 0.0027922022956938988, "optim/total_tokens": 7510949888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.7920501232147217, "created_at": "2025-01-16T22:01:28.058065+00:00"} {"global_step": 14327, "acc_step": 0, "speed/wps": 12912.945947607026, "speed/FLOPS": 202815560354364.56, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06613990664482117, "optim/lr": 0.0027921587844195244, "optim/total_tokens": 7511474176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437948, "loss/out": 2.819446086883545, "created_at": "2025-01-16T22:01:38.213067+00:00"} {"global_step": 14328, "acc_step": 0, "speed/wps": 12913.241637331746, "speed/FLOPS": 202820204567811.56, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04809930920600891, "optim/lr": 0.002792115268929255, "optim/total_tokens": 7511998464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.790658950805664, "created_at": "2025-01-16T22:01:48.366815+00:00"} {"global_step": 14329, "acc_step": 0, "speed/wps": 12905.445745233379, "speed/FLOPS": 202697759369728.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051143813878297806, "optim/lr": 0.0027920717492232335, "optim/total_tokens": 7512522752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8963561058044434, "created_at": "2025-01-16T22:01:58.526847+00:00"} {"global_step": 14330, "acc_step": 0, "speed/wps": 12910.925533349735, "speed/FLOPS": 202783826972113.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.056406453251838684, "optim/lr": 0.0027920282253016014, "optim/total_tokens": 7513047040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 2.9100217819213867, "created_at": "2025-01-16T22:02:08.687633+00:00"} {"global_step": 14331, "acc_step": 0, "speed/wps": 12911.87733658065, "speed/FLOPS": 202798776349769.9, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05539730191230774, "optim/lr": 0.0027919846971645, "optim/total_tokens": 7513571328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 2.7005832195281982, "created_at": "2025-01-16T22:02:18.844122+00:00"} {"global_step": 14332, "acc_step": 0, "speed/wps": 12910.283556296465, "speed/FLOPS": 202773743840323.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05397647246718407, "optim/lr": 0.002791941164812073, "optim/total_tokens": 7514095616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371642, "loss/out": 2.8199527263641357, "created_at": "2025-01-16T22:02:29.006836+00:00"} {"global_step": 14333, "acc_step": 0, "speed/wps": 12904.80152515741, "speed/FLOPS": 202687641008184.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0572175569832325, "optim/lr": 0.0027918976282444606, "optim/total_tokens": 7514619904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.871181011199951, "created_at": "2025-01-16T22:02:39.167725+00:00"} {"global_step": 14334, "acc_step": 0, "speed/wps": 12907.051902397443, "speed/FLOPS": 202722986275081.22, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05853525176644325, "optim/lr": 0.002791854087461806, "optim/total_tokens": 7515144192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 3.0860214233398438, "created_at": "2025-01-16T22:02:49.326554+00:00"} {"global_step": 14335, "acc_step": 0, "speed/wps": 12908.435526337838, "speed/FLOPS": 202744717990363.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060567259788513184, "optim/lr": 0.002791810542464251, "optim/total_tokens": 7515668480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8272314071655273, "created_at": "2025-01-16T22:02:59.484148+00:00"} {"global_step": 14336, "acc_step": 0, "speed/wps": 12909.689200020475, "speed/FLOPS": 202764408658277.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05709419772028923, "optim/lr": 0.002791766993251937, "optim/total_tokens": 7516192768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.811821699142456, "created_at": "2025-01-16T22:03:09.640863+00:00"} {"global_step": 14337, "acc_step": 0, "speed/wps": 12908.152715016722, "speed/FLOPS": 202740276049942.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049659039825201035, "optim/lr": 0.0027917234398250074, "optim/total_tokens": 7516717056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 2.903377056121826, "created_at": "2025-01-16T22:03:19.803845+00:00"} {"global_step": 14338, "acc_step": 0, "speed/wps": 12915.661063383895, "speed/FLOPS": 202858204978600.38, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04712289944291115, "optim/lr": 0.002791679882183603, "optim/total_tokens": 7517241344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 2.8694040775299072, "created_at": "2025-01-16T22:03:29.956354+00:00"} {"global_step": 14339, "acc_step": 0, "speed/wps": 12911.721127439143, "speed/FLOPS": 202796322870548.7, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04349333792924881, "optim/lr": 0.0027916363203278668, "optim/total_tokens": 7517765632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 3.011202573776245, "created_at": "2025-01-16T22:03:40.113744+00:00"} {"global_step": 14340, "acc_step": 0, "speed/wps": 12908.761284872313, "speed/FLOPS": 202749834475786.72, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042636554688215256, "optim/lr": 0.0027915927542579403, "optim/total_tokens": 7518289920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.7927236557006836, "created_at": "2025-01-16T22:03:50.274841+00:00"} {"global_step": 14341, "acc_step": 0, "speed/wps": 12913.118562951668, "speed/FLOPS": 202818271515550.5, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06305599212646484, "optim/lr": 0.002791549183973966, "optim/total_tokens": 7518814208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.8883328437805176, "created_at": "2025-01-16T22:04:00.440251+00:00"} {"global_step": 14342, "acc_step": 0, "speed/wps": 12914.127910858288, "speed/FLOPS": 202834124711413.34, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05354063957929611, "optim/lr": 0.0027915056094760865, "optim/total_tokens": 7519338496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.9571373462677, "created_at": "2025-01-16T22:04:10.598010+00:00"} {"global_step": 14343, "acc_step": 0, "speed/wps": 12911.453324427153, "speed/FLOPS": 202792116656245.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05538324639201164, "optim/lr": 0.002791462030764443, "optim/total_tokens": 7519862784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 2.8026857376098633, "created_at": "2025-01-16T22:04:20.756263+00:00"} {"global_step": 14344, "acc_step": 0, "speed/wps": 12909.340132434956, "speed/FLOPS": 202758926072177.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07630840688943863, "optim/lr": 0.0027914184478391784, "optim/total_tokens": 7520387072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.896510124206543, "created_at": "2025-01-16T22:04:30.914463+00:00"} {"global_step": 14345, "acc_step": 0, "speed/wps": 12910.766670518156, "speed/FLOPS": 202781331813043.34, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06813061982393265, "optim/lr": 0.0027913748607004346, "optim/total_tokens": 7520911360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.8535959720611572, "created_at": "2025-01-16T22:04:41.071173+00:00"} {"global_step": 14346, "acc_step": 0, "speed/wps": 12908.027520201456, "speed/FLOPS": 202738309693332.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05018732324242592, "optim/lr": 0.002791331269348354, "optim/total_tokens": 7521435648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.6972060203552246, "created_at": "2025-01-16T22:04:51.233864+00:00"} {"global_step": 14347, "acc_step": 0, "speed/wps": 12908.534592672991, "speed/FLOPS": 202746273963288.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0651865005493164, "optim/lr": 0.0027912876737830787, "optim/total_tokens": 7521959936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380183, "loss/out": 2.8440725803375244, "created_at": "2025-01-16T22:05:01.392154+00:00"} {"global_step": 14348, "acc_step": 0, "speed/wps": 12906.399042735053, "speed/FLOPS": 202712732216960.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0725143626332283, "optim/lr": 0.0027912440740047507, "optim/total_tokens": 7522484224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.9240078926086426, "created_at": "2025-01-16T22:05:11.551400+00:00"} {"global_step": 14349, "acc_step": 0, "speed/wps": 12908.195290899615, "speed/FLOPS": 202740944762688.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05190368741750717, "optim/lr": 0.0027912004700135127, "optim/total_tokens": 7523008512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.9037277698516846, "created_at": "2025-01-16T22:05:21.709249+00:00"} {"global_step": 14350, "acc_step": 0, "speed/wps": 12908.86108439167, "speed/FLOPS": 202751401964377.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05753179267048836, "optim/lr": 0.0027911568618095065, "optim/total_tokens": 7523532800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.817749500274658, "created_at": "2025-01-16T22:05:31.866877+00:00"} {"global_step": 14351, "acc_step": 0, "speed/wps": 12916.062052463345, "speed/FLOPS": 202864503063109.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05856037884950638, "optim/lr": 0.002791113249392875, "optim/total_tokens": 7524057088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505080, "loss/out": 2.7277443408966064, "created_at": "2025-01-16T22:05:42.019644+00:00"} {"global_step": 14352, "acc_step": 0, "speed/wps": 12911.290921209882, "speed/FLOPS": 202789565890552.84, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06302421540021896, "optim/lr": 0.00279106963276376, "optim/total_tokens": 7524581376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 2.8581151962280273, "created_at": "2025-01-16T22:05:52.179704+00:00"} {"global_step": 14353, "acc_step": 0, "speed/wps": 12909.829356474767, "speed/FLOPS": 202766610007987.44, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055653735995292664, "optim/lr": 0.0027910260119223037, "optim/total_tokens": 7525105664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.782118797302246, "created_at": "2025-01-16T22:06:02.336805+00:00"} {"global_step": 14354, "acc_step": 0, "speed/wps": 12909.081610712357, "speed/FLOPS": 202754865633278.44, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06484029442071915, "optim/lr": 0.0027909823868686487, "optim/total_tokens": 7525629952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.918980121612549, "created_at": "2025-01-16T22:06:12.493999+00:00"} {"global_step": 14355, "acc_step": 0, "speed/wps": 12903.425210494614, "speed/FLOPS": 202666024095150.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04711638018488884, "optim/lr": 0.0027909387576029376, "optim/total_tokens": 7526154240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.6158323287963867, "created_at": "2025-01-16T22:06:22.656792+00:00"} {"global_step": 14356, "acc_step": 0, "speed/wps": 12910.905288126594, "speed/FLOPS": 202783508993064.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053454674780368805, "optim/lr": 0.0027908951241253123, "optim/total_tokens": 7526678528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372786, "loss/out": 2.7889151573181152, "created_at": "2025-01-16T22:06:32.812400+00:00"} {"global_step": 14357, "acc_step": 0, "speed/wps": 12908.004863026563, "speed/FLOPS": 202737953831266.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05536150559782982, "optim/lr": 0.0027908514864359154, "optim/total_tokens": 7527202816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.804548978805542, "created_at": "2025-01-16T22:06:42.974499+00:00"} {"global_step": 14358, "acc_step": 0, "speed/wps": 12910.46872901017, "speed/FLOPS": 202776652232245.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049669694155454636, "optim/lr": 0.002790807844534889, "optim/total_tokens": 7527727104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8185577392578125, "created_at": "2025-01-16T22:06:53.132507+00:00"} {"global_step": 14359, "acc_step": 0, "speed/wps": 12912.55320449828, "speed/FLOPS": 202809391784156.12, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08716127276420593, "optim/lr": 0.0027907641984223756, "optim/total_tokens": 7528251392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.887526512145996, "created_at": "2025-01-16T22:07:03.288805+00:00"} {"global_step": 14360, "acc_step": 0, "speed/wps": 12910.58801979322, "speed/FLOPS": 202778525857916.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05255948007106781, "optim/lr": 0.002790720548098518, "optim/total_tokens": 7528775680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 2.8701353073120117, "created_at": "2025-01-16T22:07:13.451580+00:00"} {"global_step": 14361, "acc_step": 0, "speed/wps": 12904.680201924923, "speed/FLOPS": 202685735460102.3, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07150204479694366, "optim/lr": 0.0027906768935634578, "optim/total_tokens": 7529299968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 2.8565011024475098, "created_at": "2025-01-16T22:07:23.612948+00:00"} {"global_step": 14362, "acc_step": 0, "speed/wps": 12910.920541030286, "speed/FLOPS": 202783748560876.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05475575476884842, "optim/lr": 0.002790633234817338, "optim/total_tokens": 7529824256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8589420318603516, "created_at": "2025-01-16T22:07:33.768527+00:00"} {"global_step": 14363, "acc_step": 0, "speed/wps": 12907.598566304165, "speed/FLOPS": 202731572382931.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07439803332090378, "optim/lr": 0.0027905895718603013, "optim/total_tokens": 7530348544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351757, "loss/out": 2.8206639289855957, "created_at": "2025-01-16T22:07:43.927239+00:00"} {"global_step": 14364, "acc_step": 0, "speed/wps": 12902.54949103049, "speed/FLOPS": 202652269717601.66, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05262121558189392, "optim/lr": 0.00279054590469249, "optim/total_tokens": 7530872832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505601, "loss/out": 2.851977825164795, "created_at": "2025-01-16T22:07:54.092441+00:00"} {"global_step": 14365, "acc_step": 0, "speed/wps": 12908.684799947852, "speed/FLOPS": 202748633174946.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06188279762864113, "optim/lr": 0.002790502233314046, "optim/total_tokens": 7531397120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.8802661895751953, "created_at": "2025-01-16T22:08:04.252083+00:00"} {"global_step": 14366, "acc_step": 0, "speed/wps": 12909.48790015304, "speed/FLOPS": 202761246967244.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05867592990398407, "optim/lr": 0.0027904585577251118, "optim/total_tokens": 7531921408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.856325149536133, "created_at": "2025-01-16T22:08:14.412783+00:00"} {"global_step": 14367, "acc_step": 0, "speed/wps": 12907.866847530137, "speed/FLOPS": 202735786108242.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05559031665325165, "optim/lr": 0.0027904148779258307, "optim/total_tokens": 7532445696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 2.9092962741851807, "created_at": "2025-01-16T22:08:24.573619+00:00"} {"global_step": 14368, "acc_step": 0, "speed/wps": 12906.819373211078, "speed/FLOPS": 202719334084680.75, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047501035034656525, "optim/lr": 0.002790371193916345, "optim/total_tokens": 7532969984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8911280632019043, "created_at": "2025-01-16T22:08:34.732697+00:00"} {"global_step": 14369, "acc_step": 0, "speed/wps": 12903.715972134922, "speed/FLOPS": 202670590906259.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046335846185684204, "optim/lr": 0.002790327505696797, "optim/total_tokens": 7533494272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.78963041305542, "created_at": "2025-01-16T22:08:44.894593+00:00"} {"global_step": 14370, "acc_step": 0, "speed/wps": 12910.467567178548, "speed/FLOPS": 202776633984083.72, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06284850090742111, "optim/lr": 0.002790283813267329, "optim/total_tokens": 7534018560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379498, "loss/out": 2.922276020050049, "created_at": "2025-01-16T22:08:55.051452+00:00"} {"global_step": 14371, "acc_step": 0, "speed/wps": 12915.559093957301, "speed/FLOPS": 202856603408635.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043339576572179794, "optim/lr": 0.002790240116628084, "optim/total_tokens": 7534542848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.7277345657348633, "created_at": "2025-01-16T22:09:05.206818+00:00"} {"global_step": 14372, "acc_step": 0, "speed/wps": 12903.113208414974, "speed/FLOPS": 202661123673753.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05824645608663559, "optim/lr": 0.002790196415779204, "optim/total_tokens": 7535067136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.883580207824707, "created_at": "2025-01-16T22:09:15.369778+00:00"} {"global_step": 14373, "acc_step": 0, "speed/wps": 12907.921482750726, "speed/FLOPS": 202736644229458.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04917467013001442, "optim/lr": 0.002790152710720832, "optim/total_tokens": 7535591424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.840860605239868, "created_at": "2025-01-16T22:09:25.530026+00:00"} {"global_step": 14374, "acc_step": 0, "speed/wps": 12912.089529018758, "speed/FLOPS": 202802109123593.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04617321491241455, "optim/lr": 0.002790109001453111, "optim/total_tokens": 7536115712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 2.8069887161254883, "created_at": "2025-01-16T22:09:35.684998+00:00"} {"global_step": 14375, "acc_step": 0, "speed/wps": 12902.37218306608, "speed/FLOPS": 202649484852373.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05142471566796303, "optim/lr": 0.0027900652879761833, "optim/total_tokens": 7536640000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.7900476455688477, "created_at": "2025-01-16T22:09:45.850127+00:00"} {"global_step": 14376, "acc_step": 0, "speed/wps": 12905.089166806925, "speed/FLOPS": 202692158815550.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0479508601129055, "optim/lr": 0.0027900215702901907, "optim/total_tokens": 7537164288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8830928802490234, "created_at": "2025-01-16T22:09:56.017819+00:00"} {"global_step": 14377, "acc_step": 0, "speed/wps": 12910.533258959005, "speed/FLOPS": 202777665763765.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043701525777578354, "optim/lr": 0.0027899778483952767, "optim/total_tokens": 7537688576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290025, "loss/out": 2.802570104598999, "created_at": "2025-01-16T22:10:06.173579+00:00"} {"global_step": 14378, "acc_step": 0, "speed/wps": 12909.893959375555, "speed/FLOPS": 202767624685318.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04475823789834976, "optim/lr": 0.002789934122291584, "optim/total_tokens": 7538212864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.808041572570801, "created_at": "2025-01-16T22:10:16.329822+00:00"} {"global_step": 14379, "acc_step": 0, "speed/wps": 12907.48181315719, "speed/FLOPS": 202729738614322.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04542474448680878, "optim/lr": 0.0027898903919792548, "optim/total_tokens": 7538737152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420077, "loss/out": 2.8815526962280273, "created_at": "2025-01-16T22:10:26.491534+00:00"} {"global_step": 14380, "acc_step": 0, "speed/wps": 12906.96913219704, "speed/FLOPS": 202721686255345.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047759830951690674, "optim/lr": 0.0027898466574584322, "optim/total_tokens": 7539261440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.7906124591827393, "created_at": "2025-01-16T22:10:36.654050+00:00"} {"global_step": 14381, "acc_step": 0, "speed/wps": 12911.487872075986, "speed/FLOPS": 202792659274543.84, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05947732925415039, "optim/lr": 0.002789802918729259, "optim/total_tokens": 7539785728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442730, "loss/out": 2.784173011779785, "created_at": "2025-01-16T22:10:46.809369+00:00"} {"global_step": 14382, "acc_step": 0, "speed/wps": 12905.604388388452, "speed/FLOPS": 202700251078477.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050150271505117416, "optim/lr": 0.0027897591757918773, "optim/total_tokens": 7540310016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.0647220611572266, "created_at": "2025-01-16T22:10:56.969235+00:00"} {"global_step": 14383, "acc_step": 0, "speed/wps": 12905.067247702176, "speed/FLOPS": 202691814545890.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04963967204093933, "optim/lr": 0.00278971542864643, "optim/total_tokens": 7540834304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.700514793395996, "created_at": "2025-01-16T22:11:07.131019+00:00"} {"global_step": 14384, "acc_step": 0, "speed/wps": 12905.44331568758, "speed/FLOPS": 202697721210373.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06215495988726616, "optim/lr": 0.00278967167729306, "optim/total_tokens": 7541358592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 2.965625286102295, "created_at": "2025-01-16T22:11:17.291168+00:00"} {"global_step": 14385, "acc_step": 0, "speed/wps": 12909.502895114185, "speed/FLOPS": 202761482483714.56, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06902945041656494, "optim/lr": 0.0027896279217319102, "optim/total_tokens": 7541882880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.7844080924987793, "created_at": "2025-01-16T22:11:27.456025+00:00"} {"global_step": 14386, "acc_step": 0, "speed/wps": 12912.585706093647, "speed/FLOPS": 202809902266373.25, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05637355148792267, "optim/lr": 0.0027895841619631234, "optim/total_tokens": 7542407168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 2.757986545562744, "created_at": "2025-01-16T22:11:37.610159+00:00"} {"global_step": 14387, "acc_step": 0, "speed/wps": 12904.747594367267, "speed/FLOPS": 202686793951016.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06640703976154327, "optim/lr": 0.0027895403979868415, "optim/total_tokens": 7542931456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.8540306091308594, "created_at": "2025-01-16T22:11:47.771034+00:00"} {"global_step": 14388, "acc_step": 0, "speed/wps": 12907.212951332787, "speed/FLOPS": 202725515769915.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04766429215669632, "optim/lr": 0.002789496629803208, "optim/total_tokens": 7543455744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 2.7043983936309814, "created_at": "2025-01-16T22:11:57.929767+00:00"} {"global_step": 14389, "acc_step": 0, "speed/wps": 12905.494506754156, "speed/FLOPS": 202698525236417.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07424525916576385, "optim/lr": 0.002789452857412367, "optim/total_tokens": 7543980032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.820411205291748, "created_at": "2025-01-16T22:12:08.090880+00:00"} {"global_step": 14390, "acc_step": 0, "speed/wps": 12904.125272265146, "speed/FLOPS": 202677019527242.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048515792936086655, "optim/lr": 0.0027894090808144585, "optim/total_tokens": 7544504320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.950941324234009, "created_at": "2025-01-16T22:12:18.258267+00:00"} {"global_step": 14391, "acc_step": 0, "speed/wps": 12897.06442784764, "speed/FLOPS": 202566119263048.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06057709828019142, "optim/lr": 0.0027893653000096268, "optim/total_tokens": 7545028608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415004, "loss/out": 2.8246986865997314, "created_at": "2025-01-16T22:12:28.434196+00:00"} {"global_step": 14392, "acc_step": 0, "speed/wps": 12909.2096050651, "speed/FLOPS": 202756875960470.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05045225843787193, "optim/lr": 0.0027893215149980156, "optim/total_tokens": 7545552896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.9105653762817383, "created_at": "2025-01-16T22:12:38.592366+00:00"} {"global_step": 14393, "acc_step": 0, "speed/wps": 12903.925201230131, "speed/FLOPS": 202673877136710.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0556272454559803, "optim/lr": 0.002789277725779766, "optim/total_tokens": 7546077184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480466, "loss/out": 2.953989028930664, "created_at": "2025-01-16T22:12:48.759225+00:00"} {"global_step": 14394, "acc_step": 0, "speed/wps": 12906.286995843486, "speed/FLOPS": 202710972366559.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05224098265171051, "optim/lr": 0.002789233932355022, "optim/total_tokens": 7546601472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8046751022338867, "created_at": "2025-01-16T22:12:58.919060+00:00"} {"global_step": 14395, "acc_step": 0, "speed/wps": 12904.571497981233, "speed/FLOPS": 202684028115291.78, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04723680391907692, "optim/lr": 0.0027891901347239264, "optim/total_tokens": 7547125760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.8259973526000977, "created_at": "2025-01-16T22:13:09.085030+00:00"} {"global_step": 14396, "acc_step": 0, "speed/wps": 12911.541110594606, "speed/FLOPS": 202793495458636.16, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05232563987374306, "optim/lr": 0.0027891463328866225, "optim/total_tokens": 7547650048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 2.7885286808013916, "created_at": "2025-01-16T22:13:19.246424+00:00"} {"global_step": 14397, "acc_step": 0, "speed/wps": 12904.22238881652, "speed/FLOPS": 202678544876133.72, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05355126038193703, "optim/lr": 0.0027891025268432512, "optim/total_tokens": 7548174336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.8181562423706055, "created_at": "2025-01-16T22:13:29.410733+00:00"} {"global_step": 14398, "acc_step": 0, "speed/wps": 12914.182205669189, "speed/FLOPS": 202834977486027.44, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06315048784017563, "optim/lr": 0.0027890587165939583, "optim/total_tokens": 7548698624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.9465630054473877, "created_at": "2025-01-16T22:13:39.575039+00:00"} {"global_step": 14399, "acc_step": 0, "speed/wps": 12900.54898425633, "speed/FLOPS": 202620849009729.06, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05552298575639725, "optim/lr": 0.0027890149021388847, "optim/total_tokens": 7549222912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292143, "loss/out": 2.959829092025757, "created_at": "2025-01-16T22:13:49.739037+00:00"} {"global_step": 14400, "acc_step": 0, "speed/wps": 12909.287919749326, "speed/FLOPS": 202758106000202.34, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.06531041115522385, "optim/lr": 0.0027889710834781735, "optim/total_tokens": 7549747200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.9218573570251465, "created_at": "2025-01-16T22:13:59.896510+00:00"} {"global_step": 14401, "acc_step": 0, "speed/wps": 12909.369433724567, "speed/FLOPS": 202759386289195.47, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0641895979642868, "optim/lr": 0.002788927260611969, "optim/total_tokens": 7550271488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 2.7343103885650635, "created_at": "2025-01-16T22:14:10.054268+00:00"} {"global_step": 14402, "acc_step": 0, "speed/wps": 12908.356088958493, "speed/FLOPS": 202743470317162.66, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062133826315402985, "optim/lr": 0.002788883433540413, "optim/total_tokens": 7550795776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.6820807456970215, "created_at": "2025-01-16T22:14:20.212384+00:00"} {"global_step": 14403, "acc_step": 0, "speed/wps": 12907.678926028315, "speed/FLOPS": 202732834542824.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04779772087931633, "optim/lr": 0.0027888396022636487, "optim/total_tokens": 7551320064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 2.8809146881103516, "created_at": "2025-01-16T22:14:30.372719+00:00"} {"global_step": 14404, "acc_step": 0, "speed/wps": 12907.316329080526, "speed/FLOPS": 202727139459499.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05707983672618866, "optim/lr": 0.002788795766781819, "optim/total_tokens": 7551844352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.9418835639953613, "created_at": "2025-01-16T22:14:40.531152+00:00"} {"global_step": 14405, "acc_step": 0, "speed/wps": 12906.759883684905, "speed/FLOPS": 202718399719925.03, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04610424488782883, "optim/lr": 0.0027887519270950677, "optim/total_tokens": 7552368640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.9114527702331543, "created_at": "2025-01-16T22:14:50.690124+00:00"} {"global_step": 14406, "acc_step": 0, "speed/wps": 12908.841814183455, "speed/FLOPS": 202751099299277.06, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05237681046128273, "optim/lr": 0.002788708083203537, "optim/total_tokens": 7552892928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.844728946685791, "created_at": "2025-01-16T22:15:00.847634+00:00"} {"global_step": 14407, "acc_step": 0, "speed/wps": 12912.971426536335, "speed/FLOPS": 202815960535961.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045711975544691086, "optim/lr": 0.0027886642351073704, "optim/total_tokens": 7553417216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466705, "loss/out": 2.7646021842956543, "created_at": "2025-01-16T22:15:11.004855+00:00"} {"global_step": 14408, "acc_step": 0, "speed/wps": 12908.380951882906, "speed/FLOPS": 202743860823555.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06184285134077072, "optim/lr": 0.0027886203828067103, "optim/total_tokens": 7553941504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.822242498397827, "created_at": "2025-01-16T22:15:21.164161+00:00"} {"global_step": 14409, "acc_step": 0, "speed/wps": 12909.597023246675, "speed/FLOPS": 202762960895380.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04928578808903694, "optim/lr": 0.0027885765263017, "optim/total_tokens": 7554465792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.878420829772949, "created_at": "2025-01-16T22:15:31.321573+00:00"} {"global_step": 14410, "acc_step": 0, "speed/wps": 12914.111111056973, "speed/FLOPS": 202833860847448.34, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05873849242925644, "optim/lr": 0.0027885326655924833, "optim/total_tokens": 7554990080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 2.8359670639038086, "created_at": "2025-01-16T22:15:41.474516+00:00"} {"global_step": 14411, "acc_step": 0, "speed/wps": 12911.435537778585, "speed/FLOPS": 202791837292488.3, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04267377778887749, "optim/lr": 0.002788488800679203, "optim/total_tokens": 7555514368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8468337059020996, "created_at": "2025-01-16T22:15:51.630272+00:00"} {"global_step": 14412, "acc_step": 0, "speed/wps": 12912.972509437486, "speed/FLOPS": 202815977544411.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04692818596959114, "optim/lr": 0.0027884449315620017, "optim/total_tokens": 7556038656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 2.77897572517395, "created_at": "2025-01-16T22:16:01.787502+00:00"} {"global_step": 14413, "acc_step": 0, "speed/wps": 12912.68486220504, "speed/FLOPS": 202811459649357.44, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04928043857216835, "optim/lr": 0.0027884010582410234, "optim/total_tokens": 7556562944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8130459785461426, "created_at": "2025-01-16T22:16:11.941712+00:00"} {"global_step": 14414, "acc_step": 0, "speed/wps": 12912.737714888699, "speed/FLOPS": 202812289773381.5, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04081080108880997, "optim/lr": 0.00278835718071641, "optim/total_tokens": 7557087232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 2.8050448894500732, "created_at": "2025-01-16T22:16:22.097101+00:00"} {"global_step": 14415, "acc_step": 0, "speed/wps": 12912.174953358335, "speed/FLOPS": 202803450830231.75, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04552263393998146, "optim/lr": 0.0027883132989883065, "optim/total_tokens": 7557611520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 2.8397045135498047, "created_at": "2025-01-16T22:16:32.253985+00:00"} {"global_step": 14416, "acc_step": 0, "speed/wps": 12905.25631559416, "speed/FLOPS": 202694784116940.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04723787307739258, "optim/lr": 0.0027882694130568542, "optim/total_tokens": 7558135808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396559, "loss/out": 2.847019910812378, "created_at": "2025-01-16T22:16:42.415988+00:00"} {"global_step": 14417, "acc_step": 0, "speed/wps": 12913.149982092424, "speed/FLOPS": 202818764996330.5, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051019228994846344, "optim/lr": 0.002788225522922197, "optim/total_tokens": 7558660096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8542256355285645, "created_at": "2025-01-16T22:16:52.571414+00:00"} {"global_step": 14418, "acc_step": 0, "speed/wps": 12916.054773168646, "speed/FLOPS": 202864388731783.8, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04606078565120697, "optim/lr": 0.0027881816285844786, "optim/total_tokens": 7559184384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.8902664184570312, "created_at": "2025-01-16T22:17:02.723030+00:00"} {"global_step": 14419, "acc_step": 0, "speed/wps": 12904.165048214714, "speed/FLOPS": 202677644263188.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049841370433568954, "optim/lr": 0.0027881377300438418, "optim/total_tokens": 7559708672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.7803962230682373, "created_at": "2025-01-16T22:17:12.885035+00:00"} {"global_step": 14420, "acc_step": 0, "speed/wps": 12900.66155473581, "speed/FLOPS": 202622617083797.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06720411777496338, "optim/lr": 0.002788093827300429, "optim/total_tokens": 7560232960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.807858943939209, "created_at": "2025-01-16T22:17:23.049467+00:00"} {"global_step": 14421, "acc_step": 0, "speed/wps": 12904.526588977365, "speed/FLOPS": 202683322757674.6, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05228817090392113, "optim/lr": 0.0027880499203543846, "optim/total_tokens": 7560757248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 2.869880199432373, "created_at": "2025-01-16T22:17:33.212177+00:00"} {"global_step": 14422, "acc_step": 0, "speed/wps": 12905.961030793093, "speed/FLOPS": 202705852637520.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049436960369348526, "optim/lr": 0.002788006009205852, "optim/total_tokens": 7561281536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8625056743621826, "created_at": "2025-01-16T22:17:43.371972+00:00"} {"global_step": 14423, "acc_step": 0, "speed/wps": 12910.778921846873, "speed/FLOPS": 202781524236995.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04256473854184151, "optim/lr": 0.0027879620938549734, "optim/total_tokens": 7561805824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.79441499710083, "created_at": "2025-01-16T22:17:53.527878+00:00"} {"global_step": 14424, "acc_step": 0, "speed/wps": 12908.490881491229, "speed/FLOPS": 202745587419113.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04537524655461311, "optim/lr": 0.0027879181743018927, "optim/total_tokens": 7562330112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.7674951553344727, "created_at": "2025-01-16T22:18:03.685483+00:00"} {"global_step": 14425, "acc_step": 0, "speed/wps": 12904.179201403398, "speed/FLOPS": 202677866558464.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.044022850692272186, "optim/lr": 0.002787874250546753, "optim/total_tokens": 7562854400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.844348430633545, "created_at": "2025-01-16T22:18:13.848271+00:00"} {"global_step": 14426, "acc_step": 0, "speed/wps": 12911.244134224913, "speed/FLOPS": 202788831036661.84, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05240711569786072, "optim/lr": 0.0027878303225896983, "optim/total_tokens": 7563378688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.9735798835754395, "created_at": "2025-01-16T22:18:24.006270+00:00"} {"global_step": 14427, "acc_step": 0, "speed/wps": 12910.806672187284, "speed/FLOPS": 202781960094225.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04973650351166725, "optim/lr": 0.0027877863904308712, "optim/total_tokens": 7563902976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 2.693021535873413, "created_at": "2025-01-16T22:18:34.164535+00:00"} {"global_step": 14428, "acc_step": 0, "speed/wps": 12911.232105878886, "speed/FLOPS": 202788642114958.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04791777953505516, "optim/lr": 0.0027877424540704147, "optim/total_tokens": 7564427264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.819899082183838, "created_at": "2025-01-16T22:18:44.325068+00:00"} {"global_step": 14429, "acc_step": 0, "speed/wps": 12909.773816947147, "speed/FLOPS": 202765737683386.84, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048195257782936096, "optim/lr": 0.002787698513508473, "optim/total_tokens": 7564951552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.841472864151001, "created_at": "2025-01-16T22:18:54.482901+00:00"} {"global_step": 14430, "acc_step": 0, "speed/wps": 12915.972704163352, "speed/FLOPS": 202863099725281.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046503692865371704, "optim/lr": 0.0027876545687451895, "optim/total_tokens": 7565475840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 2.7321062088012695, "created_at": "2025-01-16T22:19:04.637927+00:00"} {"global_step": 14431, "acc_step": 0, "speed/wps": 12910.216329851362, "speed/FLOPS": 202772687956624.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06339515745639801, "optim/lr": 0.002787610619780706, "optim/total_tokens": 7566000128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.8613839149475098, "created_at": "2025-01-16T22:19:14.796088+00:00"} {"global_step": 14432, "acc_step": 0, "speed/wps": 12913.661867378853, "speed/FLOPS": 202826804858157.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052799519151449203, "optim/lr": 0.002787566666615168, "optim/total_tokens": 7566524416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373889, "loss/out": 2.7301230430603027, "created_at": "2025-01-16T22:19:24.955200+00:00"} {"global_step": 14433, "acc_step": 0, "speed/wps": 12911.429470713743, "speed/FLOPS": 202791742000897.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04963089898228645, "optim/lr": 0.0027875227092487185, "optim/total_tokens": 7567048704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.926619291305542, "created_at": "2025-01-16T22:19:35.111612+00:00"} {"global_step": 14434, "acc_step": 0, "speed/wps": 12918.150506924752, "speed/FLOPS": 202897305110263.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06559733301401138, "optim/lr": 0.0027874787476814995, "optim/total_tokens": 7567572992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 2.771998405456543, "created_at": "2025-01-16T22:19:45.264483+00:00"} {"global_step": 14435, "acc_step": 0, "speed/wps": 12912.487355919351, "speed/FLOPS": 202808357541738.62, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04778670147061348, "optim/lr": 0.0027874347819136554, "optim/total_tokens": 7568097280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.7634129524230957, "created_at": "2025-01-16T22:19:55.424264+00:00"} {"global_step": 14436, "acc_step": 0, "speed/wps": 12915.685609536871, "speed/FLOPS": 202858590509662.78, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04472678527235985, "optim/lr": 0.0027873908119453302, "optim/total_tokens": 7568621568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372547, "loss/out": 2.930495262145996, "created_at": "2025-01-16T22:20:05.579966+00:00"} {"global_step": 14437, "acc_step": 0, "speed/wps": 12913.740975491608, "speed/FLOPS": 202828047359774.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061130452901124954, "optim/lr": 0.002787346837776666, "optim/total_tokens": 7569145856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.654116630554199, "created_at": "2025-01-16T22:20:15.735133+00:00"} {"global_step": 14438, "acc_step": 0, "speed/wps": 12908.116076682865, "speed/FLOPS": 202739700594562.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.058695416897535324, "optim/lr": 0.0027873028594078077, "optim/total_tokens": 7569670144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.830573320388794, "created_at": "2025-01-16T22:20:25.898468+00:00"} {"global_step": 14439, "acc_step": 0, "speed/wps": 12912.42517463596, "speed/FLOPS": 202807380899236.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04865086078643799, "optim/lr": 0.002787258876838898, "optim/total_tokens": 7570194432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.7061948776245117, "created_at": "2025-01-16T22:20:36.053594+00:00"} {"global_step": 14440, "acc_step": 0, "speed/wps": 12911.336770415148, "speed/FLOPS": 202790286015324.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05211294814944267, "optim/lr": 0.00278721489007008, "optim/total_tokens": 7570718720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365072, "loss/out": 2.8129074573516846, "created_at": "2025-01-16T22:20:46.209058+00:00"} {"global_step": 14441, "acc_step": 0, "speed/wps": 12910.43732409995, "speed/FLOPS": 202776158974976.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06393329799175262, "optim/lr": 0.0027871708991014984, "optim/total_tokens": 7571243008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 2.7508294582366943, "created_at": "2025-01-16T22:20:56.365102+00:00"} {"global_step": 14442, "acc_step": 0, "speed/wps": 12914.263116662925, "speed/FLOPS": 202836248304367.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.062284015119075775, "optim/lr": 0.0027871269039332954, "optim/total_tokens": 7571767296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462167, "loss/out": 2.7626028060913086, "created_at": "2025-01-16T22:21:06.521530+00:00"} {"global_step": 14443, "acc_step": 0, "speed/wps": 12903.562534757786, "speed/FLOPS": 202668180961406.78, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04984553903341293, "optim/lr": 0.0027870829045656153, "optim/total_tokens": 7572291584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.926915168762207, "created_at": "2025-01-16T22:21:16.683141+00:00"} {"global_step": 14444, "acc_step": 0, "speed/wps": 12911.364794403831, "speed/FLOPS": 202790726170576.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06322074681520462, "optim/lr": 0.0027870389009986016, "optim/total_tokens": 7572815872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.7652759552001953, "created_at": "2025-01-16T22:21:26.838629+00:00"} {"global_step": 14445, "acc_step": 0, "speed/wps": 12910.459338432134, "speed/FLOPS": 202776504740313.8, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045266035944223404, "optim/lr": 0.002786994893232398, "optim/total_tokens": 7573340160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.773188591003418, "created_at": "2025-01-16T22:21:36.995638+00:00"} {"global_step": 14446, "acc_step": 0, "speed/wps": 12910.444115042057, "speed/FLOPS": 202776265636053.53, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050026196986436844, "optim/lr": 0.0027869508812671476, "optim/total_tokens": 7573864448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.852842330932617, "created_at": "2025-01-16T22:21:47.152399+00:00"} {"global_step": 14447, "acc_step": 0, "speed/wps": 12909.281040030659, "speed/FLOPS": 202757997944766.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05678322911262512, "optim/lr": 0.0027869068651029945, "optim/total_tokens": 7574388736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8424336910247803, "created_at": "2025-01-16T22:21:57.312084+00:00"} {"global_step": 14448, "acc_step": 0, "speed/wps": 12909.615774671538, "speed/FLOPS": 202763255412275.56, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.05708193778991699, "optim/lr": 0.0027868628447400823, "optim/total_tokens": 7574913024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8524155616760254, "created_at": "2025-01-16T22:22:07.469323+00:00"} {"global_step": 14449, "acc_step": 0, "speed/wps": 12901.962463071903, "speed/FLOPS": 202643049636851.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05006565526127815, "optim/lr": 0.0027868188201785543, "optim/total_tokens": 7575437312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 2.9020113945007324, "created_at": "2025-01-16T22:22:17.632608+00:00"} {"global_step": 14450, "acc_step": 0, "speed/wps": 12912.910300970556, "speed/FLOPS": 202815000474955.2, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05979990214109421, "optim/lr": 0.002786774791418554, "optim/total_tokens": 7575961600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.742351531982422, "created_at": "2025-01-16T22:22:27.791120+00:00"} {"global_step": 14451, "acc_step": 0, "speed/wps": 12914.194201592016, "speed/FLOPS": 202835165898479.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05931166559457779, "optim/lr": 0.002786730758460226, "optim/total_tokens": 7576485888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438471, "loss/out": 2.9041080474853516, "created_at": "2025-01-16T22:22:37.947935+00:00"} {"global_step": 14452, "acc_step": 0, "speed/wps": 12915.912657281304, "speed/FLOPS": 202862156606485.4, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05327612906694412, "optim/lr": 0.0027866867213037125, "optim/total_tokens": 7577010176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8368468284606934, "created_at": "2025-01-16T22:22:48.101373+00:00"} {"global_step": 14453, "acc_step": 0, "speed/wps": 12910.310811836502, "speed/FLOPS": 202774171926032.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0629437044262886, "optim/lr": 0.002786642679949159, "optim/total_tokens": 7577534464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.9263737201690674, "created_at": "2025-01-16T22:22:58.258734+00:00"} {"global_step": 14454, "acc_step": 0, "speed/wps": 12914.558474451349, "speed/FLOPS": 202840887304298.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043303657323122025, "optim/lr": 0.002786598634396707, "optim/total_tokens": 7578058752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.9054200649261475, "created_at": "2025-01-16T22:23:08.411698+00:00"} {"global_step": 14455, "acc_step": 0, "speed/wps": 12908.39472006449, "speed/FLOPS": 202744077071767.03, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0545119047164917, "optim/lr": 0.002786554584646502, "optim/total_tokens": 7578583040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427550, "loss/out": 2.7954750061035156, "created_at": "2025-01-16T22:23:18.569678+00:00"} {"global_step": 14456, "acc_step": 0, "speed/wps": 12910.142576383498, "speed/FLOPS": 202771529557063.3, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041648510843515396, "optim/lr": 0.002786510530698687, "optim/total_tokens": 7579107328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8564586639404297, "created_at": "2025-01-16T22:23:28.726330+00:00"} {"global_step": 14457, "acc_step": 0, "speed/wps": 12912.239387822063, "speed/FLOPS": 202804462862026.84, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05745025724172592, "optim/lr": 0.0027864664725534057, "optim/total_tokens": 7579631616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 2.8689403533935547, "created_at": "2025-01-16T22:23:38.883124+00:00"} {"global_step": 14458, "acc_step": 0, "speed/wps": 12913.376635000233, "speed/FLOPS": 202822324891701.4, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05261857807636261, "optim/lr": 0.002786422410210802, "optim/total_tokens": 7580155904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.9006404876708984, "created_at": "2025-01-16T22:23:49.040414+00:00"} {"global_step": 14459, "acc_step": 0, "speed/wps": 12913.06259820807, "speed/FLOPS": 202817392512348.7, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05086870491504669, "optim/lr": 0.00278637834367102, "optim/total_tokens": 7580680192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.773346424102783, "created_at": "2025-01-16T22:23:59.194646+00:00"} {"global_step": 14460, "acc_step": 0, "speed/wps": 12910.658237086864, "speed/FLOPS": 202779628717001.75, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05819716677069664, "optim/lr": 0.0027863342729342027, "optim/total_tokens": 7581204480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.881467580795288, "created_at": "2025-01-16T22:24:09.350615+00:00"} {"global_step": 14461, "acc_step": 0, "speed/wps": 12912.245287426358, "speed/FLOPS": 202804555523419.2, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04599129036068916, "optim/lr": 0.0027862901980004943, "optim/total_tokens": 7581728768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.838651657104492, "created_at": "2025-01-16T22:24:19.505457+00:00"} {"global_step": 14462, "acc_step": 0, "speed/wps": 12914.289060583993, "speed/FLOPS": 202836655789298.47, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07240567356348038, "optim/lr": 0.0027862461188700384, "optim/total_tokens": 7582253056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.8853209018707275, "created_at": "2025-01-16T22:24:29.661143+00:00"} {"global_step": 14463, "acc_step": 0, "speed/wps": 12913.88403895893, "speed/FLOPS": 202830294368118.8, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04955180734395981, "optim/lr": 0.0027862020355429794, "optim/total_tokens": 7582777344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.845247983932495, "created_at": "2025-01-16T22:24:39.815628+00:00"} {"global_step": 14464, "acc_step": 0, "speed/wps": 12908.925511834877, "speed/FLOPS": 202752413885905.28, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0715632364153862, "optim/lr": 0.0027861579480194607, "optim/total_tokens": 7583301632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 2.847813844680786, "created_at": "2025-01-16T22:24:49.977224+00:00"} {"global_step": 14465, "acc_step": 0, "speed/wps": 12907.933188836052, "speed/FLOPS": 202736828089614.03, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055608656257390976, "optim/lr": 0.0027861138562996265, "optim/total_tokens": 7583825920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8036279678344727, "created_at": "2025-01-16T22:25:00.137879+00:00"} {"global_step": 14466, "acc_step": 0, "speed/wps": 12913.780441040815, "speed/FLOPS": 202828667220456.38, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053338728845119476, "optim/lr": 0.0027860697603836197, "optim/total_tokens": 7584350208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 254508, "loss/out": 2.704134941101074, "created_at": "2025-01-16T22:25:10.297142+00:00"} {"global_step": 14467, "acc_step": 0, "speed/wps": 12911.223950779633, "speed/FLOPS": 202788514027918.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05815592035651207, "optim/lr": 0.002786025660271585, "optim/total_tokens": 7584874496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8094828128814697, "created_at": "2025-01-16T22:25:20.453349+00:00"} {"global_step": 14468, "acc_step": 0, "speed/wps": 12914.06092165908, "speed/FLOPS": 202833072553987.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051038697361946106, "optim/lr": 0.002785981555963666, "optim/total_tokens": 7585398784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 2.8602890968322754, "created_at": "2025-01-16T22:25:30.606481+00:00"} {"global_step": 14469, "acc_step": 0, "speed/wps": 12915.010533659013, "speed/FLOPS": 202847987515349.56, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06240406632423401, "optim/lr": 0.002785937447460007, "optim/total_tokens": 7585923072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.9407236576080322, "created_at": "2025-01-16T22:25:40.758980+00:00"} {"global_step": 14470, "acc_step": 0, "speed/wps": 12909.953989387817, "speed/FLOPS": 202768567539150.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04688519984483719, "optim/lr": 0.0027858933347607515, "optim/total_tokens": 7586447360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 2.9030332565307617, "created_at": "2025-01-16T22:25:50.916478+00:00"} {"global_step": 14471, "acc_step": 0, "speed/wps": 12909.1834411766, "speed/FLOPS": 202756465020649.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05523325130343437, "optim/lr": 0.0027858492178660433, "optim/total_tokens": 7586971648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.8482730388641357, "created_at": "2025-01-16T22:26:01.073629+00:00"} {"global_step": 14472, "acc_step": 0, "speed/wps": 12914.752730504375, "speed/FLOPS": 202843938362547.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0467262901365757, "optim/lr": 0.002785805096776027, "optim/total_tokens": 7587495936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.712292194366455, "created_at": "2025-01-16T22:26:11.234260+00:00"} {"global_step": 14473, "acc_step": 0, "speed/wps": 12905.956753599507, "speed/FLOPS": 202705785458317.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05337974429130554, "optim/lr": 0.0027857609714908456, "optim/total_tokens": 7588020224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.8633501529693604, "created_at": "2025-01-16T22:26:21.396006+00:00"} {"global_step": 14474, "acc_step": 0, "speed/wps": 12912.500050040087, "speed/FLOPS": 202808556920348.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0433313250541687, "optim/lr": 0.0027857168420106445, "optim/total_tokens": 7588544512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.9209964275360107, "created_at": "2025-01-16T22:26:31.553215+00:00"} {"global_step": 14475, "acc_step": 0, "speed/wps": 12910.529617096352, "speed/FLOPS": 202777608563308.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060031741857528687, "optim/lr": 0.0027856727083355654, "optim/total_tokens": 7589068800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8592071533203125, "created_at": "2025-01-16T22:26:41.710745+00:00"} {"global_step": 14476, "acc_step": 0, "speed/wps": 12906.519852645088, "speed/FLOPS": 202714629702607.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05721515789628029, "optim/lr": 0.002785628570465755, "optim/total_tokens": 7589593088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8987889289855957, "created_at": "2025-01-16T22:26:51.871870+00:00"} {"global_step": 14477, "acc_step": 0, "speed/wps": 12910.749539542201, "speed/FLOPS": 202781062747525.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04903829097747803, "optim/lr": 0.0027855844284013554, "optim/total_tokens": 7590117376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.8485536575317383, "created_at": "2025-01-16T22:27:02.027656+00:00"} {"global_step": 14478, "acc_step": 0, "speed/wps": 12907.693928625587, "speed/FLOPS": 202733070179230.72, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0731491893529892, "optim/lr": 0.002785540282142511, "optim/total_tokens": 7590641664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.971053123474121, "created_at": "2025-01-16T22:27:12.191061+00:00"} {"global_step": 14479, "acc_step": 0, "speed/wps": 12908.195434153575, "speed/FLOPS": 202740947012688.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0622861348092556, "optim/lr": 0.002785496131689366, "optim/total_tokens": 7591165952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 2.8856468200683594, "created_at": "2025-01-16T22:27:22.351095+00:00"} {"global_step": 14480, "acc_step": 0, "speed/wps": 12910.07013710356, "speed/FLOPS": 202770391798629.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052723903208971024, "optim/lr": 0.0027854519770420653, "optim/total_tokens": 7591690240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.7756032943725586, "created_at": "2025-01-16T22:27:32.508349+00:00"} {"global_step": 14481, "acc_step": 0, "speed/wps": 12913.99022416562, "speed/FLOPS": 202831962152703.66, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04664991796016693, "optim/lr": 0.0027854078182007517, "optim/total_tokens": 7592214528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 2.83243727684021, "created_at": "2025-01-16T22:27:42.661567+00:00"} {"global_step": 14482, "acc_step": 0, "speed/wps": 12913.202574365341, "speed/FLOPS": 202819591030246.03, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05886967480182648, "optim/lr": 0.0027853636551655693, "optim/total_tokens": 7592738816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9001293182373047, "created_at": "2025-01-16T22:27:52.819255+00:00"} {"global_step": 14483, "acc_step": 0, "speed/wps": 12913.92081550577, "speed/FLOPS": 202830871994323.22, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050051841884851456, "optim/lr": 0.002785319487936663, "optim/total_tokens": 7593263104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 2.800786256790161, "created_at": "2025-01-16T22:28:02.973064+00:00"} {"global_step": 14484, "acc_step": 0, "speed/wps": 12915.760232358758, "speed/FLOPS": 202859762563623.03, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04692565277218819, "optim/lr": 0.0027852753165141762, "optim/total_tokens": 7593787392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 2.708225965499878, "created_at": "2025-01-16T22:28:13.125715+00:00"} {"global_step": 14485, "acc_step": 0, "speed/wps": 12907.422601630356, "speed/FLOPS": 202728808615928.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04285348206758499, "optim/lr": 0.002785231140898254, "optim/total_tokens": 7594311680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.756488561630249, "created_at": "2025-01-16T22:28:23.286589+00:00"} {"global_step": 14486, "acc_step": 0, "speed/wps": 12913.09073039835, "speed/FLOPS": 202817834367054.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043008241802453995, "optim/lr": 0.0027851869610890393, "optim/total_tokens": 7594835968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8408732414245605, "created_at": "2025-01-16T22:28:33.447173+00:00"} {"global_step": 14487, "acc_step": 0, "speed/wps": 12905.750378235216, "speed/FLOPS": 202702544049631.8, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04531346634030342, "optim/lr": 0.0027851427770866767, "optim/total_tokens": 7595360256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 2.8673644065856934, "created_at": "2025-01-16T22:28:43.607043+00:00"} {"global_step": 14488, "acc_step": 0, "speed/wps": 12909.8764221577, "speed/FLOPS": 202767349239212.94, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.039675720036029816, "optim/lr": 0.002785098588891311, "optim/total_tokens": 7595884544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.830589771270752, "created_at": "2025-01-16T22:28:53.763386+00:00"} {"global_step": 14489, "acc_step": 0, "speed/wps": 12913.240980928831, "speed/FLOPS": 202820194258101.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.052876606583595276, "optim/lr": 0.002785054396503085, "optim/total_tokens": 7596408832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.793379306793213, "created_at": "2025-01-16T22:29:03.917818+00:00"} {"global_step": 14490, "acc_step": 0, "speed/wps": 12906.13402280808, "speed/FLOPS": 202708569714831.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05193012207746506, "optim/lr": 0.0027850101999221445, "optim/total_tokens": 7596933120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 2.836893081665039, "created_at": "2025-01-16T22:29:14.080745+00:00"} {"global_step": 14491, "acc_step": 0, "speed/wps": 12909.934578545126, "speed/FLOPS": 202768262665192.72, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.001, "optim/grad_norm": 0.047646574676036835, "optim/lr": 0.0027849659991486322, "optim/total_tokens": 7597457408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.7487971782684326, "created_at": "2025-01-16T22:29:24.239509+00:00"} {"global_step": 14492, "acc_step": 0, "speed/wps": 12911.636123696162, "speed/FLOPS": 202794987769957.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046941958367824554, "optim/lr": 0.0027849217941826937, "optim/total_tokens": 7597981696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391702, "loss/out": 2.8202645778656006, "created_at": "2025-01-16T22:29:34.394605+00:00"} {"global_step": 14493, "acc_step": 0, "speed/wps": 12909.78796941552, "speed/FLOPS": 202765959967350.2, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.041110727936029434, "optim/lr": 0.0027848775850244723, "optim/total_tokens": 7598505984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.917645215988159, "created_at": "2025-01-16T22:29:44.551416+00:00"} {"global_step": 14494, "acc_step": 0, "speed/wps": 12912.920472332617, "speed/FLOPS": 202815160230173.28, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0653749406337738, "optim/lr": 0.002784833371674112, "optim/total_tokens": 7599030272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.916578769683838, "created_at": "2025-01-16T22:29:54.705405+00:00"} {"global_step": 14495, "acc_step": 0, "speed/wps": 12911.77140407883, "speed/FLOPS": 202797112534262.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06062503531575203, "optim/lr": 0.0027847891541317588, "optim/total_tokens": 7599554560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.777904987335205, "created_at": "2025-01-16T22:30:04.863678+00:00"} {"global_step": 14496, "acc_step": 0, "speed/wps": 12914.116033558947, "speed/FLOPS": 202833938162106.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05728491395711899, "optim/lr": 0.002784744932397555, "optim/total_tokens": 7600078848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.7720260620117188, "created_at": "2025-01-16T22:30:15.016966+00:00"} {"global_step": 14497, "acc_step": 0, "speed/wps": 12911.645172466686, "speed/FLOPS": 202795129893333.2, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05296783521771431, "optim/lr": 0.0027847007064716457, "optim/total_tokens": 7600603136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.7435269355773926, "created_at": "2025-01-16T22:30:25.174124+00:00"} {"global_step": 14498, "acc_step": 0, "speed/wps": 12910.77004247377, "speed/FLOPS": 202781384774239.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05372202396392822, "optim/lr": 0.002784656476354175, "optim/total_tokens": 7601127424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8844127655029297, "created_at": "2025-01-16T22:30:35.335167+00:00"} {"global_step": 14499, "acc_step": 0, "speed/wps": 12910.139255679398, "speed/FLOPS": 202771477400842.3, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04996681585907936, "optim/lr": 0.002784612242045287, "optim/total_tokens": 7601651712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.8010923862457275, "created_at": "2025-01-16T22:30:45.492044+00:00"} {"global_step": 14500, "acc_step": 0, "speed/wps": 12907.879803671489, "speed/FLOPS": 202735989602245.66, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06421961635351181, "optim/lr": 0.0027845680035451267, "optim/total_tokens": 7602176000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.9741225242614746, "created_at": "2025-01-16T22:30:55.655584+00:00"} {"global_step": 14501, "acc_step": 0, "speed/wps": 12907.35643366709, "speed/FLOPS": 202727769357140.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.047299548983573914, "optim/lr": 0.0027845237608538384, "optim/total_tokens": 7602700288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 2.780245542526245, "created_at": "2025-01-16T22:31:05.821544+00:00"} {"global_step": 14502, "acc_step": 0, "speed/wps": 12913.908961101703, "speed/FLOPS": 202830685804617.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05729369819164276, "optim/lr": 0.0027844795139715657, "optim/total_tokens": 7603224576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486862, "loss/out": 2.763451099395752, "created_at": "2025-01-16T22:31:15.979043+00:00"} {"global_step": 14503, "acc_step": 0, "speed/wps": 12904.143972302512, "speed/FLOPS": 202677313237025.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04866413772106171, "optim/lr": 0.002784435262898453, "optim/total_tokens": 7603748864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8651680946350098, "created_at": "2025-01-16T22:31:26.149863+00:00"} {"global_step": 14504, "acc_step": 0, "speed/wps": 12902.264437060749, "speed/FLOPS": 202647792553301.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04546767845749855, "optim/lr": 0.0027843910076346453, "optim/total_tokens": 7604273152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.8805899620056152, "created_at": "2025-01-16T22:31:36.316668+00:00"} {"global_step": 14505, "acc_step": 0, "speed/wps": 12902.459147899097, "speed/FLOPS": 202650850754578.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05710529908537865, "optim/lr": 0.002784346748180287, "optim/total_tokens": 7604797440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.866619110107422, "created_at": "2025-01-16T22:31:46.481977+00:00"} {"global_step": 14506, "acc_step": 0, "speed/wps": 12908.523400526883, "speed/FLOPS": 202746098175254.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049980029463768005, "optim/lr": 0.0027843024845355226, "optim/total_tokens": 7605321728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.7758283615112305, "created_at": "2025-01-16T22:31:56.640035+00:00"} {"global_step": 14507, "acc_step": 0, "speed/wps": 12913.029256555532, "speed/FLOPS": 202816868835879.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051631294190883636, "optim/lr": 0.0027842582167004956, "optim/total_tokens": 7605846016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.834476947784424, "created_at": "2025-01-16T22:32:06.795476+00:00"} {"global_step": 14508, "acc_step": 0, "speed/wps": 12904.981476001502, "speed/FLOPS": 202690467383468.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04849753901362419, "optim/lr": 0.002784213944675351, "optim/total_tokens": 7606370304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 2.861971855163574, "created_at": "2025-01-16T22:32:16.958165+00:00"} {"global_step": 14509, "acc_step": 0, "speed/wps": 12902.223756140225, "speed/FLOPS": 202647153603543.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06646391749382019, "optim/lr": 0.0027841696684602333, "optim/total_tokens": 7606894592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.847823143005371, "created_at": "2025-01-16T22:32:27.121674+00:00"} {"global_step": 14510, "acc_step": 0, "speed/wps": 12909.897455219398, "speed/FLOPS": 202767679592349.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05206931009888649, "optim/lr": 0.002784125388055287, "optim/total_tokens": 7607418880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.9086642265319824, "created_at": "2025-01-16T22:32:37.284443+00:00"} {"global_step": 14511, "acc_step": 0, "speed/wps": 12907.859553792778, "speed/FLOPS": 202735671550074.25, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.08369887620210648, "optim/lr": 0.0027840811034606566, "optim/total_tokens": 7607943168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8956403732299805, "created_at": "2025-01-16T22:32:47.443766+00:00"} {"global_step": 14512, "acc_step": 0, "speed/wps": 12908.14146078573, "speed/FLOPS": 202740099286779.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07612190395593643, "optim/lr": 0.002784036814676486, "optim/total_tokens": 7608467456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.934947967529297, "created_at": "2025-01-16T22:32:57.606024+00:00"} {"global_step": 14513, "acc_step": 0, "speed/wps": 12910.229874564753, "speed/FLOPS": 202772900694960.2, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06872709840536118, "optim/lr": 0.0027839925217029208, "optim/total_tokens": 7608991744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.901585817337036, "created_at": "2025-01-16T22:33:07.764326+00:00"} {"global_step": 14514, "acc_step": 0, "speed/wps": 12903.597256694602, "speed/FLOPS": 202668726317137.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08857255429029465, "optim/lr": 0.002783948224540105, "optim/total_tokens": 7609516032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.749390125274658, "created_at": "2025-01-16T22:33:17.958827+00:00"} {"global_step": 14515, "acc_step": 0, "speed/wps": 12905.546413019221, "speed/FLOPS": 202699340495637.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06066959351301193, "optim/lr": 0.002783903923188182, "optim/total_tokens": 7610040320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.7877326011657715, "created_at": "2025-01-16T22:33:28.123813+00:00"} {"global_step": 14516, "acc_step": 0, "speed/wps": 12902.05617848121, "speed/FLOPS": 202644521566132.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054094456136226654, "optim/lr": 0.002783859617647298, "optim/total_tokens": 7610564608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291084, "loss/out": 2.873857021331787, "created_at": "2025-01-16T22:33:38.288423+00:00"} {"global_step": 14517, "acc_step": 0, "speed/wps": 12910.998866999542, "speed/FLOPS": 202784978777854.9, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06638200581073761, "optim/lr": 0.002783815307917597, "optim/total_tokens": 7611088896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.821014642715454, "created_at": "2025-01-16T22:33:48.448886+00:00"} {"global_step": 14518, "acc_step": 0, "speed/wps": 12907.231543129978, "speed/FLOPS": 202725807779637.9, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.056063294410705566, "optim/lr": 0.002783770993999223, "optim/total_tokens": 7611613184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 2.807382583618164, "created_at": "2025-01-16T22:33:58.607574+00:00"} {"global_step": 14519, "acc_step": 0, "speed/wps": 12906.201235150013, "speed/FLOPS": 202709625377021.03, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059870678931474686, "optim/lr": 0.0027837266758923215, "optim/total_tokens": 7612137472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.900169849395752, "created_at": "2025-01-16T22:34:08.771074+00:00"} {"global_step": 14520, "acc_step": 0, "speed/wps": 12907.724330508005, "speed/FLOPS": 202733547682570.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06563513725996017, "optim/lr": 0.002783682353597036, "optim/total_tokens": 7612661760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 2.836610794067383, "created_at": "2025-01-16T22:34:18.930604+00:00"} {"global_step": 14521, "acc_step": 0, "speed/wps": 12896.410749956996, "speed/FLOPS": 202555852353528.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0702265128493309, "optim/lr": 0.0027836380271135126, "optim/total_tokens": 7613186048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.9222049713134766, "created_at": "2025-01-16T22:34:29.098604+00:00"} {"global_step": 14522, "acc_step": 0, "speed/wps": 12910.162212864454, "speed/FLOPS": 202771837974980.1, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044425033032894135, "optim/lr": 0.002783593696441895, "optim/total_tokens": 7613710336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.683133840560913, "created_at": "2025-01-16T22:34:39.254921+00:00"} {"global_step": 14523, "acc_step": 0, "speed/wps": 12905.836405138349, "speed/FLOPS": 202703895220358.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05833742395043373, "optim/lr": 0.0027835493615823276, "optim/total_tokens": 7614234624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.7735910415649414, "created_at": "2025-01-16T22:34:49.420904+00:00"} {"global_step": 14524, "acc_step": 0, "speed/wps": 12908.547651977286, "speed/FLOPS": 202746479077607.94, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04774407669901848, "optim/lr": 0.0027835050225349556, "optim/total_tokens": 7614758912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.8862998485565186, "created_at": "2025-01-16T22:34:59.578904+00:00"} {"global_step": 14525, "acc_step": 0, "speed/wps": 12915.080417438732, "speed/FLOPS": 202849085136140.44, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05176544934511185, "optim/lr": 0.002783460679299923, "optim/total_tokens": 7615283200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 2.7983365058898926, "created_at": "2025-01-16T22:35:09.731334+00:00"} {"global_step": 14526, "acc_step": 0, "speed/wps": 12912.873046790703, "speed/FLOPS": 202814415346868.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05103522166609764, "optim/lr": 0.002783416331877375, "optim/total_tokens": 7615807488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8035337924957275, "created_at": "2025-01-16T22:35:19.888179+00:00"} {"global_step": 14527, "acc_step": 0, "speed/wps": 12911.54777832209, "speed/FLOPS": 202793600184458.72, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0676075890660286, "optim/lr": 0.0027833719802674564, "optim/total_tokens": 7616331776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 2.7874104976654053, "created_at": "2025-01-16T22:35:30.045887+00:00"} {"global_step": 14528, "acc_step": 0, "speed/wps": 12901.9850895773, "speed/FLOPS": 202643405017210.44, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06772159785032272, "optim/lr": 0.0027833276244703113, "optim/total_tokens": 7616856064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.876861810684204, "created_at": "2025-01-16T22:35:40.208493+00:00"} {"global_step": 14529, "acc_step": 0, "speed/wps": 12901.81017680913, "speed/FLOPS": 202640657771830.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043475210666656494, "optim/lr": 0.0027832832644860855, "optim/total_tokens": 7617380352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8858377933502197, "created_at": "2025-01-16T22:35:50.373502+00:00"} {"global_step": 14530, "acc_step": 0, "speed/wps": 12906.253260722044, "speed/FLOPS": 202710442510120.4, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07125632464885712, "optim/lr": 0.002783238900314923, "optim/total_tokens": 7617904640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 2.864675998687744, "created_at": "2025-01-16T22:36:00.533471+00:00"} {"global_step": 14531, "acc_step": 0, "speed/wps": 12908.774792196085, "speed/FLOPS": 202750046626867.56, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04645814746618271, "optim/lr": 0.0027831945319569677, "optim/total_tokens": 7618428928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9116435050964355, "created_at": "2025-01-16T22:36:10.693968+00:00"} {"global_step": 14532, "acc_step": 0, "speed/wps": 12906.030590619901, "speed/FLOPS": 202706945170185.47, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04984505847096443, "optim/lr": 0.0027831501594123664, "optim/total_tokens": 7618953216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 2.8435258865356445, "created_at": "2025-01-16T22:36:20.853891+00:00"} {"global_step": 14533, "acc_step": 0, "speed/wps": 12904.651515918395, "speed/FLOPS": 202685284906951.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051525622606277466, "optim/lr": 0.002783105782681262, "optim/total_tokens": 7619477504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8301143646240234, "created_at": "2025-01-16T22:36:31.014374+00:00"} {"global_step": 14534, "acc_step": 0, "speed/wps": 12904.489334643591, "speed/FLOPS": 202682737627169.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045174770057201385, "optim/lr": 0.0027830614017638005, "optim/total_tokens": 7620001792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 2.7634334564208984, "created_at": "2025-01-16T22:36:41.185366+00:00"} {"global_step": 14535, "acc_step": 0, "speed/wps": 12897.094893139283, "speed/FLOPS": 202566597762317.28, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05535779148340225, "optim/lr": 0.002783017016660126, "optim/total_tokens": 7620526080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.844511032104492, "created_at": "2025-01-16T22:36:51.352320+00:00"} {"global_step": 14536, "acc_step": 0, "speed/wps": 12893.901439681707, "speed/FLOPS": 202516440187501.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05154895409941673, "optim/lr": 0.0027829726273703834, "optim/total_tokens": 7621050368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.736067295074463, "created_at": "2025-01-16T22:37:01.521358+00:00"} {"global_step": 14537, "acc_step": 0, "speed/wps": 12904.176061243323, "speed/FLOPS": 202677817237935.88, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05608534812927246, "optim/lr": 0.0027829282338947174, "optim/total_tokens": 7621574656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.889410972595215, "created_at": "2025-01-16T22:37:11.683318+00:00"} {"global_step": 14538, "acc_step": 0, "speed/wps": 12898.567106603608, "speed/FLOPS": 202589720897807.8, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05582290142774582, "optim/lr": 0.0027828838362332735, "optim/total_tokens": 7622098944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.8331046104431152, "created_at": "2025-01-16T22:37:21.848534+00:00"} {"global_step": 14539, "acc_step": 0, "speed/wps": 12906.434558743806, "speed/FLOPS": 202713290044681.75, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04591281712055206, "optim/lr": 0.0027828394343861963, "optim/total_tokens": 7622623232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 2.9596471786499023, "created_at": "2025-01-16T22:37:32.011878+00:00"} {"global_step": 14540, "acc_step": 0, "speed/wps": 12904.957125533183, "speed/FLOPS": 202690084925902.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06040855124592781, "optim/lr": 0.0027827950283536304, "optim/total_tokens": 7623147520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.8154959678649902, "created_at": "2025-01-16T22:37:42.172849+00:00"} {"global_step": 14541, "acc_step": 0, "speed/wps": 12907.876989638942, "speed/FLOPS": 202735945403997.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045107875019311905, "optim/lr": 0.002782750618135721, "optim/total_tokens": 7623671808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7913992404937744, "created_at": "2025-01-16T22:37:52.335224+00:00"} {"global_step": 14542, "acc_step": 0, "speed/wps": 12905.93940105184, "speed/FLOPS": 202705512912711.47, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05146678164601326, "optim/lr": 0.002782706203732612, "optim/total_tokens": 7624196096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 2.9073193073272705, "created_at": "2025-01-16T22:38:02.494782+00:00"} {"global_step": 14543, "acc_step": 0, "speed/wps": 12905.882696110804, "speed/FLOPS": 202704622283691.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048962850123643875, "optim/lr": 0.0027826617851444504, "optim/total_tokens": 7624720384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.7575440406799316, "created_at": "2025-01-16T22:38:12.654482+00:00"} {"global_step": 14544, "acc_step": 0, "speed/wps": 12901.737309420418, "speed/FLOPS": 202639513289360.97, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045962847769260406, "optim/lr": 0.002782617362371379, "optim/total_tokens": 7625244672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.7347331047058105, "created_at": "2025-01-16T22:38:22.817205+00:00"} {"global_step": 14545, "acc_step": 0, "speed/wps": 12908.875953572302, "speed/FLOPS": 202751635505291.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04406600072979927, "optim/lr": 0.002782572935413544, "optim/total_tokens": 7625768960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.679279088973999, "created_at": "2025-01-16T22:38:32.975261+00:00"} {"global_step": 14546, "acc_step": 0, "speed/wps": 12908.642109539845, "speed/FLOPS": 202747962663426.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04422979801893234, "optim/lr": 0.0027825285042710896, "optim/total_tokens": 7626293248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.827622413635254, "created_at": "2025-01-16T22:38:43.134914+00:00"} {"global_step": 14547, "acc_step": 0, "speed/wps": 12912.538666857983, "speed/FLOPS": 202809163450538.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04825456812977791, "optim/lr": 0.0027824840689441607, "optim/total_tokens": 7626817536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.8542752265930176, "created_at": "2025-01-16T22:38:53.306282+00:00"} {"global_step": 14548, "acc_step": 0, "speed/wps": 12911.892975590075, "speed/FLOPS": 202799021981903.2, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051078200340270996, "optim/lr": 0.0027824396294329036, "optim/total_tokens": 7627341824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.906930923461914, "created_at": "2025-01-16T22:39:03.463063+00:00"} {"global_step": 14549, "acc_step": 0, "speed/wps": 12907.903721221883, "speed/FLOPS": 202736365260240.75, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057608816772699356, "optim/lr": 0.002782395185737462, "optim/total_tokens": 7627866112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388029, "loss/out": 2.812511920928955, "created_at": "2025-01-16T22:39:13.621093+00:00"} {"global_step": 14550, "acc_step": 0, "speed/wps": 12908.592446894156, "speed/FLOPS": 202747182643332.06, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045194484293460846, "optim/lr": 0.002782350737857981, "optim/total_tokens": 7628390400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.8287267684936523, "created_at": "2025-01-16T22:39:23.778599+00:00"} {"global_step": 14551, "acc_step": 0, "speed/wps": 12912.578264855887, "speed/FLOPS": 202809785391508.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05746874585747719, "optim/lr": 0.002782306285794606, "optim/total_tokens": 7628914688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.6777119636535645, "created_at": "2025-01-16T22:39:33.933762+00:00"} {"global_step": 14552, "acc_step": 0, "speed/wps": 12908.102935378784, "speed/FLOPS": 202739494192323.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05204840004444122, "optim/lr": 0.002782261829547482, "optim/total_tokens": 7629438976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.758202075958252, "created_at": "2025-01-16T22:39:44.094808+00:00"} {"global_step": 14553, "acc_step": 0, "speed/wps": 12913.677536779769, "speed/FLOPS": 202827050967631.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06207527965307236, "optim/lr": 0.002782217369116754, "optim/total_tokens": 7629963264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378157, "loss/out": 2.855153799057007, "created_at": "2025-01-16T22:39:54.249691+00:00"} {"global_step": 14554, "acc_step": 0, "speed/wps": 12910.737069158333, "speed/FLOPS": 202780866883010.78, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061836566776037216, "optim/lr": 0.002782172904502567, "optim/total_tokens": 7630487552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.8555564880371094, "created_at": "2025-01-16T22:40:04.405350+00:00"} {"global_step": 14555, "acc_step": 0, "speed/wps": 12907.11850639837, "speed/FLOPS": 202724032382439.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.047341566532850266, "optim/lr": 0.0027821284357050662, "optim/total_tokens": 7631011840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 2.7849559783935547, "created_at": "2025-01-16T22:40:14.563946+00:00"} {"global_step": 14556, "acc_step": 0, "speed/wps": 12904.424519596607, "speed/FLOPS": 202681719617791.78, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054416678845882416, "optim/lr": 0.002782083962724396, "optim/total_tokens": 7631536128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332493, "loss/out": 2.814971446990967, "created_at": "2025-01-16T22:40:24.725255+00:00"} {"global_step": 14557, "acc_step": 0, "speed/wps": 12910.215504402842, "speed/FLOPS": 202772674991820.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0432458259165287, "optim/lr": 0.0027820394855607032, "optim/total_tokens": 7632060416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.778172492980957, "created_at": "2025-01-16T22:40:34.885851+00:00"} {"global_step": 14558, "acc_step": 0, "speed/wps": 12909.71728198831, "speed/FLOPS": 202764849724171.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.053519658744335175, "optim/lr": 0.002781995004214131, "optim/total_tokens": 7632584704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 2.823183059692383, "created_at": "2025-01-16T22:40:45.044410+00:00"} {"global_step": 14559, "acc_step": 0, "speed/wps": 12906.905365835324, "speed/FLOPS": 202720684717010.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04632299393415451, "optim/lr": 0.0027819505186848256, "optim/total_tokens": 7633108992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.809826374053955, "created_at": "2025-01-16T22:40:55.207964+00:00"} {"global_step": 14560, "acc_step": 0, "speed/wps": 12911.086924147994, "speed/FLOPS": 202786361836374.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.041048817336559296, "optim/lr": 0.0027819060289729315, "optim/total_tokens": 7633633280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 2.8704428672790527, "created_at": "2025-01-16T22:41:05.363436+00:00"} {"global_step": 14561, "acc_step": 0, "speed/wps": 12910.285078122271, "speed/FLOPS": 202773767742688.66, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05762310326099396, "optim/lr": 0.0027818615350785948, "optim/total_tokens": 7634157568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.7883012294769287, "created_at": "2025-01-16T22:41:15.522494+00:00"} {"global_step": 14562, "acc_step": 0, "speed/wps": 12908.167213185006, "speed/FLOPS": 202740503763597.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04328685253858566, "optim/lr": 0.0027818170370019595, "optim/total_tokens": 7634681856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.8360085487365723, "created_at": "2025-01-16T22:41:25.683410+00:00"} {"global_step": 14563, "acc_step": 0, "speed/wps": 12909.859618021068, "speed/FLOPS": 202767085307155.8, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04971632361412048, "optim/lr": 0.002781772534743172, "optim/total_tokens": 7635206144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.785919427871704, "created_at": "2025-01-16T22:41:35.839705+00:00"} {"global_step": 14564, "acc_step": 0, "speed/wps": 12910.22932860697, "speed/FLOPS": 202772892119943.0, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048340387642383575, "optim/lr": 0.0027817280283023767, "optim/total_tokens": 7635730432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 2.7079927921295166, "created_at": "2025-01-16T22:41:45.999749+00:00"} {"global_step": 14565, "acc_step": 0, "speed/wps": 12904.793524907944, "speed/FLOPS": 202687515353272.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04612473398447037, "optim/lr": 0.002781683517679719, "optim/total_tokens": 7636254720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331010, "loss/out": 2.783898115158081, "created_at": "2025-01-16T22:41:56.161469+00:00"} {"global_step": 14566, "acc_step": 0, "speed/wps": 12909.189710980007, "speed/FLOPS": 202756563496527.22, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043239302933216095, "optim/lr": 0.002781639002875344, "optim/total_tokens": 7636779008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.7805910110473633, "created_at": "2025-01-16T22:42:06.320877+00:00"} {"global_step": 14567, "acc_step": 0, "speed/wps": 12911.401146764492, "speed/FLOPS": 202791297134353.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049631860107183456, "optim/lr": 0.0027815944838893974, "optim/total_tokens": 7637303296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 2.6956911087036133, "created_at": "2025-01-16T22:42:16.480966+00:00"} {"global_step": 14568, "acc_step": 0, "speed/wps": 12910.62489482133, "speed/FLOPS": 202779105030904.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049985431134700775, "optim/lr": 0.0027815499607220236, "optim/total_tokens": 7637827584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.826265573501587, "created_at": "2025-01-16T22:42:26.637177+00:00"} {"global_step": 14569, "acc_step": 0, "speed/wps": 12908.671582893474, "speed/FLOPS": 202748425582945.62, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04252292588353157, "optim/lr": 0.0027815054333733685, "optim/total_tokens": 7638351872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 2.7711031436920166, "created_at": "2025-01-16T22:42:36.794658+00:00"} {"global_step": 14570, "acc_step": 0, "speed/wps": 12910.682357987765, "speed/FLOPS": 202780007568896.0, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05887310206890106, "optim/lr": 0.002781460901843577, "optim/total_tokens": 7638876160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.886739730834961, "created_at": "2025-01-16T22:42:46.952213+00:00"} {"global_step": 14571, "acc_step": 0, "speed/wps": 12907.592280289156, "speed/FLOPS": 202731473652427.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05579356849193573, "optim/lr": 0.002781416366132795, "optim/total_tokens": 7639400448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8988847732543945, "created_at": "2025-01-16T22:42:57.111832+00:00"} {"global_step": 14572, "acc_step": 0, "speed/wps": 12906.705070097882, "speed/FLOPS": 202717538797219.06, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0470319502055645, "optim/lr": 0.0027813718262411674, "optim/total_tokens": 7639924736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.764833688735962, "created_at": "2025-01-16T22:43:07.270625+00:00"} {"global_step": 14573, "acc_step": 0, "speed/wps": 12908.71360216154, "speed/FLOPS": 202749085553291.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060648106038570404, "optim/lr": 0.0027813272821688396, "optim/total_tokens": 7640449024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409120, "loss/out": 2.867741823196411, "created_at": "2025-01-16T22:43:17.427832+00:00"} {"global_step": 14574, "acc_step": 0, "speed/wps": 12911.790176035804, "speed/FLOPS": 202797407373641.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05483859032392502, "optim/lr": 0.002781282733915957, "optim/total_tokens": 7640973312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.890629291534424, "created_at": "2025-01-16T22:43:27.583260+00:00"} {"global_step": 14575, "acc_step": 0, "speed/wps": 12911.339116896876, "speed/FLOPS": 202790322870044.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055385515093803406, "optim/lr": 0.0027812381814826645, "optim/total_tokens": 7641497600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.823075532913208, "created_at": "2025-01-16T22:43:37.743770+00:00"} {"global_step": 14576, "acc_step": 0, "speed/wps": 12907.320445000674, "speed/FLOPS": 202727204105681.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05862601846456528, "optim/lr": 0.0027811936248691082, "optim/total_tokens": 7642021888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.719411849975586, "created_at": "2025-01-16T22:43:47.902327+00:00"} {"global_step": 14577, "acc_step": 0, "speed/wps": 12914.03831202803, "speed/FLOPS": 202832717438662.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06872614473104477, "optim/lr": 0.002781149064075433, "optim/total_tokens": 7642546176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507980, "loss/out": 2.809166431427002, "created_at": "2025-01-16T22:43:58.055738+00:00"} {"global_step": 14578, "acc_step": 0, "speed/wps": 12908.105278317435, "speed/FLOPS": 202739530991394.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07736396044492722, "optim/lr": 0.002781104499101784, "optim/total_tokens": 7643070464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 2.88980770111084, "created_at": "2025-01-16T22:44:08.217786+00:00"} {"global_step": 14579, "acc_step": 0, "speed/wps": 12903.695294025316, "speed/FLOPS": 202670266128133.7, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05077046528458595, "optim/lr": 0.0027810599299483074, "optim/total_tokens": 7643594752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 2.794552803039551, "created_at": "2025-01-16T22:44:18.385851+00:00"} {"global_step": 14580, "acc_step": 0, "speed/wps": 12912.549399189873, "speed/FLOPS": 202809332016558.56, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.061044998466968536, "optim/lr": 0.002781015356615148, "optim/total_tokens": 7644119040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 2.7155094146728516, "created_at": "2025-01-16T22:44:28.546732+00:00"} {"global_step": 14581, "acc_step": 0, "speed/wps": 12912.042533675241, "speed/FLOPS": 202801370997145.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04806767776608467, "optim/lr": 0.0027809707791024516, "optim/total_tokens": 7644643328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.757284164428711, "created_at": "2025-01-16T22:44:38.702420+00:00"} {"global_step": 14582, "acc_step": 0, "speed/wps": 12908.353303132977, "speed/FLOPS": 202743426561944.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06086133420467377, "optim/lr": 0.002780926197410363, "optim/total_tokens": 7645167616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 2.669398307800293, "created_at": "2025-01-16T22:44:48.862847+00:00"} {"global_step": 14583, "acc_step": 0, "speed/wps": 12909.867020599051, "speed/FLOPS": 202767201574815.25, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04590361565351486, "optim/lr": 0.0027808816115390284, "optim/total_tokens": 7645691904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.7119593620300293, "created_at": "2025-01-16T22:44:59.019342+00:00"} {"global_step": 14584, "acc_step": 0, "speed/wps": 12910.202550677912, "speed/FLOPS": 202772471535770.4, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06508234888315201, "optim/lr": 0.002780837021488593, "optim/total_tokens": 7646216192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.875291109085083, "created_at": "2025-01-16T22:45:09.176371+00:00"} {"global_step": 14585, "acc_step": 0, "speed/wps": 12911.405245153608, "speed/FLOPS": 202791361505186.62, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04536264017224312, "optim/lr": 0.0027807924272592027, "optim/total_tokens": 7646740480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.7540476322174072, "created_at": "2025-01-16T22:45:19.335830+00:00"} {"global_step": 14586, "acc_step": 0, "speed/wps": 12907.369975931535, "speed/FLOPS": 202727982057012.97, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0496915802359581, "optim/lr": 0.002780747828851002, "optim/total_tokens": 7647264768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.9557135105133057, "created_at": "2025-01-16T22:45:29.495338+00:00"} {"global_step": 14587, "acc_step": 0, "speed/wps": 12909.689159757978, "speed/FLOPS": 202764408025899.78, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.047334223985672, "optim/lr": 0.002780703226264137, "optim/total_tokens": 7647789056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.7105045318603516, "created_at": "2025-01-16T22:45:39.654510+00:00"} {"global_step": 14588, "acc_step": 0, "speed/wps": 12909.470012345564, "speed/FLOPS": 202760966014647.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04858916625380516, "optim/lr": 0.002780658619498753, "optim/total_tokens": 7648313344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.77451753616333, "created_at": "2025-01-16T22:45:49.812122+00:00"} {"global_step": 14589, "acc_step": 0, "speed/wps": 12912.224739530691, "speed/FLOPS": 202804232790482.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045504000037908554, "optim/lr": 0.0027806140085549964, "optim/total_tokens": 7648837632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.8070178031921387, "created_at": "2025-01-16T22:45:59.967633+00:00"} {"global_step": 14590, "acc_step": 0, "speed/wps": 12913.646554676021, "speed/FLOPS": 202826564351118.12, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04861920326948166, "optim/lr": 0.002780569393433011, "optim/total_tokens": 7649361920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.9051568508148193, "created_at": "2025-01-16T22:46:10.121409+00:00"} {"global_step": 14591, "acc_step": 0, "speed/wps": 12912.738514595267, "speed/FLOPS": 202812302333872.06, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06240790709853172, "optim/lr": 0.0027805247741329447, "optim/total_tokens": 7649886208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393877, "loss/out": 2.9024534225463867, "created_at": "2025-01-16T22:46:20.279273+00:00"} {"global_step": 14592, "acc_step": 0, "speed/wps": 12907.560524760785, "speed/FLOPS": 202730974888217.88, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0513579286634922, "optim/lr": 0.002780480150654941, "optim/total_tokens": 7650410496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 2.863412857055664, "created_at": "2025-01-16T22:46:30.445640+00:00"} {"global_step": 14593, "acc_step": 0, "speed/wps": 12909.24666996595, "speed/FLOPS": 202757458115671.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04993857443332672, "optim/lr": 0.0027804355229991465, "optim/total_tokens": 7650934784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375916, "loss/out": 2.9083447456359863, "created_at": "2025-01-16T22:46:40.605744+00:00"} {"global_step": 14594, "acc_step": 0, "speed/wps": 12908.395045651736, "speed/FLOPS": 202744082185562.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07225874811410904, "optim/lr": 0.002780390891165707, "optim/total_tokens": 7651459072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8341989517211914, "created_at": "2025-01-16T22:46:50.767598+00:00"} {"global_step": 14595, "acc_step": 0, "speed/wps": 12911.863607167912, "speed/FLOPS": 202798560710476.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04779841750860214, "optim/lr": 0.002780346255154767, "optim/total_tokens": 7651983360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8601181507110596, "created_at": "2025-01-16T22:47:00.922403+00:00"} {"global_step": 14596, "acc_step": 0, "speed/wps": 12912.986858487306, "speed/FLOPS": 202816202915956.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.07931021600961685, "optim/lr": 0.0027803016149664734, "optim/total_tokens": 7652507648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.7839250564575195, "created_at": "2025-01-16T22:47:11.078159+00:00"} {"global_step": 14597, "acc_step": 0, "speed/wps": 12908.621416823034, "speed/FLOPS": 202747637655873.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05808897316455841, "optim/lr": 0.002780256970600971, "optim/total_tokens": 7653031936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 2.8571829795837402, "created_at": "2025-01-16T22:47:21.236573+00:00"} {"global_step": 14598, "acc_step": 0, "speed/wps": 12909.799609018026, "speed/FLOPS": 202766142783302.03, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05100341886281967, "optim/lr": 0.0027802123220584066, "optim/total_tokens": 7653556224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 2.821317195892334, "created_at": "2025-01-16T22:47:31.392957+00:00"} {"global_step": 14599, "acc_step": 0, "speed/wps": 12906.89072964381, "speed/FLOPS": 202720454835510.84, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04717843607068062, "optim/lr": 0.002780167669338924, "optim/total_tokens": 7654080512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.7916622161865234, "created_at": "2025-01-16T22:47:41.553077+00:00"} {"global_step": 14600, "acc_step": 0, "speed/wps": 12908.788478336022, "speed/FLOPS": 202750261586502.0, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048471126705408096, "optim/lr": 0.002780123012442671, "optim/total_tokens": 7654604800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333550, "loss/out": 2.785916805267334, "created_at": "2025-01-16T22:47:51.710472+00:00"} {"global_step": 14601, "acc_step": 0, "speed/wps": 12905.364776645029, "speed/FLOPS": 202696487646785.8, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044755566865205765, "optim/lr": 0.0027800783513697914, "optim/total_tokens": 7655129088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.7437238693237305, "created_at": "2025-01-16T22:48:01.878158+00:00"} {"global_step": 14602, "acc_step": 0, "speed/wps": 12911.742420009643, "speed/FLOPS": 202796657299627.44, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04532899707555771, "optim/lr": 0.002780033686120432, "optim/total_tokens": 7655653376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 2.842924118041992, "created_at": "2025-01-16T22:48:12.033257+00:00"} {"global_step": 14603, "acc_step": 0, "speed/wps": 12907.01718790756, "speed/FLOPS": 202722441036314.9, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049516744911670685, "optim/lr": 0.0027799890166947385, "optim/total_tokens": 7656177664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.9066898822784424, "created_at": "2025-01-16T22:48:22.197137+00:00"} {"global_step": 14604, "acc_step": 0, "speed/wps": 12916.148290385909, "speed/FLOPS": 202865857548187.03, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05290431156754494, "optim/lr": 0.002779944343092856, "optim/total_tokens": 7656701952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.7699010372161865, "created_at": "2025-01-16T22:48:32.352779+00:00"} {"global_step": 14605, "acc_step": 0, "speed/wps": 12905.816651631489, "speed/FLOPS": 202703584964389.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04287123680114746, "optim/lr": 0.002779899665314931, "optim/total_tokens": 7657226240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.796063184738159, "created_at": "2025-01-16T22:48:42.517411+00:00"} {"global_step": 14606, "acc_step": 0, "speed/wps": 12913.170665633914, "speed/FLOPS": 202819089859771.78, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.052680958062410355, "optim/lr": 0.0027798549833611087, "optim/total_tokens": 7657750528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.9818527698516846, "created_at": "2025-01-16T22:48:52.672016+00:00"} {"global_step": 14607, "acc_step": 0, "speed/wps": 12917.074103646519, "speed/FLOPS": 202880398717645.66, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06356661766767502, "optim/lr": 0.0027798102972315356, "optim/total_tokens": 7658274816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 2.8221230506896973, "created_at": "2025-01-16T22:49:02.824868+00:00"} {"global_step": 14608, "acc_step": 0, "speed/wps": 12907.269309288613, "speed/FLOPS": 202726400949055.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050379350781440735, "optim/lr": 0.0027797656069263564, "optim/total_tokens": 7658799104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8858509063720703, "created_at": "2025-01-16T22:49:12.984456+00:00"} {"global_step": 14609, "acc_step": 0, "speed/wps": 12909.16751389036, "speed/FLOPS": 202756214860732.3, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048450808972120285, "optim/lr": 0.002779720912445718, "optim/total_tokens": 7659323392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312410, "loss/out": 2.7457098960876465, "created_at": "2025-01-16T22:49:23.143958+00:00"} {"global_step": 14610, "acc_step": 0, "speed/wps": 12902.89683936418, "speed/FLOPS": 202657725300488.88, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04504330828785896, "optim/lr": 0.002779676213789766, "optim/total_tokens": 7659847680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8931565284729004, "created_at": "2025-01-16T22:49:33.305877+00:00"} {"global_step": 14611, "acc_step": 0, "speed/wps": 12911.898816832596, "speed/FLOPS": 202799113726643.7, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04215126112103462, "optim/lr": 0.002779631510958645, "optim/total_tokens": 7660371968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426204, "loss/out": 2.8625316619873047, "created_at": "2025-01-16T22:49:43.463954+00:00"} {"global_step": 14612, "acc_step": 0, "speed/wps": 12910.62831644615, "speed/FLOPS": 202779158772224.06, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05271461606025696, "optim/lr": 0.002779586803952503, "optim/total_tokens": 7660896256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.7507801055908203, "created_at": "2025-01-16T22:49:53.619883+00:00"} {"global_step": 14613, "acc_step": 0, "speed/wps": 12910.737686222064, "speed/FLOPS": 202780876574844.62, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04812583699822426, "optim/lr": 0.002779542092771484, "optim/total_tokens": 7661420544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.8484556674957275, "created_at": "2025-01-16T22:50:03.776515+00:00"} {"global_step": 14614, "acc_step": 0, "speed/wps": 12913.17062653453, "speed/FLOPS": 202819089245662.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.051430683583021164, "optim/lr": 0.002779497377415735, "optim/total_tokens": 7661944832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.708545684814453, "created_at": "2025-01-16T22:50:13.931768+00:00"} {"global_step": 14615, "acc_step": 0, "speed/wps": 12895.829254003776, "speed/FLOPS": 202546719160523.72, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05263676866889, "optim/lr": 0.0027794526578854006, "optim/total_tokens": 7662469120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8710596561431885, "created_at": "2025-01-16T22:50:24.101205+00:00"} {"global_step": 14616, "acc_step": 0, "speed/wps": 12918.646275971916, "speed/FLOPS": 202905091844407.2, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05860529839992523, "optim/lr": 0.002779407934180629, "optim/total_tokens": 7662993408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.755448341369629, "created_at": "2025-01-16T22:50:34.250582+00:00"} {"global_step": 14617, "acc_step": 0, "speed/wps": 12913.780678029168, "speed/FLOPS": 202828670942684.12, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05852179974317551, "optim/lr": 0.002779363206301564, "optim/total_tokens": 7663517696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.810227870941162, "created_at": "2025-01-16T22:50:44.405333+00:00"} {"global_step": 14618, "acc_step": 0, "speed/wps": 12913.67927506184, "speed/FLOPS": 202827078269739.97, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049436379224061966, "optim/lr": 0.0027793184742483526, "optim/total_tokens": 7664041984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.7436270713806152, "created_at": "2025-01-16T22:50:54.558656+00:00"} {"global_step": 14619, "acc_step": 0, "speed/wps": 12913.529105568769, "speed/FLOPS": 202824719651496.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05475273355841637, "optim/lr": 0.00277927373802114, "optim/total_tokens": 7664566272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442492, "loss/out": 2.8439700603485107, "created_at": "2025-01-16T22:51:04.713166+00:00"} {"global_step": 14620, "acc_step": 0, "speed/wps": 12908.835415658225, "speed/FLOPS": 202750998801645.84, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05859696492552757, "optim/lr": 0.0027792289976200724, "optim/total_tokens": 7665090560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.83949613571167, "created_at": "2025-01-16T22:51:14.870312+00:00"} {"global_step": 14621, "acc_step": 0, "speed/wps": 12904.84511077593, "speed/FLOPS": 202688325580215.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04206091910600662, "optim/lr": 0.002779184253045296, "optim/total_tokens": 7665614848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.027878761291504, "created_at": "2025-01-16T22:51:25.031931+00:00"} {"global_step": 14622, "acc_step": 0, "speed/wps": 12904.132420985465, "speed/FLOPS": 202677131807718.22, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04632122069597244, "optim/lr": 0.002779139504296957, "optim/total_tokens": 7666139136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 2.625711441040039, "created_at": "2025-01-16T22:51:35.194072+00:00"} {"global_step": 14623, "acc_step": 0, "speed/wps": 12905.844075304905, "speed/FLOPS": 202704015690864.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04387199506163597, "optim/lr": 0.002779094751375201, "optim/total_tokens": 7666663424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.89837646484375, "created_at": "2025-01-16T22:51:45.355756+00:00"} {"global_step": 14624, "acc_step": 0, "speed/wps": 12911.453322058123, "speed/FLOPS": 202792116619036.62, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0504775270819664, "optim/lr": 0.0027790499942801743, "optim/total_tokens": 7667187712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472295, "loss/out": 2.677936553955078, "created_at": "2025-01-16T22:51:55.510985+00:00"} {"global_step": 14625, "acc_step": 0, "speed/wps": 12908.913993397438, "speed/FLOPS": 202752232973017.3, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04709852114319801, "optim/lr": 0.0027790052330120226, "optim/total_tokens": 7667712000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490582, "loss/out": 2.8660998344421387, "created_at": "2025-01-16T22:52:05.669107+00:00"} {"global_step": 14626, "acc_step": 0, "speed/wps": 12908.225512819754, "speed/FLOPS": 202741419439473.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04005094990134239, "optim/lr": 0.0027789604675708918, "optim/total_tokens": 7668236288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389084, "loss/out": 2.6588852405548096, "created_at": "2025-01-16T22:52:15.827066+00:00"} {"global_step": 14627, "acc_step": 0, "speed/wps": 12908.83959411708, "speed/FLOPS": 202751064430083.97, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056372467428445816, "optim/lr": 0.0027789156979569287, "optim/total_tokens": 7668760576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9591434001922607, "created_at": "2025-01-16T22:52:25.984476+00:00"} {"global_step": 14628, "acc_step": 0, "speed/wps": 12910.191138915445, "speed/FLOPS": 202772292298359.47, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05274733155965805, "optim/lr": 0.0027788709241702786, "optim/total_tokens": 7669284864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.855426073074341, "created_at": "2025-01-16T22:52:36.141679+00:00"} {"global_step": 14629, "acc_step": 0, "speed/wps": 12903.37099190026, "speed/FLOPS": 202665172517621.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06655342131853104, "optim/lr": 0.0027788261462110882, "optim/total_tokens": 7669809152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.7887003421783447, "created_at": "2025-01-16T22:52:46.303200+00:00"} {"global_step": 14630, "acc_step": 0, "speed/wps": 12913.172792403839, "speed/FLOPS": 202819123263615.94, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04823685064911842, "optim/lr": 0.0027787813640795032, "optim/total_tokens": 7670333440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8926284313201904, "created_at": "2025-01-16T22:52:56.459252+00:00"} {"global_step": 14631, "acc_step": 0, "speed/wps": 12907.434057033806, "speed/FLOPS": 202728988538780.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05084647610783577, "optim/lr": 0.0027787365777756693, "optim/total_tokens": 7670857728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384815, "loss/out": 2.7779641151428223, "created_at": "2025-01-16T22:53:06.618710+00:00"} {"global_step": 14632, "acc_step": 0, "speed/wps": 12912.720428218789, "speed/FLOPS": 202812018262476.2, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04551009088754654, "optim/lr": 0.002778691787299734, "optim/total_tokens": 7671382016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8067195415496826, "created_at": "2025-01-16T22:53:16.779967+00:00"} {"global_step": 14633, "acc_step": 0, "speed/wps": 12905.88675075851, "speed/FLOPS": 202704685967505.62, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05322517082095146, "optim/lr": 0.0027786469926518423, "optim/total_tokens": 7671906304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 2.759918212890625, "created_at": "2025-01-16T22:53:26.939571+00:00"} {"global_step": 14634, "acc_step": 0, "speed/wps": 12905.198261699805, "speed/FLOPS": 202693872300755.56, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044421661645174026, "optim/lr": 0.0027786021938321403, "optim/total_tokens": 7672430592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.766547918319702, "created_at": "2025-01-16T22:53:37.100424+00:00"} {"global_step": 14635, "acc_step": 0, "speed/wps": 12905.736329295221, "speed/FLOPS": 202702323391723.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048613280057907104, "optim/lr": 0.0027785573908407748, "optim/total_tokens": 7672954880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.8788650035858154, "created_at": "2025-01-16T22:53:47.261145+00:00"} {"global_step": 14636, "acc_step": 0, "speed/wps": 12918.603184062213, "speed/FLOPS": 202904415026750.8, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0526091493666172, "optim/lr": 0.002778512583677892, "optim/total_tokens": 7673479168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.7584195137023926, "created_at": "2025-01-16T22:53:57.413997+00:00"} {"global_step": 14637, "acc_step": 0, "speed/wps": 12916.931468594012, "speed/FLOPS": 202878158438144.94, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.055553603917360306, "optim/lr": 0.0027784677723436367, "optim/total_tokens": 7674003456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8579022884368896, "created_at": "2025-01-16T22:54:07.565035+00:00"} {"global_step": 14638, "acc_step": 0, "speed/wps": 12906.87565434207, "speed/FLOPS": 202720218057181.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05845530703663826, "optim/lr": 0.0027784229568381573, "optim/total_tokens": 7674527744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.767125129699707, "created_at": "2025-01-16T22:54:17.727262+00:00"} {"global_step": 14639, "acc_step": 0, "speed/wps": 12911.498125298436, "speed/FLOPS": 202792820315491.84, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05813965946435928, "optim/lr": 0.002778378137161598, "optim/total_tokens": 7675052032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 2.8442976474761963, "created_at": "2025-01-16T22:54:27.882290+00:00"} {"global_step": 14640, "acc_step": 0, "speed/wps": 12914.855738416987, "speed/FLOPS": 202845556243362.6, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05959348380565643, "optim/lr": 0.002778333313314107, "optim/total_tokens": 7675576320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.8684566020965576, "created_at": "2025-01-16T22:54:38.034998+00:00"} {"global_step": 14641, "acc_step": 0, "speed/wps": 12907.145163816373, "speed/FLOPS": 202724451073820.12, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06670835614204407, "optim/lr": 0.002778288485295828, "optim/total_tokens": 7676100608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8396975994110107, "created_at": "2025-01-16T22:54:48.196877+00:00"} {"global_step": 14642, "acc_step": 0, "speed/wps": 12905.764515966133, "speed/FLOPS": 202702766102123.16, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05691639333963394, "optim/lr": 0.00277824365310691, "optim/total_tokens": 7676624896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 2.914341688156128, "created_at": "2025-01-16T22:54:58.356348+00:00"} {"global_step": 14643, "acc_step": 0, "speed/wps": 12914.531695151696, "speed/FLOPS": 202840466698599.1, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07299425452947617, "optim/lr": 0.002778198816747497, "optim/total_tokens": 7677149184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7911791801452637, "created_at": "2025-01-16T22:55:08.510503+00:00"} {"global_step": 14644, "acc_step": 0, "speed/wps": 12906.107125758568, "speed/FLOPS": 202708147259708.53, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06724507361650467, "optim/lr": 0.002778153976217737, "optim/total_tokens": 7677673472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409693, "loss/out": 2.8314435482025146, "created_at": "2025-01-16T22:55:18.673025+00:00"} {"global_step": 14645, "acc_step": 0, "speed/wps": 12912.426418563778, "speed/FLOPS": 202807400436832.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06146463006734848, "optim/lr": 0.0027781091315177754, "optim/total_tokens": 7678197760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8093347549438477, "created_at": "2025-01-16T22:55:28.838677+00:00"} {"global_step": 14646, "acc_step": 0, "speed/wps": 12908.833567380743, "speed/FLOPS": 202750969771908.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08018533885478973, "optim/lr": 0.0027780642826477583, "optim/total_tokens": 7678722048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367469, "loss/out": 2.7669262886047363, "created_at": "2025-01-16T22:55:38.999278+00:00"} {"global_step": 14647, "acc_step": 0, "speed/wps": 12909.573145603328, "speed/FLOPS": 202762585864180.88, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046122487634420395, "optim/lr": 0.0027780194296078324, "optim/total_tokens": 7679246336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.797835350036621, "created_at": "2025-01-16T22:55:49.158223+00:00"} {"global_step": 14648, "acc_step": 0, "speed/wps": 12911.271869982469, "speed/FLOPS": 202789266664847.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0756995901465416, "optim/lr": 0.002777974572398145, "optim/total_tokens": 7679770624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8863189220428467, "created_at": "2025-01-16T22:55:59.313414+00:00"} {"global_step": 14649, "acc_step": 0, "speed/wps": 12906.460564724546, "speed/FLOPS": 202713698504345.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055505070835351944, "optim/lr": 0.0027779297110188408, "optim/total_tokens": 7680294912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 2.770555257797241, "created_at": "2025-01-16T22:56:09.472699+00:00"} {"global_step": 14650, "acc_step": 0, "speed/wps": 12905.28654742149, "speed/FLOPS": 202695258949331.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05671996995806694, "optim/lr": 0.0027778848454700668, "optim/total_tokens": 7680819200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8603124618530273, "created_at": "2025-01-16T22:56:19.632692+00:00"} {"global_step": 14651, "acc_step": 0, "speed/wps": 12907.98796076433, "speed/FLOPS": 202737688358012.16, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06391926854848862, "optim/lr": 0.00277783997575197, "optim/total_tokens": 7681343488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 2.870077610015869, "created_at": "2025-01-16T22:56:29.793963+00:00"} {"global_step": 14652, "acc_step": 0, "speed/wps": 12913.515440182007, "speed/FLOPS": 202824505017819.47, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04587173834443092, "optim/lr": 0.0027777951018646955, "optim/total_tokens": 7681867776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.87369441986084, "created_at": "2025-01-16T22:56:39.951837+00:00"} {"global_step": 14653, "acc_step": 0, "speed/wps": 12913.547166958013, "speed/FLOPS": 202825003330433.78, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059637896716594696, "optim/lr": 0.0027777502238083905, "optim/total_tokens": 7682392064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 2.8469443321228027, "created_at": "2025-01-16T22:56:50.110409+00:00"} {"global_step": 14654, "acc_step": 0, "speed/wps": 12908.02138057039, "speed/FLOPS": 202738213261989.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06168455630540848, "optim/lr": 0.0027777053415832024, "optim/total_tokens": 7682916352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.6973400115966797, "created_at": "2025-01-16T22:57:00.268982+00:00"} {"global_step": 14655, "acc_step": 0, "speed/wps": 12908.66918291705, "speed/FLOPS": 202748387888017.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05228901654481888, "optim/lr": 0.002777660455189276, "optim/total_tokens": 7683440640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.9257383346557617, "created_at": "2025-01-16T22:57:10.426415+00:00"} {"global_step": 14656, "acc_step": 0, "speed/wps": 12912.18463430196, "speed/FLOPS": 202803602882754.4, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053275588899850845, "optim/lr": 0.0027776155646267583, "optim/total_tokens": 7683964928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.7697479724884033, "created_at": "2025-01-16T22:57:20.581038+00:00"} {"global_step": 14657, "acc_step": 0, "speed/wps": 12906.3056346832, "speed/FLOPS": 202711265115149.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05535636469721794, "optim/lr": 0.0027775706698957956, "optim/total_tokens": 7684489216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 2.867292642593384, "created_at": "2025-01-16T22:57:30.743297+00:00"} {"global_step": 14658, "acc_step": 0, "speed/wps": 12915.736241828143, "speed/FLOPS": 202859385759373.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04971160739660263, "optim/lr": 0.002777525770996535, "optim/total_tokens": 7685013504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.8059098720550537, "created_at": "2025-01-16T22:57:40.895169+00:00"} {"global_step": 14659, "acc_step": 0, "speed/wps": 12910.797353365686, "speed/FLOPS": 202781813729326.44, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047703664749860764, "optim/lr": 0.002777480867929123, "optim/total_tokens": 7685537792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.8401758670806885, "created_at": "2025-01-16T22:57:51.053877+00:00"} {"global_step": 14660, "acc_step": 0, "speed/wps": 12902.372474046959, "speed/FLOPS": 202649489422628.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05594172701239586, "optim/lr": 0.002777435960693705, "optim/total_tokens": 7686062080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.9370226860046387, "created_at": "2025-01-16T22:58:01.216947+00:00"} {"global_step": 14661, "acc_step": 0, "speed/wps": 12906.950500910387, "speed/FLOPS": 202721393625386.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07393037527799606, "optim/lr": 0.0027773910492904287, "optim/total_tokens": 7686586368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.800398349761963, "created_at": "2025-01-16T22:58:11.376993+00:00"} {"global_step": 14662, "acc_step": 0, "speed/wps": 12910.7660487141, "speed/FLOPS": 202781322046756.2, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06349777430295944, "optim/lr": 0.0027773461337194396, "optim/total_tokens": 7687110656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413096, "loss/out": 2.846120834350586, "created_at": "2025-01-16T22:58:21.534875+00:00"} {"global_step": 14663, "acc_step": 0, "speed/wps": 12911.365794120624, "speed/FLOPS": 202790741872502.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05017958581447601, "optim/lr": 0.002777301213980885, "optim/total_tokens": 7687634944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506648, "loss/out": 2.891697406768799, "created_at": "2025-01-16T22:58:31.690047+00:00"} {"global_step": 14664, "acc_step": 0, "speed/wps": 12913.316595666172, "speed/FLOPS": 202821381891457.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0447445884346962, "optim/lr": 0.002777256290074912, "optim/total_tokens": 7688159232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.798485279083252, "created_at": "2025-01-16T22:58:41.847683+00:00"} {"global_step": 14665, "acc_step": 0, "speed/wps": 12909.295305094032, "speed/FLOPS": 202758221997189.5, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050696950405836105, "optim/lr": 0.0027772113620016655, "optim/total_tokens": 7688683520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8111276626586914, "created_at": "2025-01-16T22:58:52.004610+00:00"} {"global_step": 14666, "acc_step": 0, "speed/wps": 12912.615794561192, "speed/FLOPS": 202810374847101.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04925955832004547, "optim/lr": 0.0027771664297612936, "optim/total_tokens": 7689207808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415820, "loss/out": 2.7027411460876465, "created_at": "2025-01-16T22:59:02.159224+00:00"} {"global_step": 14667, "acc_step": 0, "speed/wps": 12910.220603947326, "speed/FLOPS": 202772755087174.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04060211777687073, "optim/lr": 0.002777121493353942, "optim/total_tokens": 7689732096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8493151664733887, "created_at": "2025-01-16T22:59:12.315274+00:00"} {"global_step": 14668, "acc_step": 0, "speed/wps": 12910.159836019271, "speed/FLOPS": 202771800643360.38, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0465814508497715, "optim/lr": 0.002777076552779758, "optim/total_tokens": 7690256384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.8665170669555664, "created_at": "2025-01-16T22:59:22.474502+00:00"} {"global_step": 14669, "acc_step": 0, "speed/wps": 12902.671967752447, "speed/FLOPS": 202654193382820.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05140114948153496, "optim/lr": 0.0027770316080388876, "optim/total_tokens": 7690780672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 2.87577486038208, "created_at": "2025-01-16T22:59:32.637272+00:00"} {"global_step": 14670, "acc_step": 0, "speed/wps": 12917.609308671974, "speed/FLOPS": 202888804848019.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04576415568590164, "optim/lr": 0.0027769866591314776, "optim/total_tokens": 7691304960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 2.842268705368042, "created_at": "2025-01-16T22:59:42.792023+00:00"} {"global_step": 14671, "acc_step": 0, "speed/wps": 12908.732765425566, "speed/FLOPS": 202749386538685.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05092582106590271, "optim/lr": 0.002776941706057675, "optim/total_tokens": 7691829248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286807, "loss/out": 2.6875500679016113, "created_at": "2025-01-16T22:59:52.949785+00:00"} {"global_step": 14672, "acc_step": 0, "speed/wps": 12897.99609832779, "speed/FLOPS": 202580752428188.5, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.040894344449043274, "optim/lr": 0.002776896748817626, "optim/total_tokens": 7692353536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.820347785949707, "created_at": "2025-01-16T23:00:03.115735+00:00"} {"global_step": 14673, "acc_step": 0, "speed/wps": 12901.385369711763, "speed/FLOPS": 202633985592620.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04248279705643654, "optim/lr": 0.002776851787411478, "optim/total_tokens": 7692877824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402693, "loss/out": 2.8258132934570312, "created_at": "2025-01-16T23:00:13.283071+00:00"} {"global_step": 14674, "acc_step": 0, "speed/wps": 12900.913194909348, "speed/FLOPS": 202626569438512.62, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0583667978644371, "optim/lr": 0.0027768068218393762, "optim/total_tokens": 7693402112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8808059692382812, "created_at": "2025-01-16T23:00:23.449084+00:00"} {"global_step": 14675, "acc_step": 0, "speed/wps": 12897.909515160052, "speed/FLOPS": 202579392520561.4, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07406900078058243, "optim/lr": 0.0027767618521014692, "optim/total_tokens": 7693926400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 2.863823890686035, "created_at": "2025-01-16T23:00:33.617785+00:00"} {"global_step": 14676, "acc_step": 0, "speed/wps": 12898.035338856116, "speed/FLOPS": 202581368754607.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0026, "optim/grad_norm": 0.047822508960962296, "optim/lr": 0.0027767168781979027, "optim/total_tokens": 7694450688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 2.8565022945404053, "created_at": "2025-01-16T23:00:43.783590+00:00"} {"global_step": 14677, "acc_step": 0, "speed/wps": 12903.328168642167, "speed/FLOPS": 202664499919508.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06378877907991409, "optim/lr": 0.0027766719001288236, "optim/total_tokens": 7694974976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.7398908138275146, "created_at": "2025-01-16T23:00:53.947853+00:00"} {"global_step": 14678, "acc_step": 0, "speed/wps": 12897.968631163867, "speed/FLOPS": 202580321018635.22, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05582324042916298, "optim/lr": 0.002776626917894378, "optim/total_tokens": 7695499264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.809278726577759, "created_at": "2025-01-16T23:01:04.114087+00:00"} {"global_step": 14679, "acc_step": 0, "speed/wps": 12897.571116846268, "speed/FLOPS": 202574077510035.4, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0462515763938427, "optim/lr": 0.002776581931494714, "optim/total_tokens": 7696023552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 2.7671682834625244, "created_at": "2025-01-16T23:01:14.285641+00:00"} {"global_step": 14680, "acc_step": 0, "speed/wps": 12902.323293733354, "speed/FLOPS": 202648716978222.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05626601353287697, "optim/lr": 0.002776536940929977, "optim/total_tokens": 7696547840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 2.7884984016418457, "created_at": "2025-01-16T23:01:24.450845+00:00"} {"global_step": 14681, "acc_step": 0, "speed/wps": 12890.62504542602, "speed/FLOPS": 202464979913478.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0024, "optim/grad_norm": 0.05940235033631325, "optim/lr": 0.002776491946200315, "optim/total_tokens": 7697072128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8772106170654297, "created_at": "2025-01-16T23:01:34.627038+00:00"} {"global_step": 14682, "acc_step": 0, "speed/wps": 12883.641181954892, "speed/FLOPS": 202355288740829.9, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04993876442313194, "optim/lr": 0.002776446947305874, "optim/total_tokens": 7697596416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 2.894864082336426, "created_at": "2025-01-16T23:01:44.804829+00:00"} {"global_step": 14683, "acc_step": 0, "speed/wps": 12897.245067135666, "speed/FLOPS": 202568956451291.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04279535636305809, "optim/lr": 0.0027764019442468014, "optim/total_tokens": 7698120704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.7263941764831543, "created_at": "2025-01-16T23:01:54.971297+00:00"} {"global_step": 14684, "acc_step": 0, "speed/wps": 12896.821991801218, "speed/FLOPS": 202562311471797.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 0.051825426518917084, "optim/lr": 0.0027763569370232435, "optim/total_tokens": 7698644992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.7361948490142822, "created_at": "2025-01-16T23:02:05.139648+00:00"} {"global_step": 14685, "acc_step": 0, "speed/wps": 12894.724504993106, "speed/FLOPS": 202529367559226.9, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045158758759498596, "optim/lr": 0.0027763119256353475, "optim/total_tokens": 7699169280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 2.6979691982269287, "created_at": "2025-01-16T23:02:15.310852+00:00"} {"global_step": 14686, "acc_step": 0, "speed/wps": 12901.273510380355, "speed/FLOPS": 202632228688110.12, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.060297295451164246, "optim/lr": 0.0027762669100832595, "optim/total_tokens": 7699693568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.7737104892730713, "created_at": "2025-01-16T23:02:25.475555+00:00"} {"global_step": 14687, "acc_step": 0, "speed/wps": 12906.430913275179, "speed/FLOPS": 202713232787587.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04925285279750824, "optim/lr": 0.002776221890367128, "optim/total_tokens": 7700217856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489004, "loss/out": 2.882112503051758, "created_at": "2025-01-16T23:02:35.638417+00:00"} {"global_step": 14688, "acc_step": 0, "speed/wps": 12891.529538686073, "speed/FLOPS": 202479186223031.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.060723718255758286, "optim/lr": 0.0027761768664870973, "optim/total_tokens": 7700742144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 2.823598623275757, "created_at": "2025-01-16T23:02:45.810810+00:00"} {"global_step": 14689, "acc_step": 0, "speed/wps": 12898.885943109346, "speed/FLOPS": 202594728663257.7, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054080698639154434, "optim/lr": 0.002776131838443317, "optim/total_tokens": 7701266432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.7779064178466797, "created_at": "2025-01-16T23:02:55.978172+00:00"} {"global_step": 14690, "acc_step": 0, "speed/wps": 12892.414396781143, "speed/FLOPS": 202493084135336.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.07058493793010712, "optim/lr": 0.0027760868062359327, "optim/total_tokens": 7701790720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 2.698029041290283, "created_at": "2025-01-16T23:03:06.150392+00:00"} {"global_step": 14691, "acc_step": 0, "speed/wps": 12895.013085851933, "speed/FLOPS": 202533900118166.47, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0023, "optim/grad_norm": 0.07282257825136185, "optim/lr": 0.0027760417698650917, "optim/total_tokens": 7702315008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.929440498352051, "created_at": "2025-01-16T23:03:16.321326+00:00"} {"global_step": 14692, "acc_step": 0, "speed/wps": 12899.04682497428, "speed/FLOPS": 202597255534021.34, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05547915771603584, "optim/lr": 0.0027759967293309405, "optim/total_tokens": 7702839296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8232526779174805, "created_at": "2025-01-16T23:03:26.486745+00:00"} {"global_step": 14693, "acc_step": 0, "speed/wps": 12895.097026075931, "speed/FLOPS": 202535218514730.3, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07147279381752014, "optim/lr": 0.002775951684633626, "optim/total_tokens": 7703363584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.9225873947143555, "created_at": "2025-01-16T23:03:36.655602+00:00"} {"global_step": 14694, "acc_step": 0, "speed/wps": 12899.430884525647, "speed/FLOPS": 202603287717028.7, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04832899942994118, "optim/lr": 0.0027759066357732954, "optim/total_tokens": 7703887872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7727839946746826, "created_at": "2025-01-16T23:03:46.820880+00:00"} {"global_step": 14695, "acc_step": 0, "speed/wps": 12895.83852641571, "speed/FLOPS": 202546864796494.75, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07529445737600327, "optim/lr": 0.0027758615827500966, "optim/total_tokens": 7704412160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.829591751098633, "created_at": "2025-01-16T23:03:56.989193+00:00"} {"global_step": 14696, "acc_step": 0, "speed/wps": 12905.698521218179, "speed/FLOPS": 202701729563920.25, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06810271739959717, "optim/lr": 0.002775816525564175, "optim/total_tokens": 7704936448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.7673704624176025, "created_at": "2025-01-16T23:04:07.154164+00:00"} {"global_step": 14697, "acc_step": 0, "speed/wps": 12896.746428292185, "speed/FLOPS": 202561124643052.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05763407424092293, "optim/lr": 0.0027757714642156783, "optim/total_tokens": 7705460736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.8782763481140137, "created_at": "2025-01-16T23:04:17.323387+00:00"} {"global_step": 14698, "acc_step": 0, "speed/wps": 12905.347352318728, "speed/FLOPS": 202696213973797.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07886768877506256, "optim/lr": 0.0027757263987047537, "optim/total_tokens": 7705985024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378681, "loss/out": 2.8811287879943848, "created_at": "2025-01-16T23:04:27.483486+00:00"} {"global_step": 14699, "acc_step": 0, "speed/wps": 12905.934780656891, "speed/FLOPS": 202705440343059.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0580793134868145, "optim/lr": 0.002775681329031548, "optim/total_tokens": 7706509312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.806462287902832, "created_at": "2025-01-16T23:04:37.646014+00:00"} {"global_step": 14700, "acc_step": 0, "speed/wps": 12906.682455441745, "speed/FLOPS": 202717183602968.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0742473229765892, "optim/lr": 0.0027756362551962085, "optim/total_tokens": 7707033600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.903033494949341, "created_at": "2025-01-16T23:04:47.807075+00:00"} {"global_step": 14701, "acc_step": 0, "speed/wps": 12903.904264527599, "speed/FLOPS": 202673548297027.62, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.08904659003019333, "optim/lr": 0.0027755911771988823, "optim/total_tokens": 7707557888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 2.7881157398223877, "created_at": "2025-01-16T23:04:57.985055+00:00"} {"global_step": 14702, "acc_step": 0, "speed/wps": 12904.697618533368, "speed/FLOPS": 202686009011870.88, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0793067067861557, "optim/lr": 0.0027755460950397155, "optim/total_tokens": 7708082176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.7766966819763184, "created_at": "2025-01-16T23:05:08.145747+00:00"} {"global_step": 14703, "acc_step": 0, "speed/wps": 12904.582013553545, "speed/FLOPS": 202684193276804.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04993705451488495, "optim/lr": 0.002775501008718857, "optim/total_tokens": 7708606464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.7864999771118164, "created_at": "2025-01-16T23:05:18.310670+00:00"} {"global_step": 14704, "acc_step": 0, "speed/wps": 12895.350222735095, "speed/FLOPS": 202539195316191.12, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06361423432826996, "optim/lr": 0.002775455918236452, "optim/total_tokens": 7709130752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.777193546295166, "created_at": "2025-01-16T23:05:28.479352+00:00"} {"global_step": 14705, "acc_step": 0, "speed/wps": 12889.513547704311, "speed/FLOPS": 202447522314400.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05518621951341629, "optim/lr": 0.002775410823592649, "optim/total_tokens": 7709655040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.911651134490967, "created_at": "2025-01-16T23:05:38.654074+00:00"} {"global_step": 14706, "acc_step": 0, "speed/wps": 12895.323172113176, "speed/FLOPS": 202538770449002.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.045229654759168625, "optim/lr": 0.0027753657247875946, "optim/total_tokens": 7710179328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 2.8167552947998047, "created_at": "2025-01-16T23:05:48.824442+00:00"} {"global_step": 14707, "acc_step": 0, "speed/wps": 12905.018250376479, "speed/FLOPS": 202691044975561.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0470474474132061, "optim/lr": 0.0027753206218214354, "optim/total_tokens": 7710703616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.855278968811035, "created_at": "2025-01-16T23:05:58.987300+00:00"} {"global_step": 14708, "acc_step": 0, "speed/wps": 12904.111246976898, "speed/FLOPS": 202676799240817.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04779354855418205, "optim/lr": 0.00277527551469432, "optim/total_tokens": 7711227904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.774160385131836, "created_at": "2025-01-16T23:06:09.152974+00:00"} {"global_step": 14709, "acc_step": 0, "speed/wps": 12890.92877604693, "speed/FLOPS": 202469750420249.06, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05667737126350403, "optim/lr": 0.002775230403406394, "optim/total_tokens": 7711752192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437181, "loss/out": 2.755751132965088, "created_at": "2025-01-16T23:06:19.325413+00:00"} {"global_step": 14710, "acc_step": 0, "speed/wps": 12911.427523379029, "speed/FLOPS": 202791711415330.34, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04451940581202507, "optim/lr": 0.002775185287957805, "optim/total_tokens": 7712276480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 508224, "loss/out": 2.7913763523101807, "created_at": "2025-01-16T23:06:29.481535+00:00"} {"global_step": 14711, "acc_step": 0, "speed/wps": 12906.470221714495, "speed/FLOPS": 202713850180642.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05590684711933136, "optim/lr": 0.0027751401683487006, "optim/total_tokens": 7712800768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469331, "loss/out": 2.9160523414611816, "created_at": "2025-01-16T23:06:39.641383+00:00"} {"global_step": 14712, "acc_step": 0, "speed/wps": 12901.78120196192, "speed/FLOPS": 202640202682039.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05972833186388016, "optim/lr": 0.0027750950445792283, "optim/total_tokens": 7713325056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.8118577003479004, "created_at": "2025-01-16T23:06:49.804843+00:00"} {"global_step": 14713, "acc_step": 0, "speed/wps": 12900.330860080367, "speed/FLOPS": 202617423069811.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.046024736016988754, "optim/lr": 0.002775049916649535, "optim/total_tokens": 7713849344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.877122402191162, "created_at": "2025-01-16T23:06:59.971193+00:00"} {"global_step": 14714, "acc_step": 0, "speed/wps": 12901.072047135967, "speed/FLOPS": 202629064431017.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.062443461269140244, "optim/lr": 0.002775004784559767, "optim/total_tokens": 7714373632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.889991283416748, "created_at": "2025-01-16T23:07:10.136632+00:00"} {"global_step": 14715, "acc_step": 0, "speed/wps": 12897.876814356177, "speed/FLOPS": 202578878909500.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04790592938661575, "optim/lr": 0.0027749596483100736, "optim/total_tokens": 7714897920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.77441668510437, "created_at": "2025-01-16T23:07:20.309425+00:00"} {"global_step": 14716, "acc_step": 0, "speed/wps": 12903.517341743654, "speed/FLOPS": 202667471143018.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06078750267624855, "optim/lr": 0.0027749145079005995, "optim/total_tokens": 7715422208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.720102548599243, "created_at": "2025-01-16T23:07:30.470956+00:00"} {"global_step": 14717, "acc_step": 0, "speed/wps": 12904.37950985905, "speed/FLOPS": 202681012678013.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.046946071088314056, "optim/lr": 0.002774869363331494, "optim/total_tokens": 7715946496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.8640613555908203, "created_at": "2025-01-16T23:07:40.631893+00:00"} {"global_step": 14718, "acc_step": 0, "speed/wps": 12901.256580801632, "speed/FLOPS": 202631962785812.6, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05678706616163254, "optim/lr": 0.0027748242146029035, "optim/total_tokens": 7716470784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.937216281890869, "created_at": "2025-01-16T23:07:50.798590+00:00"} {"global_step": 14719, "acc_step": 0, "speed/wps": 12903.21631820707, "speed/FLOPS": 202662743154726.4, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08170871436595917, "optim/lr": 0.0027747790617149762, "optim/total_tokens": 7716995072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.9437203407287598, "created_at": "2025-01-16T23:08:00.965505+00:00"} {"global_step": 14720, "acc_step": 0, "speed/wps": 12901.904876430726, "speed/FLOPS": 202642145159519.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0022, "optim/grad_norm": 0.041663434356451035, "optim/lr": 0.002774733904667858, "optim/total_tokens": 7717519360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8592872619628906, "created_at": "2025-01-16T23:08:11.130083+00:00"} {"global_step": 14721, "acc_step": 0, "speed/wps": 12901.810225301715, "speed/FLOPS": 202640658533473.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08879515528678894, "optim/lr": 0.002774688743461697, "optim/total_tokens": 7718043648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8086252212524414, "created_at": "2025-01-16T23:08:21.294539+00:00"} {"global_step": 14722, "acc_step": 0, "speed/wps": 12897.294737060502, "speed/FLOPS": 202569736585714.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04659068211913109, "optim/lr": 0.002774643578096641, "optim/total_tokens": 7718567936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 2.9731225967407227, "created_at": "2025-01-16T23:08:31.471309+00:00"} {"global_step": 14723, "acc_step": 0, "speed/wps": 12909.501824636454, "speed/FLOPS": 202761465670390.84, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.08609376847743988, "optim/lr": 0.0027745984085728367, "optim/total_tokens": 7719092224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.7292661666870117, "created_at": "2025-01-16T23:08:41.632021+00:00"} {"global_step": 14724, "acc_step": 0, "speed/wps": 12911.486443543745, "speed/FLOPS": 202792636837482.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05533990263938904, "optim/lr": 0.0027745532348904312, "optim/total_tokens": 7719616512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454731, "loss/out": 2.7526824474334717, "created_at": "2025-01-16T23:08:51.791878+00:00"} {"global_step": 14725, "acc_step": 0, "speed/wps": 12908.85183823666, "speed/FLOPS": 202751256740807.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06992745399475098, "optim/lr": 0.0027745080570495736, "optim/total_tokens": 7720140800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 2.765174150466919, "created_at": "2025-01-16T23:09:01.953461+00:00"} {"global_step": 14726, "acc_step": 0, "speed/wps": 12894.806613737646, "speed/FLOPS": 202530657189889.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.058687206357717514, "optim/lr": 0.0027744628750504085, "optim/total_tokens": 7720665088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.7787983417510986, "created_at": "2025-01-16T23:09:12.123917+00:00"} {"global_step": 14727, "acc_step": 0, "speed/wps": 12897.767573715146, "speed/FLOPS": 202577163135121.2, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05967620760202408, "optim/lr": 0.0027744176888930857, "optim/total_tokens": 7721189376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289738, "loss/out": 2.841201066970825, "created_at": "2025-01-16T23:09:22.291655+00:00"} {"global_step": 14728, "acc_step": 0, "speed/wps": 12909.652730591486, "speed/FLOPS": 202763835855781.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06044944375753403, "optim/lr": 0.0027743724985777513, "optim/total_tokens": 7721713664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.929694414138794, "created_at": "2025-01-16T23:09:32.450473+00:00"} {"global_step": 14729, "acc_step": 0, "speed/wps": 12903.517992426167, "speed/FLOPS": 202667481362881.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06354252249002457, "optim/lr": 0.002774327304104554, "optim/total_tokens": 7722237952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 2.7856967449188232, "created_at": "2025-01-16T23:09:42.614878+00:00"} {"global_step": 14730, "acc_step": 0, "speed/wps": 12911.409556756527, "speed/FLOPS": 202791429224835.22, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.045000266283750534, "optim/lr": 0.0027742821054736397, "optim/total_tokens": 7722762240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.775991439819336, "created_at": "2025-01-16T23:09:52.771345+00:00"} {"global_step": 14731, "acc_step": 0, "speed/wps": 12903.777689788756, "speed/FLOPS": 202671560266822.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05846046283841133, "optim/lr": 0.002774236902685157, "optim/total_tokens": 7723286528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.874997615814209, "created_at": "2025-01-16T23:10:02.934423+00:00"} {"global_step": 14732, "acc_step": 0, "speed/wps": 12914.457874594114, "speed/FLOPS": 202839307245302.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05043996125459671, "optim/lr": 0.002774191695739253, "optim/total_tokens": 7723810816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8056952953338623, "created_at": "2025-01-16T23:10:13.090276+00:00"} {"global_step": 14733, "acc_step": 0, "speed/wps": 12910.22240761904, "speed/FLOPS": 202772783416317.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051436152309179306, "optim/lr": 0.0027741464846360755, "optim/total_tokens": 7724335104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8733067512512207, "created_at": "2025-01-16T23:10:23.247259+00:00"} {"global_step": 14734, "acc_step": 0, "speed/wps": 12916.522343161249, "speed/FLOPS": 202871732561026.12, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05330819636583328, "optim/lr": 0.002774101269375771, "optim/total_tokens": 7724859392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.90037202835083, "created_at": "2025-01-16T23:10:33.400586+00:00"} {"global_step": 14735, "acc_step": 0, "speed/wps": 12908.13292600487, "speed/FLOPS": 202739965236317.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05379185453057289, "optim/lr": 0.0027740560499584883, "optim/total_tokens": 7725383680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.0041580200195312, "created_at": "2025-01-16T23:10:43.558717+00:00"} {"global_step": 14736, "acc_step": 0, "speed/wps": 12917.869860493727, "speed/FLOPS": 202892897172415.56, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0638393685221672, "optim/lr": 0.0027740108263843745, "optim/total_tokens": 7725907968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 2.749277114868164, "created_at": "2025-01-16T23:10:53.710665+00:00"} {"global_step": 14737, "acc_step": 0, "speed/wps": 12914.906828780086, "speed/FLOPS": 202846358687720.56, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05462713539600372, "optim/lr": 0.002773965598653577, "optim/total_tokens": 7726432256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.703486919403076, "created_at": "2025-01-16T23:11:03.865448+00:00"} {"global_step": 14738, "acc_step": 0, "speed/wps": 12916.846227891681, "speed/FLOPS": 202876819615781.84, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05636604130268097, "optim/lr": 0.002773920366766243, "optim/total_tokens": 7726956544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475100, "loss/out": 2.8579418659210205, "created_at": "2025-01-16T23:11:14.019116+00:00"} {"global_step": 14739, "acc_step": 0, "speed/wps": 12915.774753547836, "speed/FLOPS": 202859990638851.7, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06625005602836609, "optim/lr": 0.0027738751307225208, "optim/total_tokens": 7727480832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.805176258087158, "created_at": "2025-01-16T23:11:24.171935+00:00"} {"global_step": 14740, "acc_step": 0, "speed/wps": 12913.554038190608, "speed/FLOPS": 202825111252583.75, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043686266988515854, "optim/lr": 0.0027738298905225583, "optim/total_tokens": 7728005120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.8838047981262207, "created_at": "2025-01-16T23:11:34.325502+00:00"} {"global_step": 14741, "acc_step": 0, "speed/wps": 12909.821053936324, "speed/FLOPS": 202766479605212.3, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.001, "optim/grad_norm": 0.053870975971221924, "optim/lr": 0.002773784646166501, "optim/total_tokens": 7728529408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8180723190307617, "created_at": "2025-01-16T23:11:44.482499+00:00"} {"global_step": 14742, "acc_step": 0, "speed/wps": 12905.852138398326, "speed/FLOPS": 202704142332826.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046712182462215424, "optim/lr": 0.002773739397654499, "optim/total_tokens": 7729053696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 2.8465349674224854, "created_at": "2025-01-16T23:11:54.645023+00:00"} {"global_step": 14743, "acc_step": 0, "speed/wps": 12914.49309959405, "speed/FLOPS": 202839860502330.4, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05595773085951805, "optim/lr": 0.002773694144986699, "optim/total_tokens": 7729577984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.758047103881836, "created_at": "2025-01-16T23:12:04.797728+00:00"} {"global_step": 14744, "acc_step": 0, "speed/wps": 12917.405139172035, "speed/FLOPS": 202885598085464.38, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07121286541223526, "optim/lr": 0.0027736488881632483, "optim/total_tokens": 7730102272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 2.8253331184387207, "created_at": "2025-01-16T23:12:14.950538+00:00"} {"global_step": 14745, "acc_step": 0, "speed/wps": 12913.637088379637, "speed/FLOPS": 202826415669925.38, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06363557279109955, "optim/lr": 0.002773603627184295, "optim/total_tokens": 7730626560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8015823364257812, "created_at": "2025-01-16T23:12:25.105252+00:00"} {"global_step": 14746, "acc_step": 0, "speed/wps": 12915.72154293096, "speed/FLOPS": 202859154892994.62, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04853285104036331, "optim/lr": 0.002773558362049986, "optim/total_tokens": 7731150848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 2.9641976356506348, "created_at": "2025-01-16T23:12:35.259698+00:00"} {"global_step": 14747, "acc_step": 0, "speed/wps": 12913.63200153966, "speed/FLOPS": 202826335774113.47, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06197504326701164, "optim/lr": 0.0027735130927604703, "optim/total_tokens": 7731675136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.742586612701416, "created_at": "2025-01-16T23:12:45.415463+00:00"} {"global_step": 14748, "acc_step": 0, "speed/wps": 12916.888391680015, "speed/FLOPS": 202877481856016.72, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05176176503300667, "optim/lr": 0.0027734678193158946, "optim/total_tokens": 7732199424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 2.7431135177612305, "created_at": "2025-01-16T23:12:55.566572+00:00"} {"global_step": 14749, "acc_step": 0, "speed/wps": 12913.714679478639, "speed/FLOPS": 202827634344756.53, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04679770767688751, "optim/lr": 0.0027734225417164068, "optim/total_tokens": 7732723712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499005, "loss/out": 2.8778011798858643, "created_at": "2025-01-16T23:13:05.720085+00:00"} {"global_step": 14750, "acc_step": 0, "speed/wps": 12908.51671105921, "speed/FLOPS": 202745993107972.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05134105682373047, "optim/lr": 0.002773377259962155, "optim/total_tokens": 7733248000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.811666250228882, "created_at": "2025-01-16T23:13:15.881214+00:00"} {"global_step": 14751, "acc_step": 0, "speed/wps": 12911.806853710306, "speed/FLOPS": 202797669319437.53, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04902593418955803, "optim/lr": 0.0027733319740532864, "optim/total_tokens": 7733772288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.791597843170166, "created_at": "2025-01-16T23:13:26.036078+00:00"} {"global_step": 14752, "acc_step": 0, "speed/wps": 12915.706393655037, "speed/FLOPS": 202858916952798.97, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05492691695690155, "optim/lr": 0.002773286683989949, "optim/total_tokens": 7734296576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 2.681210517883301, "created_at": "2025-01-16T23:13:36.188395+00:00"} {"global_step": 14753, "acc_step": 0, "speed/wps": 12908.246481306074, "speed/FLOPS": 202741748778364.72, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04784693568944931, "optim/lr": 0.0027732413897722913, "optim/total_tokens": 7734820864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 2.8025403022766113, "created_at": "2025-01-16T23:13:46.350362+00:00"} {"global_step": 14754, "acc_step": 0, "speed/wps": 12914.962964766313, "speed/FLOPS": 202847240380523.22, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04785282909870148, "optim/lr": 0.0027731960914004596, "optim/total_tokens": 7735345152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435028, "loss/out": 2.831130027770996, "created_at": "2025-01-16T23:13:56.505411+00:00"} {"global_step": 14755, "acc_step": 0, "speed/wps": 12916.149851526718, "speed/FLOPS": 202865882068048.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0463881678879261, "optim/lr": 0.002773150788874603, "optim/total_tokens": 7735869440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.905802011489868, "created_at": "2025-01-16T23:14:06.659691+00:00"} {"global_step": 14756, "acc_step": 0, "speed/wps": 12913.799626461137, "speed/FLOPS": 202828968553846.06, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06623681634664536, "optim/lr": 0.0027731054821948683, "optim/total_tokens": 7736393728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.850985527038574, "created_at": "2025-01-16T23:14:16.815055+00:00"} {"global_step": 14757, "acc_step": 0, "speed/wps": 12916.0511898289, "speed/FLOPS": 202864332450509.1, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04931730031967163, "optim/lr": 0.002773060171361404, "optim/total_tokens": 7736918016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 2.8246471881866455, "created_at": "2025-01-16T23:14:26.969513+00:00"} {"global_step": 14758, "acc_step": 0, "speed/wps": 12912.964417303068, "speed/FLOPS": 202815850446320.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05971968546509743, "optim/lr": 0.002773014856374358, "optim/total_tokens": 7737442304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424340, "loss/out": 2.9197957515716553, "created_at": "2025-01-16T23:14:37.129069+00:00"} {"global_step": 14759, "acc_step": 0, "speed/wps": 12915.289334924426, "speed/FLOPS": 202852366472335.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06918389350175858, "optim/lr": 0.0027729695372338773, "optim/total_tokens": 7737966592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9568982124328613, "created_at": "2025-01-16T23:14:47.281114+00:00"} {"global_step": 14760, "acc_step": 0, "speed/wps": 12921.20267449383, "speed/FLOPS": 202945243596053.53, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07366077601909637, "optim/lr": 0.0027729242139401104, "optim/total_tokens": 7738490880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416344, "loss/out": 2.838862895965576, "created_at": "2025-01-16T23:14:57.436399+00:00"} {"global_step": 14761, "acc_step": 0, "speed/wps": 12914.896987314725, "speed/FLOPS": 202846204113983.5, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06503912806510925, "optim/lr": 0.002772878886493206, "optim/total_tokens": 7739015168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.839918613433838, "created_at": "2025-01-16T23:15:07.589724+00:00"} {"global_step": 14762, "acc_step": 0, "speed/wps": 12913.957209515402, "speed/FLOPS": 202831443612255.1, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06679795682430267, "optim/lr": 0.00277283355489331, "optim/total_tokens": 7739539456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 2.735288619995117, "created_at": "2025-01-16T23:15:17.744188+00:00"} {"global_step": 14763, "acc_step": 0, "speed/wps": 12915.943592177973, "speed/FLOPS": 202862642481546.6, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07112551480531693, "optim/lr": 0.002772788219140572, "optim/total_tokens": 7740063744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.8113322257995605, "created_at": "2025-01-16T23:15:27.898935+00:00"} {"global_step": 14764, "acc_step": 0, "speed/wps": 12911.014433999395, "speed/FLOPS": 202785223278978.88, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05177432298660278, "optim/lr": 0.002772742879235139, "optim/total_tokens": 7740588032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.847259521484375, "created_at": "2025-01-16T23:15:38.057576+00:00"} {"global_step": 14765, "acc_step": 0, "speed/wps": 12911.958549201026, "speed/FLOPS": 202800051905570.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0654895231127739, "optim/lr": 0.0027726975351771593, "optim/total_tokens": 7741112320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.799502372741699, "created_at": "2025-01-16T23:15:48.213808+00:00"} {"global_step": 14766, "acc_step": 0, "speed/wps": 12915.653261830259, "speed/FLOPS": 202858082444480.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06571101397275925, "optim/lr": 0.002772652186966781, "optim/total_tokens": 7741636608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469813, "loss/out": 2.727454662322998, "created_at": "2025-01-16T23:15:58.367419+00:00"} {"global_step": 14767, "acc_step": 0, "speed/wps": 12913.938877990753, "speed/FLOPS": 202831155690470.34, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.045311927795410156, "optim/lr": 0.002772606834604152, "optim/total_tokens": 7742160896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.854841709136963, "created_at": "2025-01-16T23:16:08.524166+00:00"} {"global_step": 14768, "acc_step": 0, "speed/wps": 12913.166515176299, "speed/FLOPS": 202819024671131.6, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05396387726068497, "optim/lr": 0.0027725614780894203, "optim/total_tokens": 7742685184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.7974483966827393, "created_at": "2025-01-16T23:16:18.678319+00:00"} {"global_step": 14769, "acc_step": 0, "speed/wps": 12909.56764750544, "speed/FLOPS": 202762499508998.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05053563788533211, "optim/lr": 0.002772516117422733, "optim/total_tokens": 7743209472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332063, "loss/out": 2.9462945461273193, "created_at": "2025-01-16T23:16:28.838274+00:00"} {"global_step": 14770, "acc_step": 0, "speed/wps": 12917.777627330926, "speed/FLOPS": 202891448523852.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04971303790807724, "optim/lr": 0.002772470752604239, "optim/total_tokens": 7743733760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.894199848175049, "created_at": "2025-01-16T23:16:38.992585+00:00"} {"global_step": 14771, "acc_step": 0, "speed/wps": 12917.719400259808, "speed/FLOPS": 202890533987687.84, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05458119884133339, "optim/lr": 0.002772425383634086, "optim/total_tokens": 7744258048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 2.6513850688934326, "created_at": "2025-01-16T23:16:49.146122+00:00"} {"global_step": 14772, "acc_step": 0, "speed/wps": 12914.495787339118, "speed/FLOPS": 202839902717060.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050651442259550095, "optim/lr": 0.002772380010512422, "optim/total_tokens": 7744782336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.746469020843506, "created_at": "2025-01-16T23:16:59.301991+00:00"} {"global_step": 14773, "acc_step": 0, "speed/wps": 12918.554178763086, "speed/FLOPS": 202903645331187.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.067364402115345, "optim/lr": 0.0027723346332393962, "optim/total_tokens": 7745306624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.866791009902954, "created_at": "2025-01-16T23:17:09.451858+00:00"} {"global_step": 14774, "acc_step": 0, "speed/wps": 12916.29822089455, "speed/FLOPS": 202868212412851.38, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07375500351190567, "optim/lr": 0.0027722892518151545, "optim/total_tokens": 7745830912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.915968894958496, "created_at": "2025-01-16T23:17:19.604314+00:00"} {"global_step": 14775, "acc_step": 0, "speed/wps": 12918.564970901403, "speed/FLOPS": 202903814836549.9, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04424043372273445, "optim/lr": 0.0027722438662398465, "optim/total_tokens": 7746355200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.763869285583496, "created_at": "2025-01-16T23:17:29.755370+00:00"} {"global_step": 14776, "acc_step": 0, "speed/wps": 12917.246848346023, "speed/FLOPS": 202883111910527.47, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04513822868466377, "optim/lr": 0.0027721984765136203, "optim/total_tokens": 7746879488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 2.7301371097564697, "created_at": "2025-01-16T23:17:39.906079+00:00"} {"global_step": 14777, "acc_step": 0, "speed/wps": 12913.19046298616, "speed/FLOPS": 202819400804393.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047075897455215454, "optim/lr": 0.002772153082636623, "optim/total_tokens": 7747403776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 2.889059543609619, "created_at": "2025-01-16T23:17:50.062099+00:00"} {"global_step": 14778, "acc_step": 0, "speed/wps": 12920.270323253078, "speed/FLOPS": 202930599738632.7, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04518445208668709, "optim/lr": 0.0027721076846090034, "optim/total_tokens": 7747928064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469093, "loss/out": 2.8538670539855957, "created_at": "2025-01-16T23:18:00.211662+00:00"} {"global_step": 14779, "acc_step": 0, "speed/wps": 12921.417624534024, "speed/FLOPS": 202948619681807.1, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04821427911520004, "optim/lr": 0.0027720622824309094, "optim/total_tokens": 7748452352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.8292009830474854, "created_at": "2025-01-16T23:18:10.362951+00:00"} {"global_step": 14780, "acc_step": 0, "speed/wps": 12917.924580426956, "speed/FLOPS": 202893756624160.0, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.044880758970975876, "optim/lr": 0.0027720168761024896, "optim/total_tokens": 7748976640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449950, "loss/out": 2.81980037689209, "created_at": "2025-01-16T23:18:20.513885+00:00"} {"global_step": 14781, "acc_step": 0, "speed/wps": 12914.374691787267, "speed/FLOPS": 202838000745015.5, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05384274199604988, "optim/lr": 0.002771971465623892, "optim/total_tokens": 7749500928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8374128341674805, "created_at": "2025-01-16T23:18:30.667541+00:00"} {"global_step": 14782, "acc_step": 0, "speed/wps": 12914.96726215344, "speed/FLOPS": 202847307876893.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04568413645029068, "optim/lr": 0.0027719260509952636, "optim/total_tokens": 7750025216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.91937255859375, "created_at": "2025-01-16T23:18:40.821807+00:00"} {"global_step": 14783, "acc_step": 0, "speed/wps": 12914.380964264217, "speed/FLOPS": 202838099262885.28, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0557267889380455, "optim/lr": 0.002771880632216754, "optim/total_tokens": 7750549504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.799221992492676, "created_at": "2025-01-16T23:18:50.975346+00:00"} {"global_step": 14784, "acc_step": 0, "speed/wps": 12920.171804979935, "speed/FLOPS": 202929052373774.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054160911589860916, "optim/lr": 0.002771835209288511, "optim/total_tokens": 7751073792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 2.714426040649414, "created_at": "2025-01-16T23:19:01.123557+00:00"} {"global_step": 14785, "acc_step": 0, "speed/wps": 12915.573531725144, "speed/FLOPS": 202856830173618.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04929979890584946, "optim/lr": 0.0027717897822106823, "optim/total_tokens": 7751598080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.777996778488159, "created_at": "2025-01-16T23:19:11.279462+00:00"} {"global_step": 14786, "acc_step": 0, "speed/wps": 12918.191824301663, "speed/FLOPS": 202897954056443.8, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04996546357870102, "optim/lr": 0.002771744350983417, "optim/total_tokens": 7752122368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.9045634269714355, "created_at": "2025-01-16T23:19:21.434048+00:00"} {"global_step": 14787, "acc_step": 0, "speed/wps": 12913.194008008903, "speed/FLOPS": 202819456483846.75, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.001, "optim/grad_norm": 0.054145000874996185, "optim/lr": 0.0027716989156068624, "optim/total_tokens": 7752646656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.94377064704895, "created_at": "2025-01-16T23:19:31.593611+00:00"} {"global_step": 14788, "acc_step": 0, "speed/wps": 12912.610190800066, "speed/FLOPS": 202810286832333.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05509565398097038, "optim/lr": 0.0027716534760811673, "optim/total_tokens": 7753170944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.981722354888916, "created_at": "2025-01-16T23:19:41.755609+00:00"} {"global_step": 14789, "acc_step": 0, "speed/wps": 12913.314212919939, "speed/FLOPS": 202821344467153.44, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04750838503241539, "optim/lr": 0.0027716080324064803, "optim/total_tokens": 7753695232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 2.8226706981658936, "created_at": "2025-01-16T23:19:51.911929+00:00"} {"global_step": 14790, "acc_step": 0, "speed/wps": 12911.230860996693, "speed/FLOPS": 202788622562373.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047375455498695374, "optim/lr": 0.0027715625845829484, "optim/total_tokens": 7754219520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.8181681632995605, "created_at": "2025-01-16T23:20:02.067833+00:00"} {"global_step": 14791, "acc_step": 0, "speed/wps": 12916.67407004685, "speed/FLOPS": 202874115640259.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061058834195137024, "optim/lr": 0.002771517132610721, "optim/total_tokens": 7754743808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486049, "loss/out": 2.8258252143859863, "created_at": "2025-01-16T23:20:12.220040+00:00"} {"global_step": 14792, "acc_step": 0, "speed/wps": 12909.25150942482, "speed/FLOPS": 202757534126022.94, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06675460189580917, "optim/lr": 0.002771471676489946, "optim/total_tokens": 7755268096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 2.677724838256836, "created_at": "2025-01-16T23:20:22.378055+00:00"} {"global_step": 14793, "acc_step": 0, "speed/wps": 12912.367577486326, "speed/FLOPS": 202806476256855.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04707767069339752, "optim/lr": 0.0027714262162207718, "optim/total_tokens": 7755792384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 3.0395689010620117, "created_at": "2025-01-16T23:20:32.558729+00:00"} {"global_step": 14794, "acc_step": 0, "speed/wps": 12908.673575573146, "speed/FLOPS": 202748456880718.1, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0566142238676548, "optim/lr": 0.0027713807518033465, "optim/total_tokens": 7756316672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8503034114837646, "created_at": "2025-01-16T23:20:42.718822+00:00"} {"global_step": 14795, "acc_step": 0, "speed/wps": 12913.367453467396, "speed/FLOPS": 202822180683111.44, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045943733304739, "optim/lr": 0.002771335283237819, "optim/total_tokens": 7756840960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.792613983154297, "created_at": "2025-01-16T23:20:52.875526+00:00"} {"global_step": 14796, "acc_step": 0, "speed/wps": 12912.053677431864, "speed/FLOPS": 202801546025155.94, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05094888061285019, "optim/lr": 0.0027712898105243374, "optim/total_tokens": 7757365248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305632, "loss/out": 2.706545829772949, "created_at": "2025-01-16T23:21:03.032349+00:00"} {"global_step": 14797, "acc_step": 0, "speed/wps": 12913.920044091967, "speed/FLOPS": 202830859878209.4, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050750814378261566, "optim/lr": 0.0027712443336630494, "optim/total_tokens": 7757889536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470861, "loss/out": 2.7857182025909424, "created_at": "2025-01-16T23:21:13.191970+00:00"} {"global_step": 14798, "acc_step": 0, "speed/wps": 12909.581629001113, "speed/FLOPS": 202762719107600.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04869544506072998, "optim/lr": 0.0027711988526541048, "optim/total_tokens": 7758413824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 2.9096455574035645, "created_at": "2025-01-16T23:21:23.348580+00:00"} {"global_step": 14799, "acc_step": 0, "speed/wps": 12911.564628312535, "speed/FLOPS": 202793864836712.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06901823729276657, "optim/lr": 0.0027711533674976505, "optim/total_tokens": 7758938112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.7970969676971436, "created_at": "2025-01-16T23:21:33.503729+00:00"} {"global_step": 14800, "acc_step": 0, "speed/wps": 12910.455144712645, "speed/FLOPS": 202776438872186.47, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06893283873796463, "optim/lr": 0.0027711078781938354, "optim/total_tokens": 7759462400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 2.8603339195251465, "created_at": "2025-01-16T23:21:43.660959+00:00"} {"global_step": 14801, "acc_step": 0, "speed/wps": 12910.044833049547, "speed/FLOPS": 202769994363690.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.049274690449237823, "optim/lr": 0.0027710623847428083, "optim/total_tokens": 7759986688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.7198572158813477, "created_at": "2025-01-16T23:21:53.821041+00:00"} {"global_step": 14802, "acc_step": 0, "speed/wps": 12916.303685608584, "speed/FLOPS": 202868298243694.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05186215415596962, "optim/lr": 0.0027710168871447173, "optim/total_tokens": 7760510976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.7511703968048096, "created_at": "2025-01-16T23:22:03.976375+00:00"} {"global_step": 14803, "acc_step": 0, "speed/wps": 12917.318597552312, "speed/FLOPS": 202884238830405.97, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058670781552791595, "optim/lr": 0.0027709713853997105, "optim/total_tokens": 7761035264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.798781156539917, "created_at": "2025-01-16T23:22:14.126890+00:00"} {"global_step": 14804, "acc_step": 0, "speed/wps": 12920.488543392887, "speed/FLOPS": 202934027185794.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053081922233104706, "optim/lr": 0.0027709258795079374, "optim/total_tokens": 7761559552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 2.832021713256836, "created_at": "2025-01-16T23:22:24.276068+00:00"} {"global_step": 14805, "acc_step": 0, "speed/wps": 12915.92407916091, "speed/FLOPS": 202862336002799.94, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05649399384856224, "optim/lr": 0.0027708803694695455, "optim/total_tokens": 7762083840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.7729082107543945, "created_at": "2025-01-16T23:22:34.431284+00:00"} {"global_step": 14806, "acc_step": 0, "speed/wps": 12911.865824720779, "speed/FLOPS": 202798595540191.7, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05823206529021263, "optim/lr": 0.002770834855284684, "optim/total_tokens": 7762608128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404090, "loss/out": 2.767302989959717, "created_at": "2025-01-16T23:22:44.587348+00:00"} {"global_step": 14807, "acc_step": 0, "speed/wps": 12913.117910115725, "speed/FLOPS": 202818261261864.94, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05051194131374359, "optim/lr": 0.0027707893369535008, "optim/total_tokens": 7763132416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.8961355686187744, "created_at": "2025-01-16T23:22:54.741404+00:00"} {"global_step": 14808, "acc_step": 0, "speed/wps": 12906.886642405963, "speed/FLOPS": 202720390639824.0, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05068032070994377, "optim/lr": 0.002770743814476144, "optim/total_tokens": 7763656704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.815321445465088, "created_at": "2025-01-16T23:23:04.900624+00:00"} {"global_step": 14809, "acc_step": 0, "speed/wps": 12910.9518689351, "speed/FLOPS": 202784240608670.75, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04166281223297119, "optim/lr": 0.0027706982878527636, "optim/total_tokens": 7764180992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.9118144512176514, "created_at": "2025-01-16T23:23:15.058853+00:00"} {"global_step": 14810, "acc_step": 0, "speed/wps": 12914.248693674923, "speed/FLOPS": 202836021771521.6, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06395421922206879, "optim/lr": 0.002770652757083507, "optim/total_tokens": 7764705280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.752833843231201, "created_at": "2025-01-16T23:23:25.212874+00:00"} {"global_step": 14811, "acc_step": 0, "speed/wps": 12915.193386259905, "speed/FLOPS": 202850859466712.38, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04445203021168709, "optim/lr": 0.002770607222168523, "optim/total_tokens": 7765229568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8399806022644043, "created_at": "2025-01-16T23:23:35.366105+00:00"} {"global_step": 14812, "acc_step": 0, "speed/wps": 12912.035046863268, "speed/FLOPS": 202801253406475.06, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06640654802322388, "optim/lr": 0.0027705616831079608, "optim/total_tokens": 7765753856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 2.904690742492676, "created_at": "2025-01-16T23:23:45.520915+00:00"} {"global_step": 14813, "acc_step": 0, "speed/wps": 12917.040076187313, "speed/FLOPS": 202879864269640.22, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04571118950843811, "optim/lr": 0.0027705161399019674, "optim/total_tokens": 7766278144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8942153453826904, "created_at": "2025-01-16T23:23:55.682456+00:00"} {"global_step": 14814, "acc_step": 0, "speed/wps": 12913.642983336638, "speed/FLOPS": 202826508258325.6, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04326443746685982, "optim/lr": 0.0027704705925506935, "optim/total_tokens": 7766802432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7565441131591797, "created_at": "2025-01-16T23:24:05.836146+00:00"} {"global_step": 14815, "acc_step": 0, "speed/wps": 12909.760294506355, "speed/FLOPS": 202765525294872.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045742470771074295, "optim/lr": 0.0027704250410542854, "optim/total_tokens": 7767326720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8268280029296875, "created_at": "2025-01-16T23:24:15.992894+00:00"} {"global_step": 14816, "acc_step": 0, "speed/wps": 12907.501499550566, "speed/FLOPS": 202730047816182.25, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04892192780971527, "optim/lr": 0.0027703794854128934, "optim/total_tokens": 7767851008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8929529190063477, "created_at": "2025-01-16T23:24:26.158498+00:00"} {"global_step": 14817, "acc_step": 0, "speed/wps": 12910.818324286392, "speed/FLOPS": 202782143106453.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04534683749079704, "optim/lr": 0.002770333925626666, "optim/total_tokens": 7768375296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 2.7092366218566895, "created_at": "2025-01-16T23:24:36.317092+00:00"} {"global_step": 14818, "acc_step": 0, "speed/wps": 12903.385345646124, "speed/FLOPS": 202665397962924.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055079665035009384, "optim/lr": 0.0027702883616957516, "optim/total_tokens": 7768899584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8507261276245117, "created_at": "2025-01-16T23:24:46.479412+00:00"} {"global_step": 14819, "acc_step": 0, "speed/wps": 12909.487935677664, "speed/FLOPS": 202761247525207.6, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05058162659406662, "optim/lr": 0.0027702427936202975, "optim/total_tokens": 7769423872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 2.7940850257873535, "created_at": "2025-01-16T23:24:56.638488+00:00"} {"global_step": 14820, "acc_step": 0, "speed/wps": 12910.27679039354, "speed/FLOPS": 202773637572520.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05009503290057182, "optim/lr": 0.002770197221400455, "optim/total_tokens": 7769948160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.7534666061401367, "created_at": "2025-01-16T23:25:06.796846+00:00"} {"global_step": 14821, "acc_step": 0, "speed/wps": 12911.42839636295, "speed/FLOPS": 202791725126742.4, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05600018799304962, "optim/lr": 0.002770151645036371, "optim/total_tokens": 7770472448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.9116010665893555, "created_at": "2025-01-16T23:25:16.954387+00:00"} {"global_step": 14822, "acc_step": 0, "speed/wps": 12914.996605585586, "speed/FLOPS": 202847768755817.22, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11231230199337006, "optim/lr": 0.002770106064528195, "optim/total_tokens": 7770996736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8061532974243164, "created_at": "2025-01-16T23:25:27.108214+00:00"} {"global_step": 14823, "acc_step": 0, "speed/wps": 12906.946387593549, "speed/FLOPS": 202721329020093.53, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.08879311382770538, "optim/lr": 0.002770060479876074, "optim/total_tokens": 7771521024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.9086263179779053, "created_at": "2025-01-16T23:25:37.273047+00:00"} {"global_step": 14824, "acc_step": 0, "speed/wps": 12911.860944210133, "speed/FLOPS": 202798518885065.47, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07297135144472122, "optim/lr": 0.002770014891080159, "optim/total_tokens": 7772045312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 2.681582450866699, "created_at": "2025-01-16T23:25:47.428645+00:00"} {"global_step": 14825, "acc_step": 0, "speed/wps": 12918.106838497539, "speed/FLOPS": 202896619237607.3, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0605006106197834, "optim/lr": 0.0027699692981405983, "optim/total_tokens": 7772569600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.972464084625244, "created_at": "2025-01-16T23:25:57.578754+00:00"} {"global_step": 14826, "acc_step": 0, "speed/wps": 12918.325769359324, "speed/FLOPS": 202900057847634.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06481046974658966, "optim/lr": 0.0027699237010575394, "optim/total_tokens": 7773093888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.86899995803833, "created_at": "2025-01-16T23:26:07.728846+00:00"} {"global_step": 14827, "acc_step": 0, "speed/wps": 12913.985209331167, "speed/FLOPS": 202831883387837.47, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05452343076467514, "optim/lr": 0.002769878099831132, "optim/total_tokens": 7773618176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.972280979156494, "created_at": "2025-01-16T23:26:17.887506+00:00"} {"global_step": 14828, "acc_step": 0, "speed/wps": 12908.599981810014, "speed/FLOPS": 202747300989539.66, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054738521575927734, "optim/lr": 0.002769832494461525, "optim/total_tokens": 7774142464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.837048292160034, "created_at": "2025-01-16T23:26:28.044834+00:00"} {"global_step": 14829, "acc_step": 0, "speed/wps": 12913.825600513645, "speed/FLOPS": 202829376512032.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06491895020008087, "optim/lr": 0.0027697868849488666, "optim/total_tokens": 7774666752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.768707752227783, "created_at": "2025-01-16T23:26:38.201518+00:00"} {"global_step": 14830, "acc_step": 0, "speed/wps": 12903.398621676866, "speed/FLOPS": 202665606481230.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0487988106906414, "optim/lr": 0.0027697412712933056, "optim/total_tokens": 7775191040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341575, "loss/out": 2.8570196628570557, "created_at": "2025-01-16T23:26:48.362952+00:00"} {"global_step": 14831, "acc_step": 0, "speed/wps": 12914.87530864842, "speed/FLOPS": 202845863620739.56, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15328533947467804, "optim/lr": 0.0027696956534949916, "optim/total_tokens": 7775715328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9196767807006836, "created_at": "2025-01-16T23:26:58.520466+00:00"} {"global_step": 14832, "acc_step": 0, "speed/wps": 12915.82407083624, "speed/FLOPS": 202860765234635.12, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05627596005797386, "optim/lr": 0.002769650031554073, "optim/total_tokens": 7776239616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 2.9406251907348633, "created_at": "2025-01-16T23:27:08.674786+00:00"} {"global_step": 14833, "acc_step": 0, "speed/wps": 12911.091975818537, "speed/FLOPS": 202786441179802.16, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052600082010030746, "optim/lr": 0.0027696044054706984, "optim/total_tokens": 7776763904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.837881565093994, "created_at": "2025-01-16T23:27:18.834751+00:00"} {"global_step": 14834, "acc_step": 0, "speed/wps": 12916.557395679065, "speed/FLOPS": 202872283108985.66, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06416105479001999, "optim/lr": 0.002769558775245017, "optim/total_tokens": 7777288192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.9034643173217773, "created_at": "2025-01-16T23:27:28.986881+00:00"} {"global_step": 14835, "acc_step": 0, "speed/wps": 12909.0889591145, "speed/FLOPS": 202754981050031.88, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06499063223600388, "optim/lr": 0.0027695131408771775, "optim/total_tokens": 7777812480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.6558077335357666, "created_at": "2025-01-16T23:27:39.144037+00:00"} {"global_step": 14836, "acc_step": 0, "speed/wps": 12910.154574276921, "speed/FLOPS": 202771718000466.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.11322364211082458, "optim/lr": 0.0027694675023673286, "optim/total_tokens": 7778336768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444591, "loss/out": 2.8868234157562256, "created_at": "2025-01-16T23:27:49.300509+00:00"} {"global_step": 14837, "acc_step": 0, "speed/wps": 12910.619728633194, "speed/FLOPS": 202779023888820.7, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07342557609081268, "optim/lr": 0.0027694218597156196, "optim/total_tokens": 7778861056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.943772554397583, "created_at": "2025-01-16T23:27:59.457253+00:00"} {"global_step": 14838, "acc_step": 0, "speed/wps": 12906.009839997929, "speed/FLOPS": 202706619253153.03, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06872781366109848, "optim/lr": 0.0027693762129221994, "optim/total_tokens": 7779385344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 2.823798179626465, "created_at": "2025-01-16T23:28:09.618281+00:00"} {"global_step": 14839, "acc_step": 0, "speed/wps": 12908.175011617102, "speed/FLOPS": 202740626248689.75, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05444209277629852, "optim/lr": 0.002769330561987217, "optim/total_tokens": 7779909632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.807523250579834, "created_at": "2025-01-16T23:28:19.778616+00:00"} {"global_step": 14840, "acc_step": 0, "speed/wps": 12916.927437826995, "speed/FLOPS": 202878095129410.06, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05537387728691101, "optim/lr": 0.002769284906910821, "optim/total_tokens": 7780433920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424340, "loss/out": 2.8684635162353516, "created_at": "2025-01-16T23:28:29.929416+00:00"} {"global_step": 14841, "acc_step": 0, "speed/wps": 12914.304021669315, "speed/FLOPS": 202836890773702.16, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05093536153435707, "optim/lr": 0.0027692392476931604, "optim/total_tokens": 7780958208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8200273513793945, "created_at": "2025-01-16T23:28:40.083456+00:00"} {"global_step": 14842, "acc_step": 0, "speed/wps": 12914.918821308547, "speed/FLOPS": 202846547046859.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05086430162191391, "optim/lr": 0.0027691935843343844, "optim/total_tokens": 7781482496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425994, "loss/out": 2.908578872680664, "created_at": "2025-01-16T23:28:50.238269+00:00"} {"global_step": 14843, "acc_step": 0, "speed/wps": 12916.239237839418, "speed/FLOPS": 202867286002919.3, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04707227274775505, "optim/lr": 0.0027691479168346416, "optim/total_tokens": 7782006784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.7660210132598877, "created_at": "2025-01-16T23:29:00.390215+00:00"} {"global_step": 14844, "acc_step": 0, "speed/wps": 12918.3588976551, "speed/FLOPS": 202900578173042.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05313963443040848, "optim/lr": 0.002769102245194082, "optim/total_tokens": 7782531072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 2.829921245574951, "created_at": "2025-01-16T23:29:10.540485+00:00"} {"global_step": 14845, "acc_step": 0, "speed/wps": 12912.775475363795, "speed/FLOPS": 202812882853531.22, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04621998965740204, "optim/lr": 0.002769056569412853, "optim/total_tokens": 7783055360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.803354263305664, "created_at": "2025-01-16T23:29:20.694774+00:00"} {"global_step": 14846, "acc_step": 0, "speed/wps": 12913.88858328504, "speed/FLOPS": 202830365743004.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04661937803030014, "optim/lr": 0.002769010889491105, "optim/total_tokens": 7783579648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 2.8125572204589844, "created_at": "2025-01-16T23:29:30.850825+00:00"} {"global_step": 14847, "acc_step": 0, "speed/wps": 12914.526342149282, "speed/FLOPS": 202840382622340.38, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08738944679498672, "optim/lr": 0.0027689652054289863, "optim/total_tokens": 7784103936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.949948310852051, "created_at": "2025-01-16T23:29:41.007536+00:00"} {"global_step": 14848, "acc_step": 0, "speed/wps": 12917.613466728606, "speed/FLOPS": 202888870156012.38, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04688497632741928, "optim/lr": 0.0027689195172266464, "optim/total_tokens": 7784628224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484186, "loss/out": 2.8586902618408203, "created_at": "2025-01-16T23:29:51.159854+00:00"} {"global_step": 14849, "acc_step": 0, "speed/wps": 12915.861889665272, "speed/FLOPS": 202861359231313.6, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0775473341345787, "optim/lr": 0.0027688738248842337, "optim/total_tokens": 7785152512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.7045459747314453, "created_at": "2025-01-16T23:30:01.312460+00:00"} {"global_step": 14850, "acc_step": 0, "speed/wps": 12918.833765115483, "speed/FLOPS": 202908036619044.0, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05312461405992508, "optim/lr": 0.002768828128401898, "optim/total_tokens": 7785676800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 2.79223895072937, "created_at": "2025-01-16T23:30:11.461814+00:00"} {"global_step": 14851, "acc_step": 0, "speed/wps": 12911.138382674515, "speed/FLOPS": 202787170063244.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.08138968050479889, "optim/lr": 0.0027687824277797886, "optim/total_tokens": 7786201088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.794910430908203, "created_at": "2025-01-16T23:30:21.623908+00:00"} {"global_step": 14852, "acc_step": 0, "speed/wps": 12917.823898479437, "speed/FLOPS": 202892175275822.72, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0529143325984478, "optim/lr": 0.0027687367230180534, "optim/total_tokens": 7786725376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381371, "loss/out": 2.8862617015838623, "created_at": "2025-01-16T23:30:31.774386+00:00"} {"global_step": 14853, "acc_step": 0, "speed/wps": 12914.529822726172, "speed/FLOPS": 202840437289583.25, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06643415242433548, "optim/lr": 0.0027686910141168425, "optim/total_tokens": 7787249664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.7734832763671875, "created_at": "2025-01-16T23:30:41.929979+00:00"} {"global_step": 14854, "acc_step": 0, "speed/wps": 12906.297562738133, "speed/FLOPS": 202711138334160.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07113224267959595, "optim/lr": 0.0027686453010763046, "optim/total_tokens": 7787773952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.7773666381835938, "created_at": "2025-01-16T23:30:52.091419+00:00"} {"global_step": 14855, "acc_step": 0, "speed/wps": 12912.769115589597, "speed/FLOPS": 202812782964538.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06872185319662094, "optim/lr": 0.0027685995838965896, "optim/total_tokens": 7788298240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.760329246520996, "created_at": "2025-01-16T23:31:02.251245+00:00"} {"global_step": 14856, "acc_step": 0, "speed/wps": 12916.819761559422, "speed/FLOPS": 202876403925665.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05016707628965378, "optim/lr": 0.0027685538625778455, "optim/total_tokens": 7788822528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.9021763801574707, "created_at": "2025-01-16T23:31:12.404117+00:00"} {"global_step": 14857, "acc_step": 0, "speed/wps": 12911.292829415377, "speed/FLOPS": 202789595861542.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.064344123005867, "optim/lr": 0.002768508137120223, "optim/total_tokens": 7789346816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 2.715944766998291, "created_at": "2025-01-16T23:31:22.560981+00:00"} {"global_step": 14858, "acc_step": 0, "speed/wps": 12917.484913233684, "speed/FLOPS": 202886851046723.62, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05752841383218765, "optim/lr": 0.0027684624075238694, "optim/total_tokens": 7789871104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.822211742401123, "created_at": "2025-01-16T23:31:32.711453+00:00"} {"global_step": 14859, "acc_step": 0, "speed/wps": 12914.747012302018, "speed/FLOPS": 202843848550321.66, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05973008647561073, "optim/lr": 0.002768416673788935, "optim/total_tokens": 7790395392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.86877703666687, "created_at": "2025-01-16T23:31:42.865714+00:00"} {"global_step": 14860, "acc_step": 0, "speed/wps": 12914.431199997045, "speed/FLOPS": 202838888284100.47, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06128702685236931, "optim/lr": 0.002768370935915569, "optim/total_tokens": 7790919680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.789717435836792, "created_at": "2025-01-16T23:31:53.020624+00:00"} {"global_step": 14861, "acc_step": 0, "speed/wps": 12913.586183428111, "speed/FLOPS": 202825616137710.94, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05145882070064545, "optim/lr": 0.0027683251939039206, "optim/total_tokens": 7791443968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 2.7793281078338623, "created_at": "2025-01-16T23:32:03.176564+00:00"} {"global_step": 14862, "acc_step": 0, "speed/wps": 12909.669247616175, "speed/FLOPS": 202764095278350.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055033087730407715, "optim/lr": 0.0027682794477541387, "optim/total_tokens": 7791968256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472483, "loss/out": 2.874155044555664, "created_at": "2025-01-16T23:32:13.335231+00:00"} {"global_step": 14863, "acc_step": 0, "speed/wps": 12909.775224968926, "speed/FLOPS": 202765759798303.72, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05062413588166237, "optim/lr": 0.002768233697466373, "optim/total_tokens": 7792492544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 2.841667890548706, "created_at": "2025-01-16T23:32:23.493632+00:00"} {"global_step": 14864, "acc_step": 0, "speed/wps": 12913.43069955227, "speed/FLOPS": 202823174049783.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049236323684453964, "optim/lr": 0.0027681879430407727, "optim/total_tokens": 7793016832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501678, "loss/out": 2.8769350051879883, "created_at": "2025-01-16T23:32:33.650841+00:00"} {"global_step": 14865, "acc_step": 0, "speed/wps": 12908.978990901986, "speed/FLOPS": 202753253848142.16, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055917367339134216, "optim/lr": 0.0027681421844774866, "optim/total_tokens": 7793541120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.8688251972198486, "created_at": "2025-01-16T23:32:43.807948+00:00"} {"global_step": 14866, "acc_step": 0, "speed/wps": 12912.54476576788, "speed/FLOPS": 202809259242299.16, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05212350934743881, "optim/lr": 0.0027680964217766646, "optim/total_tokens": 7794065408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 2.845411539077759, "created_at": "2025-01-16T23:32:53.962475+00:00"} {"global_step": 14867, "acc_step": 0, "speed/wps": 12912.325815194681, "speed/FLOPS": 202805820322677.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048016104847192764, "optim/lr": 0.0027680506549384552, "optim/total_tokens": 7794589696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366652, "loss/out": 2.8282055854797363, "created_at": "2025-01-16T23:33:04.117055+00:00"} {"global_step": 14868, "acc_step": 0, "speed/wps": 12912.582873422762, "speed/FLOPS": 202809857775384.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06940312683582306, "optim/lr": 0.002768004883963009, "optim/total_tokens": 7795113984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.859877347946167, "created_at": "2025-01-16T23:33:14.272299+00:00"} {"global_step": 14869, "acc_step": 0, "speed/wps": 12909.462826922225, "speed/FLOPS": 202760853157700.38, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09321126341819763, "optim/lr": 0.002767959108850475, "optim/total_tokens": 7795638272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.8771414756774902, "created_at": "2025-01-16T23:33:24.432273+00:00"} {"global_step": 14870, "acc_step": 0, "speed/wps": 12907.81773778713, "speed/FLOPS": 202735014772244.38, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0543997697532177, "optim/lr": 0.002767913329601001, "optim/total_tokens": 7796162560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8705661296844482, "created_at": "2025-01-16T23:33:34.590636+00:00"} {"global_step": 14871, "acc_step": 0, "speed/wps": 12907.05962964316, "speed/FLOPS": 202723107642093.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05290978029370308, "optim/lr": 0.002767867546214738, "optim/total_tokens": 7796686848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415290, "loss/out": 2.821781873703003, "created_at": "2025-01-16T23:33:44.751888+00:00"} {"global_step": 14872, "acc_step": 0, "speed/wps": 12909.284335426826, "speed/FLOPS": 202758049703492.12, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05820508673787117, "optim/lr": 0.002767821758691835, "optim/total_tokens": 7797211136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.7967886924743652, "created_at": "2025-01-16T23:33:54.908679+00:00"} {"global_step": 14873, "acc_step": 0, "speed/wps": 12918.078661918176, "speed/FLOPS": 202896176685709.75, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05248813331127167, "optim/lr": 0.0027677759670324408, "optim/total_tokens": 7797735424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.7984204292297363, "created_at": "2025-01-16T23:34:05.058562+00:00"} {"global_step": 14874, "acc_step": 0, "speed/wps": 12906.025767777895, "speed/FLOPS": 202706869420824.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06333388388156891, "optim/lr": 0.002767730171236706, "optim/total_tokens": 7798259712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.865473747253418, "created_at": "2025-01-16T23:34:15.219016+00:00"} {"global_step": 14875, "acc_step": 0, "speed/wps": 12913.274749035489, "speed/FLOPS": 202820724632619.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05186319723725319, "optim/lr": 0.002767684371304779, "optim/total_tokens": 7798784000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.7447590827941895, "created_at": "2025-01-16T23:34:25.376943+00:00"} {"global_step": 14876, "acc_step": 0, "speed/wps": 12910.711872637065, "speed/FLOPS": 202780471137020.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07211305946111679, "optim/lr": 0.00276763856723681, "optim/total_tokens": 7799308288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295368, "loss/out": 2.692915439605713, "created_at": "2025-01-16T23:34:35.532740+00:00"} {"global_step": 14877, "acc_step": 0, "speed/wps": 12908.136141493185, "speed/FLOPS": 202740015739980.5, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06561630964279175, "optim/lr": 0.002767592759032947, "optim/total_tokens": 7799832576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8288609981536865, "created_at": "2025-01-16T23:34:45.694542+00:00"} {"global_step": 14878, "acc_step": 0, "speed/wps": 12907.243904940986, "speed/FLOPS": 202726001938866.7, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05751962214708328, "optim/lr": 0.002767546946693341, "optim/total_tokens": 7800356864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 2.7495229244232178, "created_at": "2025-01-16T23:34:55.854106+00:00"} {"global_step": 14879, "acc_step": 0, "speed/wps": 12907.476372478428, "speed/FLOPS": 202729653160986.1, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06648873537778854, "optim/lr": 0.0027675011302181413, "optim/total_tokens": 7800881152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.8070030212402344, "created_at": "2025-01-16T23:35:06.015272+00:00"} {"global_step": 14880, "acc_step": 0, "speed/wps": 12904.410765799115, "speed/FLOPS": 202681503595502.47, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05524007976055145, "optim/lr": 0.002767455309607496, "optim/total_tokens": 7801405440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.784231424331665, "created_at": "2025-01-16T23:35:16.183783+00:00"} {"global_step": 14881, "acc_step": 0, "speed/wps": 12907.397864325956, "speed/FLOPS": 202728420082570.03, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05544315278530121, "optim/lr": 0.0027674094848615565, "optim/total_tokens": 7801929728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 2.850977897644043, "created_at": "2025-01-16T23:35:26.344580+00:00"} {"global_step": 14882, "acc_step": 0, "speed/wps": 12912.440220255186, "speed/FLOPS": 202807617211361.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06317581236362457, "optim/lr": 0.002767363655980471, "optim/total_tokens": 7802454016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.873471260070801, "created_at": "2025-01-16T23:35:36.501563+00:00"} {"global_step": 14883, "acc_step": 0, "speed/wps": 12907.116511831337, "speed/FLOPS": 202724001055023.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06748971343040466, "optim/lr": 0.00276731782296439, "optim/total_tokens": 7802978304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385268, "loss/out": 2.944723129272461, "created_at": "2025-01-16T23:35:46.661173+00:00"} {"global_step": 14884, "acc_step": 0, "speed/wps": 12888.763792017993, "speed/FLOPS": 202435746371074.8, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051605746150016785, "optim/lr": 0.0027672719858134616, "optim/total_tokens": 7803502592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8617875576019287, "created_at": "2025-01-16T23:35:56.836670+00:00"} {"global_step": 14885, "acc_step": 0, "speed/wps": 12899.0581271932, "speed/FLOPS": 202597433050900.34, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06821229308843613, "optim/lr": 0.0027672261445278365, "optim/total_tokens": 7804026880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.847127914428711, "created_at": "2025-01-16T23:36:07.003381+00:00"} {"global_step": 14886, "acc_step": 0, "speed/wps": 12898.066964789296, "speed/FLOPS": 202581865483347.16, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0454552061855793, "optim/lr": 0.002767180299107664, "optim/total_tokens": 7804551168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 2.8225274085998535, "created_at": "2025-01-16T23:36:17.172137+00:00"} {"global_step": 14887, "acc_step": 0, "speed/wps": 12897.139209126368, "speed/FLOPS": 202567293805791.0, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04659119248390198, "optim/lr": 0.002767134449553094, "optim/total_tokens": 7805075456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.741241455078125, "created_at": "2025-01-16T23:36:27.339478+00:00"} {"global_step": 14888, "acc_step": 0, "speed/wps": 12910.75966604697, "speed/FLOPS": 202781221798198.2, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0494687594473362, "optim/lr": 0.0027670885958642755, "optim/total_tokens": 7805599744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.89735746383667, "created_at": "2025-01-16T23:36:37.500643+00:00"} {"global_step": 14889, "acc_step": 0, "speed/wps": 12907.172132665295, "speed/FLOPS": 202724874656651.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05119135603308678, "optim/lr": 0.0027670427380413584, "optim/total_tokens": 7806124032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.916271924972534, "created_at": "2025-01-16T23:36:47.662291+00:00"} {"global_step": 14890, "acc_step": 0, "speed/wps": 12906.151956998305, "speed/FLOPS": 202708851395933.2, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05669945850968361, "optim/lr": 0.0027669968760844926, "optim/total_tokens": 7806648320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.6956310272216797, "created_at": "2025-01-16T23:36:57.821856+00:00"} {"global_step": 14891, "acc_step": 0, "speed/wps": 12912.812606949947, "speed/FLOPS": 202813466056115.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060745786875486374, "optim/lr": 0.0027669510099938266, "optim/total_tokens": 7807172608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.047898769378662, "created_at": "2025-01-16T23:37:07.976759+00:00"} {"global_step": 14892, "acc_step": 0, "speed/wps": 12903.173751502763, "speed/FLOPS": 202662074586142.5, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05019403249025345, "optim/lr": 0.0027669051397695116, "optim/total_tokens": 7807696896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.731438398361206, "created_at": "2025-01-16T23:37:18.138601+00:00"} {"global_step": 14893, "acc_step": 0, "speed/wps": 12909.464185141878, "speed/FLOPS": 202760874490406.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05026131495833397, "optim/lr": 0.0027668592654116965, "optim/total_tokens": 7808221184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335000, "loss/out": 2.785043239593506, "created_at": "2025-01-16T23:37:28.299888+00:00"} {"global_step": 14894, "acc_step": 0, "speed/wps": 12916.958269702087, "speed/FLOPS": 202878579386375.84, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.04546613246202469, "optim/lr": 0.0027668133869205307, "optim/total_tokens": 7808745472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.7512001991271973, "created_at": "2025-01-16T23:37:38.452885+00:00"} {"global_step": 14895, "acc_step": 0, "speed/wps": 12913.738339012874, "speed/FLOPS": 202828005950253.3, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.03753949701786041, "optim/lr": 0.002766767504296164, "optim/total_tokens": 7809269760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.858708381652832, "created_at": "2025-01-16T23:37:48.616348+00:00"} {"global_step": 14896, "acc_step": 0, "speed/wps": 12916.852544262396, "speed/FLOPS": 202876918823063.5, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050272468477487564, "optim/lr": 0.002766721617538747, "optim/total_tokens": 7809794048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.8520331382751465, "created_at": "2025-01-16T23:37:58.771508+00:00"} {"global_step": 14897, "acc_step": 0, "speed/wps": 12911.259720767885, "speed/FLOPS": 202789075844737.38, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04846535995602608, "optim/lr": 0.002766675726648428, "optim/total_tokens": 7810318336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.7714202404022217, "created_at": "2025-01-16T23:38:08.929027+00:00"} {"global_step": 14898, "acc_step": 0, "speed/wps": 12916.030994907682, "speed/FLOPS": 202864015261520.6, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04928502067923546, "optim/lr": 0.0027666298316253576, "optim/total_tokens": 7810842624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.809234619140625, "created_at": "2025-01-16T23:38:19.081553+00:00"} {"global_step": 14899, "acc_step": 0, "speed/wps": 12915.705900572246, "speed/FLOPS": 202858909208256.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05878014490008354, "optim/lr": 0.0027665839324696856, "optim/total_tokens": 7811366912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 2.839418411254883, "created_at": "2025-01-16T23:38:29.233769+00:00"} {"global_step": 14900, "acc_step": 0, "speed/wps": 12911.094061632546, "speed/FLOPS": 202786473940377.38, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05302375555038452, "optim/lr": 0.002766538029181561, "optim/total_tokens": 7811891200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 2.823176145553589, "created_at": "2025-01-16T23:38:39.392006+00:00"} {"global_step": 14901, "acc_step": 0, "speed/wps": 12910.210383545229, "speed/FLOPS": 202772594561715.1, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.054997026920318604, "optim/lr": 0.0027664921217611343, "optim/total_tokens": 7812415488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.7937817573547363, "created_at": "2025-01-16T23:38:49.549816+00:00"} {"global_step": 14902, "acc_step": 0, "speed/wps": 12914.962037972224, "speed/FLOPS": 202847225823948.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06252537667751312, "optim/lr": 0.0027664462102085555, "optim/total_tokens": 7812939776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 2.8968045711517334, "created_at": "2025-01-16T23:38:59.702511+00:00"} {"global_step": 14903, "acc_step": 0, "speed/wps": 12914.521849517738, "speed/FLOPS": 202840312059388.47, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04314287751913071, "optim/lr": 0.0027664002945239734, "optim/total_tokens": 7813464064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.737903594970703, "created_at": "2025-01-16T23:39:09.855170+00:00"} {"global_step": 14904, "acc_step": 0, "speed/wps": 12915.06398256408, "speed/FLOPS": 202848827003850.12, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06742846965789795, "optim/lr": 0.002766354374707539, "optim/total_tokens": 7813988352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.933957815170288, "created_at": "2025-01-16T23:39:20.008718+00:00"} {"global_step": 14905, "acc_step": 0, "speed/wps": 12919.687309423627, "speed/FLOPS": 202921442705296.2, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.061712268739938736, "optim/lr": 0.0027663084507594006, "optim/total_tokens": 7814512640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 2.8053581714630127, "created_at": "2025-01-16T23:39:30.157541+00:00"} {"global_step": 14906, "acc_step": 0, "speed/wps": 12917.163514607935, "speed/FLOPS": 202881803039658.25, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050683408975601196, "optim/lr": 0.00276626252267971, "optim/total_tokens": 7815036928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.8584139347076416, "created_at": "2025-01-16T23:39:40.308390+00:00"} {"global_step": 14907, "acc_step": 0, "speed/wps": 12908.106552194491, "speed/FLOPS": 202739550999384.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051309943199157715, "optim/lr": 0.002766216590468615, "optim/total_tokens": 7815561216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.8557796478271484, "created_at": "2025-01-16T23:39:50.467754+00:00"} {"global_step": 14908, "acc_step": 0, "speed/wps": 12911.64300592524, "speed/FLOPS": 202795095864822.62, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.001, "optim/grad_norm": 0.050505802035331726, "optim/lr": 0.0027661706541262667, "optim/total_tokens": 7816085504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.787322998046875, "created_at": "2025-01-16T23:40:00.626482+00:00"} {"global_step": 14909, "acc_step": 0, "speed/wps": 12910.851524238093, "speed/FLOPS": 202782664557316.16, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06011241301894188, "optim/lr": 0.0027661247136528152, "optim/total_tokens": 7816609792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.825390338897705, "created_at": "2025-01-16T23:40:10.784555+00:00"} {"global_step": 14910, "acc_step": 0, "speed/wps": 12914.773068383878, "speed/FLOPS": 202844257796892.1, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05430815368890762, "optim/lr": 0.0027660787690484095, "optim/total_tokens": 7817134080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 2.7719545364379883, "created_at": "2025-01-16T23:40:20.940254+00:00"} {"global_step": 14911, "acc_step": 0, "speed/wps": 12912.359391335593, "speed/FLOPS": 202806347682108.94, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04982961341738701, "optim/lr": 0.0027660328203132, "optim/total_tokens": 7817658368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.870086908340454, "created_at": "2025-01-16T23:40:31.095910+00:00"} {"global_step": 14912, "acc_step": 0, "speed/wps": 12911.824925681824, "speed/FLOPS": 202797953164583.72, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05914716422557831, "optim/lr": 0.0027659868674473364, "optim/total_tokens": 7818182656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.7862846851348877, "created_at": "2025-01-16T23:40:41.252908+00:00"} {"global_step": 14913, "acc_step": 0, "speed/wps": 12913.616299071722, "speed/FLOPS": 202826089145277.06, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06034598872065544, "optim/lr": 0.002765940910450969, "optim/total_tokens": 7818706944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.744267463684082, "created_at": "2025-01-16T23:40:51.410182+00:00"} {"global_step": 14914, "acc_step": 0, "speed/wps": 12912.99581318256, "speed/FLOPS": 202816343561750.8, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05581791326403618, "optim/lr": 0.0027658949493242473, "optim/total_tokens": 7819231232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 2.8438639640808105, "created_at": "2025-01-16T23:41:01.565961+00:00"} {"global_step": 14915, "acc_step": 0, "speed/wps": 12911.84912675595, "speed/FLOPS": 202798333275708.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04296988248825073, "optim/lr": 0.002765848984067321, "optim/total_tokens": 7819755520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.632371187210083, "created_at": "2025-01-16T23:41:11.722503+00:00"} {"global_step": 14916, "acc_step": 0, "speed/wps": 12910.892526076535, "speed/FLOPS": 202783308547531.34, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.054620254784822464, "optim/lr": 0.0027658030146803414, "optim/total_tokens": 7820279808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.7303824424743652, "created_at": "2025-01-16T23:41:21.889727+00:00"} {"global_step": 14917, "acc_step": 0, "speed/wps": 12909.386582352694, "speed/FLOPS": 202759655631964.88, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04368966817855835, "optim/lr": 0.0027657570411634574, "optim/total_tokens": 7820804096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.867164134979248, "created_at": "2025-01-16T23:41:32.055928+00:00"} {"global_step": 14918, "acc_step": 0, "speed/wps": 12919.866369808373, "speed/FLOPS": 202924255094693.38, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054026734083890915, "optim/lr": 0.0027657110635168187, "optim/total_tokens": 7821328384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470104, "loss/out": 2.792344570159912, "created_at": "2025-01-16T23:41:42.204495+00:00"} {"global_step": 14919, "acc_step": 0, "speed/wps": 12909.465626252319, "speed/FLOPS": 202760897125026.2, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0491652637720108, "optim/lr": 0.002765665081740576, "optim/total_tokens": 7821852672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.858888626098633, "created_at": "2025-01-16T23:41:52.367240+00:00"} {"global_step": 14920, "acc_step": 0, "speed/wps": 12908.168017061205, "speed/FLOPS": 202740516389577.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0464058592915535, "optim/lr": 0.002765619095834879, "optim/total_tokens": 7822376960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.8079943656921387, "created_at": "2025-01-16T23:42:02.525429+00:00"} {"global_step": 14921, "acc_step": 0, "speed/wps": 12913.270347323154, "speed/FLOPS": 202820655497678.28, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05098576843738556, "optim/lr": 0.0027655731057998787, "optim/total_tokens": 7822901248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.857463836669922, "created_at": "2025-01-16T23:42:12.683591+00:00"} {"global_step": 14922, "acc_step": 0, "speed/wps": 12913.809417671308, "speed/FLOPS": 202829122338256.3, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0561089813709259, "optim/lr": 0.0027655271116357235, "optim/total_tokens": 7823425536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.813593626022339, "created_at": "2025-01-16T23:42:22.837103+00:00"} {"global_step": 14923, "acc_step": 0, "speed/wps": 12909.600567480871, "speed/FLOPS": 202763016562449.28, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060680389404296875, "optim/lr": 0.0027654811133425647, "optim/total_tokens": 7823949824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370824, "loss/out": 2.7998318672180176, "created_at": "2025-01-16T23:42:32.994709+00:00"} {"global_step": 14924, "acc_step": 0, "speed/wps": 12917.553067869561, "speed/FLOPS": 202887921508934.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06101379171013832, "optim/lr": 0.002765435110920552, "optim/total_tokens": 7824474112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.7600579261779785, "created_at": "2025-01-16T23:42:43.147664+00:00"} {"global_step": 14925, "acc_step": 0, "speed/wps": 12918.07018750664, "speed/FLOPS": 202896043583431.53, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061644136905670166, "optim/lr": 0.002765389104369835, "optim/total_tokens": 7824998400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.7762582302093506, "created_at": "2025-01-16T23:42:53.303537+00:00"} {"global_step": 14926, "acc_step": 0, "speed/wps": 12916.602446546383, "speed/FLOPS": 202872990694763.25, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06263421475887299, "optim/lr": 0.0027653430936905644, "optim/total_tokens": 7825522688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.632730007171631, "created_at": "2025-01-16T23:43:03.455409+00:00"} {"global_step": 14927, "acc_step": 0, "speed/wps": 12908.917744465269, "speed/FLOPS": 202752291888692.1, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.050303976982831955, "optim/lr": 0.0027652970788828906, "optim/total_tokens": 7826046976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 2.8560118675231934, "created_at": "2025-01-16T23:43:13.614982+00:00"} {"global_step": 14928, "acc_step": 0, "speed/wps": 12911.798012014302, "speed/FLOPS": 202797530448452.03, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04766247421503067, "optim/lr": 0.002765251059946963, "optim/total_tokens": 7826571264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.7319183349609375, "created_at": "2025-01-16T23:43:23.771502+00:00"} {"global_step": 14929, "acc_step": 0, "speed/wps": 12916.328382380396, "speed/FLOPS": 202868686140432.78, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06317553669214249, "optim/lr": 0.0027652050368829322, "optim/total_tokens": 7827095552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 2.86209774017334, "created_at": "2025-01-16T23:43:33.925924+00:00"} {"global_step": 14930, "acc_step": 0, "speed/wps": 12909.29854839623, "speed/FLOPS": 202758272937707.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05779809132218361, "optim/lr": 0.0027651590096909474, "optim/total_tokens": 7827619840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.7320871353149414, "created_at": "2025-01-16T23:43:44.088337+00:00"} {"global_step": 14931, "acc_step": 0, "speed/wps": 12914.191859997505, "speed/FLOPS": 202835129120520.1, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048746418207883835, "optim/lr": 0.0027651129783711607, "optim/total_tokens": 7828144128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 2.8601725101470947, "created_at": "2025-01-16T23:43:54.242403+00:00"} {"global_step": 14932, "acc_step": 0, "speed/wps": 12913.3646962849, "speed/FLOPS": 202822137377771.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05378373712301254, "optim/lr": 0.0027650669429237205, "optim/total_tokens": 7828668416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.0304861068725586, "created_at": "2025-01-16T23:44:04.395953+00:00"} {"global_step": 14933, "acc_step": 0, "speed/wps": 12903.61545830927, "speed/FLOPS": 202669012198507.44, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04448489099740982, "optim/lr": 0.0027650209033487777, "optim/total_tokens": 7829192704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447797, "loss/out": 2.778770923614502, "created_at": "2025-01-16T23:44:14.559500+00:00"} {"global_step": 14934, "acc_step": 0, "speed/wps": 12911.590580345186, "speed/FLOPS": 202794272449047.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05724494531750679, "optim/lr": 0.002764974859646483, "optim/total_tokens": 7829716992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.753739833831787, "created_at": "2025-01-16T23:44:24.714667+00:00"} {"global_step": 14935, "acc_step": 0, "speed/wps": 12907.229483422316, "speed/FLOPS": 202725775429098.75, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05085708945989609, "optim/lr": 0.002764928811816986, "optim/total_tokens": 7830241280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.864474296569824, "created_at": "2025-01-16T23:44:34.873175+00:00"} {"global_step": 14936, "acc_step": 0, "speed/wps": 12914.521136100935, "speed/FLOPS": 202840300854197.22, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046395204961299896, "optim/lr": 0.0027648827598604365, "optim/total_tokens": 7830765568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.828035831451416, "created_at": "2025-01-16T23:44:45.026360+00:00"} {"global_step": 14937, "acc_step": 0, "speed/wps": 12913.275063020466, "speed/FLOPS": 202820729564184.62, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06673955917358398, "optim/lr": 0.0027648367037769856, "optim/total_tokens": 7831289856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412572, "loss/out": 2.8397955894470215, "created_at": "2025-01-16T23:44:55.180395+00:00"} {"global_step": 14938, "acc_step": 0, "speed/wps": 12911.542742878297, "speed/FLOPS": 202793521095894.53, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054371923208236694, "optim/lr": 0.0027647906435667832, "optim/total_tokens": 7831814144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8252713680267334, "created_at": "2025-01-16T23:45:05.338802+00:00"} {"global_step": 14939, "acc_step": 0, "speed/wps": 12912.676465953164, "speed/FLOPS": 202811327774684.06, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0516299307346344, "optim/lr": 0.00276474457922998, "optim/total_tokens": 7832338432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 2.919562816619873, "created_at": "2025-01-16T23:45:15.493365+00:00"} {"global_step": 14940, "acc_step": 0, "speed/wps": 12909.624187074603, "speed/FLOPS": 202763387540625.56, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05092938616871834, "optim/lr": 0.0027646985107667253, "optim/total_tokens": 7832862720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.7733967304229736, "created_at": "2025-01-16T23:45:25.655196+00:00"} {"global_step": 14941, "acc_step": 0, "speed/wps": 12913.627657644542, "speed/FLOPS": 202826267547271.5, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.042858343571424484, "optim/lr": 0.0027646524381771706, "optim/total_tokens": 7833387008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.946854591369629, "created_at": "2025-01-16T23:45:35.809973+00:00"} {"global_step": 14942, "acc_step": 0, "speed/wps": 12912.056171067716, "speed/FLOPS": 202801585191133.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04931353032588959, "optim/lr": 0.0027646063614614655, "optim/total_tokens": 7833911296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 2.842742681503296, "created_at": "2025-01-16T23:45:45.966322+00:00"} {"global_step": 14943, "acc_step": 0, "speed/wps": 12912.861642154397, "speed/FLOPS": 202814236221383.22, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04740513488650322, "optim/lr": 0.0027645602806197606, "optim/total_tokens": 7834435584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.843365430831909, "created_at": "2025-01-16T23:45:56.125553+00:00"} {"global_step": 14944, "acc_step": 0, "speed/wps": 12913.132244085991, "speed/FLOPS": 202818486396564.9, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054700206965208054, "optim/lr": 0.0027645141956522063, "optim/total_tokens": 7834959872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.8337254524230957, "created_at": "2025-01-16T23:46:06.280379+00:00"} {"global_step": 14945, "acc_step": 0, "speed/wps": 12912.997074992298, "speed/FLOPS": 202816363380206.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05882367864251137, "optim/lr": 0.002764468106558952, "optim/total_tokens": 7835484160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.9195120334625244, "created_at": "2025-01-16T23:46:16.445630+00:00"} {"global_step": 14946, "acc_step": 0, "speed/wps": 12913.108676818782, "speed/FLOPS": 202818116240248.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04665769636631012, "optim/lr": 0.0027644220133401496, "optim/total_tokens": 7836008448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349184, "loss/out": 2.8366775512695312, "created_at": "2025-01-16T23:46:26.599728+00:00"} {"global_step": 14947, "acc_step": 0, "speed/wps": 12910.880816991825, "speed/FLOPS": 202783124640266.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.059159379452466965, "optim/lr": 0.0027643759159959487, "optim/total_tokens": 7836532736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.9353108406066895, "created_at": "2025-01-16T23:46:36.758132+00:00"} {"global_step": 14948, "acc_step": 0, "speed/wps": 12914.38396244636, "speed/FLOPS": 202838146353455.78, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04348449781537056, "optim/lr": 0.0027643298145264995, "optim/total_tokens": 7837057024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.798581123352051, "created_at": "2025-01-16T23:46:46.911922+00:00"} {"global_step": 14949, "acc_step": 0, "speed/wps": 12913.020552991975, "speed/FLOPS": 202816732134453.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0717010647058487, "optim/lr": 0.0027642837089319524, "optim/total_tokens": 7837581312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.884849786758423, "created_at": "2025-01-16T23:46:57.065828+00:00"} {"global_step": 14950, "acc_step": 0, "speed/wps": 12916.894223217028, "speed/FLOPS": 202877573448318.9, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04707483947277069, "optim/lr": 0.0027642375992124584, "optim/total_tokens": 7838105600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 2.933821201324463, "created_at": "2025-01-16T23:47:07.221930+00:00"} {"global_step": 14951, "acc_step": 0, "speed/wps": 12915.01969263994, "speed/FLOPS": 202848131369730.78, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0594477616250515, "optim/lr": 0.0027641914853681684, "optim/total_tokens": 7838629888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 2.882761001586914, "created_at": "2025-01-16T23:47:17.377618+00:00"} {"global_step": 14952, "acc_step": 0, "speed/wps": 12911.258867942917, "speed/FLOPS": 202789062449949.34, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060487035661935806, "optim/lr": 0.0027641453673992312, "optim/total_tokens": 7839154176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401645, "loss/out": 2.914215564727783, "created_at": "2025-01-16T23:47:27.532852+00:00"} {"global_step": 14953, "acc_step": 0, "speed/wps": 12916.214889912728, "speed/FLOPS": 202866903585273.06, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04889588803052902, "optim/lr": 0.0027640992453057986, "optim/total_tokens": 7839678464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.841998815536499, "created_at": "2025-01-16T23:47:37.686286+00:00"} {"global_step": 14954, "acc_step": 0, "speed/wps": 12913.31614542067, "speed/FLOPS": 202821374819732.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05818666145205498, "optim/lr": 0.0027640531190880205, "optim/total_tokens": 7840202752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472818, "loss/out": 2.8569905757904053, "created_at": "2025-01-16T23:47:47.840473+00:00"} {"global_step": 14955, "acc_step": 0, "speed/wps": 12916.336059096106, "speed/FLOPS": 202868806713801.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05437598377466202, "optim/lr": 0.0027640069887460476, "optim/total_tokens": 7840727040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.854804515838623, "created_at": "2025-01-16T23:47:57.993531+00:00"} {"global_step": 14956, "acc_step": 0, "speed/wps": 12912.046040159783, "speed/FLOPS": 202801426071303.2, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05057529732584953, "optim/lr": 0.00276396085428003, "optim/total_tokens": 7841251328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.789228916168213, "created_at": "2025-01-16T23:48:08.148157+00:00"} {"global_step": 14957, "acc_step": 0, "speed/wps": 12909.640415085812, "speed/FLOPS": 202763642423841.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04544049873948097, "optim/lr": 0.0027639147156901193, "optim/total_tokens": 7841775616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.8697214126586914, "created_at": "2025-01-16T23:48:18.310460+00:00"} {"global_step": 14958, "acc_step": 0, "speed/wps": 12913.708218060336, "speed/FLOPS": 202827532859303.16, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0409550741314888, "optim/lr": 0.002763868572976465, "optim/total_tokens": 7842299904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 2.882272958755493, "created_at": "2025-01-16T23:48:28.464237+00:00"} {"global_step": 14959, "acc_step": 0, "speed/wps": 12918.161709842447, "speed/FLOPS": 202897481067480.25, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050547607243061066, "optim/lr": 0.0027638224261392176, "optim/total_tokens": 7842824192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.840445041656494, "created_at": "2025-01-16T23:48:38.614589+00:00"} {"global_step": 14960, "acc_step": 0, "speed/wps": 12914.389035658127, "speed/FLOPS": 202838226035217.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04243917018175125, "optim/lr": 0.0027637762751785283, "optim/total_tokens": 7843348480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.7984402179718018, "created_at": "2025-01-16T23:48:48.771573+00:00"} {"global_step": 14961, "acc_step": 0, "speed/wps": 12910.947468806211, "speed/FLOPS": 202784171498600.12, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04303722456097603, "optim/lr": 0.0027637301200945474, "optim/total_tokens": 7843872768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.9125349521636963, "created_at": "2025-01-16T23:48:58.930671+00:00"} {"global_step": 14962, "acc_step": 0, "speed/wps": 12912.875100006504, "speed/FLOPS": 202814447595443.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052989412099123, "optim/lr": 0.002763683960887426, "optim/total_tokens": 7844397056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.833162784576416, "created_at": "2025-01-16T23:49:09.085915+00:00"} {"global_step": 14963, "acc_step": 0, "speed/wps": 12910.542475545473, "speed/FLOPS": 202777810522921.2, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05583497881889343, "optim/lr": 0.0027636377975573134, "optim/total_tokens": 7844921344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8192763328552246, "created_at": "2025-01-16T23:49:19.244635+00:00"} {"global_step": 14964, "acc_step": 0, "speed/wps": 12907.76293060447, "speed/FLOPS": 202734153950127.7, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057722508907318115, "optim/lr": 0.002763591630104361, "optim/total_tokens": 7845445632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.72509765625, "created_at": "2025-01-16T23:49:29.403656+00:00"} {"global_step": 14965, "acc_step": 0, "speed/wps": 12914.63224164451, "speed/FLOPS": 202842045919434.06, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04851992055773735, "optim/lr": 0.0027635454585287203, "optim/total_tokens": 7845969920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.799179792404175, "created_at": "2025-01-16T23:49:39.556472+00:00"} {"global_step": 14966, "acc_step": 0, "speed/wps": 12914.639648511698, "speed/FLOPS": 202842162254461.38, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06102576479315758, "optim/lr": 0.0027634992828305402, "optim/total_tokens": 7846494208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.793506145477295, "created_at": "2025-01-16T23:49:49.712164+00:00"} {"global_step": 14967, "acc_step": 0, "speed/wps": 12913.682418663806, "speed/FLOPS": 202827127644328.6, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.0574212409555912, "optim/lr": 0.0027634531030099724, "optim/total_tokens": 7847018496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 2.983983039855957, "created_at": "2025-01-16T23:49:59.867808+00:00"} {"global_step": 14968, "acc_step": 0, "speed/wps": 12915.466059372193, "speed/FLOPS": 202855142172632.97, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06096808612346649, "optim/lr": 0.0027634069190671682, "optim/total_tokens": 7847542784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.773578405380249, "created_at": "2025-01-16T23:50:10.022164+00:00"} {"global_step": 14969, "acc_step": 0, "speed/wps": 12914.516483487641, "speed/FLOPS": 202840227778512.1, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05013423413038254, "optim/lr": 0.0027633607310022764, "optim/total_tokens": 7848067072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 2.9056339263916016, "created_at": "2025-01-16T23:50:20.175198+00:00"} {"global_step": 14970, "acc_step": 0, "speed/wps": 12912.461677501313, "speed/FLOPS": 202807954226897.5, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04493982717394829, "optim/lr": 0.0027633145388154495, "optim/total_tokens": 7848591360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.878654956817627, "created_at": "2025-01-16T23:50:30.331876+00:00"} {"global_step": 14971, "acc_step": 0, "speed/wps": 12913.122072390306, "speed/FLOPS": 202818326636106.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05090278387069702, "optim/lr": 0.002763268342506837, "optim/total_tokens": 7849115648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8086023330688477, "created_at": "2025-01-16T23:50:40.486092+00:00"} {"global_step": 14972, "acc_step": 0, "speed/wps": 12913.42748379536, "speed/FLOPS": 202823123541902.4, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05448091775178909, "optim/lr": 0.0027632221420765906, "optim/total_tokens": 7849639936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.867852210998535, "created_at": "2025-01-16T23:50:50.644174+00:00"} {"global_step": 14973, "acc_step": 0, "speed/wps": 12911.806028058385, "speed/FLOPS": 202797656351439.56, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0481262244284153, "optim/lr": 0.00276317593752486, "optim/total_tokens": 7850164224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 2.806004524230957, "created_at": "2025-01-16T23:51:00.799483+00:00"} {"global_step": 14974, "acc_step": 0, "speed/wps": 12909.171930567536, "speed/FLOPS": 202756284230716.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05434459075331688, "optim/lr": 0.002763129728851797, "optim/total_tokens": 7850688512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.726046562194824, "created_at": "2025-01-16T23:51:10.958639+00:00"} {"global_step": 14975, "acc_step": 0, "speed/wps": 12911.76229353366, "speed/FLOPS": 202796969440631.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04879763722419739, "optim/lr": 0.0027630835160575517, "optim/total_tokens": 7851212800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 2.807060718536377, "created_at": "2025-01-16T23:51:21.117188+00:00"} {"global_step": 14976, "acc_step": 0, "speed/wps": 12911.983825219117, "speed/FLOPS": 202800448900167.75, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04926471412181854, "optim/lr": 0.0027630372991422755, "optim/total_tokens": 7851737088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.762453556060791, "created_at": "2025-01-16T23:51:31.274480+00:00"} {"global_step": 14977, "acc_step": 0, "speed/wps": 12909.186374188692, "speed/FLOPS": 202756511087634.38, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04776204749941826, "optim/lr": 0.0027629910781061186, "optim/total_tokens": 7852261376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 2.84491229057312, "created_at": "2025-01-16T23:51:41.432082+00:00"} {"global_step": 14978, "acc_step": 0, "speed/wps": 12910.91380407009, "speed/FLOPS": 202783642747659.12, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.06900844722986221, "optim/lr": 0.002762944852949231, "optim/total_tokens": 7852785664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471672, "loss/out": 2.8669207096099854, "created_at": "2025-01-16T23:51:51.589586+00:00"} {"global_step": 14979, "acc_step": 0, "speed/wps": 12913.778009540809, "speed/FLOPS": 202828629030407.53, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05961567535996437, "optim/lr": 0.002762898623671765, "optim/total_tokens": 7853309952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402693, "loss/out": 2.9981322288513184, "created_at": "2025-01-16T23:52:01.743565+00:00"} {"global_step": 14980, "acc_step": 0, "speed/wps": 12911.943199076271, "speed/FLOPS": 202799810810768.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.060108721256256104, "optim/lr": 0.0027628523902738715, "optim/total_tokens": 7853834240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.8669145107269287, "created_at": "2025-01-16T23:52:11.898304+00:00"} {"global_step": 14981, "acc_step": 0, "speed/wps": 12913.662520269798, "speed/FLOPS": 202826815112707.44, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05967172607779503, "optim/lr": 0.0027628061527557005, "optim/total_tokens": 7854358528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 2.8970632553100586, "created_at": "2025-01-16T23:52:22.053275+00:00"} {"global_step": 14982, "acc_step": 0, "speed/wps": 12908.927019131923, "speed/FLOPS": 202752437560076.62, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05312643200159073, "optim/lr": 0.0027627599111174025, "optim/total_tokens": 7854882816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.970449924468994, "created_at": "2025-01-16T23:52:32.210499+00:00"} {"global_step": 14983, "acc_step": 0, "speed/wps": 12905.916830501244, "speed/FLOPS": 202705158411199.2, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06732705235481262, "optim/lr": 0.0027627136653591293, "optim/total_tokens": 7855407104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.8399810791015625, "created_at": "2025-01-16T23:52:42.373943+00:00"} {"global_step": 14984, "acc_step": 0, "speed/wps": 12916.025995083091, "speed/FLOPS": 202863936732404.94, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.061718638986349106, "optim/lr": 0.0027626674154810314, "optim/total_tokens": 7855931392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.9264397621154785, "created_at": "2025-01-16T23:52:52.529697+00:00"} {"global_step": 14985, "acc_step": 0, "speed/wps": 12918.02456081149, "speed/FLOPS": 202895326953486.53, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08859964460134506, "optim/lr": 0.00276262116148326, "optim/total_tokens": 7856455680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.742133617401123, "created_at": "2025-01-16T23:53:02.679734+00:00"} {"global_step": 14986, "acc_step": 0, "speed/wps": 12918.254994332978, "speed/FLOPS": 202898946228590.34, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049063943326473236, "optim/lr": 0.002762574903365966, "optim/total_tokens": 7856979968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.820563554763794, "created_at": "2025-01-16T23:53:12.833495+00:00"} {"global_step": 14987, "acc_step": 0, "speed/wps": 12910.390990974041, "speed/FLOPS": 202775431249565.22, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07440149784088135, "optim/lr": 0.0027625286411292996, "optim/total_tokens": 7857504256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.764207363128662, "created_at": "2025-01-16T23:53:22.989565+00:00"} {"global_step": 14988, "acc_step": 0, "speed/wps": 12911.37033785909, "speed/FLOPS": 202790813238158.56, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.055707648396492004, "optim/lr": 0.0027624823747734125, "optim/total_tokens": 7858028544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 2.845520496368408, "created_at": "2025-01-16T23:53:33.155104+00:00"} {"global_step": 14989, "acc_step": 0, "speed/wps": 12912.662931507204, "speed/FLOPS": 202811115197612.0, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05452895909547806, "optim/lr": 0.0027624361042984553, "optim/total_tokens": 7858552832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.887373924255371, "created_at": "2025-01-16T23:53:43.312418+00:00"} {"global_step": 14990, "acc_step": 0, "speed/wps": 12913.60881754406, "speed/FLOPS": 202825971637604.44, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06285431981086731, "optim/lr": 0.002762389829704579, "optim/total_tokens": 7859077120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339985, "loss/out": 2.818493366241455, "created_at": "2025-01-16T23:53:53.466742+00:00"} {"global_step": 14991, "acc_step": 0, "speed/wps": 12920.114337966039, "speed/FLOPS": 202928149775353.47, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04570618271827698, "optim/lr": 0.002762343550991935, "optim/total_tokens": 7859601408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 2.8755221366882324, "created_at": "2025-01-16T23:54:03.618233+00:00"} {"global_step": 14992, "acc_step": 0, "speed/wps": 12915.94673451177, "speed/FLOPS": 202862691836216.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05090775713324547, "optim/lr": 0.0027622972681606737, "optim/total_tokens": 7860125696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342101, "loss/out": 2.7888922691345215, "created_at": "2025-01-16T23:54:13.770896+00:00"} {"global_step": 14993, "acc_step": 0, "speed/wps": 12912.807944904736, "speed/FLOPS": 202813392832289.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050166454166173935, "optim/lr": 0.002762250981210946, "optim/total_tokens": 7860649984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8050241470336914, "created_at": "2025-01-16T23:54:23.925144+00:00"} {"global_step": 14994, "acc_step": 0, "speed/wps": 12916.008173312137, "speed/FLOPS": 202863656817002.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044940054416656494, "optim/lr": 0.002762204690142903, "optim/total_tokens": 7861174272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459256, "loss/out": 2.8484320640563965, "created_at": "2025-01-16T23:54:34.077430+00:00"} {"global_step": 14995, "acc_step": 0, "speed/wps": 12909.13489096174, "speed/FLOPS": 202755702472809.75, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04465121030807495, "optim/lr": 0.002762158394956697, "optim/total_tokens": 7861698560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8436999320983887, "created_at": "2025-01-16T23:54:44.237932+00:00"} {"global_step": 14996, "acc_step": 0, "speed/wps": 12908.17837511562, "speed/FLOPS": 202740679077055.9, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04568792134523392, "optim/lr": 0.0027621120956524774, "optim/total_tokens": 7862222848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.969329357147217, "created_at": "2025-01-16T23:54:54.395865+00:00"} {"global_step": 14997, "acc_step": 0, "speed/wps": 12911.856813547816, "speed/FLOPS": 202798454007337.66, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041517190635204315, "optim/lr": 0.002762065792230396, "optim/total_tokens": 7862747136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.7482829093933105, "created_at": "2025-01-16T23:55:04.550902+00:00"} {"global_step": 14998, "acc_step": 0, "speed/wps": 12913.432930677109, "speed/FLOPS": 202823209092665.1, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05027094483375549, "optim/lr": 0.0027620194846906037, "optim/total_tokens": 7863271424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.795898914337158, "created_at": "2025-01-16T23:55:14.704748+00:00"} {"global_step": 14999, "acc_step": 0, "speed/wps": 12914.587012535038, "speed/FLOPS": 202841335534118.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04660032317042351, "optim/lr": 0.0027619731730332514, "optim/total_tokens": 7863795712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.8192341327667236, "created_at": "2025-01-16T23:55:24.859498+00:00"} {"global_step": 15000, "acc_step": 0, "speed/wps": 12913.89628317006, "speed/FLOPS": 202830486680279.9, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057695258408784866, "optim/lr": 0.002761926857258491, "optim/total_tokens": 7864320000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.766085624694824, "created_at": "2025-01-16T23:55:35.047707+00:00"} {"global_step": 15001, "acc_step": 0, "speed/wps": 5275.260410351102, "speed/FLOPS": 82855213712005.38, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.2433, "optim/grad_norm": 0.05575466528534889, "optim/lr": 0.0027618805373664725, "optim/total_tokens": 7864844288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466705, "loss/out": 2.979332447052002, "created_at": "2025-01-16T23:55:59.900091+00:00"} {"global_step": 15002, "acc_step": 0, "speed/wps": 12947.48360373577, "speed/FLOPS": 203358021703579.47, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04502605274319649, "optim/lr": 0.002761834213357348, "optim/total_tokens": 7865368576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.8871607780456543, "created_at": "2025-01-16T23:56:10.028686+00:00"} {"global_step": 15003, "acc_step": 0, "speed/wps": 12930.55103082654, "speed/FLOPS": 203092072378246.53, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04667390137910843, "optim/lr": 0.002761787885231268, "optim/total_tokens": 7865892864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.769275188446045, "created_at": "2025-01-16T23:56:20.171683+00:00"} {"global_step": 15004, "acc_step": 0, "speed/wps": 12925.134472529575, "speed/FLOPS": 203006997887061.6, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.050616975873708725, "optim/lr": 0.0027617415529883837, "optim/total_tokens": 7866417152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.891780138015747, "created_at": "2025-01-16T23:56:30.320214+00:00"} {"global_step": 15005, "acc_step": 0, "speed/wps": 12917.279109188814, "speed/FLOPS": 202883618611394.53, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04599890485405922, "optim/lr": 0.002761695216628847, "optim/total_tokens": 7866941440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.779918909072876, "created_at": "2025-01-16T23:56:40.470827+00:00"} {"global_step": 15006, "acc_step": 0, "speed/wps": 12912.651477483192, "speed/FLOPS": 202810935296425.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05006812885403633, "optim/lr": 0.0027616488761528085, "optim/total_tokens": 7867465728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 2.789128303527832, "created_at": "2025-01-16T23:56:50.627062+00:00"} {"global_step": 15007, "acc_step": 0, "speed/wps": 12910.56989680377, "speed/FLOPS": 202778241211463.5, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04423744976520538, "optim/lr": 0.002761602531560419, "optim/total_tokens": 7867990016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416874, "loss/out": 2.9065096378326416, "created_at": "2025-01-16T23:57:00.787132+00:00"} {"global_step": 15008, "acc_step": 0, "speed/wps": 12907.322995996503, "speed/FLOPS": 202727244172576.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05937622860074043, "optim/lr": 0.00276155618285183, "optim/total_tokens": 7868514304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 2.7736377716064453, "created_at": "2025-01-16T23:57:10.946095+00:00"} {"global_step": 15009, "acc_step": 0, "speed/wps": 12892.065480955887, "speed/FLOPS": 202487603932840.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04664527624845505, "optim/lr": 0.0027615098300271938, "optim/total_tokens": 7869038592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.743821382522583, "created_at": "2025-01-16T23:57:21.119537+00:00"} {"global_step": 15010, "acc_step": 0, "speed/wps": 12899.517114505532, "speed/FLOPS": 202604642077355.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07715605199337006, "optim/lr": 0.0027614634730866593, "optim/total_tokens": 7869562880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 2.8562417030334473, "created_at": "2025-01-16T23:57:31.286448+00:00"} {"global_step": 15011, "acc_step": 0, "speed/wps": 12890.49482776041, "speed/FLOPS": 202462934666101.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05129428207874298, "optim/lr": 0.0027614171120303806, "optim/total_tokens": 7870087168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8644533157348633, "created_at": "2025-01-16T23:57:41.461438+00:00"} {"global_step": 15012, "acc_step": 0, "speed/wps": 12902.71597074999, "speed/FLOPS": 202654884510363.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06703410297632217, "optim/lr": 0.0027613707468585064, "optim/total_tokens": 7870611456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.90162992477417, "created_at": "2025-01-16T23:57:51.625615+00:00"} {"global_step": 15013, "acc_step": 0, "speed/wps": 12892.903552956377, "speed/FLOPS": 202500767005239.78, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06033935025334358, "optim/lr": 0.0027613243775711895, "optim/total_tokens": 7871135744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.7835493087768555, "created_at": "2025-01-16T23:58:01.798329+00:00"} {"global_step": 15014, "acc_step": 0, "speed/wps": 12895.702873079372, "speed/FLOPS": 202544734174440.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04014161601662636, "optim/lr": 0.0027612780041685807, "optim/total_tokens": 7871660032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 2.7570371627807617, "created_at": "2025-01-16T23:58:11.968265+00:00"} {"global_step": 15015, "acc_step": 0, "speed/wps": 12894.112667214988, "speed/FLOPS": 202519757806168.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06132955104112625, "optim/lr": 0.0027612316266508313, "optim/total_tokens": 7872184320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.864725112915039, "created_at": "2025-01-16T23:58:22.137089+00:00"} {"global_step": 15016, "acc_step": 0, "speed/wps": 12891.248324927656, "speed/FLOPS": 202474769374527.03, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.046207495033741, "optim/lr": 0.002761185245018093, "optim/total_tokens": 7872708608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8326642513275146, "created_at": "2025-01-16T23:58:32.314848+00:00"} {"global_step": 15017, "acc_step": 0, "speed/wps": 12893.351836588521, "speed/FLOPS": 202507807915688.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058339208364486694, "optim/lr": 0.002761138859270516, "optim/total_tokens": 7873232896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349184, "loss/out": 2.821742057800293, "created_at": "2025-01-16T23:58:42.484408+00:00"} {"global_step": 15018, "acc_step": 0, "speed/wps": 12888.216234320165, "speed/FLOPS": 202427146225007.28, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048741135746240616, "optim/lr": 0.0027610924694082537, "optim/total_tokens": 7873757184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.7023003101348877, "created_at": "2025-01-16T23:58:52.657965+00:00"} {"global_step": 15019, "acc_step": 0, "speed/wps": 12892.638307463636, "speed/FLOPS": 202496600960290.78, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.0490504615008831, "optim/lr": 0.002761046075431455, "optim/total_tokens": 7874281472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 2.8647732734680176, "created_at": "2025-01-16T23:59:02.828314+00:00"} {"global_step": 15020, "acc_step": 0, "speed/wps": 12892.759710278773, "speed/FLOPS": 202498507758327.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048469409346580505, "optim/lr": 0.002760999677340273, "optim/total_tokens": 7874805760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.868834972381592, "created_at": "2025-01-16T23:59:13.002600+00:00"} {"global_step": 15021, "acc_step": 0, "speed/wps": 12895.402264840017, "speed/FLOPS": 202540012708962.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05476900562644005, "optim/lr": 0.0027609532751348585, "optim/total_tokens": 7875330048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362676, "loss/out": 2.7904021739959717, "created_at": "2025-01-16T23:59:23.173439+00:00"} {"global_step": 15022, "acc_step": 0, "speed/wps": 12891.522341336986, "speed/FLOPS": 202479073178773.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04921621456742287, "optim/lr": 0.0027609068688153634, "optim/total_tokens": 7875854336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8023581504821777, "created_at": "2025-01-16T23:59:33.345394+00:00"} {"global_step": 15023, "acc_step": 0, "speed/wps": 12896.170458476501, "speed/FLOPS": 202552078245631.94, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04642094671726227, "optim/lr": 0.002760860458381938, "optim/total_tokens": 7876378624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 2.862407684326172, "created_at": "2025-01-16T23:59:43.513677+00:00"} {"global_step": 15024, "acc_step": 0, "speed/wps": 12893.513953017304, "speed/FLOPS": 202510354176974.22, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054534442722797394, "optim/lr": 0.0027608140438347343, "optim/total_tokens": 7876902912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485284, "loss/out": 2.7151055335998535, "created_at": "2025-01-16T23:59:53.686034+00:00"} {"global_step": 15025, "acc_step": 0, "speed/wps": 12885.62888552379, "speed/FLOPS": 202386508356770.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048715293407440186, "optim/lr": 0.0027607676251739043, "optim/total_tokens": 7877427200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 2.7296619415283203, "created_at": "2025-01-17T00:00:03.862145+00:00"} {"global_step": 15026, "acc_step": 0, "speed/wps": 12888.118342357395, "speed/FLOPS": 202425608697214.3, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05515548586845398, "optim/lr": 0.0027607212023995987, "optim/total_tokens": 7877951488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.8938002586364746, "created_at": "2025-01-17T00:00:14.038516+00:00"} {"global_step": 15027, "acc_step": 0, "speed/wps": 12890.602422146332, "speed/FLOPS": 202464624583782.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04486924037337303, "optim/lr": 0.002760674775511969, "optim/total_tokens": 7878475776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.886176824569702, "created_at": "2025-01-17T00:00:24.210487+00:00"} {"global_step": 15028, "acc_step": 0, "speed/wps": 12891.456933802834, "speed/FLOPS": 202478045863570.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.07532721757888794, "optim/lr": 0.0027606283445111676, "optim/total_tokens": 7879000064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.7987537384033203, "created_at": "2025-01-17T00:00:34.384707+00:00"} {"global_step": 15029, "acc_step": 0, "speed/wps": 12895.185942821323, "speed/FLOPS": 202536615074401.1, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04240074008703232, "optim/lr": 0.002760581909397345, "optim/total_tokens": 7879524352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.741497039794922, "created_at": "2025-01-17T00:00:44.553058+00:00"} {"global_step": 15030, "acc_step": 0, "speed/wps": 12885.274238922533, "speed/FLOPS": 202380938144562.16, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08803379535675049, "optim/lr": 0.0027605354701706526, "optim/total_tokens": 7880048640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.8287441730499268, "created_at": "2025-01-17T00:00:54.738885+00:00"} {"global_step": 15031, "acc_step": 0, "speed/wps": 12889.943784318753, "speed/FLOPS": 202454279771637.6, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05742930620908737, "optim/lr": 0.0027604890268312426, "optim/total_tokens": 7880572928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.7486376762390137, "created_at": "2025-01-17T00:01:04.912502+00:00"} {"global_step": 15032, "acc_step": 0, "speed/wps": 12892.869410595853, "speed/FLOPS": 202500230752551.22, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06307724118232727, "optim/lr": 0.0027604425793792665, "optim/total_tokens": 7881097216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.813295364379883, "created_at": "2025-01-17T00:01:15.082272+00:00"} {"global_step": 15033, "acc_step": 0, "speed/wps": 12886.310947191878, "speed/FLOPS": 202397221072518.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05251796171069145, "optim/lr": 0.0027603961278148755, "optim/total_tokens": 7881621504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419875, "loss/out": 2.790069341659546, "created_at": "2025-01-17T00:01:25.259409+00:00"} {"global_step": 15034, "acc_step": 0, "speed/wps": 12895.928596060074, "speed/FLOPS": 202548279464028.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05563764646649361, "optim/lr": 0.0027603496721382206, "optim/total_tokens": 7882145792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458985, "loss/out": 2.8944385051727295, "created_at": "2025-01-17T00:01:35.431523+00:00"} {"global_step": 15035, "acc_step": 0, "speed/wps": 12890.774770066202, "speed/FLOPS": 202467331544695.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04110820218920708, "optim/lr": 0.0027603032123494544, "optim/total_tokens": 7882670080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.806735038757324, "created_at": "2025-01-17T00:01:45.604756+00:00"} {"global_step": 15036, "acc_step": 0, "speed/wps": 12895.693566595393, "speed/FLOPS": 202544588003321.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06268826872110367, "optim/lr": 0.0027602567484487283, "optim/total_tokens": 7883194368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 2.8739354610443115, "created_at": "2025-01-17T00:01:55.777376+00:00"} {"global_step": 15037, "acc_step": 0, "speed/wps": 12893.388151372606, "speed/FLOPS": 202508378289274.66, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05384337529540062, "optim/lr": 0.0027602102804361933, "optim/total_tokens": 7883718656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314812, "loss/out": 2.9222028255462646, "created_at": "2025-01-17T00:02:05.946705+00:00"} {"global_step": 15038, "acc_step": 0, "speed/wps": 12890.466468073675, "speed/FLOPS": 202462489238251.38, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04837234690785408, "optim/lr": 0.0027601638083120015, "optim/total_tokens": 7884242944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.866652011871338, "created_at": "2025-01-17T00:02:16.118781+00:00"} {"global_step": 15039, "acc_step": 0, "speed/wps": 12894.34893148544, "speed/FLOPS": 202523468661196.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06933985650539398, "optim/lr": 0.002760117332076305, "optim/total_tokens": 7884767232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.7937235832214355, "created_at": "2025-01-17T00:02:26.289960+00:00"} {"global_step": 15040, "acc_step": 0, "speed/wps": 12896.010562975942, "speed/FLOPS": 202549566867076.22, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047922614961862564, "optim/lr": 0.0027600708517292547, "optim/total_tokens": 7885291520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448897, "loss/out": 2.8145828247070312, "created_at": "2025-01-17T00:02:36.457359+00:00"} {"global_step": 15041, "acc_step": 0, "speed/wps": 12889.707590550368, "speed/FLOPS": 202450570023940.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05796235799789429, "optim/lr": 0.0027600243672710016, "optim/total_tokens": 7885815808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 2.874202013015747, "created_at": "2025-01-17T00:02:46.632957+00:00"} {"global_step": 15042, "acc_step": 0, "speed/wps": 12892.100595717886, "speed/FLOPS": 202488155458430.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046927351504564285, "optim/lr": 0.002759977878701699, "optim/total_tokens": 7886340096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.745694398880005, "created_at": "2025-01-17T00:02:56.803771+00:00"} {"global_step": 15043, "acc_step": 0, "speed/wps": 12898.402589859452, "speed/FLOPS": 202587136936271.12, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04877733066678047, "optim/lr": 0.0027599313860214975, "optim/total_tokens": 7886864384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392822, "loss/out": 2.778347969055176, "created_at": "2025-01-17T00:03:06.969496+00:00"} {"global_step": 15044, "acc_step": 0, "speed/wps": 12891.164245874817, "speed/FLOPS": 202473448797465.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05277915298938751, "optim/lr": 0.0027598848892305488, "optim/total_tokens": 7887388672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.936995029449463, "created_at": "2025-01-17T00:03:17.143161+00:00"} {"global_step": 15045, "acc_step": 0, "speed/wps": 12893.102657532358, "speed/FLOPS": 202503894216203.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04928825423121452, "optim/lr": 0.002759838388329005, "optim/total_tokens": 7887912960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.649893283843994, "created_at": "2025-01-17T00:03:27.316214+00:00"} {"global_step": 15046, "acc_step": 0, "speed/wps": 12890.971182887843, "speed/FLOPS": 202470416477958.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04932928457856178, "optim/lr": 0.0027597918833170178, "optim/total_tokens": 7888437248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.805248737335205, "created_at": "2025-01-17T00:03:37.489619+00:00"} {"global_step": 15047, "acc_step": 0, "speed/wps": 12893.64487664994, "speed/FLOPS": 202512410512532.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05075124278664589, "optim/lr": 0.0027597453741947386, "optim/total_tokens": 7888961536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.7945444583892822, "created_at": "2025-01-17T00:03:47.660204+00:00"} {"global_step": 15048, "acc_step": 0, "speed/wps": 12894.278361049212, "speed/FLOPS": 202522360255522.03, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04475080594420433, "optim/lr": 0.0027596988609623198, "optim/total_tokens": 7889485824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.8880233764648438, "created_at": "2025-01-17T00:03:57.832181+00:00"} {"global_step": 15049, "acc_step": 0, "speed/wps": 12893.735671664404, "speed/FLOPS": 202513836573000.28, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.049680497497320175, "optim/lr": 0.002759652343619912, "optim/total_tokens": 7890010112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.8145594596862793, "created_at": "2025-01-17T00:04:08.006169+00:00"} {"global_step": 15050, "acc_step": 0, "speed/wps": 12892.679791324386, "speed/FLOPS": 202497252521328.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05672984570264816, "optim/lr": 0.0027596058221676686, "optim/total_tokens": 7890534400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.8400769233703613, "created_at": "2025-01-17T00:04:18.182800+00:00"} {"global_step": 15051, "acc_step": 0, "speed/wps": 12892.870568081158, "speed/FLOPS": 202500248932448.5, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04540933668613434, "optim/lr": 0.0027595592966057395, "optim/total_tokens": 7891058688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 2.7986040115356445, "created_at": "2025-01-17T00:04:28.355182+00:00"} {"global_step": 15052, "acc_step": 0, "speed/wps": 12892.981160855077, "speed/FLOPS": 202501985943933.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050818849354982376, "optim/lr": 0.002759512766934278, "optim/total_tokens": 7891582976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.7785720825195312, "created_at": "2025-01-17T00:04:38.525008+00:00"} {"global_step": 15053, "acc_step": 0, "speed/wps": 12889.80539096354, "speed/FLOPS": 202452106113821.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04565184563398361, "optim/lr": 0.002759466233153435, "optim/total_tokens": 7892107264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.828115940093994, "created_at": "2025-01-17T00:04:48.700738+00:00"} {"global_step": 15054, "acc_step": 0, "speed/wps": 12893.913492460577, "speed/FLOPS": 202516629492955.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059444691985845566, "optim/lr": 0.0027594196952633633, "optim/total_tokens": 7892631552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 2.66865611076355, "created_at": "2025-01-17T00:04:58.869665+00:00"} {"global_step": 15055, "acc_step": 0, "speed/wps": 12897.923654630014, "speed/FLOPS": 202579614600366.84, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04611196368932724, "optim/lr": 0.0027593731532642132, "optim/total_tokens": 7893155840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 2.820706367492676, "created_at": "2025-01-17T00:05:09.038919+00:00"} {"global_step": 15056, "acc_step": 0, "speed/wps": 12892.050807543872, "speed/FLOPS": 202487373466741.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050685230642557144, "optim/lr": 0.002759326607156138, "optim/total_tokens": 7893680128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 2.5813934803009033, "created_at": "2025-01-17T00:05:19.211065+00:00"} {"global_step": 15057, "acc_step": 0, "speed/wps": 12891.541143026127, "speed/FLOPS": 202479368485137.75, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047580111771821976, "optim/lr": 0.002759280056939289, "optim/total_tokens": 7894204416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 2.9466586112976074, "created_at": "2025-01-17T00:05:29.385386+00:00"} {"global_step": 15058, "acc_step": 0, "speed/wps": 12895.846247330634, "speed/FLOPS": 202546986064073.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045607976615428925, "optim/lr": 0.002759233502613818, "optim/total_tokens": 7894728704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424126, "loss/out": 2.898413896560669, "created_at": "2025-01-17T00:05:39.553473+00:00"} {"global_step": 15059, "acc_step": 0, "speed/wps": 12888.860068412443, "speed/FLOPS": 202437258524147.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05014682188630104, "optim/lr": 0.002759186944179878, "optim/total_tokens": 7895252992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 2.8433620929718018, "created_at": "2025-01-17T00:05:49.726270+00:00"} {"global_step": 15060, "acc_step": 0, "speed/wps": 12894.53523401294, "speed/FLOPS": 202526394798397.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04635339230298996, "optim/lr": 0.002759140381637619, "optim/total_tokens": 7895777280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8843560218811035, "created_at": "2025-01-17T00:05:59.894695+00:00"} {"global_step": 15061, "acc_step": 0, "speed/wps": 12892.453939947409, "speed/FLOPS": 202493705215100.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05506022647023201, "optim/lr": 0.0027590938149871934, "optim/total_tokens": 7896301568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.9492316246032715, "created_at": "2025-01-17T00:06:10.064725+00:00"} {"global_step": 15062, "acc_step": 0, "speed/wps": 12883.410584078581, "speed/FLOPS": 202351666884307.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05916180834174156, "optim/lr": 0.0027590472442287546, "optim/total_tokens": 7896825856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 2.842555522918701, "created_at": "2025-01-17T00:06:20.243451+00:00"} {"global_step": 15063, "acc_step": 0, "speed/wps": 12892.723357696837, "speed/FLOPS": 202497936791074.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050071168690919876, "optim/lr": 0.002759000669362453, "optim/total_tokens": 7897350144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.837392807006836, "created_at": "2025-01-17T00:06:30.416628+00:00"} {"global_step": 15064, "acc_step": 0, "speed/wps": 12890.370612783923, "speed/FLOPS": 202460983699206.84, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07569106668233871, "optim/lr": 0.0027589540903884407, "optim/total_tokens": 7897874432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 2.846674919128418, "created_at": "2025-01-17T00:06:40.591388+00:00"} {"global_step": 15065, "acc_step": 0, "speed/wps": 12893.892459472405, "speed/FLOPS": 202516299140973.94, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.001, "optim/grad_norm": 0.07381441444158554, "optim/lr": 0.0027589075073068702, "optim/total_tokens": 7898398720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.6928882598876953, "created_at": "2025-01-17T00:06:50.760268+00:00"} {"global_step": 15066, "acc_step": 0, "speed/wps": 12892.904073827496, "speed/FLOPS": 202500775186236.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058238495141267776, "optim/lr": 0.0027588609201178935, "optim/total_tokens": 7898923008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413993, "loss/out": 2.7600762844085693, "created_at": "2025-01-17T00:07:00.930877+00:00"} {"global_step": 15067, "acc_step": 0, "speed/wps": 12886.601089075159, "speed/FLOPS": 202401778149492.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06933754682540894, "optim/lr": 0.0027588143288216627, "optim/total_tokens": 7899447296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 2.948734760284424, "created_at": "2025-01-17T00:07:11.106441+00:00"} {"global_step": 15068, "acc_step": 0, "speed/wps": 12893.830279958023, "speed/FLOPS": 202515322526256.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0482928566634655, "optim/lr": 0.0027587677334183296, "optim/total_tokens": 7899971584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8313345909118652, "created_at": "2025-01-17T00:07:21.276650+00:00"} {"global_step": 15069, "acc_step": 0, "speed/wps": 12893.344413958768, "speed/FLOPS": 202507691333088.84, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07102789729833603, "optim/lr": 0.0027587211339080457, "optim/total_tokens": 7900495872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 2.8035380840301514, "created_at": "2025-01-17T00:07:31.448678+00:00"} {"global_step": 15070, "acc_step": 0, "speed/wps": 12897.828385792718, "speed/FLOPS": 202578118272364.0, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04252173751592636, "optim/lr": 0.0027586745302909635, "optim/total_tokens": 7901020160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453646, "loss/out": 2.779682159423828, "created_at": "2025-01-17T00:07:41.616075+00:00"} {"global_step": 15071, "acc_step": 0, "speed/wps": 12892.404034485051, "speed/FLOPS": 202492921381237.28, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05735507234930992, "optim/lr": 0.002758627922567235, "optim/total_tokens": 7901544448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.8625893592834473, "created_at": "2025-01-17T00:07:51.786420+00:00"} {"global_step": 15072, "acc_step": 0, "speed/wps": 12894.127917088434, "speed/FLOPS": 202519997326386.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.001, "optim/grad_norm": 0.046813108026981354, "optim/lr": 0.002758581310737013, "optim/total_tokens": 7902068736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.7458887100219727, "created_at": "2025-01-17T00:08:01.955264+00:00"} {"global_step": 15073, "acc_step": 0, "speed/wps": 12889.968829008658, "speed/FLOPS": 202454673132907.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060213830322027206, "optim/lr": 0.002758534694800448, "optim/total_tokens": 7902593024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415290, "loss/out": 2.80395245552063, "created_at": "2025-01-17T00:08:12.129503+00:00"} {"global_step": 15074, "acc_step": 0, "speed/wps": 12890.821489107002, "speed/FLOPS": 202468065331429.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05657820776104927, "optim/lr": 0.002758488074757694, "optim/total_tokens": 7903117312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302403, "loss/out": 2.8927221298217773, "created_at": "2025-01-17T00:08:22.304961+00:00"} {"global_step": 15075, "acc_step": 0, "speed/wps": 12890.754028328061, "speed/FLOPS": 202467005767195.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05244983360171318, "optim/lr": 0.0027584414506089013, "optim/total_tokens": 7903641600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 2.778202533721924, "created_at": "2025-01-17T00:08:32.483235+00:00"} {"global_step": 15076, "acc_step": 0, "speed/wps": 12889.672425074832, "speed/FLOPS": 202450017701825.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05035334452986717, "optim/lr": 0.0027583948223542233, "optim/total_tokens": 7904165888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.851803779602051, "created_at": "2025-01-17T00:08:42.656140+00:00"} {"global_step": 15077, "acc_step": 0, "speed/wps": 12889.617154467616, "speed/FLOPS": 202449149600989.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.056932173669338226, "optim/lr": 0.0027583481899938114, "optim/total_tokens": 7904690176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420921, "loss/out": 2.817384719848633, "created_at": "2025-01-17T00:08:52.828696+00:00"} {"global_step": 15078, "acc_step": 0, "speed/wps": 12887.157422534168, "speed/FLOPS": 202410516130948.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.05645102262496948, "optim/lr": 0.002758301553527818, "optim/total_tokens": 7905214464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 2.911020040512085, "created_at": "2025-01-17T00:09:03.002960+00:00"} {"global_step": 15079, "acc_step": 0, "speed/wps": 12892.926219747393, "speed/FLOPS": 202501123018340.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050113242119550705, "optim/lr": 0.0027582549129563953, "optim/total_tokens": 7905738752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8615610599517822, "created_at": "2025-01-17T00:09:13.172811+00:00"} {"global_step": 15080, "acc_step": 0, "speed/wps": 12891.129999369843, "speed/FLOPS": 202472910909045.25, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0740862563252449, "optim/lr": 0.002758208268279695, "optim/total_tokens": 7906263040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 2.9501113891601562, "created_at": "2025-01-17T00:09:23.361664+00:00"} {"global_step": 15081, "acc_step": 0, "speed/wps": 12894.300805851572, "speed/FLOPS": 202522712781985.5, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04778970777988434, "optim/lr": 0.002758161619497871, "optim/total_tokens": 7906787328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 2.8023924827575684, "created_at": "2025-01-17T00:09:33.530870+00:00"} {"global_step": 15082, "acc_step": 0, "speed/wps": 12892.467941087814, "speed/FLOPS": 202493925122250.56, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05003183335065842, "optim/lr": 0.002758114966611073, "optim/total_tokens": 7907311616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.949774742126465, "created_at": "2025-01-17T00:09:43.700930+00:00"} {"global_step": 15083, "acc_step": 0, "speed/wps": 12884.71894331461, "speed/FLOPS": 202372216463983.6, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0027, "optim/grad_norm": 0.04132714495062828, "optim/lr": 0.0027580683096194544, "optim/total_tokens": 7907835904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.904831886291504, "created_at": "2025-01-17T00:09:53.882996+00:00"} {"global_step": 15084, "acc_step": 0, "speed/wps": 12889.38051700848, "speed/FLOPS": 202445432884519.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04232097044587135, "optim/lr": 0.0027580216485231686, "optim/total_tokens": 7908360192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8354997634887695, "created_at": "2025-01-17T00:10:04.055655+00:00"} {"global_step": 15085, "acc_step": 0, "speed/wps": 12883.432164317563, "speed/FLOPS": 202352005831615.3, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04515965282917023, "optim/lr": 0.002757974983322366, "optim/total_tokens": 7908884480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8764638900756836, "created_at": "2025-01-17T00:10:14.234346+00:00"} {"global_step": 15086, "acc_step": 0, "speed/wps": 12886.824626774654, "speed/FLOPS": 202405289116235.88, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046395543962717056, "optim/lr": 0.0027579283140172, "optim/total_tokens": 7909408768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 2.8598031997680664, "created_at": "2025-01-17T00:10:24.411383+00:00"} {"global_step": 15087, "acc_step": 0, "speed/wps": 12891.539567753747, "speed/FLOPS": 202479343743320.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049699995666742325, "optim/lr": 0.0027578816406078214, "optim/total_tokens": 7909933056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.7835121154785156, "created_at": "2025-01-17T00:10:34.582446+00:00"} {"global_step": 15088, "acc_step": 0, "speed/wps": 12894.09467196181, "speed/FLOPS": 202519475165989.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04590191692113876, "optim/lr": 0.0027578349630943844, "optim/total_tokens": 7910457344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8062615394592285, "created_at": "2025-01-17T00:10:44.753795+00:00"} {"global_step": 15089, "acc_step": 0, "speed/wps": 12885.494126434585, "speed/FLOPS": 202384391780096.03, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05882817506790161, "optim/lr": 0.00275778828147704, "optim/total_tokens": 7910981632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 2.9095816612243652, "created_at": "2025-01-17T00:10:54.937879+00:00"} {"global_step": 15090, "acc_step": 0, "speed/wps": 12890.48473183931, "speed/FLOPS": 202462776095787.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.041033580899238586, "optim/lr": 0.0027577415957559413, "optim/total_tokens": 7911505920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 2.7822062969207764, "created_at": "2025-01-17T00:11:05.112055+00:00"} {"global_step": 15091, "acc_step": 0, "speed/wps": 12896.131582624452, "speed/FLOPS": 202551467646954.5, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048715002834796906, "optim/lr": 0.00275769490593124, "optim/total_tokens": 7912030208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.8843135833740234, "created_at": "2025-01-17T00:11:15.279176+00:00"} {"global_step": 15092, "acc_step": 0, "speed/wps": 12889.765031762814, "speed/FLOPS": 202451472217114.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05820383504033089, "optim/lr": 0.0027576482120030884, "optim/total_tokens": 7912554496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.796924114227295, "created_at": "2025-01-17T00:11:25.451782+00:00"} {"global_step": 15093, "acc_step": 0, "speed/wps": 12884.197575981776, "speed/FLOPS": 202364027673587.66, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05494948849081993, "optim/lr": 0.0027576015139716393, "optim/total_tokens": 7913078784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.919917345046997, "created_at": "2025-01-17T00:11:35.630110+00:00"} {"global_step": 15094, "acc_step": 0, "speed/wps": 12892.964056965766, "speed/FLOPS": 202501717303848.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04216683655977249, "optim/lr": 0.002757554811837045, "optim/total_tokens": 7913603072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 2.765169143676758, "created_at": "2025-01-17T00:11:45.800448+00:00"} {"global_step": 15095, "acc_step": 0, "speed/wps": 12890.00797702662, "speed/FLOPS": 202455288006324.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05810555815696716, "optim/lr": 0.0027575081055994577, "optim/total_tokens": 7914127360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 2.8067314624786377, "created_at": "2025-01-17T00:11:55.974404+00:00"} {"global_step": 15096, "acc_step": 0, "speed/wps": 12893.10582773996, "speed/FLOPS": 202503944008670.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06709439307451248, "optim/lr": 0.0027574613952590296, "optim/total_tokens": 7914651648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 2.7814998626708984, "created_at": "2025-01-17T00:12:06.150051+00:00"} {"global_step": 15097, "acc_step": 0, "speed/wps": 12887.688456541086, "speed/FLOPS": 202418856749744.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04074440523982048, "optim/lr": 0.002757414680815913, "optim/total_tokens": 7915175936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427570, "loss/out": 2.8099546432495117, "created_at": "2025-01-17T00:12:16.325058+00:00"} {"global_step": 15098, "acc_step": 0, "speed/wps": 12891.223483087311, "speed/FLOPS": 202474379199288.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07749870419502258, "optim/lr": 0.0027573679622702615, "optim/total_tokens": 7915700224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 2.920463800430298, "created_at": "2025-01-17T00:12:26.497521+00:00"} {"global_step": 15099, "acc_step": 0, "speed/wps": 12892.681612538603, "speed/FLOPS": 202497281126000.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06256221979856491, "optim/lr": 0.0027573212396222262, "optim/total_tokens": 7916224512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.783257484436035, "created_at": "2025-01-17T00:12:36.668004+00:00"} {"global_step": 15100, "acc_step": 0, "speed/wps": 12890.705705189863, "speed/FLOPS": 202466246785907.34, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0499889999628067, "optim/lr": 0.0027572745128719603, "optim/total_tokens": 7916748800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.865955352783203, "created_at": "2025-01-17T00:12:46.845973+00:00"} {"global_step": 15101, "acc_step": 0, "speed/wps": 12886.997044794662, "speed/FLOPS": 202407997178167.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04859744384884834, "optim/lr": 0.002757227782019616, "optim/total_tokens": 7917273088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 3.125302314758301, "created_at": "2025-01-17T00:12:57.020785+00:00"} {"global_step": 15102, "acc_step": 0, "speed/wps": 12887.414944139717, "speed/FLOPS": 202414560861635.25, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057056497782468796, "optim/lr": 0.002757181047065345, "optim/total_tokens": 7917797376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.748434066772461, "created_at": "2025-01-17T00:13:07.196473+00:00"} {"global_step": 15103, "acc_step": 0, "speed/wps": 12890.111533889118, "speed/FLOPS": 202456914509152.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04745384678244591, "optim/lr": 0.0027571343080093013, "optim/total_tokens": 7918321664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.868373155593872, "created_at": "2025-01-17T00:13:17.370352+00:00"} {"global_step": 15104, "acc_step": 0, "speed/wps": 12894.697184083096, "speed/FLOPS": 202528938446792.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04733691364526749, "optim/lr": 0.002757087564851636, "optim/total_tokens": 7918845952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.84621000289917, "created_at": "2025-01-17T00:13:27.538762+00:00"} {"global_step": 15105, "acc_step": 0, "speed/wps": 12890.125568902253, "speed/FLOPS": 202457134948319.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049750495702028275, "optim/lr": 0.002757040817592503, "optim/total_tokens": 7919370240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 2.7986412048339844, "created_at": "2025-01-17T00:13:37.714884+00:00"} {"global_step": 15106, "acc_step": 0, "speed/wps": 12892.98950080403, "speed/FLOPS": 202502116934291.88, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.049662746489048004, "optim/lr": 0.0027569940662320537, "optim/total_tokens": 7919894528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.822506904602051, "created_at": "2025-01-17T00:13:47.887463+00:00"} {"global_step": 15107, "acc_step": 0, "speed/wps": 12896.841778056987, "speed/FLOPS": 202562622242133.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04475075751543045, "optim/lr": 0.0027569473107704407, "optim/total_tokens": 7920418816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.007322311401367, "created_at": "2025-01-17T00:13:58.058399+00:00"} {"global_step": 15108, "acc_step": 0, "speed/wps": 12894.187613914997, "speed/FLOPS": 202520934947079.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04289258271455765, "optim/lr": 0.002756900551207817, "optim/total_tokens": 7920943104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 2.8222615718841553, "created_at": "2025-01-17T00:14:08.229363+00:00"} {"global_step": 15109, "acc_step": 0, "speed/wps": 12892.673992260496, "speed/FLOPS": 202497161439061.5, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0516132116317749, "optim/lr": 0.0027568537875443345, "optim/total_tokens": 7921467392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.937592029571533, "created_at": "2025-01-17T00:14:18.402047+00:00"} {"global_step": 15110, "acc_step": 0, "speed/wps": 12889.363151164127, "speed/FLOPS": 202445160130070.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04901864379644394, "optim/lr": 0.0027568070197801476, "optim/total_tokens": 7921991680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.824429512023926, "created_at": "2025-01-17T00:14:28.577731+00:00"} {"global_step": 15111, "acc_step": 0, "speed/wps": 12887.879213069744, "speed/FLOPS": 202421852843153.7, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05356421321630478, "optim/lr": 0.002756760247915406, "optim/total_tokens": 7922515968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.8198060989379883, "created_at": "2025-01-17T00:14:38.754849+00:00"} {"global_step": 15112, "acc_step": 0, "speed/wps": 12891.830840272658, "speed/FLOPS": 202483918578480.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05021347850561142, "optim/lr": 0.002756713471950265, "optim/total_tokens": 7923040256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.916405200958252, "created_at": "2025-01-17T00:14:48.925584+00:00"} {"global_step": 15113, "acc_step": 0, "speed/wps": 12891.057219509066, "speed/FLOPS": 202471767801321.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05934601649641991, "optim/lr": 0.002756666691884875, "optim/total_tokens": 7923564544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 2.7761993408203125, "created_at": "2025-01-17T00:14:59.098935+00:00"} {"global_step": 15114, "acc_step": 0, "speed/wps": 12891.048748632315, "speed/FLOPS": 202471634754562.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04939550161361694, "optim/lr": 0.0027566199077193906, "optim/total_tokens": 7924088832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.820089817047119, "created_at": "2025-01-17T00:15:09.270124+00:00"} {"global_step": 15115, "acc_step": 0, "speed/wps": 12889.971919757281, "speed/FLOPS": 202454721677361.8, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0491090789437294, "optim/lr": 0.0027565731194539637, "optim/total_tokens": 7924613120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 2.9374008178710938, "created_at": "2025-01-17T00:15:19.445839+00:00"} {"global_step": 15116, "acc_step": 0, "speed/wps": 12896.022557068898, "speed/FLOPS": 202549755250787.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045868124812841415, "optim/lr": 0.002756526327088746, "optim/total_tokens": 7925137408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 2.928638458251953, "created_at": "2025-01-17T00:15:29.617462+00:00"} {"global_step": 15117, "acc_step": 0, "speed/wps": 12892.086321391715, "speed/FLOPS": 202487931260522.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05066535249352455, "optim/lr": 0.0027564795306238912, "optim/total_tokens": 7925661696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8950858116149902, "created_at": "2025-01-17T00:15:39.790481+00:00"} {"global_step": 15118, "acc_step": 0, "speed/wps": 12885.644019773787, "speed/FLOPS": 202386746060963.22, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04693286120891571, "optim/lr": 0.0027564327300595515, "optim/total_tokens": 7926185984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 2.8092477321624756, "created_at": "2025-01-17T00:15:49.966040+00:00"} {"global_step": 15119, "acc_step": 0, "speed/wps": 12893.94689376646, "speed/FLOPS": 202517154106362.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.060748763382434845, "optim/lr": 0.0027563859253958802, "optim/total_tokens": 7926710272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.7972607612609863, "created_at": "2025-01-17T00:16:00.137173+00:00"} {"global_step": 15120, "acc_step": 0, "speed/wps": 12892.62949554365, "speed/FLOPS": 202496462556978.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04527096450328827, "optim/lr": 0.0027563391166330296, "optim/total_tokens": 7927234560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 2.80194354057312, "created_at": "2025-01-17T00:16:10.309942+00:00"} {"global_step": 15121, "acc_step": 0, "speed/wps": 12887.395000342014, "speed/FLOPS": 202414247616886.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05674002692103386, "optim/lr": 0.002756292303771153, "optim/total_tokens": 7927758848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 2.912050485610962, "created_at": "2025-01-17T00:16:20.485576+00:00"} {"global_step": 15122, "acc_step": 0, "speed/wps": 12881.183707670429, "speed/FLOPS": 202316690730268.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06007293611764908, "optim/lr": 0.002756245486810402, "optim/total_tokens": 7928283136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.891207695007324, "created_at": "2025-01-17T00:16:30.666025+00:00"} {"global_step": 15123, "acc_step": 0, "speed/wps": 12878.024926795595, "speed/FLOPS": 202267077736009.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04669296741485596, "optim/lr": 0.0027561986657509293, "optim/total_tokens": 7928807424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.76920223236084, "created_at": "2025-01-17T00:16:40.849685+00:00"} {"global_step": 15124, "acc_step": 0, "speed/wps": 12879.204004811656, "speed/FLOPS": 202285596776475.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.050969354808330536, "optim/lr": 0.0027561518405928892, "optim/total_tokens": 7929331712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.698751449584961, "created_at": "2025-01-17T00:16:51.030226+00:00"} {"global_step": 15125, "acc_step": 0, "speed/wps": 12889.122901498487, "speed/FLOPS": 202441386678934.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0444771982729435, "optim/lr": 0.002756105011336434, "optim/total_tokens": 7929856000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 2.7856814861297607, "created_at": "2025-01-17T00:17:01.203254+00:00"} {"global_step": 15126, "acc_step": 0, "speed/wps": 12897.383370904077, "speed/FLOPS": 202571128702023.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055011019110679626, "optim/lr": 0.002756058177981715, "optim/total_tokens": 7930380288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.7308349609375, "created_at": "2025-01-17T00:17:11.372541+00:00"} {"global_step": 15127, "acc_step": 0, "speed/wps": 12890.766748235043, "speed/FLOPS": 202467205550813.84, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.042218830436468124, "optim/lr": 0.0027560113405288864, "optim/total_tokens": 7930904576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.891972064971924, "created_at": "2025-01-17T00:17:21.544005+00:00"} {"global_step": 15128, "acc_step": 0, "speed/wps": 12891.340400453424, "speed/FLOPS": 202476215547184.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05179551988840103, "optim/lr": 0.0027559644989781007, "optim/total_tokens": 7931428864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 2.7531094551086426, "created_at": "2025-01-17T00:17:31.714973+00:00"} {"global_step": 15129, "acc_step": 0, "speed/wps": 12894.496235701696, "speed/FLOPS": 202525782276330.44, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04369889572262764, "optim/lr": 0.0027559176533295112, "optim/total_tokens": 7931953152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 2.768263578414917, "created_at": "2025-01-17T00:17:41.890524+00:00"} {"global_step": 15130, "acc_step": 0, "speed/wps": 12893.11467102105, "speed/FLOPS": 202504082904552.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047955796122550964, "optim/lr": 0.00275587080358327, "optim/total_tokens": 7932477440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.6925692558288574, "created_at": "2025-01-17T00:17:52.070151+00:00"} {"global_step": 15131, "acc_step": 0, "speed/wps": 12890.744838792392, "speed/FLOPS": 202466861432910.28, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049444347620010376, "optim/lr": 0.00275582394973953, "optim/total_tokens": 7933001728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 2.807483673095703, "created_at": "2025-01-17T00:18:02.242523+00:00"} {"global_step": 15132, "acc_step": 0, "speed/wps": 12891.537376071075, "speed/FLOPS": 202479309319932.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06131665036082268, "optim/lr": 0.0027557770917984448, "optim/total_tokens": 7933526016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.8458924293518066, "created_at": "2025-01-17T00:18:12.417547+00:00"} {"global_step": 15133, "acc_step": 0, "speed/wps": 12890.469725535097, "speed/FLOPS": 202462540401159.22, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.057829465717077255, "optim/lr": 0.0027557302297601663, "optim/total_tokens": 7934050304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 2.8420166969299316, "created_at": "2025-01-17T00:18:22.589110+00:00"} {"global_step": 15134, "acc_step": 0, "speed/wps": 12886.993514140884, "speed/FLOPS": 202407941724398.1, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05893611162900925, "optim/lr": 0.002755683363624848, "optim/total_tokens": 7934574592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8389053344726562, "created_at": "2025-01-17T00:18:32.766934+00:00"} {"global_step": 15135, "acc_step": 0, "speed/wps": 12891.579703074425, "speed/FLOPS": 202479974123683.34, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06654606759548187, "optim/lr": 0.0027556364933926424, "optim/total_tokens": 7935098880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.899121046066284, "created_at": "2025-01-17T00:18:42.937726+00:00"} {"global_step": 15136, "acc_step": 0, "speed/wps": 12889.147587690502, "speed/FLOPS": 202441774409502.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.06162998452782631, "optim/lr": 0.0027555896190637036, "optim/total_tokens": 7935623168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394694, "loss/out": 2.8947553634643555, "created_at": "2025-01-17T00:18:53.112932+00:00"} {"global_step": 15137, "acc_step": 0, "speed/wps": 12886.763192722825, "speed/FLOPS": 202404324210032.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07064922899007797, "optim/lr": 0.002755542740638183, "optim/total_tokens": 7936147456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313754, "loss/out": 2.8504693508148193, "created_at": "2025-01-17T00:19:03.291500+00:00"} {"global_step": 15138, "acc_step": 0, "speed/wps": 12890.56743597506, "speed/FLOPS": 202464075077887.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051561612635850906, "optim/lr": 0.0027554958581162336, "optim/total_tokens": 7936671744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 2.914306163787842, "created_at": "2025-01-17T00:19:13.468041+00:00"} {"global_step": 15139, "acc_step": 0, "speed/wps": 12892.061459788145, "speed/FLOPS": 202487540774875.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05784439668059349, "optim/lr": 0.0027554489714980094, "optim/total_tokens": 7937196032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.756800651550293, "created_at": "2025-01-17T00:19:23.643711+00:00"} {"global_step": 15140, "acc_step": 0, "speed/wps": 12895.974746265629, "speed/FLOPS": 202549004316423.44, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044858697801828384, "optim/lr": 0.0027554020807836633, "optim/total_tokens": 7937720320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.805788040161133, "created_at": "2025-01-17T00:19:33.815079+00:00"} {"global_step": 15141, "acc_step": 0, "speed/wps": 12890.339829781304, "speed/FLOPS": 202460500209850.44, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05638856068253517, "optim/lr": 0.002755355185973348, "optim/total_tokens": 7938244608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.881068706512451, "created_at": "2025-01-17T00:19:43.990049+00:00"} {"global_step": 15142, "acc_step": 0, "speed/wps": 12893.558191243159, "speed/FLOPS": 202511048999100.97, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05268529802560806, "optim/lr": 0.0027553082870672156, "optim/total_tokens": 7938768896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 2.8069355487823486, "created_at": "2025-01-17T00:19:54.159902+00:00"} {"global_step": 15143, "acc_step": 0, "speed/wps": 12890.43140589743, "speed/FLOPS": 202461938538592.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05705057829618454, "optim/lr": 0.0027552613840654207, "optim/total_tokens": 7939293184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.732203245162964, "created_at": "2025-01-17T00:20:04.331623+00:00"} {"global_step": 15144, "acc_step": 0, "speed/wps": 12889.464326074027, "speed/FLOPS": 202446749221059.47, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05337439849972725, "optim/lr": 0.0027552144769681154, "optim/total_tokens": 7939817472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.77363657951355, "created_at": "2025-01-17T00:20:14.505477+00:00"} {"global_step": 15145, "acc_step": 0, "speed/wps": 12890.3789493559, "speed/FLOPS": 202461114636525.38, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.046113912016153336, "optim/lr": 0.0027551675657754527, "optim/total_tokens": 7940341760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 2.803138256072998, "created_at": "2025-01-17T00:20:24.679267+00:00"} {"global_step": 15146, "acc_step": 0, "speed/wps": 12893.142078536928, "speed/FLOPS": 202504513377250.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04459817707538605, "optim/lr": 0.002755120650487586, "optim/total_tokens": 7940866048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.9171743392944336, "created_at": "2025-01-17T00:20:34.850190+00:00"} {"global_step": 15147, "acc_step": 0, "speed/wps": 12894.15161004572, "speed/FLOPS": 202520369456838.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045973144471645355, "optim/lr": 0.002755073731104668, "optim/total_tokens": 7941390336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.868716239929199, "created_at": "2025-01-17T00:20:45.019270+00:00"} {"global_step": 15148, "acc_step": 0, "speed/wps": 12890.13584371296, "speed/FLOPS": 202457296328340.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04551166668534279, "optim/lr": 0.002755026807626852, "optim/total_tokens": 7941914624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 2.8743743896484375, "created_at": "2025-01-17T00:20:55.196086+00:00"} {"global_step": 15149, "acc_step": 0, "speed/wps": 12889.358782264559, "speed/FLOPS": 202445091510499.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.039990395307540894, "optim/lr": 0.0027549798800542914, "optim/total_tokens": 7942438912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 2.878946304321289, "created_at": "2025-01-17T00:21:05.395003+00:00"} {"global_step": 15150, "acc_step": 0, "speed/wps": 12890.132096468826, "speed/FLOPS": 202457237472722.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.044045962393283844, "optim/lr": 0.0027549329483871395, "optim/total_tokens": 7942963200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.8664090633392334, "created_at": "2025-01-17T00:21:15.568969+00:00"} {"global_step": 15151, "acc_step": 0, "speed/wps": 12890.18330663267, "speed/FLOPS": 202458041798715.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053220827132463455, "optim/lr": 0.002754886012625548, "optim/total_tokens": 7943487488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 2.889019012451172, "created_at": "2025-01-17T00:21:25.746262+00:00"} {"global_step": 15152, "acc_step": 0, "speed/wps": 12887.286241644728, "speed/FLOPS": 202412539412095.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041641492396593094, "optim/lr": 0.002754839072769671, "optim/total_tokens": 7944011776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.772782325744629, "created_at": "2025-01-17T00:21:35.920692+00:00"} {"global_step": 15153, "acc_step": 0, "speed/wps": 12889.375900198987, "speed/FLOPS": 202445360371182.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05059783533215523, "optim/lr": 0.0027547921288196622, "optim/total_tokens": 7944536064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.7767574787139893, "created_at": "2025-01-17T00:21:46.095263+00:00"} {"global_step": 15154, "acc_step": 0, "speed/wps": 12890.491622235859, "speed/FLOPS": 202462884318933.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05495697259902954, "optim/lr": 0.0027547451807756742, "optim/total_tokens": 7945060352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.874926805496216, "created_at": "2025-01-17T00:21:56.269572+00:00"} {"global_step": 15155, "acc_step": 0, "speed/wps": 12898.227575622343, "speed/FLOPS": 202584388097182.56, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07162902504205704, "optim/lr": 0.00275469822863786, "optim/total_tokens": 7945584640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.6635313034057617, "created_at": "2025-01-17T00:22:06.436531+00:00"} {"global_step": 15156, "acc_step": 0, "speed/wps": 12888.043106227036, "speed/FLOPS": 202424427010401.78, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05815522372722626, "optim/lr": 0.0027546512724063725, "optim/total_tokens": 7946108928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8407297134399414, "created_at": "2025-01-17T00:22:16.612690+00:00"} {"global_step": 15157, "acc_step": 0, "speed/wps": 12890.518496741872, "speed/FLOPS": 202463306419980.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07194000482559204, "optim/lr": 0.002754604312081366, "optim/total_tokens": 7946633216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345322, "loss/out": 2.8943591117858887, "created_at": "2025-01-17T00:22:26.785135+00:00"} {"global_step": 15158, "acc_step": 0, "speed/wps": 12896.64803562035, "speed/FLOPS": 202559579250935.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0644056424498558, "optim/lr": 0.0027545573476629925, "optim/total_tokens": 7947157504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8039066791534424, "created_at": "2025-01-17T00:22:36.952022+00:00"} {"global_step": 15159, "acc_step": 0, "speed/wps": 12891.847606955027, "speed/FLOPS": 202484181922266.72, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054258082062006, "optim/lr": 0.002754510379151406, "optim/total_tokens": 7947681792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.7398133277893066, "created_at": "2025-01-17T00:22:47.122699+00:00"} {"global_step": 15160, "acc_step": 0, "speed/wps": 12887.2504792793, "speed/FLOPS": 202411977715004.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06492608785629272, "optim/lr": 0.0027544634065467593, "optim/total_tokens": 7948206080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 2.8219590187072754, "created_at": "2025-01-17T00:22:57.297682+00:00"} {"global_step": 15161, "acc_step": 0, "speed/wps": 12882.418544318742, "speed/FLOPS": 202336085536674.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07170268148183823, "optim/lr": 0.0027544164298492066, "optim/total_tokens": 7948730368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 2.7855920791625977, "created_at": "2025-01-17T00:23:07.475951+00:00"} {"global_step": 15162, "acc_step": 0, "speed/wps": 12876.769651129905, "speed/FLOPS": 202247361906750.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05135554075241089, "optim/lr": 0.0027543694490589003, "optim/total_tokens": 7949254656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.780099391937256, "created_at": "2025-01-17T00:23:17.659564+00:00"} {"global_step": 15163, "acc_step": 0, "speed/wps": 12884.555010834158, "speed/FLOPS": 202369641679111.25, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0901956707239151, "optim/lr": 0.0027543224641759937, "optim/total_tokens": 7949778944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.7804458141326904, "created_at": "2025-01-17T00:23:27.837072+00:00"} {"global_step": 15164, "acc_step": 0, "speed/wps": 12886.40724122168, "speed/FLOPS": 202398733502578.56, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04923434555530548, "optim/lr": 0.0027542754752006406, "optim/total_tokens": 7950303232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8584513664245605, "created_at": "2025-01-17T00:23:38.014848+00:00"} {"global_step": 15165, "acc_step": 0, "speed/wps": 12891.392838841975, "speed/FLOPS": 202477039164134.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07247758656740189, "optim/lr": 0.0027542284821329933, "optim/total_tokens": 7950827520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 2.7385339736938477, "created_at": "2025-01-17T00:23:48.188033+00:00"} {"global_step": 15166, "acc_step": 0, "speed/wps": 12893.177136562266, "speed/FLOPS": 202505064011713.53, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05601690709590912, "optim/lr": 0.0027541814849732066, "optim/total_tokens": 7951351808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 2.7001330852508545, "created_at": "2025-01-17T00:23:58.359507+00:00"} {"global_step": 15167, "acc_step": 0, "speed/wps": 12893.260287481611, "speed/FLOPS": 202506370011163.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06376564502716064, "optim/lr": 0.0027541344837214324, "optim/total_tokens": 7951876096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 2.9648022651672363, "created_at": "2025-01-17T00:24:08.531382+00:00"} {"global_step": 15168, "acc_step": 0, "speed/wps": 12882.541277109007, "speed/FLOPS": 202338013223797.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0017, "optim/grad_norm": 0.06848369538784027, "optim/lr": 0.002754087478377825, "optim/total_tokens": 7952400384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.7730772495269775, "created_at": "2025-01-17T00:24:18.709681+00:00"} {"global_step": 15169, "acc_step": 0, "speed/wps": 12889.88055440641, "speed/FLOPS": 202453286658977.25, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04597403109073639, "optim/lr": 0.0027540404689425364, "optim/total_tokens": 7952924672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343445, "loss/out": 2.810120105743408, "created_at": "2025-01-17T00:24:28.884364+00:00"} {"global_step": 15170, "acc_step": 0, "speed/wps": 12886.789432188081, "speed/FLOPS": 202404736336891.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06382911652326584, "optim/lr": 0.002753993455415722, "optim/total_tokens": 7953448960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.8089516162872314, "created_at": "2025-01-17T00:24:39.063035+00:00"} {"global_step": 15171, "acc_step": 0, "speed/wps": 12890.368926816564, "speed/FLOPS": 202460957218772.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04973319172859192, "optim/lr": 0.0027539464377975345, "optim/total_tokens": 7953973248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.8365252017974854, "created_at": "2025-01-17T00:24:49.235765+00:00"} {"global_step": 15172, "acc_step": 0, "speed/wps": 12889.850685132198, "speed/FLOPS": 202452817520981.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048398084938526154, "optim/lr": 0.002753899416088127, "optim/total_tokens": 7954497536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.830599069595337, "created_at": "2025-01-17T00:24:59.411509+00:00"} {"global_step": 15173, "acc_step": 0, "speed/wps": 12894.940181797845, "speed/FLOPS": 202532755059816.34, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048388246446847916, "optim/lr": 0.002753852390287652, "optim/total_tokens": 7955021824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.8204221725463867, "created_at": "2025-01-17T00:25:09.579735+00:00"} {"global_step": 15174, "acc_step": 0, "speed/wps": 12892.277907284495, "speed/FLOPS": 202490940380235.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059797175228595734, "optim/lr": 0.0027538053603962643, "optim/total_tokens": 7955546112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.848236083984375, "created_at": "2025-01-17T00:25:19.750254+00:00"} {"global_step": 15175, "acc_step": 0, "speed/wps": 12891.397780645995, "speed/FLOPS": 202477116781957.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06175084784626961, "optim/lr": 0.002753758326414117, "optim/total_tokens": 7956070400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.796269655227661, "created_at": "2025-01-17T00:25:29.923199+00:00"} {"global_step": 15176, "acc_step": 0, "speed/wps": 12893.83288941359, "speed/FLOPS": 202515363511342.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04657768830657005, "optim/lr": 0.0027537112883413636, "optim/total_tokens": 7956594688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.763936996459961, "created_at": "2025-01-17T00:25:40.092501+00:00"} {"global_step": 15177, "acc_step": 0, "speed/wps": 12890.256893620977, "speed/FLOPS": 202459197583486.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05546823516488075, "optim/lr": 0.0027536642461781576, "optim/total_tokens": 7957118976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285262, "loss/out": 2.8192386627197266, "created_at": "2025-01-17T00:25:50.270611+00:00"} {"global_step": 15178, "acc_step": 0, "speed/wps": 12897.095614087139, "speed/FLOPS": 202566609085794.03, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05942295119166374, "optim/lr": 0.002753617199924652, "optim/total_tokens": 7957643264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 2.7940828800201416, "created_at": "2025-01-17T00:26:00.437249+00:00"} {"global_step": 15179, "acc_step": 0, "speed/wps": 12887.3252919435, "speed/FLOPS": 202413152750698.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0444079227745533, "optim/lr": 0.002753570149581, "optim/total_tokens": 7958167552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477844, "loss/out": 2.8127448558807373, "created_at": "2025-01-17T00:26:10.614054+00:00"} {"global_step": 15180, "acc_step": 0, "speed/wps": 12891.770879487303, "speed/FLOPS": 202482976811951.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04582348093390465, "optim/lr": 0.0027535230951473565, "optim/total_tokens": 7958691840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.766303539276123, "created_at": "2025-01-17T00:26:20.784656+00:00"} {"global_step": 15181, "acc_step": 0, "speed/wps": 12890.250620943561, "speed/FLOPS": 202459099062467.9, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047446321696043015, "optim/lr": 0.002753476036623874, "optim/total_tokens": 7959216128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.803668260574341, "created_at": "2025-01-17T00:26:30.957088+00:00"} {"global_step": 15182, "acc_step": 0, "speed/wps": 12888.943500068115, "speed/FLOPS": 202438568932946.72, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04885772243142128, "optim/lr": 0.0027534289740107063, "optim/total_tokens": 7959740416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.5835728645324707, "created_at": "2025-01-17T00:26:41.133670+00:00"} {"global_step": 15183, "acc_step": 0, "speed/wps": 12889.97533398552, "speed/FLOPS": 202454775302507.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048991329967975616, "optim/lr": 0.002753381907308007, "optim/total_tokens": 7960264704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.7702062129974365, "created_at": "2025-01-17T00:26:51.308947+00:00"} {"global_step": 15184, "acc_step": 0, "speed/wps": 12891.254415550336, "speed/FLOPS": 202474865036125.62, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052285343408584595, "optim/lr": 0.002753334836515929, "optim/total_tokens": 7960788992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 2.769148349761963, "created_at": "2025-01-17T00:27:01.479970+00:00"} {"global_step": 15185, "acc_step": 0, "speed/wps": 12885.90773282591, "speed/FLOPS": 202390888036826.53, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06553669273853302, "optim/lr": 0.002753287761634627, "optim/total_tokens": 7961313280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 2.8972537517547607, "created_at": "2025-01-17T00:27:11.656411+00:00"} {"global_step": 15186, "acc_step": 0, "speed/wps": 12887.800749580909, "speed/FLOPS": 202420620466241.4, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05512167513370514, "optim/lr": 0.0027532406826642542, "optim/total_tokens": 7961837568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 2.7378244400024414, "created_at": "2025-01-17T00:27:21.836188+00:00"} {"global_step": 15187, "acc_step": 0, "speed/wps": 12887.824403949156, "speed/FLOPS": 202420991990599.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05122525990009308, "optim/lr": 0.002753193599604964, "optim/total_tokens": 7962361856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475100, "loss/out": 2.8158578872680664, "created_at": "2025-01-17T00:27:32.017192+00:00"} {"global_step": 15188, "acc_step": 0, "speed/wps": 12887.656639552108, "speed/FLOPS": 202418357020211.16, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07392976433038712, "optim/lr": 0.00275314651245691, "optim/total_tokens": 7962886144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460071, "loss/out": 2.861253499984741, "created_at": "2025-01-17T00:27:42.190980+00:00"} {"global_step": 15189, "acc_step": 0, "speed/wps": 12892.444225972884, "speed/FLOPS": 202493552643782.6, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06587514281272888, "optim/lr": 0.002753099421220246, "optim/total_tokens": 7963410432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 2.7548911571502686, "created_at": "2025-01-17T00:27:52.362318+00:00"} {"global_step": 15190, "acc_step": 0, "speed/wps": 12896.450956976152, "speed/FLOPS": 202556483860014.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.053432244807481766, "optim/lr": 0.0027530523258951253, "optim/total_tokens": 7963934720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.816134214401245, "created_at": "2025-01-17T00:28:02.529850+00:00"} {"global_step": 15191, "acc_step": 0, "speed/wps": 12891.514153255115, "speed/FLOPS": 202478944573696.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06231717765331268, "optim/lr": 0.0027530052264817017, "optim/total_tokens": 7964459008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.7877559661865234, "created_at": "2025-01-17T00:28:12.701164+00:00"} {"global_step": 15192, "acc_step": 0, "speed/wps": 12889.435967379972, "speed/FLOPS": 202446303808800.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04728500545024872, "optim/lr": 0.0027529581229801293, "optim/total_tokens": 7964983296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8446240425109863, "created_at": "2025-01-17T00:28:22.878824+00:00"} {"global_step": 15193, "acc_step": 0, "speed/wps": 12890.396235291733, "speed/FLOPS": 202461386135901.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07648290693759918, "optim/lr": 0.0027529110153905615, "optim/total_tokens": 7965507584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 2.7234058380126953, "created_at": "2025-01-17T00:28:33.054326+00:00"} {"global_step": 15194, "acc_step": 0, "speed/wps": 12889.161659558715, "speed/FLOPS": 202441995427529.12, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05135844647884369, "optim/lr": 0.002752863903713152, "optim/total_tokens": 7966031872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.903163194656372, "created_at": "2025-01-17T00:28:43.226998+00:00"} {"global_step": 15195, "acc_step": 0, "speed/wps": 12893.025434402422, "speed/FLOPS": 202502681320832.56, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05856978893280029, "optim/lr": 0.0027528167879480544, "optim/total_tokens": 7966556160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.877200126647949, "created_at": "2025-01-17T00:28:53.400494+00:00"} {"global_step": 15196, "acc_step": 0, "speed/wps": 12885.675839803722, "speed/FLOPS": 202387245838258.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04643373563885689, "optim/lr": 0.002752769668095422, "optim/total_tokens": 7967080448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476020, "loss/out": 2.7512385845184326, "created_at": "2025-01-17T00:29:03.579542+00:00"} {"global_step": 15197, "acc_step": 0, "speed/wps": 12888.124523261653, "speed/FLOPS": 202425705776809.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047788139432668686, "optim/lr": 0.0027527225441554095, "optim/total_tokens": 7967604736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 2.814052104949951, "created_at": "2025-01-17T00:29:13.755374+00:00"} {"global_step": 15198, "acc_step": 0, "speed/wps": 12887.793378187984, "speed/FLOPS": 202420504688386.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052296750247478485, "optim/lr": 0.0027526754161281706, "optim/total_tokens": 7968129024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 2.8371121883392334, "created_at": "2025-01-17T00:29:23.929934+00:00"} {"global_step": 15199, "acc_step": 0, "speed/wps": 12889.336361746573, "speed/FLOPS": 202444739365455.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1865997314453125, "optim/lr": 0.0027526282840138576, "optim/total_tokens": 7968653312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.7074506282806396, "created_at": "2025-01-17T00:29:34.107513+00:00"} {"global_step": 15200, "acc_step": 0, "speed/wps": 12886.335759483358, "speed/FLOPS": 202397610783652.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061840832233428955, "optim/lr": 0.002752581147812626, "optim/total_tokens": 7969177600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 2.8288016319274902, "created_at": "2025-01-17T00:29:44.282657+00:00"} {"global_step": 15201, "acc_step": 0, "speed/wps": 12887.639849544876, "speed/FLOPS": 202418093310075.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05920623615384102, "optim/lr": 0.0027525340075246285, "optim/total_tokens": 7969701888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.8549466133117676, "created_at": "2025-01-17T00:29:54.457562+00:00"} {"global_step": 15202, "acc_step": 0, "speed/wps": 12888.39100648308, "speed/FLOPS": 202429891261987.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06529519706964493, "optim/lr": 0.0027524868631500204, "optim/total_tokens": 7970226176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.869072437286377, "created_at": "2025-01-17T00:30:04.630773+00:00"} {"global_step": 15203, "acc_step": 0, "speed/wps": 12889.130871706124, "speed/FLOPS": 202441511861997.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058031488209962845, "optim/lr": 0.002752439714688953, "optim/total_tokens": 7970750464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.845799446105957, "created_at": "2025-01-17T00:30:14.803615+00:00"} {"global_step": 15204, "acc_step": 0, "speed/wps": 12892.359997865022, "speed/FLOPS": 202492229725607.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06437979638576508, "optim/lr": 0.0027523925621415826, "optim/total_tokens": 7971274752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 2.748821258544922, "created_at": "2025-01-17T00:30:24.973998+00:00"} {"global_step": 15205, "acc_step": 0, "speed/wps": 12888.053778994672, "speed/FLOPS": 202424594640883.4, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047124605625867844, "optim/lr": 0.002752345405508062, "optim/total_tokens": 7971799040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 2.71091890335083, "created_at": "2025-01-17T00:30:35.147826+00:00"} {"global_step": 15206, "acc_step": 0, "speed/wps": 12888.83949454606, "speed/FLOPS": 202436935383304.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07238958030939102, "optim/lr": 0.0027522982447885443, "optim/total_tokens": 7972323328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.8070812225341797, "created_at": "2025-01-17T00:30:45.326335+00:00"} {"global_step": 15207, "acc_step": 0, "speed/wps": 12886.230802573882, "speed/FLOPS": 202395962291162.78, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07455193996429443, "optim/lr": 0.0027522510799831844, "optim/total_tokens": 7972847616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.8391809463500977, "created_at": "2025-01-17T00:30:55.501298+00:00"} {"global_step": 15208, "acc_step": 0, "speed/wps": 12894.005774990368, "speed/FLOPS": 202518078916895.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06112575903534889, "optim/lr": 0.002752203911092136, "optim/total_tokens": 7973371904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 2.7653512954711914, "created_at": "2025-01-17T00:31:05.674266+00:00"} {"global_step": 15209, "acc_step": 0, "speed/wps": 12889.797857903732, "speed/FLOPS": 202451987796765.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050110504031181335, "optim/lr": 0.002752156738115553, "optim/total_tokens": 7973896192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 2.727663993835449, "created_at": "2025-01-17T00:31:15.846532+00:00"} {"global_step": 15210, "acc_step": 0, "speed/wps": 12893.60245694114, "speed/FLOPS": 202511744252715.2, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05797915533185005, "optim/lr": 0.0027521095610535894, "optim/total_tokens": 7974420480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.7846055030822754, "created_at": "2025-01-17T00:31:26.016064+00:00"} {"global_step": 15211, "acc_step": 0, "speed/wps": 12893.835374835357, "speed/FLOPS": 202515402548306.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04838390648365021, "optim/lr": 0.0027520623799063984, "optim/total_tokens": 7974944768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 2.7122390270233154, "created_at": "2025-01-17T00:31:36.185457+00:00"} {"global_step": 15212, "acc_step": 0, "speed/wps": 12889.765055373588, "speed/FLOPS": 202451472587954.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.07191500067710876, "optim/lr": 0.002752015194674135, "optim/total_tokens": 7975469056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.7715044021606445, "created_at": "2025-01-17T00:31:46.359819+00:00"} {"global_step": 15213, "acc_step": 0, "speed/wps": 12891.943860931686, "speed/FLOPS": 202485693723236.97, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05672946572303772, "optim/lr": 0.0027519680053569523, "optim/total_tokens": 7975993344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 2.8339152336120605, "created_at": "2025-01-17T00:31:56.533957+00:00"} {"global_step": 15214, "acc_step": 0, "speed/wps": 12883.938347682395, "speed/FLOPS": 202359956136928.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051900263875722885, "optim/lr": 0.0027519208119550042, "optim/total_tokens": 7976517632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 2.7498891353607178, "created_at": "2025-01-17T00:32:06.715134+00:00"} {"global_step": 15215, "acc_step": 0, "speed/wps": 12885.383901553103, "speed/FLOPS": 202382660546867.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05094950646162033, "optim/lr": 0.002751873614468446, "optim/total_tokens": 7977041920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 2.7681174278259277, "created_at": "2025-01-17T00:32:16.898880+00:00"} {"global_step": 15216, "acc_step": 0, "speed/wps": 12885.699652723419, "speed/FLOPS": 202387619852884.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06012263149023056, "optim/lr": 0.0027518264128974304, "optim/total_tokens": 7977566208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370920, "loss/out": 2.755457878112793, "created_at": "2025-01-17T00:32:27.076390+00:00"} {"global_step": 15217, "acc_step": 0, "speed/wps": 12889.927030852084, "speed/FLOPS": 202454016635421.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05123870074748993, "optim/lr": 0.002751779207242112, "optim/total_tokens": 7978090496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 2.6909871101379395, "created_at": "2025-01-17T00:32:37.251445+00:00"} {"global_step": 15218, "acc_step": 0, "speed/wps": 12884.972085302756, "speed/FLOPS": 202376192406760.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06015567481517792, "optim/lr": 0.002751731997502644, "optim/total_tokens": 7978614784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 2.766173839569092, "created_at": "2025-01-17T00:32:47.428659+00:00"} {"global_step": 15219, "acc_step": 0, "speed/wps": 12886.081599961632, "speed/FLOPS": 202393618859111.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06297710537910461, "optim/lr": 0.0027516847836791818, "optim/total_tokens": 7979139072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.777848243713379, "created_at": "2025-01-17T00:32:57.604196+00:00"} {"global_step": 15220, "acc_step": 0, "speed/wps": 12884.454186639214, "speed/FLOPS": 202368058096582.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04883377626538277, "optim/lr": 0.0027516375657718785, "optim/total_tokens": 7979663360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8113932609558105, "created_at": "2025-01-17T00:33:07.780577+00:00"} {"global_step": 15221, "acc_step": 0, "speed/wps": 12887.068674294946, "speed/FLOPS": 202409122217898.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04655137658119202, "optim/lr": 0.002751590343780888, "optim/total_tokens": 7980187648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 2.9140145778656006, "created_at": "2025-01-17T00:33:17.954959+00:00"} {"global_step": 15222, "acc_step": 0, "speed/wps": 12886.845813711108, "speed/FLOPS": 202405621886186.78, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06105515733361244, "optim/lr": 0.0027515431177063646, "optim/total_tokens": 7980711936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.804079294204712, "created_at": "2025-01-17T00:33:28.129432+00:00"} {"global_step": 15223, "acc_step": 0, "speed/wps": 12883.570523459808, "speed/FLOPS": 202354178952069.7, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060463301837444305, "optim/lr": 0.0027514958875484623, "optim/total_tokens": 7981236224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 2.7985291481018066, "created_at": "2025-01-17T00:33:38.308079+00:00"} {"global_step": 15224, "acc_step": 0, "speed/wps": 12886.110601084552, "speed/FLOPS": 202394074361598.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04978802055120468, "optim/lr": 0.002751448653307336, "optim/total_tokens": 7981760512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.949981689453125, "created_at": "2025-01-17T00:33:48.483132+00:00"} {"global_step": 15225, "acc_step": 0, "speed/wps": 12881.700270497815, "speed/FLOPS": 202324804059304.7, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06310448795557022, "optim/lr": 0.0027514014149831382, "optim/total_tokens": 7982284800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411310, "loss/out": 2.8054375648498535, "created_at": "2025-01-17T00:33:58.662918+00:00"} {"global_step": 15226, "acc_step": 0, "speed/wps": 12894.644914926768, "speed/FLOPS": 202528117487866.88, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05789002403616905, "optim/lr": 0.0027513541725760248, "optim/total_tokens": 7982809088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.642331123352051, "created_at": "2025-01-17T00:34:08.834734+00:00"} {"global_step": 15227, "acc_step": 0, "speed/wps": 12892.528761071939, "speed/FLOPS": 202494880383676.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04566560313105583, "optim/lr": 0.002751306926086149, "optim/total_tokens": 7983333376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.717931032180786, "created_at": "2025-01-17T00:34:19.009953+00:00"} {"global_step": 15228, "acc_step": 0, "speed/wps": 12891.056947930741, "speed/FLOPS": 202471763535810.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05780649185180664, "optim/lr": 0.0027512596755136652, "optim/total_tokens": 7983857664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 2.839773654937744, "created_at": "2025-01-17T00:34:29.185308+00:00"} {"global_step": 15229, "acc_step": 0, "speed/wps": 12889.809290308254, "speed/FLOPS": 202452167358388.56, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050045400857925415, "optim/lr": 0.0027512124208587275, "optim/total_tokens": 7984381952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290845, "loss/out": 2.6752960681915283, "created_at": "2025-01-17T00:34:39.357498+00:00"} {"global_step": 15230, "acc_step": 0, "speed/wps": 12894.369519400376, "speed/FLOPS": 202523792022691.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05662723630666733, "optim/lr": 0.0027511651621214897, "optim/total_tokens": 7984906240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.855337619781494, "created_at": "2025-01-17T00:34:49.529187+00:00"} {"global_step": 15231, "acc_step": 0, "speed/wps": 12890.361479285353, "speed/FLOPS": 202460840245061.06, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057413723319768906, "optim/lr": 0.0027511178993021067, "optim/total_tokens": 7985430528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 2.8109543323516846, "created_at": "2025-01-17T00:34:59.701600+00:00"} {"global_step": 15232, "acc_step": 0, "speed/wps": 12891.424004743722, "speed/FLOPS": 202477528667447.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05160411447286606, "optim/lr": 0.002751070632400732, "optim/total_tokens": 7985954816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.877908229827881, "created_at": "2025-01-17T00:35:09.877617+00:00"} {"global_step": 15233, "acc_step": 0, "speed/wps": 12892.85188652995, "speed/FLOPS": 202499955513015.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06215549260377884, "optim/lr": 0.00275102336141752, "optim/total_tokens": 7986479104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 2.8465824127197266, "created_at": "2025-01-17T00:35:20.047385+00:00"} {"global_step": 15234, "acc_step": 0, "speed/wps": 12887.875382163666, "speed/FLOPS": 202421792673509.5, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06152955815196037, "optim/lr": 0.002750976086352625, "optim/total_tokens": 7987003392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8083934783935547, "created_at": "2025-01-17T00:35:30.222875+00:00"} {"global_step": 15235, "acc_step": 0, "speed/wps": 12902.524289787534, "speed/FLOPS": 202651873897450.88, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05453650280833244, "optim/lr": 0.0027509288072062018, "optim/total_tokens": 7987527680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447559, "loss/out": 2.77323842048645, "created_at": "2025-01-17T00:35:40.387640+00:00"} {"global_step": 15236, "acc_step": 0, "speed/wps": 12892.61599600322, "speed/FLOPS": 202496250528145.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045711785554885864, "optim/lr": 0.0027508815239784042, "optim/total_tokens": 7988051968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 2.8328568935394287, "created_at": "2025-01-17T00:35:50.557994+00:00"} {"global_step": 15237, "acc_step": 0, "speed/wps": 12891.916471500035, "speed/FLOPS": 202485263534575.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05010780319571495, "optim/lr": 0.002750834236669386, "optim/total_tokens": 7988576256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.872002124786377, "created_at": "2025-01-17T00:36:00.732202+00:00"} {"global_step": 15238, "acc_step": 0, "speed/wps": 12888.623328914317, "speed/FLOPS": 202433540205015.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05823739618062973, "optim/lr": 0.0027507869452793026, "optim/total_tokens": 7989100544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.894894599914551, "created_at": "2025-01-17T00:36:10.906075+00:00"} {"global_step": 15239, "acc_step": 0, "speed/wps": 12891.883352325915, "speed/FLOPS": 202484743352435.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0538962297141552, "optim/lr": 0.002750739649808307, "optim/total_tokens": 7989624832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.807877540588379, "created_at": "2025-01-17T00:36:21.086137+00:00"} {"global_step": 15240, "acc_step": 0, "speed/wps": 12897.14033783121, "speed/FLOPS": 202567311533651.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.049013178795576096, "optim/lr": 0.002750692350256554, "optim/total_tokens": 7990149120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 2.805950164794922, "created_at": "2025-01-17T00:36:31.254914+00:00"} {"global_step": 15241, "acc_step": 0, "speed/wps": 12896.395649492973, "speed/FLOPS": 202555615179990.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.061642225831747055, "optim/lr": 0.0027506450466241988, "optim/total_tokens": 7990673408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.8927667140960693, "created_at": "2025-01-17T00:36:41.423684+00:00"} {"global_step": 15242, "acc_step": 0, "speed/wps": 12888.307892391405, "speed/FLOPS": 202428585840967.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.04777277633547783, "optim/lr": 0.0027505977389113943, "optim/total_tokens": 7991197696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.7593955993652344, "created_at": "2025-01-17T00:36:51.599590+00:00"} {"global_step": 15243, "acc_step": 0, "speed/wps": 12890.849075792232, "speed/FLOPS": 202468498618229.4, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04338625818490982, "optim/lr": 0.002750550427118296, "optim/total_tokens": 7991721984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 2.909149169921875, "created_at": "2025-01-17T00:37:01.774173+00:00"} {"global_step": 15244, "acc_step": 0, "speed/wps": 12897.14654275665, "speed/FLOPS": 202567408990532.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047255195677280426, "optim/lr": 0.002750503111245057, "optim/total_tokens": 7992246272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 2.924842119216919, "created_at": "2025-01-17T00:37:11.940991+00:00"} {"global_step": 15245, "acc_step": 0, "speed/wps": 12888.234046752534, "speed/FLOPS": 202427425993734.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04598868265748024, "optim/lr": 0.002750455791291833, "optim/total_tokens": 7992770560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.7209880352020264, "created_at": "2025-01-17T00:37:22.114442+00:00"} {"global_step": 15246, "acc_step": 0, "speed/wps": 12891.8291326668, "speed/FLOPS": 202483891758183.78, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046069879084825516, "optim/lr": 0.002750408467258778, "optim/total_tokens": 7993294848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.836002826690674, "created_at": "2025-01-17T00:37:32.287417+00:00"} {"global_step": 15247, "acc_step": 0, "speed/wps": 12891.232268288468, "speed/FLOPS": 202474517182944.47, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048704203218221664, "optim/lr": 0.0027503611391460465, "optim/total_tokens": 7993819136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 2.780305862426758, "created_at": "2025-01-17T00:37:42.461916+00:00"} {"global_step": 15248, "acc_step": 0, "speed/wps": 12891.858529294674, "speed/FLOPS": 202484353472619.72, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049000728875398636, "optim/lr": 0.0027503138069537923, "optim/total_tokens": 7994343424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.751152753829956, "created_at": "2025-01-17T00:37:52.636952+00:00"} {"global_step": 15249, "acc_step": 0, "speed/wps": 12889.191099503181, "speed/FLOPS": 202442457822311.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05579526722431183, "optim/lr": 0.0027502664706821702, "optim/total_tokens": 7994867712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.76627779006958, "created_at": "2025-01-17T00:38:02.811504+00:00"} {"global_step": 15250, "acc_step": 0, "speed/wps": 12887.83311842772, "speed/FLOPS": 202421128863460.03, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05402437224984169, "optim/lr": 0.002750219130331335, "optim/total_tokens": 7995392000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.5807652473449707, "created_at": "2025-01-17T00:38:12.988415+00:00"} {"global_step": 15251, "acc_step": 0, "speed/wps": 12893.15496970066, "speed/FLOPS": 202504715850691.53, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04635195806622505, "optim/lr": 0.0027501717859014403, "optim/total_tokens": 7995916288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.793846607208252, "created_at": "2025-01-17T00:38:23.160653+00:00"} {"global_step": 15252, "acc_step": 0, "speed/wps": 12896.970997674836, "speed/FLOPS": 202564651813797.94, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05995112657546997, "optim/lr": 0.0027501244373926413, "optim/total_tokens": 7996440576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286610, "loss/out": 2.8179256916046143, "created_at": "2025-01-17T00:38:33.328428+00:00"} {"global_step": 15253, "acc_step": 0, "speed/wps": 12887.298499195424, "speed/FLOPS": 202412731933773.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050611186772584915, "optim/lr": 0.0027500770848050923, "optim/total_tokens": 7996964864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8409581184387207, "created_at": "2025-01-17T00:38:43.502610+00:00"} {"global_step": 15254, "acc_step": 0, "speed/wps": 12890.446246819545, "speed/FLOPS": 202462171635668.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04264276102185249, "optim/lr": 0.002750029728138948, "optim/total_tokens": 7997489152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.8631887435913086, "created_at": "2025-01-17T00:38:53.678554+00:00"} {"global_step": 15255, "acc_step": 0, "speed/wps": 12890.487003449694, "speed/FLOPS": 202462811774550.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06229591742157936, "optim/lr": 0.0027499823673943625, "optim/total_tokens": 7998013440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 2.8022193908691406, "created_at": "2025-01-17T00:39:03.855632+00:00"} {"global_step": 15256, "acc_step": 0, "speed/wps": 12892.279488642318, "speed/FLOPS": 202490965217633.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.12073801457881927, "optim/lr": 0.0027499350025714906, "optim/total_tokens": 7998537728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 2.8237709999084473, "created_at": "2025-01-17T00:39:14.025921+00:00"} {"global_step": 15257, "acc_step": 0, "speed/wps": 12894.673002368389, "speed/FLOPS": 202528558639733.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0649324581027031, "optim/lr": 0.002749887633670486, "optim/total_tokens": 7999062016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.7859010696411133, "created_at": "2025-01-17T00:39:24.194236+00:00"} {"global_step": 15258, "acc_step": 0, "speed/wps": 12890.83628369795, "speed/FLOPS": 202468297700810.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06093691661953926, "optim/lr": 0.0027498402606915048, "optim/total_tokens": 7999586304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 2.8374276161193848, "created_at": "2025-01-17T00:39:34.370117+00:00"} {"global_step": 15259, "acc_step": 0, "speed/wps": 12887.341991443922, "speed/FLOPS": 202413415039300.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07147469371557236, "optim/lr": 0.002749792883634701, "optim/total_tokens": 8000110592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.846181869506836, "created_at": "2025-01-17T00:39:44.545998+00:00"} {"global_step": 15260, "acc_step": 0, "speed/wps": 12884.045931721703, "speed/FLOPS": 202361645892102.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04626918584108353, "optim/lr": 0.0027497455025002285, "optim/total_tokens": 8000634880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 2.6454033851623535, "created_at": "2025-01-17T00:39:54.724943+00:00"} {"global_step": 15261, "acc_step": 0, "speed/wps": 12872.496516421632, "speed/FLOPS": 202180246454254.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06782728433609009, "optim/lr": 0.002749698117288242, "optim/total_tokens": 8001159168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.7323594093322754, "created_at": "2025-01-17T00:40:04.912572+00:00"} {"global_step": 15262, "acc_step": 0, "speed/wps": 12881.363166674979, "speed/FLOPS": 202319509380537.84, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44769415259361267, "optim/lr": 0.002749650727998897, "optim/total_tokens": 8001683456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.7660038471221924, "created_at": "2025-01-17T00:40:15.093086+00:00"} {"global_step": 15263, "acc_step": 0, "speed/wps": 12884.702521922056, "speed/FLOPS": 202371958543448.16, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07165628671646118, "optim/lr": 0.002749603334632347, "optim/total_tokens": 8002207744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.715681552886963, "created_at": "2025-01-17T00:40:25.270821+00:00"} {"global_step": 15264, "acc_step": 0, "speed/wps": 12887.771746750763, "speed/FLOPS": 202420164936939.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08738432079553604, "optim/lr": 0.0027495559371887475, "optim/total_tokens": 8002732032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.864206314086914, "created_at": "2025-01-17T00:40:35.444642+00:00"} {"global_step": 15265, "acc_step": 0, "speed/wps": 12886.115970672121, "speed/FLOPS": 202394158698350.06, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09357073903083801, "optim/lr": 0.0027495085356682527, "optim/total_tokens": 8003256320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 2.9349417686462402, "created_at": "2025-01-17T00:40:45.622626+00:00"} {"global_step": 15266, "acc_step": 0, "speed/wps": 12886.871278247105, "speed/FLOPS": 202406021841716.56, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0722474679350853, "optim/lr": 0.0027494611300710172, "optim/total_tokens": 8003780608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.866701126098633, "created_at": "2025-01-17T00:40:55.797761+00:00"} {"global_step": 15267, "acc_step": 0, "speed/wps": 12895.218709144963, "speed/FLOPS": 202537129714539.5, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06804600358009338, "optim/lr": 0.002749413720397196, "optim/total_tokens": 8004304896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.85477876663208, "created_at": "2025-01-17T00:41:05.970043+00:00"} {"global_step": 15268, "acc_step": 0, "speed/wps": 12889.59146422649, "speed/FLOPS": 202448746100450.06, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06487145274877548, "optim/lr": 0.0027493663066469436, "optim/total_tokens": 8004829184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387505, "loss/out": 2.9000258445739746, "created_at": "2025-01-17T00:41:16.142912+00:00"} {"global_step": 15269, "acc_step": 0, "speed/wps": 12884.496988514178, "speed/FLOPS": 202368730358844.94, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0711141973733902, "optim/lr": 0.0027493188888204146, "optim/total_tokens": 8005353472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 2.7600817680358887, "created_at": "2025-01-17T00:41:26.326415+00:00"} {"global_step": 15270, "acc_step": 0, "speed/wps": 12888.625309507546, "speed/FLOPS": 202433571312953.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06583938747644424, "optim/lr": 0.0027492714669177637, "optim/total_tokens": 8005877760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.834017753601074, "created_at": "2025-01-17T00:41:36.499850+00:00"} {"global_step": 15271, "acc_step": 0, "speed/wps": 12875.78043373924, "speed/FLOPS": 202231824888303.12, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.001, "optim/grad_norm": 10.355244636535645, "optim/lr": 0.002749224040939146, "optim/total_tokens": 8006402048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.999094009399414, "created_at": "2025-01-17T00:41:46.687457+00:00"} {"global_step": 15272, "acc_step": 0, "speed/wps": 12881.32776847236, "speed/FLOPS": 202318953403123.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08499336987733841, "optim/lr": 0.002749176610884716, "optim/total_tokens": 8006926336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 2.932738780975342, "created_at": "2025-01-17T00:41:56.867104+00:00"} {"global_step": 15273, "acc_step": 0, "speed/wps": 12888.17140013123, "speed/FLOPS": 202426442042461.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05958491563796997, "optim/lr": 0.0027491291767546286, "optim/total_tokens": 8007450624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.816300868988037, "created_at": "2025-01-17T00:42:07.040742+00:00"} {"global_step": 15274, "acc_step": 0, "speed/wps": 12887.833728583282, "speed/FLOPS": 202421138446791.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06435991078615189, "optim/lr": 0.002749081738549038, "optim/total_tokens": 8007974912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 2.7477712631225586, "created_at": "2025-01-17T00:42:17.214904+00:00"} {"global_step": 15275, "acc_step": 0, "speed/wps": 12890.3622042032, "speed/FLOPS": 202460851630892.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17494341731071472, "optim/lr": 0.0027490342962680995, "optim/total_tokens": 8008499200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.85319185256958, "created_at": "2025-01-17T00:42:27.392733+00:00"} {"global_step": 15276, "acc_step": 0, "speed/wps": 12887.994040992891, "speed/FLOPS": 202423656373476.8, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06293713301420212, "optim/lr": 0.0027489868499119678, "optim/total_tokens": 8009023488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 2.821673631668091, "created_at": "2025-01-17T00:42:37.566920+00:00"} {"global_step": 15277, "acc_step": 0, "speed/wps": 12890.12816623307, "speed/FLOPS": 202457175742969.22, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05834242329001427, "optim/lr": 0.002748939399480797, "optim/total_tokens": 8009547776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.765994071960449, "created_at": "2025-01-17T00:42:47.738791+00:00"} {"global_step": 15278, "acc_step": 0, "speed/wps": 12890.953601401634, "speed/FLOPS": 202470140336558.3, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05845656245946884, "optim/lr": 0.0027488919449747437, "optim/total_tokens": 8010072064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350412, "loss/out": 2.8756778240203857, "created_at": "2025-01-17T00:42:57.912491+00:00"} {"global_step": 15279, "acc_step": 0, "speed/wps": 12893.31506252646, "speed/FLOPS": 202507230328511.34, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06614059209823608, "optim/lr": 0.002748844486393961, "optim/total_tokens": 8010596352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 2.8309805393218994, "created_at": "2025-01-17T00:43:08.083301+00:00"} {"global_step": 15280, "acc_step": 0, "speed/wps": 12887.534667228734, "speed/FLOPS": 202416441277265.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04852813482284546, "optim/lr": 0.0027487970237386043, "optim/total_tokens": 8011120640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7883384227752686, "created_at": "2025-01-17T00:43:18.265694+00:00"} {"global_step": 15281, "acc_step": 0, "speed/wps": 12887.159060459097, "speed/FLOPS": 202410541856810.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05931849032640457, "optim/lr": 0.002748749557008829, "optim/total_tokens": 8011644928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.77809476852417, "created_at": "2025-01-17T00:43:28.441523+00:00"} {"global_step": 15282, "acc_step": 0, "speed/wps": 12889.048697119288, "speed/FLOPS": 202440221197191.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04840558022260666, "optim/lr": 0.0027487020862047885, "optim/total_tokens": 8012169216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 2.723074436187744, "created_at": "2025-01-17T00:43:38.614834+00:00"} {"global_step": 15283, "acc_step": 0, "speed/wps": 12888.029477023423, "speed/FLOPS": 202424212945030.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04881679266691208, "optim/lr": 0.0027486546113266396, "optim/total_tokens": 8012693504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 2.823908805847168, "created_at": "2025-01-17T00:43:48.788424+00:00"} {"global_step": 15284, "acc_step": 0, "speed/wps": 12892.51108675871, "speed/FLOPS": 202494602784300.34, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04431973397731781, "optim/lr": 0.002748607132374536, "optim/total_tokens": 8013217792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.7540080547332764, "created_at": "2025-01-17T00:43:58.958874+00:00"} {"global_step": 15285, "acc_step": 0, "speed/wps": 12892.770799516493, "speed/FLOPS": 202498681930044.06, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055615656077861786, "optim/lr": 0.0027485596493486324, "optim/total_tokens": 8013742080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.797323226928711, "created_at": "2025-01-17T00:44:09.129040+00:00"} {"global_step": 15286, "acc_step": 0, "speed/wps": 12892.984870778637, "speed/FLOPS": 202502044213380.75, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04637965187430382, "optim/lr": 0.0027485121622490847, "optim/total_tokens": 8014266368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 2.7968664169311523, "created_at": "2025-01-17T00:44:19.298827+00:00"} {"global_step": 15287, "acc_step": 0, "speed/wps": 12886.112696532828, "speed/FLOPS": 202394107273493.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04870151728391647, "optim/lr": 0.0027484646710760467, "optim/total_tokens": 8014790656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.750702381134033, "created_at": "2025-01-17T00:44:29.478779+00:00"} {"global_step": 15288, "acc_step": 0, "speed/wps": 12895.805810116171, "speed/FLOPS": 202546350942053.6, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05093216896057129, "optim/lr": 0.002748417175829674, "optim/total_tokens": 8015314944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.762071132659912, "created_at": "2025-01-17T00:44:39.647461+00:00"} {"global_step": 15289, "acc_step": 0, "speed/wps": 12890.886648140626, "speed/FLOPS": 202469088743590.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04333789646625519, "optim/lr": 0.002748369676510122, "optim/total_tokens": 8015839232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465606, "loss/out": 2.8431556224823, "created_at": "2025-01-17T00:44:49.818954+00:00"} {"global_step": 15290, "acc_step": 0, "speed/wps": 12893.095883640186, "speed/FLOPS": 202503787822919.06, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05792423337697983, "optim/lr": 0.0027483221731175456, "optim/total_tokens": 8016363520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 2.903813123703003, "created_at": "2025-01-17T00:44:59.992069+00:00"} {"global_step": 15291, "acc_step": 0, "speed/wps": 12893.076850635836, "speed/FLOPS": 202503488883431.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04906465485692024, "optim/lr": 0.0027482746656520987, "optim/total_tokens": 8016887808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.8634657859802246, "created_at": "2025-01-17T00:45:10.162215+00:00"} {"global_step": 15292, "acc_step": 0, "speed/wps": 12893.435396097833, "speed/FLOPS": 202509120332605.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045684121549129486, "optim/lr": 0.002748227154113937, "optim/total_tokens": 8017412096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.7888448238372803, "created_at": "2025-01-17T00:45:20.332628+00:00"} {"global_step": 15293, "acc_step": 0, "speed/wps": 12892.520272907575, "speed/FLOPS": 202494747065391.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05787830799818039, "optim/lr": 0.0027481796385032155, "optim/total_tokens": 8017936384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 2.736438512802124, "created_at": "2025-01-17T00:45:30.502682+00:00"} {"global_step": 15294, "acc_step": 0, "speed/wps": 12889.297300439706, "speed/FLOPS": 202444125853955.2, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04834561049938202, "optim/lr": 0.0027481321188200895, "optim/total_tokens": 8018460672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.775432825088501, "created_at": "2025-01-17T00:45:40.678142+00:00"} {"global_step": 15295, "acc_step": 0, "speed/wps": 12893.102683517658, "speed/FLOPS": 202503894624338.75, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048786453902721405, "optim/lr": 0.0027480845950647137, "optim/total_tokens": 8018984960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.7559242248535156, "created_at": "2025-01-17T00:45:50.850626+00:00"} {"global_step": 15296, "acc_step": 0, "speed/wps": 12898.835510230218, "speed/FLOPS": 202593936545588.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04769157990813255, "optim/lr": 0.0027480370672372427, "optim/total_tokens": 8019509248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.8013172149658203, "created_at": "2025-01-17T00:46:01.018431+00:00"} {"global_step": 15297, "acc_step": 0, "speed/wps": 12895.262992153088, "speed/FLOPS": 202537825240033.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044742193073034286, "optim/lr": 0.002747989535337833, "optim/total_tokens": 8020033536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 2.8037753105163574, "created_at": "2025-01-17T00:46:11.189390+00:00"} {"global_step": 15298, "acc_step": 0, "speed/wps": 12891.749566529512, "speed/FLOPS": 202482642062662.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05233697593212128, "optim/lr": 0.002747941999366638, "optim/total_tokens": 8020557824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.9149317741394043, "created_at": "2025-01-17T00:46:21.364046+00:00"} {"global_step": 15299, "acc_step": 0, "speed/wps": 12890.877725263663, "speed/FLOPS": 202468948597546.72, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.044699039310216904, "optim/lr": 0.002747894459323814, "optim/total_tokens": 8021082112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 2.8485753536224365, "created_at": "2025-01-17T00:46:31.539781+00:00"} {"global_step": 15300, "acc_step": 0, "speed/wps": 12890.457086499526, "speed/FLOPS": 202462341887737.72, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05245489999651909, "optim/lr": 0.002747846915209515, "optim/total_tokens": 8021606400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 2.8143649101257324, "created_at": "2025-01-17T00:46:41.713041+00:00"} {"global_step": 15301, "acc_step": 0, "speed/wps": 12893.201343377194, "speed/FLOPS": 202505444213005.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04844881594181061, "optim/lr": 0.0027477993670238974, "optim/total_tokens": 8022130688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 2.9005908966064453, "created_at": "2025-01-17T00:46:51.889765+00:00"} {"global_step": 15302, "acc_step": 0, "speed/wps": 12893.38093895483, "speed/FLOPS": 202508265008342.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05061299726366997, "optim/lr": 0.0027477518147671154, "optim/total_tokens": 8022654976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.8831334114074707, "created_at": "2025-01-17T00:47:02.059499+00:00"} {"global_step": 15303, "acc_step": 0, "speed/wps": 12892.15605650482, "speed/FLOPS": 202489026546300.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.001, "optim/grad_norm": 0.061006058007478714, "optim/lr": 0.0027477042584393247, "optim/total_tokens": 8023179264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.763063430786133, "created_at": "2025-01-17T00:47:12.230725+00:00"} {"global_step": 15304, "acc_step": 0, "speed/wps": 12887.560750517558, "speed/FLOPS": 202416850951158.2, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04366537928581238, "optim/lr": 0.0027476566980406806, "optim/total_tokens": 8023703552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 2.8030967712402344, "created_at": "2025-01-17T00:47:22.407839+00:00"} {"global_step": 15305, "acc_step": 0, "speed/wps": 12891.16809999069, "speed/FLOPS": 202473509331651.38, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0455721877515316, "optim/lr": 0.0027476091335713374, "optim/total_tokens": 8024227840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 2.797752857208252, "created_at": "2025-01-17T00:47:32.579122+00:00"} {"global_step": 15306, "acc_step": 0, "speed/wps": 12890.807649760209, "speed/FLOPS": 202467847965471.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.055611129850149155, "optim/lr": 0.002747561565031451, "optim/total_tokens": 8024752128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.831787586212158, "created_at": "2025-01-17T00:47:42.750613+00:00"} {"global_step": 15307, "acc_step": 0, "speed/wps": 12894.07759701387, "speed/FLOPS": 202519206980468.28, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05675887688994408, "optim/lr": 0.0027475139924211765, "optim/total_tokens": 8025276416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.727205991744995, "created_at": "2025-01-17T00:47:52.921868+00:00"} {"global_step": 15308, "acc_step": 0, "speed/wps": 12893.997078151078, "speed/FLOPS": 202517942321083.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04985078051686287, "optim/lr": 0.0027474664157406685, "optim/total_tokens": 8025800704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 2.931966781616211, "created_at": "2025-01-17T00:48:03.096107+00:00"} {"global_step": 15309, "acc_step": 0, "speed/wps": 12885.628510353406, "speed/FLOPS": 202386502464203.97, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05885282903909683, "optim/lr": 0.0027474188349900833, "optim/total_tokens": 8026324992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8523783683776855, "created_at": "2025-01-17T00:48:13.271736+00:00"} {"global_step": 15310, "acc_step": 0, "speed/wps": 12895.28006905914, "speed/FLOPS": 202538093456309.0, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.042559683322906494, "optim/lr": 0.0027473712501695753, "optim/total_tokens": 8026849280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.8923134803771973, "created_at": "2025-01-17T00:48:23.440673+00:00"} {"global_step": 15311, "acc_step": 0, "speed/wps": 12893.810106031617, "speed/FLOPS": 202515005667020.88, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0635935440659523, "optim/lr": 0.0027473236612793003, "optim/total_tokens": 8027373568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.77908992767334, "created_at": "2025-01-17T00:48:33.618112+00:00"} {"global_step": 15312, "acc_step": 0, "speed/wps": 12890.380612893505, "speed/FLOPS": 202461140764669.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06301477551460266, "optim/lr": 0.0027472760683194125, "optim/total_tokens": 8027897856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 2.817640781402588, "created_at": "2025-01-17T00:48:43.791601+00:00"} {"global_step": 15313, "acc_step": 0, "speed/wps": 12888.424786384501, "speed/FLOPS": 202430421821758.25, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.039347968995571136, "optim/lr": 0.002747228471290069, "optim/total_tokens": 8028422144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.770695209503174, "created_at": "2025-01-17T00:48:53.965178+00:00"} {"global_step": 15314, "acc_step": 0, "speed/wps": 12889.401222454038, "speed/FLOPS": 202445758091994.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047129541635513306, "optim/lr": 0.0027471808701914242, "optim/total_tokens": 8028946432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 2.910781145095825, "created_at": "2025-01-17T00:49:04.137691+00:00"} {"global_step": 15315, "acc_step": 0, "speed/wps": 12891.173663902226, "speed/FLOPS": 202473596720527.66, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04343010112643242, "optim/lr": 0.0027471332650236325, "optim/total_tokens": 8029470720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 2.746183395385742, "created_at": "2025-01-17T00:49:14.310970+00:00"} {"global_step": 15316, "acc_step": 0, "speed/wps": 12893.089741669131, "speed/FLOPS": 202503691354823.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.37891414761543274, "optim/lr": 0.0027470856557868504, "optim/total_tokens": 8029995008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 2.8973255157470703, "created_at": "2025-01-17T00:49:24.481757+00:00"} {"global_step": 15317, "acc_step": 0, "speed/wps": 12900.390188662393, "speed/FLOPS": 202618354906718.3, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3595644235610962, "optim/lr": 0.0027470380424812323, "optim/total_tokens": 8030519296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 2.8884897232055664, "created_at": "2025-01-17T00:49:34.648101+00:00"} {"global_step": 15318, "acc_step": 0, "speed/wps": 12889.415889835287, "speed/FLOPS": 202445988463371.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18888011574745178, "optim/lr": 0.0027469904251069344, "optim/total_tokens": 8031043584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.841444492340088, "created_at": "2025-01-17T00:49:44.822732+00:00"} {"global_step": 15319, "acc_step": 0, "speed/wps": 12887.925958332646, "speed/FLOPS": 202422587041742.28, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3257773816585541, "optim/lr": 0.0027469428036641123, "optim/total_tokens": 8031567872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.7935023307800293, "created_at": "2025-01-17T00:49:54.999609+00:00"} {"global_step": 15320, "acc_step": 0, "speed/wps": 12891.470099096987, "speed/FLOPS": 202478252642606.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14658677577972412, "optim/lr": 0.0027468951781529204, "optim/total_tokens": 8032092160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 2.7602901458740234, "created_at": "2025-01-17T00:50:05.170599+00:00"} {"global_step": 15321, "acc_step": 0, "speed/wps": 12890.845387170231, "speed/FLOPS": 202468440683352.2, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1563521772623062, "optim/lr": 0.002746847548573514, "optim/total_tokens": 8032616448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.9093308448791504, "created_at": "2025-01-17T00:50:15.344359+00:00"} {"global_step": 15322, "acc_step": 0, "speed/wps": 12887.809445144594, "speed/FLOPS": 202420757042018.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26749685406684875, "optim/lr": 0.0027467999149260497, "optim/total_tokens": 8033140736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 2.8734283447265625, "created_at": "2025-01-17T00:50:25.519185+00:00"} {"global_step": 15323, "acc_step": 0, "speed/wps": 12892.245940083476, "speed/FLOPS": 202490438291416.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18167239427566528, "optim/lr": 0.002746752277210682, "optim/total_tokens": 8033665024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.7737159729003906, "created_at": "2025-01-17T00:50:35.693563+00:00"} {"global_step": 15324, "acc_step": 0, "speed/wps": 12892.03122257661, "speed/FLOPS": 202487065857917.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06360422074794769, "optim/lr": 0.0027467046354275665, "optim/total_tokens": 8034189312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 2.9391887187957764, "created_at": "2025-01-17T00:50:45.864262+00:00"} {"global_step": 15325, "acc_step": 0, "speed/wps": 12890.293180347282, "speed/FLOPS": 202459767516386.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08880582451820374, "optim/lr": 0.002746656989576859, "optim/total_tokens": 8034713600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.9006552696228027, "created_at": "2025-01-17T00:50:56.036136+00:00"} {"global_step": 15326, "acc_step": 0, "speed/wps": 12897.694852452409, "speed/FLOPS": 202576020947760.4, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06747116893529892, "optim/lr": 0.002746609339658714, "optim/total_tokens": 8035237888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 2.8500452041625977, "created_at": "2025-01-17T00:51:06.202916+00:00"} {"global_step": 15327, "acc_step": 0, "speed/wps": 12895.61940352027, "speed/FLOPS": 202543423170315.47, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05190776288509369, "optim/lr": 0.0027465616856732883, "optim/total_tokens": 8035762176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.9916396141052246, "created_at": "2025-01-17T00:51:16.371543+00:00"} {"global_step": 15328, "acc_step": 0, "speed/wps": 12890.6318332519, "speed/FLOPS": 202465086525610.97, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05565759539604187, "optim/lr": 0.0027465140276207362, "optim/total_tokens": 8036286464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 2.79787015914917, "created_at": "2025-01-17T00:51:26.543225+00:00"} {"global_step": 15329, "acc_step": 0, "speed/wps": 12895.726789245065, "speed/FLOPS": 202545109810686.84, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.08042051643133163, "optim/lr": 0.002746466365501214, "optim/total_tokens": 8036810752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9266409873962402, "created_at": "2025-01-17T00:51:36.719497+00:00"} {"global_step": 15330, "acc_step": 0, "speed/wps": 12892.285159803287, "speed/FLOPS": 202491054291009.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05228053033351898, "optim/lr": 0.002746418699314877, "optim/total_tokens": 8037335040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.7754368782043457, "created_at": "2025-01-17T00:51:46.891719+00:00"} {"global_step": 15331, "acc_step": 0, "speed/wps": 12895.134457777905, "speed/FLOPS": 202535806431046.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05448329076170921, "optim/lr": 0.0027463710290618797, "optim/total_tokens": 8037859328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 2.8954713344573975, "created_at": "2025-01-17T00:51:57.063427+00:00"} {"global_step": 15332, "acc_step": 0, "speed/wps": 12893.191277469949, "speed/FLOPS": 202505286114100.34, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05527767539024353, "optim/lr": 0.0027463233547423793, "optim/total_tokens": 8038383616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.860095500946045, "created_at": "2025-01-17T00:52:07.235208+00:00"} {"global_step": 15333, "acc_step": 0, "speed/wps": 12891.10302548975, "speed/FLOPS": 202472487247192.38, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.047901496291160583, "optim/lr": 0.0027462756763565305, "optim/total_tokens": 8038907904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 2.900132179260254, "created_at": "2025-01-17T00:52:17.407346+00:00"} {"global_step": 15334, "acc_step": 0, "speed/wps": 12888.354774135742, "speed/FLOPS": 202429322183184.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048392362892627716, "optim/lr": 0.002746227993904489, "optim/total_tokens": 8039432192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 2.8343515396118164, "created_at": "2025-01-17T00:52:27.580635+00:00"} {"global_step": 15335, "acc_step": 0, "speed/wps": 12891.183074858602, "speed/FLOPS": 202473744532529.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04664321616292, "optim/lr": 0.00274618030738641, "optim/total_tokens": 8039956480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.8877315521240234, "created_at": "2025-01-17T00:52:37.754571+00:00"} {"global_step": 15336, "acc_step": 0, "speed/wps": 12897.660203643007, "speed/FLOPS": 202575476740596.28, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047871001064777374, "optim/lr": 0.0027461326168024495, "optim/total_tokens": 8040480768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.877180337905884, "created_at": "2025-01-17T00:52:47.923097+00:00"} {"global_step": 15337, "acc_step": 0, "speed/wps": 12892.187944134268, "speed/FLOPS": 202489527385338.97, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.050096724182367325, "optim/lr": 0.0027460849221527636, "optim/total_tokens": 8041005056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.009321451187134, "created_at": "2025-01-17T00:52:58.093438+00:00"} {"global_step": 15338, "acc_step": 0, "speed/wps": 12893.22312179261, "speed/FLOPS": 202505786272946.4, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06510275602340698, "optim/lr": 0.0027460372234375066, "optim/total_tokens": 8041529344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 2.8939309120178223, "created_at": "2025-01-17T00:53:08.264472+00:00"} {"global_step": 15339, "acc_step": 0, "speed/wps": 12890.816083717438, "speed/FLOPS": 202467980432358.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06032438203692436, "optim/lr": 0.0027459895206568352, "optim/total_tokens": 8042053632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.787166118621826, "created_at": "2025-01-17T00:53:18.437445+00:00"} {"global_step": 15340, "acc_step": 0, "speed/wps": 12892.038475998816, "speed/FLOPS": 202487179782880.97, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.056456804275512695, "optim/lr": 0.0027459418138109045, "optim/total_tokens": 8042577920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 2.8686721324920654, "created_at": "2025-01-17T00:53:28.609603+00:00"} {"global_step": 15341, "acc_step": 0, "speed/wps": 12897.094724130207, "speed/FLOPS": 202566595107797.5, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04451143741607666, "optim/lr": 0.00274589410289987, "optim/total_tokens": 8043102208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371356, "loss/out": 2.788936138153076, "created_at": "2025-01-17T00:53:38.776135+00:00"} {"global_step": 15342, "acc_step": 0, "speed/wps": 12889.842011605402, "speed/FLOPS": 202452681291324.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04121597111225128, "optim/lr": 0.0027458463879238885, "optim/total_tokens": 8043626496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.868750810623169, "created_at": "2025-01-17T00:53:48.948309+00:00"} {"global_step": 15343, "acc_step": 0, "speed/wps": 12894.94632080609, "speed/FLOPS": 202532851481376.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042721547186374664, "optim/lr": 0.0027457986688831145, "optim/total_tokens": 8044150784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.915790557861328, "created_at": "2025-01-17T00:53:59.119052+00:00"} {"global_step": 15344, "acc_step": 0, "speed/wps": 12894.027997936992, "speed/FLOPS": 202518427958809.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045482780784368515, "optim/lr": 0.002745750945777704, "optim/total_tokens": 8044675072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 2.8164682388305664, "created_at": "2025-01-17T00:54:09.291055+00:00"} {"global_step": 15345, "acc_step": 0, "speed/wps": 12892.870903515714, "speed/FLOPS": 202500254200909.12, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049393102526664734, "optim/lr": 0.002745703218607813, "optim/total_tokens": 8045199360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422788, "loss/out": 2.724483013153076, "created_at": "2025-01-17T00:54:19.461671+00:00"} {"global_step": 15346, "acc_step": 0, "speed/wps": 12895.334027086159, "speed/FLOPS": 202538940941269.44, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048089370131492615, "optim/lr": 0.0027456554873735964, "optim/total_tokens": 8045723648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.806455612182617, "created_at": "2025-01-17T00:54:29.633294+00:00"} {"global_step": 15347, "acc_step": 0, "speed/wps": 12896.831901467182, "speed/FLOPS": 202562467116718.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05090012773871422, "optim/lr": 0.0027456077520752104, "optim/total_tokens": 8046247936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.836109161376953, "created_at": "2025-01-17T00:54:39.802956+00:00"} {"global_step": 15348, "acc_step": 0, "speed/wps": 12894.663770619849, "speed/FLOPS": 202528413642437.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05297758802771568, "optim/lr": 0.0027455600127128115, "optim/total_tokens": 8046772224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 2.8277101516723633, "created_at": "2025-01-17T00:54:49.971438+00:00"} {"global_step": 15349, "acc_step": 0, "speed/wps": 12893.2583834214, "speed/FLOPS": 202506340105281.12, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047475386410951614, "optim/lr": 0.0027455122692865547, "optim/total_tokens": 8047296512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.775074005126953, "created_at": "2025-01-17T00:55:00.140924+00:00"} {"global_step": 15350, "acc_step": 0, "speed/wps": 12895.049387545992, "speed/FLOPS": 202534470286155.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046908508986234665, "optim/lr": 0.002745464521796596, "optim/total_tokens": 8047820800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.859248638153076, "created_at": "2025-01-17T00:55:10.309836+00:00"} {"global_step": 15351, "acc_step": 0, "speed/wps": 12878.087044699654, "speed/FLOPS": 202268053383052.03, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04434169456362724, "optim/lr": 0.002745416770243091, "optim/total_tokens": 8048345088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.8412976264953613, "created_at": "2025-01-17T00:55:20.498933+00:00"} {"global_step": 15352, "acc_step": 0, "speed/wps": 12891.057741411789, "speed/FLOPS": 202471775998521.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04420623183250427, "optim/lr": 0.0027453690146261953, "optim/total_tokens": 8048869376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.8424930572509766, "created_at": "2025-01-17T00:55:30.670138+00:00"} {"global_step": 15353, "acc_step": 0, "speed/wps": 12893.721914654447, "speed/FLOPS": 202513620500254.78, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04803356155753136, "optim/lr": 0.0027453212549460653, "optim/total_tokens": 8049393664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 2.793412923812866, "created_at": "2025-01-17T00:55:40.839641+00:00"} {"global_step": 15354, "acc_step": 0, "speed/wps": 12895.122941489519, "speed/FLOPS": 202535625551912.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06925344467163086, "optim/lr": 0.0027452734912028565, "optim/total_tokens": 8049917952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.8455910682678223, "created_at": "2025-01-17T00:55:51.010276+00:00"} {"global_step": 15355, "acc_step": 0, "speed/wps": 12889.136439761354, "speed/FLOPS": 202441599315955.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04786752536892891, "optim/lr": 0.002745225723396725, "optim/total_tokens": 8050442240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.787933349609375, "created_at": "2025-01-17T00:56:01.183057+00:00"} {"global_step": 15356, "acc_step": 0, "speed/wps": 12894.063261605781, "speed/FLOPS": 202518981823185.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04842602089047432, "optim/lr": 0.0027451779515278254, "optim/total_tokens": 8050966528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361573, "loss/out": 2.788902759552002, "created_at": "2025-01-17T00:56:11.354451+00:00"} {"global_step": 15357, "acc_step": 0, "speed/wps": 12896.017586903139, "speed/FLOPS": 202549677187504.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04659253731369972, "optim/lr": 0.0027451301755963154, "optim/total_tokens": 8051490816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.660494804382324, "created_at": "2025-01-17T00:56:21.525268+00:00"} {"global_step": 15358, "acc_step": 0, "speed/wps": 12888.502329750418, "speed/FLOPS": 202431639746875.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05312936007976532, "optim/lr": 0.00274508239560235, "optim/total_tokens": 8052015104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 2.785499334335327, "created_at": "2025-01-17T00:56:31.700179+00:00"} {"global_step": 15359, "acc_step": 0, "speed/wps": 12895.386306555629, "speed/FLOPS": 202539762062177.0, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05656421184539795, "optim/lr": 0.0027450346115460845, "optim/total_tokens": 8052539392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.817777633666992, "created_at": "2025-01-17T00:56:41.868072+00:00"} {"global_step": 15360, "acc_step": 0, "speed/wps": 12888.421686935433, "speed/FLOPS": 202430373140651.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0460355170071125, "optim/lr": 0.0027449868234276763, "optim/total_tokens": 8053063680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 2.8332901000976562, "created_at": "2025-01-17T00:56:52.044485+00:00"} {"global_step": 15361, "acc_step": 0, "speed/wps": 12892.738013742523, "speed/FLOPS": 202498166984411.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047803688794374466, "optim/lr": 0.00274493903124728, "optim/total_tokens": 8053587968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.716440200805664, "created_at": "2025-01-17T00:57:02.216384+00:00"} {"global_step": 15362, "acc_step": 0, "speed/wps": 12897.339530737336, "speed/FLOPS": 202570440131961.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05142831429839134, "optim/lr": 0.0027448912350050525, "optim/total_tokens": 8054112256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.919252395629883, "created_at": "2025-01-17T00:57:12.384454+00:00"} {"global_step": 15363, "acc_step": 0, "speed/wps": 12892.81396965534, "speed/FLOPS": 202499359976396.66, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04224523529410362, "optim/lr": 0.0027448434347011486, "optim/total_tokens": 8054636544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 2.8263213634490967, "created_at": "2025-01-17T00:57:22.557771+00:00"} {"global_step": 15364, "acc_step": 0, "speed/wps": 12888.6274506263, "speed/FLOPS": 202433604942165.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05937021225690842, "optim/lr": 0.002744795630335725, "optim/total_tokens": 8055160832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357304, "loss/out": 2.755988597869873, "created_at": "2025-01-17T00:57:32.732067+00:00"} {"global_step": 15365, "acc_step": 0, "speed/wps": 12892.73026347874, "speed/FLOPS": 202498045255868.78, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.045236073434352875, "optim/lr": 0.0027447478219089377, "optim/total_tokens": 8055685120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 2.783961772918701, "created_at": "2025-01-17T00:57:42.901966+00:00"} {"global_step": 15366, "acc_step": 0, "speed/wps": 12892.304854194454, "speed/FLOPS": 202491363618485.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06858107447624207, "optim/lr": 0.0027447000094209424, "optim/total_tokens": 8056209408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.791049003601074, "created_at": "2025-01-17T00:57:53.074332+00:00"} {"global_step": 15367, "acc_step": 0, "speed/wps": 12890.10167941086, "speed/FLOPS": 202456759731029.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05197811499238014, "optim/lr": 0.0027446521928718953, "optim/total_tokens": 8056733696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 2.793390989303589, "created_at": "2025-01-17T00:58:03.247249+00:00"} {"global_step": 15368, "acc_step": 0, "speed/wps": 12895.69724735783, "speed/FLOPS": 202544645814753.0, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04797791689634323, "optim/lr": 0.0027446043722619525, "optim/total_tokens": 8057257984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.7941439151763916, "created_at": "2025-01-17T00:58:13.415549+00:00"} {"global_step": 15369, "acc_step": 0, "speed/wps": 12893.875399235516, "speed/FLOPS": 202516031186510.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051875680685043335, "optim/lr": 0.00274455654759127, "optim/total_tokens": 8057782272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 2.748717784881592, "created_at": "2025-01-17T00:58:23.585073+00:00"} {"global_step": 15370, "acc_step": 0, "speed/wps": 12892.827652680811, "speed/FLOPS": 202499574887114.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0492560938000679, "optim/lr": 0.0027445087188600033, "optim/total_tokens": 8058306560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 2.7589635848999023, "created_at": "2025-01-17T00:58:33.755091+00:00"} {"global_step": 15371, "acc_step": 0, "speed/wps": 12891.772810269042, "speed/FLOPS": 202483007137531.9, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06131172552704811, "optim/lr": 0.0027444608860683092, "optim/total_tokens": 8058830848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 2.7700083255767822, "created_at": "2025-01-17T00:58:43.928011+00:00"} {"global_step": 15372, "acc_step": 0, "speed/wps": 12894.991861418315, "speed/FLOPS": 202533566759271.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.048245884478092194, "optim/lr": 0.002744413049216343, "optim/total_tokens": 8059355136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.8353450298309326, "created_at": "2025-01-17T00:58:54.096039+00:00"} {"global_step": 15373, "acc_step": 0, "speed/wps": 12887.526029862134, "speed/FLOPS": 202416305615553.6, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05924566090106964, "optim/lr": 0.0027443652083042613, "optim/total_tokens": 8059879424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.7866291999816895, "created_at": "2025-01-17T00:59:04.272651+00:00"} {"global_step": 15374, "acc_step": 0, "speed/wps": 12891.437299863563, "speed/FLOPS": 202477737485574.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05250521004199982, "optim/lr": 0.0027443173633322202, "optim/total_tokens": 8060403712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397900, "loss/out": 2.7464261054992676, "created_at": "2025-01-17T00:59:14.449354+00:00"} {"global_step": 15375, "acc_step": 0, "speed/wps": 12890.259054171924, "speed/FLOPS": 202459231517907.8, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05394011363387108, "optim/lr": 0.002744269514300376, "optim/total_tokens": 8060928000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 2.806633472442627, "created_at": "2025-01-17T00:59:24.623346+00:00"} {"global_step": 15376, "acc_step": 0, "speed/wps": 12891.130196561557, "speed/FLOPS": 202472914006212.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05859021097421646, "optim/lr": 0.0027442216612088843, "optim/total_tokens": 8061452288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 2.8463377952575684, "created_at": "2025-01-17T00:59:34.797685+00:00"} {"global_step": 15377, "acc_step": 0, "speed/wps": 12891.079589336778, "speed/FLOPS": 202472119150205.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046299368143081665, "optim/lr": 0.002744173804057901, "optim/total_tokens": 8061976576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 2.733668088912964, "created_at": "2025-01-17T00:59:44.973332+00:00"} {"global_step": 15378, "acc_step": 0, "speed/wps": 12892.96594558442, "speed/FLOPS": 202501746967199.62, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046466365456581116, "optim/lr": 0.002744125942847583, "optim/total_tokens": 8062500864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 2.835038185119629, "created_at": "2025-01-17T00:59:55.174745+00:00"} {"global_step": 15379, "acc_step": 0, "speed/wps": 12898.701817720734, "speed/FLOPS": 202591836721014.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04922854155302048, "optim/lr": 0.0027440780775780864, "optim/total_tokens": 8063025152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 2.694774866104126, "created_at": "2025-01-17T01:00:05.339865+00:00"} {"global_step": 15380, "acc_step": 0, "speed/wps": 12892.332753110264, "speed/FLOPS": 202491801809295.38, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.0535690039396286, "optim/lr": 0.0027440302082495666, "optim/total_tokens": 8063549440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458694, "loss/out": 2.7726964950561523, "created_at": "2025-01-17T01:00:15.510147+00:00"} {"global_step": 15381, "acc_step": 0, "speed/wps": 12896.756887388168, "speed/FLOPS": 202561288917527.44, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04877844825387001, "optim/lr": 0.002743982334862181, "optim/total_tokens": 8064073728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.7064666748046875, "created_at": "2025-01-17T01:00:25.682723+00:00"} {"global_step": 15382, "acc_step": 0, "speed/wps": 12894.366856565419, "speed/FLOPS": 202523750199209.66, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04676789790391922, "optim/lr": 0.0027439344574160846, "optim/total_tokens": 8064598016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 2.8317031860351562, "created_at": "2025-01-17T01:00:35.851621+00:00"} {"global_step": 15383, "acc_step": 0, "speed/wps": 12890.02761015411, "speed/FLOPS": 202455596371570.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049554094672203064, "optim/lr": 0.002743886575911434, "optim/total_tokens": 8065122304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405960, "loss/out": 2.8557753562927246, "created_at": "2025-01-17T01:00:46.023733+00:00"} {"global_step": 15384, "acc_step": 0, "speed/wps": 12891.499749152485, "speed/FLOPS": 202478718337471.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.038097649812698364, "optim/lr": 0.002743838690348386, "optim/total_tokens": 8065646592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.867420196533203, "created_at": "2025-01-17T01:00:56.197236+00:00"} {"global_step": 15385, "acc_step": 0, "speed/wps": 12892.563757890466, "speed/FLOPS": 202495430056802.28, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04468751698732376, "optim/lr": 0.0027437908007270956, "optim/total_tokens": 8066170880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 2.823965311050415, "created_at": "2025-01-17T01:01:06.367780+00:00"} {"global_step": 15386, "acc_step": 0, "speed/wps": 12891.194190904325, "speed/FLOPS": 202473919125302.7, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04143757000565529, "optim/lr": 0.0027437429070477203, "optim/total_tokens": 8066695168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.8337113857269287, "created_at": "2025-01-17T01:01:16.543056+00:00"} {"global_step": 15387, "acc_step": 0, "speed/wps": 12885.446880901478, "speed/FLOPS": 202383649724076.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0572178028523922, "optim/lr": 0.0027436950093104153, "optim/total_tokens": 8067219456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345322, "loss/out": 2.7782938480377197, "created_at": "2025-01-17T01:01:26.720635+00:00"} {"global_step": 15388, "acc_step": 0, "speed/wps": 12886.242586159511, "speed/FLOPS": 202396147368567.44, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04597654566168785, "optim/lr": 0.0027436471075153383, "optim/total_tokens": 8067743744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 2.8448598384857178, "created_at": "2025-01-17T01:01:36.898660+00:00"} {"global_step": 15389, "acc_step": 0, "speed/wps": 12881.856762202453, "speed/FLOPS": 202327261976568.16, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04772447794675827, "optim/lr": 0.0027435992016626443, "optim/total_tokens": 8068268032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 2.8573246002197266, "created_at": "2025-01-17T01:01:47.081990+00:00"} {"global_step": 15390, "acc_step": 0, "speed/wps": 12882.47433691359, "speed/FLOPS": 202336961836043.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.046745169907808304, "optim/lr": 0.0027435512917524897, "optim/total_tokens": 8068792320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.85587215423584, "created_at": "2025-01-17T01:01:57.267321+00:00"} {"global_step": 15391, "acc_step": 0, "speed/wps": 12890.373678931575, "speed/FLOPS": 202461031857269.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04468347132205963, "optim/lr": 0.0027435033777850316, "optim/total_tokens": 8069316608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.905174732208252, "created_at": "2025-01-17T01:02:07.441519+00:00"} {"global_step": 15392, "acc_step": 0, "speed/wps": 12890.55250031926, "speed/FLOPS": 202463840492888.97, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06326578557491302, "optim/lr": 0.0027434554597604256, "optim/total_tokens": 8069840896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 2.994882583618164, "created_at": "2025-01-17T01:02:17.622246+00:00"} {"global_step": 15393, "acc_step": 0, "speed/wps": 12892.987378314936, "speed/FLOPS": 202502083597684.06, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04853196442127228, "optim/lr": 0.0027434075376788283, "optim/total_tokens": 8070365184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 2.802100896835327, "created_at": "2025-01-17T01:02:27.796054+00:00"} {"global_step": 15394, "acc_step": 0, "speed/wps": 12891.78277122255, "speed/FLOPS": 202483163587994.53, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052863746881484985, "optim/lr": 0.0027433596115403963, "optim/total_tokens": 8070889472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.860301971435547, "created_at": "2025-01-17T01:02:37.967497+00:00"} {"global_step": 15395, "acc_step": 0, "speed/wps": 12893.503742512657, "speed/FLOPS": 202510193806968.03, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04502401873469353, "optim/lr": 0.0027433116813452853, "optim/total_tokens": 8071413760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 2.7649292945861816, "created_at": "2025-01-17T01:02:48.137991+00:00"} {"global_step": 15396, "acc_step": 0, "speed/wps": 12891.057538318417, "speed/FLOPS": 202471772808660.66, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05384863540530205, "optim/lr": 0.0027432637470936523, "optim/total_tokens": 8071938048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7426159381866455, "created_at": "2025-01-17T01:02:58.312429+00:00"} {"global_step": 15397, "acc_step": 0, "speed/wps": 12893.321509459252, "speed/FLOPS": 202507331586449.84, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03989117592573166, "optim/lr": 0.0027432158087856537, "optim/total_tokens": 8072462336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 2.871527910232544, "created_at": "2025-01-17T01:03:08.484538+00:00"} {"global_step": 15398, "acc_step": 0, "speed/wps": 12890.145800956383, "speed/FLOPS": 202457452720531.5, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.070499487221241, "optim/lr": 0.002743167866421446, "optim/total_tokens": 8072986624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8660120964050293, "created_at": "2025-01-17T01:03:18.661377+00:00"} {"global_step": 15399, "acc_step": 0, "speed/wps": 12895.568239956057, "speed/FLOPS": 202542619576233.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06137708202004433, "optim/lr": 0.002743119920001185, "optim/total_tokens": 8073510912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.9313223361968994, "created_at": "2025-01-17T01:03:28.834715+00:00"} {"global_step": 15400, "acc_step": 0, "speed/wps": 12892.10615216762, "speed/FLOPS": 202488242730108.84, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.058969948440790176, "optim/lr": 0.0027430719695250272, "optim/total_tokens": 8074035200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377871, "loss/out": 2.8614795207977295, "created_at": "2025-01-17T01:03:39.006551+00:00"} {"global_step": 15401, "acc_step": 0, "speed/wps": 12897.63447660285, "speed/FLOPS": 202575072662078.0, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05237517133355141, "optim/lr": 0.0027430240149931293, "optim/total_tokens": 8074559488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 2.8186333179473877, "created_at": "2025-01-17T01:03:49.173968+00:00"} {"global_step": 15402, "acc_step": 0, "speed/wps": 12891.232282458157, "speed/FLOPS": 202474517405498.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05998969450592995, "optim/lr": 0.0027429760564056478, "optim/total_tokens": 8075083776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8851826190948486, "created_at": "2025-01-17T01:03:59.345195+00:00"} {"global_step": 15403, "acc_step": 0, "speed/wps": 12893.80967013942, "speed/FLOPS": 202514998820734.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04544082656502724, "optim/lr": 0.0027429280937627397, "optim/total_tokens": 8075608064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.960381031036377, "created_at": "2025-01-17T01:04:09.516356+00:00"} {"global_step": 15404, "acc_step": 0, "speed/wps": 12884.283215530917, "speed/FLOPS": 202365372760387.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0683416873216629, "optim/lr": 0.0027428801270645608, "optim/total_tokens": 8076132352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.9553918838500977, "created_at": "2025-01-17T01:04:19.696653+00:00"} {"global_step": 15405, "acc_step": 0, "speed/wps": 12895.039709831104, "speed/FLOPS": 202534318284344.53, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05188240855932236, "optim/lr": 0.0027428321563112666, "optim/total_tokens": 8076656640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.813077926635742, "created_at": "2025-01-17T01:04:29.865930+00:00"} {"global_step": 15406, "acc_step": 0, "speed/wps": 12889.44063617358, "speed/FLOPS": 202446377138619.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059255827218294144, "optim/lr": 0.002742784181503016, "optim/total_tokens": 8077180928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 2.8314738273620605, "created_at": "2025-01-17T01:04:40.041359+00:00"} {"global_step": 15407, "acc_step": 0, "speed/wps": 12890.537506844745, "speed/FLOPS": 202463604999768.53, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06033545359969139, "optim/lr": 0.0027427362026399642, "optim/total_tokens": 8077705216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 2.8780412673950195, "created_at": "2025-01-17T01:04:50.220586+00:00"} {"global_step": 15408, "acc_step": 0, "speed/wps": 12892.986817278703, "speed/FLOPS": 202502074785839.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0698559582233429, "optim/lr": 0.002742688219722267, "optim/total_tokens": 8078229504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377871, "loss/out": 2.845113515853882, "created_at": "2025-01-17T01:05:00.391380+00:00"} {"global_step": 15409, "acc_step": 0, "speed/wps": 12892.80843621738, "speed/FLOPS": 202499273066149.72, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0553557388484478, "optim/lr": 0.0027426402327500824, "optim/total_tokens": 8078753792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 2.654489517211914, "created_at": "2025-01-17T01:05:10.561651+00:00"} {"global_step": 15410, "acc_step": 0, "speed/wps": 12890.749814340465, "speed/FLOPS": 202466939580729.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04851216450333595, "optim/lr": 0.002742592241723566, "optim/total_tokens": 8079278080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 2.7977213859558105, "created_at": "2025-01-17T01:05:20.736577+00:00"} {"global_step": 15411, "acc_step": 0, "speed/wps": 12892.373429551419, "speed/FLOPS": 202492440688699.0, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0515635646879673, "optim/lr": 0.002742544246642875, "optim/total_tokens": 8079802368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 2.8066325187683105, "created_at": "2025-01-17T01:05:30.906855+00:00"} {"global_step": 15412, "acc_step": 0, "speed/wps": 12892.374803073728, "speed/FLOPS": 202492462261754.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05369825288653374, "optim/lr": 0.002742496247508166, "optim/total_tokens": 8080326656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314202, "loss/out": 2.9948196411132812, "created_at": "2025-01-17T01:05:41.077219+00:00"} {"global_step": 15413, "acc_step": 0, "speed/wps": 12891.239459409619, "speed/FLOPS": 202474630129383.78, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04750794917345047, "optim/lr": 0.0027424482443195947, "optim/total_tokens": 8080850944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.889256715774536, "created_at": "2025-01-17T01:05:51.249295+00:00"} {"global_step": 15414, "acc_step": 0, "speed/wps": 12894.491491179228, "speed/FLOPS": 202525707757085.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04127126187086105, "optim/lr": 0.0027424002370773186, "optim/total_tokens": 8081375232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.731550931930542, "created_at": "2025-01-17T01:06:01.419350+00:00"} {"global_step": 15415, "acc_step": 0, "speed/wps": 12892.348046014262, "speed/FLOPS": 202492042005367.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.001, "optim/grad_norm": 0.048292458057403564, "optim/lr": 0.0027423522257814937, "optim/total_tokens": 8081899520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.783222198486328, "created_at": "2025-01-17T01:06:11.591647+00:00"} {"global_step": 15416, "acc_step": 0, "speed/wps": 12886.874771073608, "speed/FLOPS": 202406076701356.44, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048193253576755524, "optim/lr": 0.002742304210432277, "optim/total_tokens": 8082423808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337297, "loss/out": 2.827897548675537, "created_at": "2025-01-17T01:06:21.771490+00:00"} {"global_step": 15417, "acc_step": 0, "speed/wps": 12894.91060216261, "speed/FLOPS": 202532290470998.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.06157882884144783, "optim/lr": 0.002742256191029826, "optim/total_tokens": 8082948096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.8610291481018066, "created_at": "2025-01-17T01:06:31.942961+00:00"} {"global_step": 15418, "acc_step": 0, "speed/wps": 12894.417766049053, "speed/FLOPS": 202524549802607.1, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.057929906994104385, "optim/lr": 0.002742208167574296, "optim/total_tokens": 8083472384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.8433337211608887, "created_at": "2025-01-17T01:06:42.116892+00:00"} {"global_step": 15419, "acc_step": 0, "speed/wps": 12893.597998936659, "speed/FLOPS": 202511674233628.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05207280442118645, "optim/lr": 0.0027421601400658443, "optim/total_tokens": 8083996672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 2.762523651123047, "created_at": "2025-01-17T01:06:52.288207+00:00"} {"global_step": 15420, "acc_step": 0, "speed/wps": 12893.35662752022, "speed/FLOPS": 202507883163854.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04983123019337654, "optim/lr": 0.0027421121085046273, "optim/total_tokens": 8084520960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.8353915214538574, "created_at": "2025-01-17T01:07:02.458620+00:00"} {"global_step": 15421, "acc_step": 0, "speed/wps": 12891.46065701945, "speed/FLOPS": 202478104341803.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.047853853553533554, "optim/lr": 0.002742064072890802, "optim/total_tokens": 8085045248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.762167453765869, "created_at": "2025-01-17T01:07:12.629584+00:00"} {"global_step": 15422, "acc_step": 0, "speed/wps": 12888.942048190844, "speed/FLOPS": 202438546129219.06, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05194208398461342, "optim/lr": 0.002742016033224525, "optim/total_tokens": 8085569536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.6975886821746826, "created_at": "2025-01-17T01:07:22.802659+00:00"} {"global_step": 15423, "acc_step": 0, "speed/wps": 12891.20431155281, "speed/FLOPS": 202474078083994.38, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05131091922521591, "optim/lr": 0.0027419679895059534, "optim/total_tokens": 8086093824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.8816311359405518, "created_at": "2025-01-17T01:07:32.975366+00:00"} {"global_step": 15424, "acc_step": 0, "speed/wps": 12888.701180378355, "speed/FLOPS": 202434762969236.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06857272237539291, "optim/lr": 0.0027419199417352434, "optim/total_tokens": 8086618112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313676, "loss/out": 2.916240692138672, "created_at": "2025-01-17T01:07:43.148650+00:00"} {"global_step": 15425, "acc_step": 0, "speed/wps": 12891.689431728277, "speed/FLOPS": 202481697562975.06, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0579461008310318, "optim/lr": 0.002741871889912552, "optim/total_tokens": 8087142400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505601, "loss/out": 2.80826997756958, "created_at": "2025-01-17T01:07:53.320240+00:00"} {"global_step": 15426, "acc_step": 0, "speed/wps": 12892.509324644738, "speed/FLOPS": 202494575107879.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06402312219142914, "optim/lr": 0.0027418238340380357, "optim/total_tokens": 8087666688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 2.829171657562256, "created_at": "2025-01-17T01:08:03.492964+00:00"} {"global_step": 15427, "acc_step": 0, "speed/wps": 12892.105450672958, "speed/FLOPS": 202488231712171.22, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.05634426698088646, "optim/lr": 0.002741775774111852, "optim/total_tokens": 8088190976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 2.7694735527038574, "created_at": "2025-01-17T01:08:13.663859+00:00"} {"global_step": 15428, "acc_step": 0, "speed/wps": 12889.51766998569, "speed/FLOPS": 202447587060494.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.059060756117105484, "optim/lr": 0.002741727710134157, "optim/total_tokens": 8088715264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.7646830081939697, "created_at": "2025-01-17T01:08:23.836479+00:00"} {"global_step": 15429, "acc_step": 0, "speed/wps": 12893.255227312911, "speed/FLOPS": 202506290534260.34, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05459100380539894, "optim/lr": 0.002741679642105108, "optim/total_tokens": 8089239552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.913797616958618, "created_at": "2025-01-17T01:08:34.009205+00:00"} {"global_step": 15430, "acc_step": 0, "speed/wps": 12895.188802128958, "speed/FLOPS": 202536659983756.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046570274978876114, "optim/lr": 0.002741631570024862, "optim/total_tokens": 8089763840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.8008921146392822, "created_at": "2025-01-17T01:08:44.177294+00:00"} {"global_step": 15431, "acc_step": 0, "speed/wps": 12895.922608551531, "speed/FLOPS": 202548185421978.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06741158664226532, "optim/lr": 0.002741583493893575, "optim/total_tokens": 8090288128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375182, "loss/out": 2.857616901397705, "created_at": "2025-01-17T01:08:54.344771+00:00"} {"global_step": 15432, "acc_step": 0, "speed/wps": 12894.869814010566, "speed/FLOPS": 202531649837021.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05377737432718277, "optim/lr": 0.0027415354137114043, "optim/total_tokens": 8090812416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.922426462173462, "created_at": "2025-01-17T01:09:04.516008+00:00"} {"global_step": 15433, "acc_step": 0, "speed/wps": 12891.366172059052, "speed/FLOPS": 202476620325664.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06202784553170204, "optim/lr": 0.002741487329478507, "optim/total_tokens": 8091336704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.900578498840332, "created_at": "2025-01-17T01:09:14.689319+00:00"} {"global_step": 15434, "acc_step": 0, "speed/wps": 12894.776335292752, "speed/FLOPS": 202530181625305.16, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05326011776924133, "optim/lr": 0.0027414392411950392, "optim/total_tokens": 8091860992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387212, "loss/out": 2.5799560546875, "created_at": "2025-01-17T01:09:24.861884+00:00"} {"global_step": 15435, "acc_step": 0, "speed/wps": 12894.894626191426, "speed/FLOPS": 202532039546417.56, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05864706262946129, "optim/lr": 0.002741391148861159, "optim/total_tokens": 8092385280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 2.781696081161499, "created_at": "2025-01-17T01:09:35.030082+00:00"} {"global_step": 15436, "acc_step": 0, "speed/wps": 12889.304045551457, "speed/FLOPS": 202444231795204.0, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052941929548978806, "optim/lr": 0.0027413430524770225, "optim/total_tokens": 8092909568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 2.7903146743774414, "created_at": "2025-01-17T01:09:45.203160+00:00"} {"global_step": 15437, "acc_step": 0, "speed/wps": 12890.930599127989, "speed/FLOPS": 202469779054242.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.054060548543930054, "optim/lr": 0.002741294952042787, "optim/total_tokens": 8093433856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.7548744678497314, "created_at": "2025-01-17T01:09:55.374422+00:00"} {"global_step": 15438, "acc_step": 0, "speed/wps": 12893.612587277572, "speed/FLOPS": 202511903363569.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05474219471216202, "optim/lr": 0.0027412468475586094, "optim/total_tokens": 8093958144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420607, "loss/out": 2.9361562728881836, "created_at": "2025-01-17T01:10:05.549693+00:00"} {"global_step": 15439, "acc_step": 0, "speed/wps": 12892.017130270542, "speed/FLOPS": 202486844518886.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.06283954530954361, "optim/lr": 0.002741198739024646, "optim/total_tokens": 8094482432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.7689332962036133, "created_at": "2025-01-17T01:10:15.723130+00:00"} {"global_step": 15440, "acc_step": 0, "speed/wps": 12891.91743986497, "speed/FLOPS": 202485278744077.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04423955827951431, "optim/lr": 0.0027411506264410543, "optim/total_tokens": 8095006720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 2.851081132888794, "created_at": "2025-01-17T01:10:25.895981+00:00"} {"global_step": 15441, "acc_step": 0, "speed/wps": 12892.220438949491, "speed/FLOPS": 202490037761064.62, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.059663672000169754, "optim/lr": 0.0027411025098079914, "optim/total_tokens": 8095531008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 2.826296329498291, "created_at": "2025-01-17T01:10:36.068716+00:00"} {"global_step": 15442, "acc_step": 0, "speed/wps": 12896.742811930739, "speed/FLOPS": 202561067843126.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05398789420723915, "optim/lr": 0.0027410543891256143, "optim/total_tokens": 8096055296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 2.7313785552978516, "created_at": "2025-01-17T01:10:46.236736+00:00"} {"global_step": 15443, "acc_step": 0, "speed/wps": 12895.63109321098, "speed/FLOPS": 202543606772971.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0499105304479599, "optim/lr": 0.00274100626439408, "optim/total_tokens": 8096579584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 2.8921420574188232, "created_at": "2025-01-17T01:10:56.404485+00:00"} {"global_step": 15444, "acc_step": 0, "speed/wps": 12891.69418245518, "speed/FLOPS": 202481772179669.25, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06514687091112137, "optim/lr": 0.0027409581356135446, "optim/total_tokens": 8097103872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 2.824706792831421, "created_at": "2025-01-17T01:11:06.576033+00:00"} {"global_step": 15445, "acc_step": 0, "speed/wps": 12896.68457224537, "speed/FLOPS": 202560153108838.4, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05567207559943199, "optim/lr": 0.0027409100027841657, "optim/total_tokens": 8097628160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 2.9131863117218018, "created_at": "2025-01-17T01:11:16.743897+00:00"} {"global_step": 15446, "acc_step": 0, "speed/wps": 12899.0691409645, "speed/FLOPS": 202597606037313.12, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06413688510656357, "optim/lr": 0.0027408618659061014, "optim/total_tokens": 8098152448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8903424739837646, "created_at": "2025-01-17T01:11:26.910458+00:00"} {"global_step": 15447, "acc_step": 0, "speed/wps": 12893.608724612433, "speed/FLOPS": 202511842695105.53, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07039824873209, "optim/lr": 0.0027408137249795073, "optim/total_tokens": 8098676736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 2.792234182357788, "created_at": "2025-01-17T01:11:37.079941+00:00"} {"global_step": 15448, "acc_step": 0, "speed/wps": 12888.973336808636, "speed/FLOPS": 202439037559956.56, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 0.047336939722299576, "optim/lr": 0.002740765580004542, "optim/total_tokens": 8099201024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.788586139678955, "created_at": "2025-01-17T01:11:47.254334+00:00"} {"global_step": 15449, "acc_step": 0, "speed/wps": 12889.875204111631, "speed/FLOPS": 202453202625245.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.07065068930387497, "optim/lr": 0.00274071743098136, "optim/total_tokens": 8099725312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 2.822948455810547, "created_at": "2025-01-17T01:11:57.429731+00:00"} {"global_step": 15450, "acc_step": 0, "speed/wps": 12891.304434918255, "speed/FLOPS": 202475650659032.6, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0524442084133625, "optim/lr": 0.0027406692779101214, "optim/total_tokens": 8100249600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 2.7737269401550293, "created_at": "2025-01-17T01:12:07.600659+00:00"} {"global_step": 15451, "acc_step": 0, "speed/wps": 12890.389106505258, "speed/FLOPS": 202461274168513.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05343160033226013, "optim/lr": 0.0027406211207909813, "optim/total_tokens": 8100773888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.7936649322509766, "created_at": "2025-01-17T01:12:17.780421+00:00"} {"global_step": 15452, "acc_step": 0, "speed/wps": 12891.344816748744, "speed/FLOPS": 202476284911170.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05025910958647728, "optim/lr": 0.0027405729596240978, "optim/total_tokens": 8101298176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 2.8090944290161133, "created_at": "2025-01-17T01:12:27.953001+00:00"} {"global_step": 15453, "acc_step": 0, "speed/wps": 12896.48127401198, "speed/FLOPS": 202556960030721.94, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.051487259566783905, "optim/lr": 0.0027405247944096273, "optim/total_tokens": 8101822464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285549, "loss/out": 2.785128116607666, "created_at": "2025-01-17T01:12:38.119959+00:00"} {"global_step": 15454, "acc_step": 0, "speed/wps": 12892.652475535018, "speed/FLOPS": 202496823489320.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05136768892407417, "optim/lr": 0.0027404766251477274, "optim/total_tokens": 8102346752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.8063817024230957, "created_at": "2025-01-17T01:12:48.293363+00:00"} {"global_step": 15455, "acc_step": 0, "speed/wps": 12891.894714031618, "speed/FLOPS": 202484921803636.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046770788729190826, "optim/lr": 0.0027404284518385553, "optim/total_tokens": 8102871040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 2.7141356468200684, "created_at": "2025-01-17T01:12:58.466681+00:00"} {"global_step": 15456, "acc_step": 0, "speed/wps": 12893.654611333859, "speed/FLOPS": 202512563409120.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.049565933644771576, "optim/lr": 0.0027403802744822682, "optim/total_tokens": 8103395328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.8643269538879395, "created_at": "2025-01-17T01:13:08.636124+00:00"} {"global_step": 15457, "acc_step": 0, "speed/wps": 12889.63511479535, "speed/FLOPS": 202449431692616.16, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05395049601793289, "optim/lr": 0.0027403320930790226, "optim/total_tokens": 8103919616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 2.858858108520508, "created_at": "2025-01-17T01:13:18.809134+00:00"} {"global_step": 15458, "acc_step": 0, "speed/wps": 12889.747722728573, "speed/FLOPS": 202451200354946.94, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.049913499504327774, "optim/lr": 0.002740283907628977, "optim/total_tokens": 8104443904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 2.8008944988250732, "created_at": "2025-01-17T01:13:28.984356+00:00"} {"global_step": 15459, "acc_step": 0, "speed/wps": 12892.941811701752, "speed/FLOPS": 202501367911408.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04374363645911217, "optim/lr": 0.0027402357181322873, "optim/total_tokens": 8104968192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.9352102279663086, "created_at": "2025-01-17T01:13:39.159507+00:00"} {"global_step": 15460, "acc_step": 0, "speed/wps": 12892.874279122181, "speed/FLOPS": 202500307219447.25, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0483398362994194, "optim/lr": 0.0027401875245891116, "optim/total_tokens": 8105492480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487669, "loss/out": 2.710319757461548, "created_at": "2025-01-17T01:13:49.340122+00:00"} {"global_step": 15461, "acc_step": 0, "speed/wps": 12896.319299214265, "speed/FLOPS": 202554415993947.12, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05370514839887619, "optim/lr": 0.002740139326999606, "optim/total_tokens": 8106016768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308621, "loss/out": 2.8145229816436768, "created_at": "2025-01-17T01:13:59.510959+00:00"} {"global_step": 15462, "acc_step": 0, "speed/wps": 12888.025316746629, "speed/FLOPS": 202424147602166.66, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05757088214159012, "optim/lr": 0.00274009112536393, "optim/total_tokens": 8106541056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.9607489109039307, "created_at": "2025-01-17T01:14:09.685622+00:00"} {"global_step": 15463, "acc_step": 0, "speed/wps": 12886.21226060809, "speed/FLOPS": 202395671064110.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0625164583325386, "optim/lr": 0.0027400429196822386, "optim/total_tokens": 8107065344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 2.8096299171447754, "created_at": "2025-01-17T01:14:19.863605+00:00"} {"global_step": 15464, "acc_step": 0, "speed/wps": 12894.987650567045, "speed/FLOPS": 202533500622065.56, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07420236617326736, "optim/lr": 0.00273999470995469, "optim/total_tokens": 8107589632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 2.8738975524902344, "created_at": "2025-01-17T01:14:30.034757+00:00"} {"global_step": 15465, "acc_step": 0, "speed/wps": 12892.347854690348, "speed/FLOPS": 202492039000362.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050890691578388214, "optim/lr": 0.0027399464961814415, "optim/total_tokens": 8108113920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351470, "loss/out": 2.798053503036499, "created_at": "2025-01-17T01:14:40.207380+00:00"} {"global_step": 15466, "acc_step": 0, "speed/wps": 12891.935733723034, "speed/FLOPS": 202485566074257.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06483311951160431, "optim/lr": 0.0027398982783626504, "optim/total_tokens": 8108638208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.7586722373962402, "created_at": "2025-01-17T01:14:50.383055+00:00"} {"global_step": 15467, "acc_step": 0, "speed/wps": 12895.474147322038, "speed/FLOPS": 202541141722119.3, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04952754080295563, "optim/lr": 0.0027398500564984737, "optim/total_tokens": 8109162496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 2.857286214828491, "created_at": "2025-01-17T01:15:00.555190+00:00"} {"global_step": 15468, "acc_step": 0, "speed/wps": 12890.994074314525, "speed/FLOPS": 202470776019270.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04943319782614708, "optim/lr": 0.002739801830589069, "optim/total_tokens": 8109686784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.8630316257476807, "created_at": "2025-01-17T01:15:10.731156+00:00"} {"global_step": 15469, "acc_step": 0, "speed/wps": 12890.872361152888, "speed/FLOPS": 202468864346815.94, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05155297368764877, "optim/lr": 0.0027397536006345934, "optim/total_tokens": 8110211072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415046, "loss/out": 2.8023111820220947, "created_at": "2025-01-17T01:15:20.903524+00:00"} {"global_step": 15470, "acc_step": 0, "speed/wps": 12893.2743895927, "speed/FLOPS": 202506591504196.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0481623113155365, "optim/lr": 0.002739705366635205, "optim/total_tokens": 8110735360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 2.708730697631836, "created_at": "2025-01-17T01:15:31.079550+00:00"} {"global_step": 15471, "acc_step": 0, "speed/wps": 12887.85422134642, "speed/FLOPS": 202421460313796.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.051460422575473785, "optim/lr": 0.0027396571285910605, "optim/total_tokens": 8111259648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.807924747467041, "created_at": "2025-01-17T01:15:41.253407+00:00"} {"global_step": 15472, "acc_step": 0, "speed/wps": 12891.048884421301, "speed/FLOPS": 202471636887314.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047152817249298096, "optim/lr": 0.0027396088865023174, "optim/total_tokens": 8111783936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.7720232009887695, "created_at": "2025-01-17T01:15:51.424590+00:00"} {"global_step": 15473, "acc_step": 0, "speed/wps": 12888.538892235003, "speed/FLOPS": 202432214010938.25, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05914850905537605, "optim/lr": 0.0027395606403691334, "optim/total_tokens": 8112308224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 2.8351619243621826, "created_at": "2025-01-17T01:16:01.600456+00:00"} {"global_step": 15474, "acc_step": 0, "speed/wps": 12894.575041832639, "speed/FLOPS": 202527020034907.88, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06258536875247955, "optim/lr": 0.0027395123901916654, "optim/total_tokens": 8112832512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 2.7292165756225586, "created_at": "2025-01-17T01:16:11.776006+00:00"} {"global_step": 15475, "acc_step": 0, "speed/wps": 12892.997234843659, "speed/FLOPS": 202502238408011.94, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05008900910615921, "optim/lr": 0.002739464135970071, "optim/total_tokens": 8113356800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 2.7607409954071045, "created_at": "2025-01-17T01:16:21.945824+00:00"} {"global_step": 15476, "acc_step": 0, "speed/wps": 12890.711684312888, "speed/FLOPS": 202466340696250.62, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04617771506309509, "optim/lr": 0.0027394158777045074, "optim/total_tokens": 8113881088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 2.8452887535095215, "created_at": "2025-01-17T01:16:32.117595+00:00"} {"global_step": 15477, "acc_step": 0, "speed/wps": 12892.696372424032, "speed/FLOPS": 202497512950283.4, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052177052944898605, "optim/lr": 0.0027393676153951332, "optim/total_tokens": 8114405376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 2.762394905090332, "created_at": "2025-01-17T01:16:42.288332+00:00"} {"global_step": 15478, "acc_step": 0, "speed/wps": 12892.087549594262, "speed/FLOPS": 202487950551131.62, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05632060766220093, "optim/lr": 0.0027393193490421043, "optim/total_tokens": 8114929664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 2.8763880729675293, "created_at": "2025-01-17T01:16:52.459748+00:00"} {"global_step": 15479, "acc_step": 0, "speed/wps": 12894.43540778578, "speed/FLOPS": 202524826890324.7, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05604061111807823, "optim/lr": 0.0027392710786455794, "optim/total_tokens": 8115453952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.85884428024292, "created_at": "2025-01-17T01:17:02.630460+00:00"} {"global_step": 15480, "acc_step": 0, "speed/wps": 12895.038805979826, "speed/FLOPS": 202534304088118.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05458958446979523, "optim/lr": 0.0027392228042057154, "optim/total_tokens": 8115978240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 2.8341150283813477, "created_at": "2025-01-17T01:17:12.800283+00:00"} {"global_step": 15481, "acc_step": 0, "speed/wps": 12889.7027279569, "speed/FLOPS": 202450493650228.5, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05841191112995148, "optim/lr": 0.0027391745257226694, "optim/total_tokens": 8116502528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 2.826240301132202, "created_at": "2025-01-17T01:17:22.977483+00:00"} {"global_step": 15482, "acc_step": 0, "speed/wps": 12886.252484293702, "speed/FLOPS": 202396302832366.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05215325579047203, "optim/lr": 0.0027391262431966, "optim/total_tokens": 8117026816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.817808151245117, "created_at": "2025-01-17T01:17:33.153079+00:00"} {"global_step": 15483, "acc_step": 0, "speed/wps": 12887.994208583319, "speed/FLOPS": 202423659005714.78, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0492849238216877, "optim/lr": 0.002739077956627664, "optim/total_tokens": 8117551104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 2.858638048171997, "created_at": "2025-01-17T01:17:43.329184+00:00"} {"global_step": 15484, "acc_step": 0, "speed/wps": 12889.4428200605, "speed/FLOPS": 202446411439564.7, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05682389438152313, "optim/lr": 0.002739029666016019, "optim/total_tokens": 8118075392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 2.8203811645507812, "created_at": "2025-01-17T01:17:53.503699+00:00"} {"global_step": 15485, "acc_step": 0, "speed/wps": 12894.32402453191, "speed/FLOPS": 202523077463266.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052398499101400375, "optim/lr": 0.0027389813713618225, "optim/total_tokens": 8118599680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287670, "loss/out": 2.6684927940368652, "created_at": "2025-01-17T01:18:03.672405+00:00"} {"global_step": 15486, "acc_step": 0, "speed/wps": 12892.691851268051, "speed/FLOPS": 202497441939316.0, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06796716898679733, "optim/lr": 0.002738933072665233, "optim/total_tokens": 8119123968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 2.789926528930664, "created_at": "2025-01-17T01:18:13.842347+00:00"} {"global_step": 15487, "acc_step": 0, "speed/wps": 12893.607276408913, "speed/FLOPS": 202511819949079.22, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046457141637802124, "optim/lr": 0.0027388847699264065, "optim/total_tokens": 8119648256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 2.7875490188598633, "created_at": "2025-01-17T01:18:24.011561+00:00"} {"global_step": 15488, "acc_step": 0, "speed/wps": 12891.507251137647, "speed/FLOPS": 202478836166456.88, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.04895887151360512, "optim/lr": 0.0027388364631455017, "optim/total_tokens": 8120172544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 2.800346612930298, "created_at": "2025-01-17T01:18:34.185665+00:00"} {"global_step": 15489, "acc_step": 0, "speed/wps": 12883.120726183672, "speed/FLOPS": 202347114267761.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.03930891677737236, "optim/lr": 0.002738788152322676, "optim/total_tokens": 8120696832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 2.796431064605713, "created_at": "2025-01-17T01:18:44.367864+00:00"} {"global_step": 15490, "acc_step": 0, "speed/wps": 12891.092758583349, "speed/FLOPS": 202472325991319.16, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04368705675005913, "optim/lr": 0.0027387398374580864, "optim/total_tokens": 8121221120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 2.841785430908203, "created_at": "2025-01-17T01:18:54.543153+00:00"} {"global_step": 15491, "acc_step": 0, "speed/wps": 12897.864271901353, "speed/FLOPS": 202578681913012.8, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.05245466157793999, "optim/lr": 0.0027386915185518912, "optim/total_tokens": 8121745408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8620738983154297, "created_at": "2025-01-17T01:19:04.711480+00:00"} {"global_step": 15492, "acc_step": 0, "speed/wps": 12894.588095354022, "speed/FLOPS": 202527225058398.62, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05559708550572395, "optim/lr": 0.002738643195604248, "optim/total_tokens": 8122269696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424894, "loss/out": 2.626847982406616, "created_at": "2025-01-17T01:19:14.883039+00:00"} {"global_step": 15493, "acc_step": 0, "speed/wps": 12894.477957022767, "speed/FLOPS": 202525495184560.34, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048345841467380524, "optim/lr": 0.0027385948686153147, "optim/total_tokens": 8122793984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 2.9457437992095947, "created_at": "2025-01-17T01:19:25.051696+00:00"} {"global_step": 15494, "acc_step": 0, "speed/wps": 12895.214222847411, "speed/FLOPS": 202537059251071.62, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05965528264641762, "optim/lr": 0.002738546537585249, "optim/total_tokens": 8123318272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.8025527000427246, "created_at": "2025-01-17T01:19:35.222994+00:00"} {"global_step": 15495, "acc_step": 0, "speed/wps": 12891.299301893008, "speed/FLOPS": 202475570037817.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049048855900764465, "optim/lr": 0.0027384982025142075, "optim/total_tokens": 8123842560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 2.9072163105010986, "created_at": "2025-01-17T01:19:45.396720+00:00"} {"global_step": 15496, "acc_step": 0, "speed/wps": 12890.21933320786, "speed/FLOPS": 202458607645584.88, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057452209293842316, "optim/lr": 0.0027384498634023488, "optim/total_tokens": 8124366848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.858431339263916, "created_at": "2025-01-17T01:19:55.569424+00:00"} {"global_step": 15497, "acc_step": 0, "speed/wps": 12894.697877589735, "speed/FLOPS": 202528949339267.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04816195368766785, "optim/lr": 0.0027384015202498303, "optim/total_tokens": 8124891136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 2.91494083404541, "created_at": "2025-01-17T01:20:05.739114+00:00"} {"global_step": 15498, "acc_step": 0, "speed/wps": 12894.369040940597, "speed/FLOPS": 202523784507823.47, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0533326119184494, "optim/lr": 0.0027383531730568104, "optim/total_tokens": 8125415424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 2.9116406440734863, "created_at": "2025-01-17T01:20:15.912833+00:00"} {"global_step": 15499, "acc_step": 0, "speed/wps": 12891.772970872318, "speed/FLOPS": 202483009660027.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05925559997558594, "optim/lr": 0.002738304821823446, "optim/total_tokens": 8125939712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 2.8081183433532715, "created_at": "2025-01-17T01:20:26.088082+00:00"} {"global_step": 15500, "acc_step": 0, "speed/wps": 12892.379006313635, "speed/FLOPS": 202492528279412.88, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.04682892933487892, "optim/lr": 0.002738256466549895, "optim/total_tokens": 8126464000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 2.8445582389831543, "created_at": "2025-01-17T01:20:36.261030+00:00"} {"global_step": 15501, "acc_step": 0, "speed/wps": 12896.094539820628, "speed/FLOPS": 202550885838818.28, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04432873800396919, "optim/lr": 0.0027382081072363155, "optim/total_tokens": 8126988288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 2.7769298553466797, "created_at": "2025-01-17T01:20:46.430631+00:00"} {"global_step": 15502, "acc_step": 0, "speed/wps": 12892.858664890156, "speed/FLOPS": 202500061976477.12, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04542629048228264, "optim/lr": 0.002738159743882865, "optim/total_tokens": 8127512576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 2.8881125450134277, "created_at": "2025-01-17T01:20:56.608310+00:00"} {"global_step": 15503, "acc_step": 0, "speed/wps": 12892.820113725853, "speed/FLOPS": 202499456477466.88, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04922688379883766, "optim/lr": 0.002738111376489702, "optim/total_tokens": 8128036864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.881380796432495, "created_at": "2025-01-17T01:21:06.779781+00:00"} {"global_step": 15504, "acc_step": 0, "speed/wps": 12895.027953866864, "speed/FLOPS": 202534133640771.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04342033714056015, "optim/lr": 0.0027380630050569833, "optim/total_tokens": 8128561152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289738, "loss/out": 2.8661017417907715, "created_at": "2025-01-17T01:21:16.947722+00:00"} {"global_step": 15505, "acc_step": 0, "speed/wps": 12891.544193198553, "speed/FLOPS": 202479416392287.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04149024933576584, "optim/lr": 0.0027380146295848673, "optim/total_tokens": 8129085440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484477, "loss/out": 2.844841480255127, "created_at": "2025-01-17T01:21:27.120730+00:00"} {"global_step": 15506, "acc_step": 0, "speed/wps": 12890.831535964784, "speed/FLOPS": 202468223131137.1, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.046848881989717484, "optim/lr": 0.0027379662500735115, "optim/total_tokens": 8129609728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.7491140365600586, "created_at": "2025-01-17T01:21:37.296342+00:00"} {"global_step": 15507, "acc_step": 0, "speed/wps": 12884.979093671698, "speed/FLOPS": 202376302482825.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05550934001803398, "optim/lr": 0.002737917866523074, "optim/total_tokens": 8130134016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.8963563442230225, "created_at": "2025-01-17T01:21:47.478857+00:00"} {"global_step": 15508, "acc_step": 0, "speed/wps": 12894.724226171558, "speed/FLOPS": 202529363179951.38, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.060269810259342194, "optim/lr": 0.0027378694789337125, "optim/total_tokens": 8130658304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 2.7933554649353027, "created_at": "2025-01-17T01:21:57.649986+00:00"} {"global_step": 15509, "acc_step": 0, "speed/wps": 12891.395896029248, "speed/FLOPS": 202477087181461.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0558924563229084, "optim/lr": 0.002737821087305585, "optim/total_tokens": 8131182592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 2.7232911586761475, "created_at": "2025-01-17T01:22:07.821569+00:00"} {"global_step": 15510, "acc_step": 0, "speed/wps": 12890.72376300843, "speed/FLOPS": 202466530408761.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07178229093551636, "optim/lr": 0.0027377726916388494, "optim/total_tokens": 8131706880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480757, "loss/out": 2.7676453590393066, "created_at": "2025-01-17T01:22:17.993260+00:00"} {"global_step": 15511, "acc_step": 0, "speed/wps": 12896.255024722295, "speed/FLOPS": 202553406474728.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07283272594213486, "optim/lr": 0.0027377242919336634, "optim/total_tokens": 8132231168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 2.703381061553955, "created_at": "2025-01-17T01:22:28.161568+00:00"} {"global_step": 15512, "acc_step": 0, "speed/wps": 12890.034719651347, "speed/FLOPS": 202455708035994.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047152623534202576, "optim/lr": 0.0027376758881901856, "optim/total_tokens": 8132755456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 2.7671892642974854, "created_at": "2025-01-17T01:22:38.334581+00:00"} {"global_step": 15513, "acc_step": 0, "speed/wps": 12887.530621743934, "speed/FLOPS": 202416377737367.16, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0670902207493782, "optim/lr": 0.0027376274804085726, "optim/total_tokens": 8133279744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.855116844177246, "created_at": "2025-01-17T01:22:48.524058+00:00"} {"global_step": 15514, "acc_step": 0, "speed/wps": 12892.069294277042, "speed/FLOPS": 202487663826289.25, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04574037343263626, "optim/lr": 0.0027375790685889843, "optim/total_tokens": 8133804032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.8740286827087402, "created_at": "2025-01-17T01:22:58.695530+00:00"} {"global_step": 15515, "acc_step": 0, "speed/wps": 12895.39887373082, "speed/FLOPS": 202539959446932.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04850661754608154, "optim/lr": 0.0027375306527315764, "optim/total_tokens": 8134328320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 2.8366403579711914, "created_at": "2025-01-17T01:23:08.864626+00:00"} {"global_step": 15516, "acc_step": 0, "speed/wps": 12893.331533046934, "speed/FLOPS": 202507489020668.25, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04285222291946411, "optim/lr": 0.002737482232836508, "optim/total_tokens": 8134852608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301523, "loss/out": 2.864337921142578, "created_at": "2025-01-17T01:23:19.033916+00:00"} {"global_step": 15517, "acc_step": 0, "speed/wps": 12890.836706401047, "speed/FLOPS": 202468304339943.53, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.0519600510597229, "optim/lr": 0.002737433808903938, "optim/total_tokens": 8135376896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 2.7430458068847656, "created_at": "2025-01-17T01:23:29.205857+00:00"} {"global_step": 15518, "acc_step": 0, "speed/wps": 12890.49335782354, "speed/FLOPS": 202462911578723.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06614671647548676, "optim/lr": 0.0027373853809340228, "optim/total_tokens": 8135901184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 2.7940521240234375, "created_at": "2025-01-17T01:23:39.377494+00:00"} {"global_step": 15519, "acc_step": 0, "speed/wps": 12889.642938071936, "speed/FLOPS": 202449554567925.25, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04736986756324768, "optim/lr": 0.0027373369489269205, "optim/total_tokens": 8136425472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 2.8215560913085938, "created_at": "2025-01-17T01:23:49.549755+00:00"} {"global_step": 15520, "acc_step": 0, "speed/wps": 12891.228727048014, "speed/FLOPS": 202474461562896.97, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06521274149417877, "optim/lr": 0.00273728851288279, "optim/total_tokens": 8136949760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 2.8094093799591064, "created_at": "2025-01-17T01:23:59.722679+00:00"} {"global_step": 15521, "acc_step": 0, "speed/wps": 12891.638058443814, "speed/FLOPS": 202480890674948.28, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04900682345032692, "optim/lr": 0.0027372400728017895, "optim/total_tokens": 8137474048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 2.874035120010376, "created_at": "2025-01-17T01:24:09.896647+00:00"} {"global_step": 15522, "acc_step": 0, "speed/wps": 12889.345420603871, "speed/FLOPS": 202444881647257.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07551822066307068, "optim/lr": 0.002737191628684076, "optim/total_tokens": 8137998336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.799072265625, "created_at": "2025-01-17T01:24:20.069876+00:00"} {"global_step": 15523, "acc_step": 0, "speed/wps": 12888.6669623032, "speed/FLOPS": 202434225527346.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.051864128559827805, "optim/lr": 0.0027371431805298086, "optim/total_tokens": 8138522624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 2.7927615642547607, "created_at": "2025-01-17T01:24:30.244040+00:00"} {"global_step": 15524, "acc_step": 0, "speed/wps": 12894.34564371008, "speed/FLOPS": 202523417022167.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06942771375179291, "optim/lr": 0.002737094728339144, "optim/total_tokens": 8139046912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 2.809206008911133, "created_at": "2025-01-17T01:24:40.414441+00:00"} {"global_step": 15525, "acc_step": 0, "speed/wps": 12890.415742090929, "speed/FLOPS": 202461692516987.53, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.050067514181137085, "optim/lr": 0.0027370462721122415, "optim/total_tokens": 8139571200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 2.935152053833008, "created_at": "2025-01-17T01:24:50.592123+00:00"} {"global_step": 15526, "acc_step": 0, "speed/wps": 12899.263619410161, "speed/FLOPS": 202600660588543.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0641087219119072, "optim/lr": 0.002736997811849259, "optim/total_tokens": 8140095488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 2.760125160217285, "created_at": "2025-01-17T01:25:00.757963+00:00"} {"global_step": 15527, "acc_step": 0, "speed/wps": 12892.800509885814, "speed/FLOPS": 202499148572220.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06914452463388443, "optim/lr": 0.002736949347550354, "optim/total_tokens": 8140619776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 2.9369845390319824, "created_at": "2025-01-17T01:25:10.927775+00:00"} {"global_step": 15528, "acc_step": 0, "speed/wps": 12893.39833216407, "speed/FLOPS": 202508538192594.47, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.0447729155421257, "optim/lr": 0.0027369008792156855, "optim/total_tokens": 8141144064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 2.710322141647339, "created_at": "2025-01-17T01:25:21.098845+00:00"} {"global_step": 15529, "acc_step": 0, "speed/wps": 12893.200385449883, "speed/FLOPS": 202505429167440.88, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.07986187189817429, "optim/lr": 0.002736852406845411, "optim/total_tokens": 8141668352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 2.7922987937927246, "created_at": "2025-01-17T01:25:31.271881+00:00"} {"global_step": 15530, "acc_step": 0, "speed/wps": 12888.12179691209, "speed/FLOPS": 202425662955742.88, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.05670136213302612, "optim/lr": 0.002736803930439689, "optim/total_tokens": 8142192640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 2.8068268299102783, "created_at": "2025-01-17T01:25:41.447668+00:00"} {"global_step": 15531, "acc_step": 0, "speed/wps": 12896.224336507712, "speed/FLOPS": 202552924474148.28, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07765533030033112, "optim/lr": 0.0027367554499986776, "optim/total_tokens": 8142716928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 2.8486247062683105, "created_at": "2025-01-17T01:25:51.620011+00:00"} {"global_step": 15532, "acc_step": 0, "speed/wps": 12892.499566888288, "speed/FLOPS": 202494421848905.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06025718152523041, "optim/lr": 0.002736706965522535, "optim/total_tokens": 8143241216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 2.851935386657715, "created_at": "2025-01-17T01:26:01.791907+00:00"} {"global_step": 15533, "acc_step": 0, "speed/wps": 12891.161402521204, "speed/FLOPS": 202473404138689.8, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.062416285276412964, "optim/lr": 0.002736658477011419, "optim/total_tokens": 8143765504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 2.931865692138672, "created_at": "2025-01-17T01:26:11.966397+00:00"} {"global_step": 15534, "acc_step": 0, "speed/wps": 12889.007139829451, "speed/FLOPS": 202439568482848.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05732012912631035, "optim/lr": 0.002736609984465488, "optim/total_tokens": 8144289792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.8630142211914062, "created_at": "2025-01-17T01:26:22.146056+00:00"} {"global_step": 15535, "acc_step": 0, "speed/wps": 12886.958997145992, "speed/FLOPS": 202407399587561.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.055209774523973465, "optim/lr": 0.0027365614878849, "optim/total_tokens": 8144814080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345728, "loss/out": 2.891965866088867, "created_at": "2025-01-17T01:26:32.322923+00:00"} {"global_step": 15536, "acc_step": 0, "speed/wps": 12887.216962788865, "speed/FLOPS": 202411451292465.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05080493539571762, "optim/lr": 0.0027365129872698145, "optim/total_tokens": 8145338368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.745035171508789, "created_at": "2025-01-17T01:26:42.502893+00:00"} {"global_step": 15537, "acc_step": 0, "speed/wps": 12889.076015346238, "speed/FLOPS": 202440650267484.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.041471581906080246, "optim/lr": 0.0027364644826203877, "optim/total_tokens": 8145862656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.792860507965088, "created_at": "2025-01-17T01:26:52.677664+00:00"} {"global_step": 15538, "acc_step": 0, "speed/wps": 12890.268892361975, "speed/FLOPS": 202459386040201.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.048275839537382126, "optim/lr": 0.0027364159739367796, "optim/total_tokens": 8146386944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8079147338867188, "created_at": "2025-01-17T01:27:02.849563+00:00"} {"global_step": 15539, "acc_step": 0, "speed/wps": 12894.027993211714, "speed/FLOPS": 202518427884592.66, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04275369644165039, "optim/lr": 0.0027363674612191475, "optim/total_tokens": 8146911232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332493, "loss/out": 2.6419315338134766, "created_at": "2025-01-17T01:27:13.018773+00:00"} {"global_step": 15540, "acc_step": 0, "speed/wps": 12885.88225675995, "speed/FLOPS": 202390487900202.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.048976149410009384, "optim/lr": 0.0027363189444676497, "optim/total_tokens": 8147435520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 2.947890281677246, "created_at": "2025-01-17T01:27:23.196619+00:00"} {"global_step": 15541, "acc_step": 0, "speed/wps": 12887.810358604327, "speed/FLOPS": 202420771389158.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05326665937900543, "optim/lr": 0.0027362704236824453, "optim/total_tokens": 8147959808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.7825660705566406, "created_at": "2025-01-17T01:27:33.371523+00:00"} {"global_step": 15542, "acc_step": 0, "speed/wps": 12893.207425214945, "speed/FLOPS": 202505539736624.8, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04812421649694443, "optim/lr": 0.002736221898863692, "optim/total_tokens": 8148484096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497956, "loss/out": 2.687443494796753, "created_at": "2025-01-17T01:27:43.547053+00:00"} {"global_step": 15543, "acc_step": 0, "speed/wps": 12891.578872922122, "speed/FLOPS": 202479961085000.7, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04429977387189865, "optim/lr": 0.0027361733700115476, "optim/total_tokens": 8149008384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 2.9047505855560303, "created_at": "2025-01-17T01:27:53.720712+00:00"} {"global_step": 15544, "acc_step": 0, "speed/wps": 12894.14008251803, "speed/FLOPS": 202520188401175.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048385024070739746, "optim/lr": 0.0027361248371261717, "optim/total_tokens": 8149532672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 2.8166444301605225, "created_at": "2025-01-17T01:28:03.891017+00:00"} {"global_step": 15545, "acc_step": 0, "speed/wps": 12890.486130933969, "speed/FLOPS": 202462798070491.88, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05039152130484581, "optim/lr": 0.002736076300207722, "optim/total_tokens": 8150056960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.8588919639587402, "created_at": "2025-01-17T01:28:14.066229+00:00"} {"global_step": 15546, "acc_step": 0, "speed/wps": 12886.77617615311, "speed/FLOPS": 202404528132646.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053915005177259445, "optim/lr": 0.0027360277592563564, "optim/total_tokens": 8150581248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 2.666815757751465, "created_at": "2025-01-17T01:28:24.240810+00:00"} {"global_step": 15547, "acc_step": 0, "speed/wps": 12891.94794343487, "speed/FLOPS": 202485757844559.44, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06153951212763786, "optim/lr": 0.002735979214272234, "optim/total_tokens": 8151105536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291904, "loss/out": 2.731001853942871, "created_at": "2025-01-17T01:28:34.413028+00:00"} {"global_step": 15548, "acc_step": 0, "speed/wps": 12893.090971244023, "speed/FLOPS": 202503710666986.97, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.051542963832616806, "optim/lr": 0.0027359306652555127, "optim/total_tokens": 8151629824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 2.7976326942443848, "created_at": "2025-01-17T01:28:44.582699+00:00"} {"global_step": 15549, "acc_step": 0, "speed/wps": 12891.506820124569, "speed/FLOPS": 202478829396804.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053072601556777954, "optim/lr": 0.0027358821122063517, "optim/total_tokens": 8152154112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.8717596530914307, "created_at": "2025-01-17T01:28:54.753826+00:00"} {"global_step": 15550, "acc_step": 0, "speed/wps": 12896.827814688366, "speed/FLOPS": 202562402928241.38, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05161352455615997, "optim/lr": 0.002735833555124908, "optim/total_tokens": 8152678400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.0021603107452393, "created_at": "2025-01-17T01:29:04.920584+00:00"} {"global_step": 15551, "acc_step": 0, "speed/wps": 12892.425953111171, "speed/FLOPS": 202493265643379.84, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05167045071721077, "optim/lr": 0.0027357849940113413, "optim/total_tokens": 8153202688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 2.702712059020996, "created_at": "2025-01-17T01:29:15.093630+00:00"} {"global_step": 15552, "acc_step": 0, "speed/wps": 12890.481801415866, "speed/FLOPS": 202462730069460.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0498296357691288, "optim/lr": 0.002735736428865809, "optim/total_tokens": 8153726976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 2.8373591899871826, "created_at": "2025-01-17T01:29:25.267526+00:00"} {"global_step": 15553, "acc_step": 0, "speed/wps": 12884.920312555241, "speed/FLOPS": 202375379244618.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.04604891315102577, "optim/lr": 0.0027356878596884703, "optim/total_tokens": 8154251264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.8250622749328613, "created_at": "2025-01-17T01:29:35.447907+00:00"} {"global_step": 15554, "acc_step": 0, "speed/wps": 12893.501410777748, "speed/FLOPS": 202510157183867.16, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05476701259613037, "optim/lr": 0.002735639286479484, "optim/total_tokens": 8154775552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434219, "loss/out": 2.772141695022583, "created_at": "2025-01-17T01:29:45.617392+00:00"} {"global_step": 15555, "acc_step": 0, "speed/wps": 12892.736346053156, "speed/FLOPS": 202498140791058.3, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.09491986036300659, "optim/lr": 0.002735590709239008, "optim/total_tokens": 8155299840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 2.869985818862915, "created_at": "2025-01-17T01:29:55.790538+00:00"} {"global_step": 15556, "acc_step": 0, "speed/wps": 12888.10627090442, "speed/FLOPS": 202425419098457.44, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06297473609447479, "optim/lr": 0.0027355421279672003, "optim/total_tokens": 8155824128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 2.8533642292022705, "created_at": "2025-01-17T01:30:05.964229+00:00"} {"global_step": 15557, "acc_step": 0, "speed/wps": 12887.686993144971, "speed/FLOPS": 202418833765097.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05514078959822655, "optim/lr": 0.0027354935426642203, "optim/total_tokens": 8156348416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8672401905059814, "created_at": "2025-01-17T01:30:16.142599+00:00"} {"global_step": 15558, "acc_step": 0, "speed/wps": 12890.3096502319, "speed/FLOPS": 202460026198556.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04810650274157524, "optim/lr": 0.0027354449533302265, "optim/total_tokens": 8156872704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368810, "loss/out": 2.8332252502441406, "created_at": "2025-01-17T01:30:26.316323+00:00"} {"global_step": 15559, "acc_step": 0, "speed/wps": 12893.584085115786, "speed/FLOPS": 202511455697952.47, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05093982070684433, "optim/lr": 0.0027353963599653767, "optim/total_tokens": 8157396992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 2.773092269897461, "created_at": "2025-01-17T01:30:36.485833+00:00"} {"global_step": 15560, "acc_step": 0, "speed/wps": 12890.09353563512, "speed/FLOPS": 202456631821841.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05338031053543091, "optim/lr": 0.00273534776256983, "optim/total_tokens": 8157921280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 2.88436222076416, "created_at": "2025-01-17T01:30:46.663146+00:00"} {"global_step": 15561, "acc_step": 0, "speed/wps": 12886.612553568493, "speed/FLOPS": 202401958215114.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05322306975722313, "optim/lr": 0.0027352991611437447, "optim/total_tokens": 8158445568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.7494735717773438, "created_at": "2025-01-17T01:30:56.840782+00:00"} {"global_step": 15562, "acc_step": 0, "speed/wps": 12890.849421748244, "speed/FLOPS": 202468504051943.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.043322350829839706, "optim/lr": 0.00273525055568728, "optim/total_tokens": 8158969856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 2.7899441719055176, "created_at": "2025-01-17T01:31:07.017044+00:00"} {"global_step": 15563, "acc_step": 0, "speed/wps": 12891.37219312394, "speed/FLOPS": 202476714894762.4, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04673808440566063, "optim/lr": 0.002735201946200593, "optim/total_tokens": 8159494144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283675, "loss/out": 2.8928868770599365, "created_at": "2025-01-17T01:31:17.190706+00:00"} {"global_step": 15564, "acc_step": 0, "speed/wps": 12889.68517708256, "speed/FLOPS": 202450217989629.56, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.047321587800979614, "optim/lr": 0.002735153332683844, "optim/total_tokens": 8160018432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.9960341453552246, "created_at": "2025-01-17T01:31:27.366333+00:00"} {"global_step": 15565, "acc_step": 0, "speed/wps": 12892.53091057408, "speed/FLOPS": 202494914144561.5, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05108985677361488, "optim/lr": 0.002735104715137191, "optim/total_tokens": 8160542720, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 2.8219656944274902, "created_at": "2025-01-17T01:31:37.537571+00:00"} {"global_step": 15566, "acc_step": 0, "speed/wps": 12893.197998309817, "speed/FLOPS": 202505391674125.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.048636648803949356, "optim/lr": 0.002735056093560792, "optim/total_tokens": 8161067008, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309387, "loss/out": 2.9098153114318848, "created_at": "2025-01-17T01:31:47.708641+00:00"} {"global_step": 15567, "acc_step": 0, "speed/wps": 12887.41788615013, "speed/FLOPS": 202414607069951.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052925266325473785, "optim/lr": 0.0027350074679548067, "optim/total_tokens": 8161591296, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 2.7623605728149414, "created_at": "2025-01-17T01:31:57.884539+00:00"} {"global_step": 15568, "acc_step": 0, "speed/wps": 12896.927963970158, "speed/FLOPS": 202563975910331.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05626780167222023, "optim/lr": 0.0027349588383193926, "optim/total_tokens": 8162115584, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7434849739074707, "created_at": "2025-01-17T01:32:08.051473+00:00"} {"global_step": 15569, "acc_step": 0, "speed/wps": 12892.475173742816, "speed/FLOPS": 202494038721036.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05033554509282112, "optim/lr": 0.002734910204654709, "optim/total_tokens": 8162639872, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 2.8209543228149414, "created_at": "2025-01-17T01:32:18.221759+00:00"} {"global_step": 15570, "acc_step": 0, "speed/wps": 12890.095225072238, "speed/FLOPS": 202456658356772.47, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052999529987573624, "optim/lr": 0.002734861566960915, "optim/total_tokens": 8163164160, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349594, "loss/out": 2.7651729583740234, "created_at": "2025-01-17T01:32:28.397373+00:00"} {"global_step": 15571, "acc_step": 0, "speed/wps": 12891.04937916555, "speed/FLOPS": 202471644657953.03, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.04912367835640907, "optim/lr": 0.0027348129252381683, "optim/total_tokens": 8163688448, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 2.9155895709991455, "created_at": "2025-01-17T01:32:38.571380+00:00"} {"global_step": 15572, "acc_step": 0, "speed/wps": 12889.329392313815, "speed/FLOPS": 202444629900936.94, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05338592082262039, "optim/lr": 0.0027347642794866285, "optim/total_tokens": 8164212736, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.764845371246338, "created_at": "2025-01-17T01:32:48.751759+00:00"} {"global_step": 15573, "acc_step": 0, "speed/wps": 12893.392703728236, "speed/FLOPS": 202508449790275.4, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04928595945239067, "optim/lr": 0.0027347156297064536, "optim/total_tokens": 8164737024, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 2.7092394828796387, "created_at": "2025-01-17T01:32:58.927366+00:00"} {"global_step": 15574, "acc_step": 0, "speed/wps": 12886.610751767912, "speed/FLOPS": 202401929915359.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.054177820682525635, "optim/lr": 0.002734666975897803, "optim/total_tokens": 8165261312, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 2.8489861488342285, "created_at": "2025-01-17T01:33:09.106160+00:00"} {"global_step": 15575, "acc_step": 0, "speed/wps": 12895.054341605884, "speed/FLOPS": 202534548096473.6, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05235463008284569, "optim/lr": 0.0027346183180608353, "optim/total_tokens": 8165785600, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 2.846189260482788, "created_at": "2025-01-17T01:33:19.274191+00:00"} {"global_step": 15576, "acc_step": 0, "speed/wps": 12891.360594992198, "speed/FLOPS": 202476532730165.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0441698282957077, "optim/lr": 0.0027345696561957085, "optim/total_tokens": 8166309888, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 2.8070201873779297, "created_at": "2025-01-17T01:33:29.445840+00:00"} {"global_step": 15577, "acc_step": 0, "speed/wps": 12887.569469360116, "speed/FLOPS": 202416987892561.53, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.06578733026981354, "optim/lr": 0.002734520990302582, "optim/total_tokens": 8166834176, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 2.9469685554504395, "created_at": "2025-01-17T01:33:39.621848+00:00"} {"global_step": 15578, "acc_step": 0, "speed/wps": 12886.506841707022, "speed/FLOPS": 202400297865066.25, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05349466949701309, "optim/lr": 0.002734472320381615, "optim/total_tokens": 8167358464, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 2.7642056941986084, "created_at": "2025-01-17T01:33:49.797270+00:00"} {"global_step": 15579, "acc_step": 0, "speed/wps": 12891.05680859926, "speed/FLOPS": 202471761347418.56, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052001092582941055, "optim/lr": 0.0027344236464329655, "optim/total_tokens": 8167882752, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 2.711491346359253, "created_at": "2025-01-17T01:33:59.968495+00:00"} {"global_step": 15580, "acc_step": 0, "speed/wps": 12890.272755387545, "speed/FLOPS": 202459446714326.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04608329012989998, "optim/lr": 0.0027343749684567925, "optim/total_tokens": 8168407040, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.817473888397217, "created_at": "2025-01-17T01:34:10.143599+00:00"} {"global_step": 15581, "acc_step": 0, "speed/wps": 12892.76331850133, "speed/FLOPS": 202498564430420.9, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.0608498714864254, "optim/lr": 0.002734326286453255, "optim/total_tokens": 8168931328, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485528, "loss/out": 2.93914794921875, "created_at": "2025-01-17T01:34:20.313839+00:00"} {"global_step": 15582, "acc_step": 0, "speed/wps": 12889.331703652639, "speed/FLOPS": 202444666203689.28, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05748859420418739, "optim/lr": 0.002734277600422512, "optim/total_tokens": 8169455616, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 2.8397269248962402, "created_at": "2025-01-17T01:34:30.486544+00:00"} {"global_step": 15583, "acc_step": 0, "speed/wps": 12896.444724357698, "speed/FLOPS": 202556385968176.94, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06447145342826843, "optim/lr": 0.0027342289103647215, "optim/total_tokens": 8169979904, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.832822322845459, "created_at": "2025-01-17T01:34:40.658250+00:00"} {"global_step": 15584, "acc_step": 0, "speed/wps": 12895.45011275382, "speed/FLOPS": 202540764226198.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.052845560014247894, "optim/lr": 0.0027341802162800435, "optim/total_tokens": 8170504192, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.8915224075317383, "created_at": "2025-01-17T01:34:50.828868+00:00"} {"global_step": 15585, "acc_step": 0, "speed/wps": 12883.434882786503, "speed/FLOPS": 202352048528905.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05332813039422035, "optim/lr": 0.0027341315181686364, "optim/total_tokens": 8171028480, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 2.911522388458252, "created_at": "2025-01-17T01:35:01.008743+00:00"} {"global_step": 15586, "acc_step": 0, "speed/wps": 12873.560395616598, "speed/FLOPS": 202196956138934.94, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04985614866018295, "optim/lr": 0.0027340828160306585, "optim/total_tokens": 8171552768, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 2.7565340995788574, "created_at": "2025-01-17T01:35:11.194958+00:00"} {"global_step": 15587, "acc_step": 0, "speed/wps": 12888.765940264891, "speed/FLOPS": 202435780112244.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053617559373378754, "optim/lr": 0.00273403410986627, "optim/total_tokens": 8172077056, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 2.726759910583496, "created_at": "2025-01-17T01:35:21.368199+00:00"} {"global_step": 15588, "acc_step": 0, "speed/wps": 12885.143652743094, "speed/FLOPS": 202378887109169.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05151743069291115, "optim/lr": 0.0027339853996756285, "optim/total_tokens": 8172601344, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 2.8514997959136963, "created_at": "2025-01-17T01:35:31.544563+00:00"} {"global_step": 15589, "acc_step": 0, "speed/wps": 12889.016525195206, "speed/FLOPS": 202439715892914.72, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05010683462023735, "optim/lr": 0.0027339366854588934, "optim/total_tokens": 8173125632, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.8488924503326416, "created_at": "2025-01-17T01:35:41.719508+00:00"} {"global_step": 15590, "acc_step": 0, "speed/wps": 12881.74171347573, "speed/FLOPS": 202325454978221.47, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.053576890379190445, "optim/lr": 0.002733887967216224, "optim/total_tokens": 8173649920, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 2.8391528129577637, "created_at": "2025-01-17T01:35:51.899220+00:00"} {"global_step": 15591, "acc_step": 0, "speed/wps": 12889.745450198203, "speed/FLOPS": 202451164661734.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.047747500240802765, "optim/lr": 0.0027338392449477787, "optim/total_tokens": 8174174208, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 2.839096784591675, "created_at": "2025-01-17T01:36:02.074899+00:00"} {"global_step": 15592, "acc_step": 0, "speed/wps": 12879.413374414069, "speed/FLOPS": 202288885213784.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04944965988397598, "optim/lr": 0.0027337905186537175, "optim/total_tokens": 8174698496, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 2.7807629108428955, "created_at": "2025-01-17T01:36:12.260575+00:00"} {"global_step": 15593, "acc_step": 0, "speed/wps": 12881.849425905622, "speed/FLOPS": 202327146749945.3, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04304027184844017, "optim/lr": 0.002733741788334198, "optim/total_tokens": 8175222784, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 2.7490808963775635, "created_at": "2025-01-17T01:36:22.441311+00:00"} {"global_step": 15594, "acc_step": 0, "speed/wps": 12886.882566219581, "speed/FLOPS": 202406199134835.6, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.049349527806043625, "optim/lr": 0.0027336930539893795, "optim/total_tokens": 8175747072, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 2.839799404144287, "created_at": "2025-01-17T01:36:32.619597+00:00"} {"global_step": 15595, "acc_step": 0, "speed/wps": 12888.664648842103, "speed/FLOPS": 202434189191260.44, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04492108151316643, "optim/lr": 0.0027336443156194216, "optim/total_tokens": 8176271360, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315023, "loss/out": 2.8427226543426514, "created_at": "2025-01-17T01:36:42.794954+00:00"} {"global_step": 15596, "acc_step": 0, "speed/wps": 12891.27562414003, "speed/FLOPS": 202475198146170.4, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.05684753879904747, "optim/lr": 0.002733595573224483, "optim/total_tokens": 8176795648, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 2.9040870666503906, "created_at": "2025-01-17T01:36:52.966975+00:00"} {"global_step": 15597, "acc_step": 0, "speed/wps": 12889.725855707238, "speed/FLOPS": 202450856903328.44, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.046991508454084396, "optim/lr": 0.0027335468268047227, "optim/total_tokens": 8177319936, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 2.8118443489074707, "created_at": "2025-01-17T01:37:03.145303+00:00"} {"global_step": 15598, "acc_step": 0, "speed/wps": 12892.59517393286, "speed/FLOPS": 202495923488918.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.042870983481407166, "optim/lr": 0.0027334980763603, "optim/total_tokens": 8177844224, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411556, "loss/out": 2.7329399585723877, "created_at": "2025-01-17T01:37:13.319365+00:00"} {"global_step": 15599, "acc_step": 0, "speed/wps": 12889.017805930996, "speed/FLOPS": 202439736008630.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04607103392481804, "optim/lr": 0.002733449321891374, "optim/total_tokens": 8178368512, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 2.8830184936523438, "created_at": "2025-01-17T01:37:23.496554+00:00"} {"global_step": 15600, "acc_step": 0, "speed/wps": 12891.120919111314, "speed/FLOPS": 202472768291107.47, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05578695982694626, "optim/lr": 0.002733400563398103, "optim/total_tokens": 8178892800, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.7806315422058105, "created_at": "2025-01-17T01:37:33.672409+00:00"} {"global_step": 15601, "acc_step": 0, "speed/wps": 12889.725765986848, "speed/FLOPS": 202450855494146.44, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043721217662096024, "optim/lr": 0.0027333518008806467, "optim/total_tokens": 8179417088, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 2.846463203430176, "created_at": "2025-01-17T01:37:43.845609+00:00"} {"global_step": 15602, "acc_step": 0, "speed/wps": 12886.737485537245, "speed/FLOPS": 202403920443357.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.001, "optim/grad_norm": 0.05345042422413826, "optim/lr": 0.002733303034339164, "optim/total_tokens": 8179941376, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 2.900698661804199, "created_at": "2025-01-17T01:37:54.020146+00:00"} {"global_step": 15603, "acc_step": 0, "speed/wps": 12891.314077506753, "speed/FLOPS": 202475802109135.28, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04793217405676842, "optim/lr": 0.0027332542637738145, "optim/total_tokens": 8180465664, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416384, "loss/out": 2.858978271484375, "created_at": "2025-01-17T01:38:04.193226+00:00"} {"global_step": 15604, "acc_step": 0, "speed/wps": 12891.69284804732, "speed/FLOPS": 202481751220960.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05719141289591789, "optim/lr": 0.0027332054891847563, "optim/total_tokens": 8180989952, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 2.8909871578216553, "created_at": "2025-01-17T01:38:14.369172+00:00"} {"global_step": 15605, "acc_step": 0, "speed/wps": 12890.493498323516, "speed/FLOPS": 202462913785468.6, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.046069368720054626, "optim/lr": 0.00273315671057215, "optim/total_tokens": 8181514240, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 2.8401927947998047, "created_at": "2025-01-17T01:38:24.548349+00:00"} {"global_step": 15606, "acc_step": 0, "speed/wps": 12891.656199174817, "speed/FLOPS": 202481175600056.66, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052736442536115646, "optim/lr": 0.0027331079279361535, "optim/total_tokens": 8182038528, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 2.779850482940674, "created_at": "2025-01-17T01:38:34.719906+00:00"} {"global_step": 15607, "acc_step": 0, "speed/wps": 12893.67103074667, "speed/FLOPS": 202512821298561.3, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.001, "optim/grad_norm": 0.052300527691841125, "optim/lr": 0.0027330591412769263, "optim/total_tokens": 8182562816, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 2.782317638397217, "created_at": "2025-01-17T01:38:44.890259+00:00"} {"global_step": 15608, "acc_step": 0, "speed/wps": 12893.603362953558, "speed/FLOPS": 202511758482885.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04377890005707741, "optim/lr": 0.0027330103505946277, "optim/total_tokens": 8183087104, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 2.76823353767395, "created_at": "2025-01-17T01:38:55.060120+00:00"} {"global_step": 15609, "acc_step": 0, "speed/wps": 12891.574757588318, "speed/FLOPS": 202479896448028.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.047318004071712494, "optim/lr": 0.002732961555889417, "optim/total_tokens": 8183611392, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 2.5892510414123535, "created_at": "2025-01-17T01:39:05.233938+00:00"} {"global_step": 15610, "acc_step": 0, "speed/wps": 12892.160206458111, "speed/FLOPS": 202489091727019.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.04721352085471153, "optim/lr": 0.002732912757161453, "optim/total_tokens": 8184135680, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 2.9043283462524414, "created_at": "2025-01-17T01:39:15.406294+00:00"} {"global_step": 15611, "acc_step": 0, "speed/wps": 12890.755683697436, "speed/FLOPS": 202467031767046.47, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06117241084575653, "optim/lr": 0.0027328639544108952, "optim/total_tokens": 8184659968, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 2.666281223297119, "created_at": "2025-01-17T01:39:25.578045+00:00"} {"global_step": 15612, "acc_step": 0, "speed/wps": 12893.252090104805, "speed/FLOPS": 202506241260096.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.056598030030727386, "optim/lr": 0.002732815147637903, "optim/total_tokens": 8185184256, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 2.724032402038574, "created_at": "2025-01-17T01:39:35.750370+00:00"} {"global_step": 15613, "acc_step": 0, "speed/wps": 12888.545127837142, "speed/FLOPS": 202432311949638.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07322513312101364, "optim/lr": 0.002732766336842635, "optim/total_tokens": 8185708544, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.7975759506225586, "created_at": "2025-01-17T01:39:45.924696+00:00"} {"global_step": 15614, "acc_step": 0, "speed/wps": 12888.803436731441, "speed/FLOPS": 202436369045777.06, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04232748597860336, "optim/lr": 0.0027327175220252515, "optim/total_tokens": 8186232832, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 2.814298152923584, "created_at": "2025-01-17T01:39:56.100354+00:00"} {"global_step": 15615, "acc_step": 0, "speed/wps": 12883.460461217683, "speed/FLOPS": 202352450273315.72, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.06524571031332016, "optim/lr": 0.0027326687031859107, "optim/total_tokens": 8186757120, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 2.810389518737793, "created_at": "2025-01-17T01:40:06.283050+00:00"} {"global_step": 15616, "acc_step": 0, "speed/wps": 12892.703093929029, "speed/FLOPS": 202497618520755.7, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053176045417785645, "optim/lr": 0.0027326198803247726, "optim/total_tokens": 8187281408, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 2.7848353385925293, "created_at": "2025-01-17T01:40:16.453344+00:00"} {"global_step": 15617, "acc_step": 0, "speed/wps": 12895.036826959496, "speed/FLOPS": 202534273004884.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.07319773733615875, "optim/lr": 0.002732571053441996, "optim/total_tokens": 8187805696, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 2.9024980068206787, "created_at": "2025-01-17T01:40:26.629596+00:00"} {"global_step": 15618, "acc_step": 0, "speed/wps": 12894.38136867485, "speed/FLOPS": 202523978131830.0, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.07017345726490021, "optim/lr": 0.0027325222225377407, "optim/total_tokens": 8188329984, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 2.9134297370910645, "created_at": "2025-01-17T01:40:36.798655+00:00"} {"global_step": 15619, "acc_step": 0, "speed/wps": 12891.641666038575, "speed/FLOPS": 202480947337181.34, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057300616055727005, "optim/lr": 0.0027324733876121655, "optim/total_tokens": 8188854272, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 2.7963743209838867, "created_at": "2025-01-17T01:40:46.969461+00:00"} {"global_step": 15620, "acc_step": 0, "speed/wps": 12886.843430103529, "speed/FLOPS": 202405584448354.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.062056250870227814, "optim/lr": 0.00273242454866543, "optim/total_tokens": 8189378560, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 2.807779312133789, "created_at": "2025-01-17T01:40:57.143934+00:00"} {"global_step": 15621, "acc_step": 0, "speed/wps": 12891.555931002926, "speed/FLOPS": 202479600750634.2, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052998658269643784, "optim/lr": 0.0027323757056976934, "optim/total_tokens": 8189902848, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 2.863422155380249, "created_at": "2025-01-17T01:41:07.314776+00:00"} {"global_step": 15622, "acc_step": 0, "speed/wps": 12892.889399687603, "speed/FLOPS": 202500544708705.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06529546529054642, "optim/lr": 0.0027323268587091157, "optim/total_tokens": 8190427136, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472818, "loss/out": 2.855388641357422, "created_at": "2025-01-17T01:41:17.484574+00:00"} {"global_step": 15623, "acc_step": 0, "speed/wps": 12888.095946218547, "speed/FLOPS": 202425256935078.22, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05131484940648079, "optim/lr": 0.002732278007699855, "optim/total_tokens": 8190951424, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.89505672454834, "created_at": "2025-01-17T01:41:27.658878+00:00"} {"global_step": 15624, "acc_step": 0, "speed/wps": 12888.11231252926, "speed/FLOPS": 202425513990477.62, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.060725416988134384, "optim/lr": 0.002732229152670072, "optim/total_tokens": 8191475712, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 2.7230453491210938, "created_at": "2025-01-17T01:41:37.837170+00:00"} {"global_step": 15625, "acc_step": 0, "speed/wps": 12896.257769876413, "speed/FLOPS": 202553449591146.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06819374114274979, "optim/lr": 0.002732180293619925, "optim/total_tokens": 8192000000, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 2.886061191558838, "created_at": "2025-01-17T01:41:48.004510+00:00"} {"global_step": 15626, "acc_step": 0, "speed/wps": 12890.832489997927, "speed/FLOPS": 202468238115538.6, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05653160810470581, "optim/lr": 0.0027321314305495748, "optim/total_tokens": 8192524288, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340917, "loss/out": 2.8015966415405273, "created_at": "2025-01-17T01:41:58.179098+00:00"} {"global_step": 15627, "acc_step": 0, "speed/wps": 12889.007521098603, "speed/FLOPS": 202439574471204.62, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06900998204946518, "optim/lr": 0.0027320825634591793, "optim/total_tokens": 8193048576, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 2.8105154037475586, "created_at": "2025-01-17T01:42:08.352451+00:00"} {"global_step": 15628, "acc_step": 0, "speed/wps": 12889.740673755163, "speed/FLOPS": 202451089641133.22, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04175309091806412, "optim/lr": 0.002732033692348899, "optim/total_tokens": 8193572864, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 2.7588555812835693, "created_at": "2025-01-17T01:42:18.528041+00:00"} {"global_step": 15629, "acc_step": 0, "speed/wps": 12887.955282873216, "speed/FLOPS": 202423047623948.1, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06214722990989685, "optim/lr": 0.002731984817218893, "optim/total_tokens": 8194097152, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 2.6933510303497314, "created_at": "2025-01-17T01:42:28.704820+00:00"} {"global_step": 15630, "acc_step": 0, "speed/wps": 12889.048993401557, "speed/FLOPS": 202440225850711.44, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.001, "optim/grad_norm": 0.0431663915514946, "optim/lr": 0.0027319359380693206, "optim/total_tokens": 8194621440, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 2.860684394836426, "created_at": "2025-01-17T01:42:38.880218+00:00"} {"global_step": 15631, "acc_step": 0, "speed/wps": 12895.5412888552, "speed/FLOPS": 202542196272159.75, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06900075078010559, "optim/lr": 0.002731887054900341, "optim/total_tokens": 8195145728, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 2.865210771560669, "created_at": "2025-01-17T01:42:49.047848+00:00"} {"global_step": 15632, "acc_step": 0, "speed/wps": 12891.164909481615, "speed/FLOPS": 202473459220322.12, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05412069708108902, "optim/lr": 0.0027318381677121145, "optim/total_tokens": 8195670016, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 2.718916654586792, "created_at": "2025-01-17T01:42:59.218982+00:00"} {"global_step": 15633, "acc_step": 0, "speed/wps": 12887.890368265786, "speed/FLOPS": 202422028050836.38, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04542464762926102, "optim/lr": 0.0027317892765048, "optim/total_tokens": 8196194304, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 2.879166603088379, "created_at": "2025-01-17T01:43:09.395022+00:00"} {"global_step": 15634, "acc_step": 0, "speed/wps": 12890.766390476247, "speed/FLOPS": 202467199931720.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05634210258722305, "optim/lr": 0.0027317403812785576, "optim/total_tokens": 8196718592, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 2.758902072906494, "created_at": "2025-01-17T01:43:19.566992+00:00"} {"global_step": 15635, "acc_step": 0, "speed/wps": 12892.788857215797, "speed/FLOPS": 202498965551025.5, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.043691158294677734, "optim/lr": 0.002731691482033546, "optim/total_tokens": 8197242880, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 2.836005687713623, "created_at": "2025-01-17T01:43:29.737402+00:00"} {"global_step": 15636, "acc_step": 0, "speed/wps": 12888.72891379534, "speed/FLOPS": 202435198560660.3, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05404653027653694, "optim/lr": 0.002731642578769926, "optim/total_tokens": 8197767168, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 2.8297243118286133, "created_at": "2025-01-17T01:43:39.912290+00:00"} {"global_step": 15637, "acc_step": 0, "speed/wps": 12891.943484212547, "speed/FLOPS": 202485687806345.22, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.001, "optim/grad_norm": 0.055194173008203506, "optim/lr": 0.0027315936714878554, "optim/total_tokens": 8198291456, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.854848861694336, "created_at": "2025-01-17T01:43:50.082792+00:00"} {"global_step": 15638, "acc_step": 0, "speed/wps": 12893.51492162224, "speed/FLOPS": 202510369390245.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0503416545689106, "optim/lr": 0.0027315447601874952, "optim/total_tokens": 8198815744, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.922077178955078, "created_at": "2025-01-17T01:44:00.252423+00:00"} {"global_step": 15639, "acc_step": 0, "speed/wps": 12890.816398972669, "speed/FLOPS": 202467985383875.53, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05403986945748329, "optim/lr": 0.002731495844869004, "optim/total_tokens": 8199340032, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 2.7378196716308594, "created_at": "2025-01-17T01:44:10.430123+00:00"} {"global_step": 15640, "acc_step": 0, "speed/wps": 12893.051580080491, "speed/FLOPS": 202503091974634.56, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04176592454314232, "optim/lr": 0.002731446925532542, "optim/total_tokens": 8199864320, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 2.8144993782043457, "created_at": "2025-01-17T01:44:20.602016+00:00"} {"global_step": 15641, "acc_step": 0, "speed/wps": 12891.400253320682, "speed/FLOPS": 202477155618711.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05170656368136406, "optim/lr": 0.002731398002178269, "optim/total_tokens": 8200388608, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.7992238998413086, "created_at": "2025-01-17T01:44:30.775782+00:00"} {"global_step": 15642, "acc_step": 0, "speed/wps": 12892.697973962806, "speed/FLOPS": 202497538104650.6, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04499552398920059, "optim/lr": 0.002731349074806345, "optim/total_tokens": 8200912896, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 2.727173328399658, "created_at": "2025-01-17T01:44:40.947814+00:00"} {"global_step": 15643, "acc_step": 0, "speed/wps": 12889.765567727378, "speed/FLOPS": 202451480635174.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.05449531227350235, "optim/lr": 0.0027313001434169275, "optim/total_tokens": 8201437184, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 2.8283848762512207, "created_at": "2025-01-17T01:44:51.120367+00:00"} {"global_step": 15644, "acc_step": 0, "speed/wps": 12888.99705686213, "speed/FLOPS": 202439410115991.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05319009721279144, "optim/lr": 0.0027312512080101786, "optim/total_tokens": 8201961472, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 2.8342225551605225, "created_at": "2025-01-17T01:45:01.295281+00:00"} {"global_step": 15645, "acc_step": 0, "speed/wps": 12891.74000477165, "speed/FLOPS": 202482491882115.88, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04787098243832588, "optim/lr": 0.002731202268586257, "optim/total_tokens": 8202485760, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 2.8261606693267822, "created_at": "2025-01-17T01:45:11.469686+00:00"} {"global_step": 15646, "acc_step": 0, "speed/wps": 12894.68053049459, "speed/FLOPS": 202528676879300.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05356667563319206, "optim/lr": 0.0027311533251453213, "optim/total_tokens": 8203010048, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 2.7329461574554443, "created_at": "2025-01-17T01:45:21.639237+00:00"} {"global_step": 15647, "acc_step": 0, "speed/wps": 12890.460499804121, "speed/FLOPS": 202462395498376.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04756736382842064, "optim/lr": 0.002731104377687533, "optim/total_tokens": 8203534336, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 2.873100519180298, "created_at": "2025-01-17T01:45:31.813090+00:00"} {"global_step": 15648, "acc_step": 0, "speed/wps": 12894.55430204014, "speed/FLOPS": 202526694287967.22, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04695689678192139, "optim/lr": 0.0027310554262130514, "optim/total_tokens": 8204058624, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 2.9416263103485107, "created_at": "2025-01-17T01:45:41.981678+00:00"} {"global_step": 15649, "acc_step": 0, "speed/wps": 12892.308445628354, "speed/FLOPS": 202491420026890.0, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05654459074139595, "optim/lr": 0.0027310064707220354, "optim/total_tokens": 8204582912, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397614, "loss/out": 2.800316333770752, "created_at": "2025-01-17T01:45:52.161041+00:00"} {"global_step": 15650, "acc_step": 0, "speed/wps": 12888.344982537588, "speed/FLOPS": 202429168392680.5, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05876224488019943, "optim/lr": 0.002730957511214645, "optim/total_tokens": 8205107200, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 2.890981674194336, "created_at": "2025-01-17T01:46:02.334550+00:00"} {"global_step": 15651, "acc_step": 0, "speed/wps": 12891.265106626246, "speed/FLOPS": 202475032954163.88, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05048024281859398, "optim/lr": 0.0027309085476910403, "optim/total_tokens": 8205631488, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 2.826439380645752, "created_at": "2025-01-17T01:46:12.510275+00:00"} {"global_step": 15652, "acc_step": 0, "speed/wps": 12892.75001710301, "speed/FLOPS": 202498355513682.28, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05365677550435066, "optim/lr": 0.002730859580151381, "optim/total_tokens": 8206155776, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 2.8856611251831055, "created_at": "2025-01-17T01:46:22.683170+00:00"} {"global_step": 15653, "acc_step": 0, "speed/wps": 12891.947354146574, "speed/FLOPS": 202485748588976.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04633171856403351, "optim/lr": 0.0027308106085958263, "optim/total_tokens": 8206680064, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425994, "loss/out": 2.6203083992004395, "created_at": "2025-01-17T01:46:32.854709+00:00"} {"global_step": 15654, "acc_step": 0, "speed/wps": 12891.833391054035, "speed/FLOPS": 202483958642006.94, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.057586487382650375, "optim/lr": 0.002730761633024537, "optim/total_tokens": 8207204352, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 2.8537840843200684, "created_at": "2025-01-17T01:46:43.025277+00:00"} {"global_step": 15655, "acc_step": 0, "speed/wps": 12887.286556727331, "speed/FLOPS": 202412544360900.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05776430666446686, "optim/lr": 0.0027307126534376717, "optim/total_tokens": 8207728640, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 2.9128429889678955, "created_at": "2025-01-17T01:46:53.204067+00:00"} {"global_step": 15656, "acc_step": 0, "speed/wps": 12896.356039600509, "speed/FLOPS": 202554993052199.66, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.044534407556056976, "optim/lr": 0.0027306636698353906, "optim/total_tokens": 8208252928, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 2.875372886657715, "created_at": "2025-01-17T01:47:03.373258+00:00"} {"global_step": 15657, "acc_step": 0, "speed/wps": 12892.446802979654, "speed/FLOPS": 202493593119215.1, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.05676976963877678, "optim/lr": 0.0027306146822178544, "optim/total_tokens": 8208777216, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377259, "loss/out": 2.7472519874572754, "created_at": "2025-01-17T01:47:13.544777+00:00"} {"global_step": 15658, "acc_step": 0, "speed/wps": 12888.61341533435, "speed/FLOPS": 202433384498618.97, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04510679841041565, "optim/lr": 0.002730565690585222, "optim/total_tokens": 8209301504, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 2.977280855178833, "created_at": "2025-01-17T01:47:23.717907+00:00"} {"global_step": 15659, "acc_step": 0, "speed/wps": 12888.252294750639, "speed/FLOPS": 202427712603620.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.0484725646674633, "optim/lr": 0.002730516694937653, "optim/total_tokens": 8209825792, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 2.755765438079834, "created_at": "2025-01-17T01:47:33.892066+00:00"} {"global_step": 15660, "acc_step": 0, "speed/wps": 12893.545520163332, "speed/FLOPS": 202510849982380.4, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05240274593234062, "optim/lr": 0.0027304676952753086, "optim/total_tokens": 8210350080, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8646464347839355, "created_at": "2025-01-17T01:47:44.061604+00:00"} {"global_step": 15661, "acc_step": 0, "speed/wps": 12894.807372236492, "speed/FLOPS": 202530669103155.8, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04955459386110306, "optim/lr": 0.0027304186915983472, "optim/total_tokens": 8210874368, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 2.9192023277282715, "created_at": "2025-01-17T01:47:54.229956+00:00"} {"global_step": 15662, "acc_step": 0, "speed/wps": 12898.456816357875, "speed/FLOPS": 202587988637943.88, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.06518033891916275, "optim/lr": 0.00273036968390693, "optim/total_tokens": 8211398656, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 2.82086181640625, "created_at": "2025-01-17T01:48:04.398620+00:00"} {"global_step": 15663, "acc_step": 0, "speed/wps": 12890.0554094899, "speed/FLOPS": 202456032998339.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.050291553139686584, "optim/lr": 0.0027303206722012153, "optim/total_tokens": 8211922944, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 2.7423367500305176, "created_at": "2025-01-17T01:48:14.572194+00:00"} {"global_step": 15664, "acc_step": 0, "speed/wps": 12890.86364728666, "speed/FLOPS": 202468727483572.72, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04722921922802925, "optim/lr": 0.002730271656481364, "optim/total_tokens": 8212447232, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 2.812509775161743, "created_at": "2025-01-17T01:48:24.744969+00:00"} {"global_step": 15665, "acc_step": 0, "speed/wps": 12896.604312722706, "speed/FLOPS": 202558892522746.78, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04486597701907158, "optim/lr": 0.0027302226367475362, "optim/total_tokens": 8212971520, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 2.824282646179199, "created_at": "2025-01-17T01:48:34.914895+00:00"} {"global_step": 15666, "acc_step": 0, "speed/wps": 12892.80608819861, "speed/FLOPS": 202499236187288.4, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05477377399802208, "optim/lr": 0.0027301736129998915, "optim/total_tokens": 8213495808, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.876457929611206, "created_at": "2025-01-17T01:48:45.084746+00:00"} {"global_step": 15667, "acc_step": 0, "speed/wps": 12891.4017022093, "speed/FLOPS": 202477178375497.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04807887226343155, "optim/lr": 0.0027301245852385902, "optim/total_tokens": 8214020096, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 2.7784314155578613, "created_at": "2025-01-17T01:48:55.255855+00:00"} {"global_step": 15668, "acc_step": 0, "speed/wps": 12892.350423391506, "speed/FLOPS": 202492079345343.8, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.053956903517246246, "optim/lr": 0.002730075553463792, "optim/total_tokens": 8214544384, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471145, "loss/out": 2.7935526371002197, "created_at": "2025-01-17T01:49:05.427171+00:00"} {"global_step": 15669, "acc_step": 0, "speed/wps": 12886.999328156338, "speed/FLOPS": 202408033041500.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.052695825695991516, "optim/lr": 0.0027300265176756562, "optim/total_tokens": 8215068672, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304572, "loss/out": 2.7656750679016113, "created_at": "2025-01-17T01:49:15.605727+00:00"} {"global_step": 15670, "acc_step": 0, "speed/wps": 12885.554908953085, "speed/FLOPS": 202385346453073.0, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.049703001976013184, "optim/lr": 0.002729977477874344, "optim/total_tokens": 8215592960, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 2.858315944671631, "created_at": "2025-01-17T01:49:25.783401+00:00"} {"global_step": 15671, "acc_step": 0, "speed/wps": 12893.050246572751, "speed/FLOPS": 202503071030063.1, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04764198139309883, "optim/lr": 0.0027299284340600144, "optim/total_tokens": 8216117248, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 2.880326271057129, "created_at": "2025-01-17T01:49:35.954790+00:00"} {"global_step": 15672, "acc_step": 0, "speed/wps": 12887.003730857548, "speed/FLOPS": 202408102191972.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.052156805992126465, "optim/lr": 0.0027298793862328284, "optim/total_tokens": 8216641536, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 2.7175302505493164, "created_at": "2025-01-17T01:49:46.130624+00:00"} {"global_step": 15673, "acc_step": 0, "speed/wps": 12892.519450901436, "speed/FLOPS": 202494734154655.4, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04256422072649002, "optim/lr": 0.002729830334392945, "optim/total_tokens": 8217165824, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 2.7284791469573975, "created_at": "2025-01-17T01:49:56.306138+00:00"} {"global_step": 15674, "acc_step": 0, "speed/wps": 12893.363172533996, "speed/FLOPS": 202507985962289.78, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05823356285691261, "optim/lr": 0.002729781278540525, "optim/total_tokens": 8217690112, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 2.927535057067871, "created_at": "2025-01-17T01:50:06.478621+00:00"} {"global_step": 15675, "acc_step": 0, "speed/wps": 12894.070402725974, "speed/FLOPS": 202519093984290.9, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045078668743371964, "optim/lr": 0.002729732218675728, "optim/total_tokens": 8218214400, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 2.71598744392395, "created_at": "2025-01-17T01:50:16.647400+00:00"} {"global_step": 15676, "acc_step": 0, "speed/wps": 12888.792580946643, "speed/FLOPS": 202436198540759.38, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05978422984480858, "optim/lr": 0.0027296831547987143, "optim/total_tokens": 8218738688, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499580, "loss/out": 2.810084104537964, "created_at": "2025-01-17T01:50:26.820564+00:00"} {"global_step": 15677, "acc_step": 0, "speed/wps": 12893.63286689815, "speed/FLOPS": 202512221882877.7, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05051886662840843, "optim/lr": 0.0027296340869096443, "optim/total_tokens": 8219262976, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 2.8166606426239014, "created_at": "2025-01-17T01:50:36.992300+00:00"} {"global_step": 15678, "acc_step": 0, "speed/wps": 12893.369817956334, "speed/FLOPS": 202508090337779.34, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.06207716837525368, "optim/lr": 0.002729585015008677, "optim/total_tokens": 8219787264, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 2.820002794265747, "created_at": "2025-01-17T01:50:47.163824+00:00"} {"global_step": 15679, "acc_step": 0, "speed/wps": 12897.906577813761, "speed/FLOPS": 202579346385501.6, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.05638669803738594, "optim/lr": 0.0027295359390959732, "optim/total_tokens": 8220311552, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 2.85786771774292, "created_at": "2025-01-17T01:50:57.329670+00:00"} {"global_step": 15680, "acc_step": 0, "speed/wps": 12890.59836056703, "speed/FLOPS": 202464560791098.78, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.0645846426486969, "optim/lr": 0.0027294868591716936, "optim/total_tokens": 8220835840, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 2.8120009899139404, "created_at": "2025-01-17T01:51:07.506584+00:00"} {"global_step": 15681, "acc_step": 0, "speed/wps": 12885.684514803366, "speed/FLOPS": 202387382091048.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.05680136755108833, "optim/lr": 0.0027294377752359973, "optim/total_tokens": 8221360128, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 2.8675742149353027, "created_at": "2025-01-17T01:51:17.683203+00:00"} {"global_step": 15682, "acc_step": 0, "speed/wps": 12891.613718155097, "speed/FLOPS": 202480508377266.9, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05078636482357979, "optim/lr": 0.0027293886872890457, "optim/total_tokens": 8221884416, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 2.5893115997314453, "created_at": "2025-01-17T01:51:27.858579+00:00"} {"global_step": 15683, "acc_step": 0, "speed/wps": 12894.117282693058, "speed/FLOPS": 202519830298593.78, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.054678890854120255, "optim/lr": 0.002729339595330997, "optim/total_tokens": 8222408704, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 2.891946315765381, "created_at": "2025-01-17T01:51:38.032419+00:00"} {"global_step": 15684, "acc_step": 0, "speed/wps": 12894.298786891204, "speed/FLOPS": 202522681071438.56, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.04517795890569687, "optim/lr": 0.0027292904993620136, "optim/total_tokens": 8222932992, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 2.8172335624694824, "created_at": "2025-01-17T01:51:48.201216+00:00"} {"global_step": 15685, "acc_step": 0, "speed/wps": 12891.415979742833, "speed/FLOPS": 202477402623781.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.04486420750617981, "optim/lr": 0.002729241399382254, "optim/total_tokens": 8223457280, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 2.7975986003875732, "created_at": "2025-01-17T01:51:58.376302+00:00"} {"global_step": 15686, "acc_step": 0, "speed/wps": 12888.95633444133, "speed/FLOPS": 202438770514414.3, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.038339197635650635, "optim/lr": 0.002729192295391879, "optim/total_tokens": 8223981568, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 2.6987807750701904, "created_at": "2025-01-17T01:52:08.551870+00:00"} {"global_step": 15687, "acc_step": 0, "speed/wps": 12885.920967017097, "speed/FLOPS": 202391095897984.8, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.045563969761133194, "optim/lr": 0.002729143187391049, "optim/total_tokens": 8224505856, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 2.8371095657348633, "created_at": "2025-01-17T01:52:18.730249+00:00"} {"global_step": 15688, "acc_step": 0, "speed/wps": 12893.22202448377, "speed/FLOPS": 202505769038203.22, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.04864810034632683, "optim/lr": 0.0027290940753799246, "optim/total_tokens": 8225030144, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485812, "loss/out": 2.7475380897521973, "created_at": "2025-01-17T01:52:28.900915+00:00"} {"global_step": 15689, "acc_step": 0, "speed/wps": 12891.84754318534, "speed/FLOPS": 202484180920676.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.05948198214173317, "optim/lr": 0.0027290449593586645, "optim/total_tokens": 8225554432, "memory/max_active_gib": 59.214274406433105, "memory/max_active_pct": 74.81178213269484, "memory/max_reserved_gib": 62.30859375, "memory/max_reserved_pct": 78.72116963934577, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 2.789217472076416, "created_at": "2025-01-17T01:52:39.071673+00:00"}